CROJFE - Volume 29, Issue 2 by Studio Bernardic

Issue 2

2008

Editorial

110 Years of University Forestry Education in the Republic of Croatia 1. International Scientific Conference »Challenges in Forestry and Wood Technology in the 21st century« On the occasion of the 110th anniversary of university forestry education in the Republic of Croatia, the International Scientific Conference »Challenges in Forestry and Wood Technology in the 21st Century« was held at the Faculty of Forestry, University of Zagreb, on October 17 2008. After the opening ceremony and introductory paper by the Faculty Dean, Assoc. Prof. Dr. Sc. Andrija Bogner, entitled »110 Years of University Forestry Education in the Republic of Croatia«, the activities were performed in two sections – Forestry Section and Wood Technology Section. 22 call for papers were presented by forestry and wood technology specialists, professors at the Faculties of Forestry from Croatia, several European countries and United States of America.

1.1 Forestry Section In the Forestry Section 10 papers were presented divided into four groups. Moderators of the first group of papers were Assoc. Prof. Dr. Sc. Igor Ani}, and Assist. Prof. Dr. Sc. Janez Kr~, and the following papers were presented: Þ Or{ani}, M., Pentek. T.: The Past, Present and Future of University Forestry Education in the Republic of Croatia Þ Mati}, S.: Treatments in the Forest Environment and Climate Change as Significant Factors Contributing to Forest Dieback and Degradation Þ Stampfer, K., Kanzian, C., Por{insky, T., Horvat, D.: Forest Biomass Utilization in Austria – State of the Art and Perspectives The second group of papers was modeled by Assoc. Prof. Dr. Sc. Karl Stampfer and Assist. Prof. Dr. Sc. Ivica Tikvi}, and the following papers were presented: Þ Diaci, J.: Close-to-Nature Silviculture as a Solution for Increased Societal Demands on Forests within a Changing Environment Þ Klimo, E.: The Effects of Norway Spruce Monocultures on Soil Properties and their Conversion to Mixed Forests Þ Kr~, J., Ko{ir, B., Poto~nik, I., Pentek, T., [u{njar, M.: Forestry Engineering in Central Europe – Present Status and Future Development Croat. j. for. eng. 29(2008)2

Assoc. Prof. Dr. Sc. Renata Pernar and Assoc. Prof. Dr. Sc. Davorin Kajba managed the third group of papers and namely: Þ Heinze, B.: Conservation of Genetic Resources and Breeding for an Uncertain Future: Support Offered by Molecular Biology Þ Dakskobler, I.: Phytocoenological Research in Forest Ecosystems at the Beginning of the 21st Century Þ Bon~ina, A., ^avlovi}, J.: Perspectives of Forest Management Planning Moderators of the last paper as well as of the discussion on all papers presented at the Conference were Assoc. Prof. Dr. Sc. Josip Margaleti} and Assist. Prof. Dr. Sc. Tomislav Por{insky. The last paper was: Þ U{~upli}, M., Dautba{i}, M., Tre{ti}, T., Mujezinovi}, O.: Current Health Status of Bosnia and Herzegovina Forests at the Beginning of the 21st Century The discussion was held by: Academician Slavko Mati}, Prof. Dr. Sc. Emil Klimo from Mendel University in Brno, Assoc. Prof. Dr. Sc. Karl Stampfer from BOKU in Vienna, Assist. Prof. Dr. Sc. Janez Kr~ from Biotechnical Faculty, University of Ljubljana, Prof. Dr. Sc. Faruk Meki} and Academician Midhat U{~upli} from Faculty of Forestry, University of Sarajevo.

1.2 Wood Technology Section In Wood Technology Section twelve papers were presented: Þ Ljuljka, B., Despot, R.: Foundation and Development of Wood Science and Technology Department at the Faculty of Forestry of the University of Zagreb Þ ^underlik, I.: Experience with the Bologna Process at the Faculty of Wood Sciences and Technology at the Technical University in Zvolen Þ Molnar, S.: Development Directions of Wood Sciences and Technology Þ Guzenda, R.: Current Problems of the Polish Woodworking Industry – Perspectives and Threats Þ Vlosky, R. P.: An Integrated Market-Based Methodology for Forest Products Sector Development Þ Petri~, M.: Implementation Process of the EU VOC Directive into the Slovenian Furniture Industry

109

T. PENTEK and T. POR[INSKY

110 Years of University Forestry Education in the Republic of Croatia (109–112)

Þ Trposki, Z.: Improvement of Output Parameters through a Decrease of Costs of the Bandsaw Þ Babiak, M.: Trends in Wood Properties Research Þ Horman, I.: Numerical Analysis of a Phenomenon in Wood Caused by Heat, Moisture or External Load Þ Rapp, O. A.: Quality Control of Thermally Modified Timber – a New Method for Testing TMT Þ Smardzewski, J.: Anthropotechnical Aspects of Furniture Design Þ Welzbacher, R. C., Brischke, C., Rapp, O. A.: Performance of Thermally Modified Timber (TMT) in Outdoor Applications – Durability, Abrasion and Optical Appearance As the Conference was held on the occasion of a very significant anniversary of the Faculty of Forestry of the University of Zagreb, the celebration of its 110th birthday, it is the right time to recall its famous past, to consider its present and to try to predict future events so as to get the Faculty ready for them.

2. Past, present and future of university forestry education in Croatia 2.1 Historical review of forestry education in Croatia Croatia makes part of the ancient Mediterranean civilization that has been developed in these regions for more than a thousand years. This is why the effects of reduced forest cover have been first observed in the Mediterranean parts of Croatia. This encouraged the residents of coastal regions to protect forests by the first terms of their town statutes, starting with the 12th century (Nin – 1103, Kor~ula – 1214, Split – 1240, Dubrovnik – 1272, Trogir – 1322, Krk – 1388, etc.). At that time, knowledge was transferred verbally, and also in writing. The beginning of forestry education in our country was first recorded with respect to forestry school of 1646 in Blato on the Island of Kor~ula (this is the time when the terms of the Venetian Senate applied for Istria, Kvarner and Dalmatia). The first forestry schools in Europe were established at the beginning of the 14th century and namely: 1807 in Würtenberg (Germany), 1813 in Mariabrunn (Austria), 1816 in Tharandt (Germany), 1824 in Nancy (France), 1828 in Stockholm (Sweden), 1846 in Bánska [tiavnica (Check), 1872 in Florence (Italy) and in Vienna (Austria), 1885 in Zurich (Switzerland), etc. The first forestry experts from Croatia were educated in Marianbrunn, Tharandt and Bánska [tiavnica. There were few of them in these schools, and yet they had a considerable intellectual, professional and patriotic role in the life of Croatia of that time. They greatly affected many events related to forestry profession and public life in Croatia, e.g. foundation of the Croatian-Slavonian Economy Society 1841 and within it the Forestry Depart-

110

ment, which gained independence in 1846 under the name Croatian-Slavonian Forestry Society. It has been active ever since and today it is known as the Croatian Forestry Society.

2.1.1 Forestry education at the School of Agriculture and Forestry in Krièvci (1860 – 1898) As a result of comprehensive activities of the members of the Croatian-Slavonian Forestry Society, and especially of Franjo [porer, Dragutin Kos and Ante Tomi}, the School of Agriculture and Forestry was established in Krièvci in 1860. Three development phases can be observed in the school activities: Þ Phase 1 of Krièvci School (1860–1877) – the aim was to provide scientific education and practical training for young people who were to work as foresters with land owners or municipals or to get a state job. Þ Phase 2 of Krièvci School (1877–1894) – in 1877 the School of Agriculture and Forestry in Krièvci was restructured for the first time, and in accordance with the new law it was named Royal School of Agriculture and Forestry in Krièvci. Þ Phase 3 of Krièvci School (1894–1898) – with the development of forestry science and the ever increasing need of forestry profession for properly qualified experts that would be acknowledged worldwide, education of forestry experts became an interesting topic of discussion. Forestry Department of School of Agriculture and Forestry was globally considered as secondary school level, and hence forestry experts decided that it was the right time to found a Academy of Forestry at the University of Zagreb. The Forest Act of 1894 played a significant role for the termination of activities of the Forestry Department of the Royal School of Agriculture and Forestry in Krièvci and for opening the Academy of Forestry at the Faculty of Philosophy of the University of Zagreb To be specific, according to paragraph 6 of this act on organization of the forest and technical department, all jobs above the 10th class of civil servants required academic education or university degree. All further activities were, therefore, focused on the foundation of a high-level forestry school – Academy of Forestry which was established in October 1898.

2.1.2 University-level forestry education at the Academy of Forestry within the Faculty of Philosophy in Zagreb (1898 – 1919) The Academy of Forestry was established on October 20 1898 (this date is still celebrated as the Day of the Faculty of Forestry of the University of Zagreb). With this Academy, forestry education in Croatia gained university status and it was the fourth high school of the University of Zagreb. Croat. j. for. eng. 29(2008)2

110 Years of University Forestry Education in the Republic of Croatia (109–112)

Several significant events for the University of Zagreb and forestry profession occurred that same day: opening ceremony of the newly built Forestry Center and Forestry Museum, regular 23rd General Meeting of the Croatian-Slavonian Forestry Society, and installation ceremony of the new rector of the University of Zagreb for the academic year 1898/99. At the beginning of the Academy of Forestry, when it was a three-year study and when the Academy was strictly connected with its parent body, Faculty of Philosophy, the dean of the Faculty of Philosophy was at the same time the dean of the Academy of Forestry. When the study became a four-year study and when a two-year Geodetic Course was established in 1908, the Academy became independent and started to choose its chairmen (who acted as deans) among teachers from the Academy of Forestry. After World War 1, in 1918 Croatian forestry was in a really difficult position. Political dependence on Austria and Hungary was terminated. To that time, forest administration was in the hands of foreigners, and in Croatia there was a serious lack of professional staff. There were even rumors among foresters that it was impossible to move on without foreign experts and that foreign forester who used to manage the state forests should be kept. At that time, Dr. Andrija Petra~i} and Dr. \uro Nenadi}, who taught at the Academy of Forestry, considered decisively that it was high time to gain independence and educate forestry experts with the highest qualification at a Croatian faculty. They immediately took measures to restructure the Academy of Forestry into an independent Department of the Faculty of Philosophy.

2.1.3 University-level forestry education at the Faculty of Agriculture and Forestry, University of Zagreb (1919 – 1960) On August 31 1919 the Decree was issued on establishment of the Faculty of Agriculture and Forestry of the University of Zagreb, which started its activities in the academic year 1919/20. The Faculty of Agriculture and Forestry consisted of two departments: Agriculture and Forestry, and the seat of the Forestry Department was in the building of the Forestry Society. On January 27 1920 Dr. Andrija Petra~i} and Dr. \uro Nenadi} were the first two regular professors appointed for the Forestry Department, and on March 18 and 19 the first two full professors were appointed for the Agriculture Department. On April 10 1920 the teaching council appointed Prof. Dr. Sc. Andrija Petra~i} the first dean of the Faculty of Agriculture and Forestry. The first teaching curriculum was approved on July 7 1921, and the subjects were divided as follows: mandatory subjects with obligatory graduation (state) exams, mandatory subjects with obligatory individual exams, mandatory subjects with obligatory lectures and/or practical training and optional subjects that were only recommended to students. Croat. j. for. eng. 29(2008)2

T. PENTEK and T. POR[INSKY

The first scientific basis was changed in the years to come, and however until 1947 the changes were only minor. More serious changes were introduced in the exam procedure so that all exams were taken individually. The development of the Faculty was neither harmonic nor without problems. The Faculty of Agriculture and Forestry had modest financial resources, and as early as 1926 there were signs in daily newspapers that the authorities planned to close the Faculty of Agriculture and Forestry in Zagreb. Such misgivings were confirmed by the fact that in 1926 and 1927 the competent Ministry of Education in Belgrade provided no funds for the Faculty and hence the number of assistants was reduced as part of economy measures. Pretty significant changes in the teaching curriculum occurred in 1947 and 1951/52. On March 11 1947 the teaching program was divided into two sections or groups at the Forestry Department (this was the beginning of the Wood-Technology Department): Forest Management (Biology Section – B) and Forest Industry (Technical Section– T). Biology Section was aimed at educating specialists for forest silviculture and management, and Technical Section at educating specialists for different technical operations in forestry as well as wood industry experts. The said teaching curriculum was developed at the request of the operating stuff who were the first to make complaints on this curriculum. Therefore numerous meetings were held within the Faculty, and then with the representatives of the operating stuff. In September 1951 an inter-faculty conference was held in Sarajevo between forestry faculties i.e. forestry departments on the territory of the former state. It was concluded that a forestry engineer had to acquire wide and comprehensive education in three groups of teaching subjects, and namely biological, technical and economic-organizational subjects, and at one or two faculties a special section was to be introduced for educating wood-industry experts. The slogan of the representatives of the forest operating stuff was »up to and from the axle of public traffic«. The Forestry Department of the Faculty of Agriculture and Forestry in Zagreb was divided into two sections: Forest Management Section and Wood Industry Section.

2.1.4 University-level forestry education at the Faculty of Forestry, University of Zagreb (1960–2006) Due to ever increasing development of all sectors of economy since 1945, and especially agriculture and forestry, an ever increasing need arose for separating agriculture and forestry into two independent faculties. Consequently, as of January 1 1960 both Departments of the then Faculty of Agriculture and Forestry became independent faculties pursuant to the Act announced in the Official Gazette »NN« of December 8 1959. The Faculty of Forestry of the University of Zagreb consisted of two departments: Forest Management De-

111

T. PENTEK and T. POR[INSKY

110 Years of University Forestry Education in the Republic of Croatia (109–112)

partment and Wood Industry Department. During the years, teaching curricula were changed (some subjects disappeared, some were divided into two new subjects, schedules of lectures, practical trainings and field trainings changed, etc.), as well as the way and conditions of taking exams, organizational structure within the Faculty (institutes, department chairs, seats) and names of departments, etc.

2.2 University forestry education at the Faculty of Forestry, University of Zagreb in accordance with the Bologna Declaration (from academic year 2005/2006) In the academic year 2006/07 a new way of studying started at the Faculty of Forestry of the University of Zagreb, and the new teaching plans and programs were created in accordance with the Bologna Declaration and the then efforts to restructure the university education in Croatia within a pretty strict framework that had to be observed. The Faculty of Forestry of the University of Zagreb consists today of the Forestry Department and Wood Technology Department. The Forestry Department is located in the new building of the Faculty of Forestry that has been almost completely finished after twelve years (the cornerstone of the new building of the Forestry Department was laid in 1996 on the Day of the Faculty of Forestry). Thus preconditions were created for further development of the Forestry Department of the Faculty of Forestry. With the aim of organizing and upgrading the teaching and scientific and research activities at the Forestry Department, the following institutes and pertaining laboratories have been established: Þ Institute of Ecology and Silviculture (Ecology and Pedology Laboratory and Laboratory for Forest Seeding and Nursery) Þ Institute of Forest Management and Remote Sensing (Laboratory for Measuring Forest Resources and Laboratory for Remote Sensing and GIS) Þ Institute of Forestry Genetics, Dendrology and Botanic (Laboratory for Molecular Biology and Physiology of Plants) Þ Institute of Forest Engineering (Laboratory for Technical and Technology Measurements in Forestry and Laboratory for Forest Biomass) Þ Institute of Forest Protection and Wildlife Management (Laboratory for Tree Pathology and Laboratory for Forestry Zoology) Þ Institute of Training and Research Forest Centers (Research Forest Facility Lipovljani, Research Forest Facility Velika, Research Forest Facility Zalesina, Research Forest Facility Rab, Research Forest Facility Zagreb, State Hunting Ground Opeke II/39 and State Hunting Ground Kalifront VIII/6).

112

2.3 Future of university forestry education in Croatia In order to provide in future the place that the university forestry education in the Republic of Croatia obviously deserves and this has always been a leadership position in the region and one of the leading positions among the university-level institutions in Europe, the following steps should be taken: Þ make a detailed analysis – objectively, critically and promptly, of the current system of university forestry education in the Republic of Croatia and determine problems and deficiencies; Þ define different possibilities for solving the determined problems, deficiencies, omissions and imperfections taking into consideration the European and global directives for university-level education and forestry, on one hand, and on the other trying to preserve a century-long tradition, with all the features specific for the Faculty of Forestry in Zagreb and the Croatian forestry profession; Þ select the most acceptable and the best solution, from all standpoints, of individually determined problems; Þ appoint responsible persons for obligations, activities and tasks, and set deadlines for their implementation; Þ establish and provide continuous monitoring of implementation of individual activities and define models and criteria for assessing their success.

4. Instead of conclusion 110 years of organized high-level forestry education at the University of Zagreb is a very significant accomplishment of which only few forestry faculties in Europe can boast of. University education and forestry profession in Croatia have been multiply, powerfully and permanently interconnected for more than a century. This is one of the key reasons of the actual beauty, naturality and quality of our Croatian low-lying, hilly, mountainous, Mediterranean and sub-Mediterranean forests. We must, we want and we wish to leave immeasurable natural resources that we inherited from our grand-grand-fathers, grand fathers and fathers to our children, grand children and grand-grand-children in a better state than they were when we took them over, because this is our moral and professional obligation. In doing so, we shall rely on the help of awe-inspiring forestry tradition, forestry tradition enriched with modern knowledge of forestry science and applied in forestry operations by forestry engineers and highly qualified forestry specialists and experts, who graduated and acquired their post-graduate diplomas from our Faculty of Forestry in Zagreb.

Tibor Pentek and Tomislav Por{insky Croat. j. for. eng. 29(2008)2

Uvodnik

110. obljetnica visoko{kolske {umarske nastave u Republici Hrvatskoj 1. Me|unarodno znanstveno savjetovanje »Izazovi u {umarstvu i drvnoj tehnologiji u 21. stolje}u« Na [umarskom je fakultetu Sveu~ili{ta u Zagrebu 17. listopada 2008. godine, povodom proslave 110. obljetnice visoko{kolske {umarske nastave u Republici Hrvatskoj, odr`ano me|unarodno znanstveno savjetovanje »Izazovi u {umarstvu i drvnoj tehnologiji u 21. stolje}u«. Nakon otvaranja savjetovanja i uvodnoga referata dekana Fakulteta izv. prof. dr. sc. Andrije Bognera pod naslovom »110 godina visoko{kolske {umarske i 60 godina visoko{kolske drvarske nastave u Republici Hrvatskoj« rad se odvijao u dvije sekcije, u [umarskoj sekciji i u Drvnotehnolo{koj sekciji. Podnesena su 22 pozivna referata {umarskih i drvnotehnolo{kih stru~njaka, profesora {umarskih fakulteta iz Hrvatske, nekoliko europskih zemalja i Sjedinjenih Ameri~kih Dr`ava.

1.1 [umarska sekcija U [umarskoj je sekciji izlo`eno deset referata podijeljenih u ~etiri skupine. Moderatori su prve skupine referata bili izv. prof. dr. sc. Igor Ani}, doc. dr. sc. Janez Kr~, a izlo`eni su ovi referati:

Þ Or{ani}, M., Pentek. T.: Pro{lost, sada{njost i budu}nost sveu~ili{ne {umarske nastave u Republici Hrvatskoj

Þ Mati}, S.: Zahvati u okoli{u {uma i klimatske promjene kao va`ni ~imbenici njihova su{enja i propadanja

Þ Stampfer, K., Kanzian, C., Por{insky, T., Horvat, D.: Uporaba {umske biomase u Austriji – pregled sada{njega stanja i mogu}a o~ekivanja. Drugu su skupinu referata modelirali izv. prof. dr. sc. Karl Stampfer i izv. prof. dr. sc. Ivica Tikvi}, a izneseni su ovi referati:

Þ Diaci, J.: Prirodno uzgajanje {uma kao odgovor na pove}ane potrebe za op}ekorisnim funkcijama {uma u promjenjivom okoli{u

Þ Klimo, E.: Utjecaj smrekovih monokultura na zna~ajke tla i njihova konverzija u mje{ovite {ume

Þ Kr~, J., Ko{ir, B., Poto~nik, I., Pentek, T., [u{njar, M.: [umarsko in`enjerstvo srednje Europe – stanje i budu}i razvoj. Croat. j. for. eng. 29(2008)2

Izv. prof. dr. sc. Renata Pernar i izv. prof. dr. sc. Davorin Kajba vodili su tre}u skupinu referata u kojoj smo ~uli ova izlaganja: Þ Heinze, B.: Za{tita genofonda i oplemenjivanje u budu}nosti: mogu}nosti molekularne biologije Þ Dakskobler, I.: Fitocenolo{ka istra`ivanja {umskih ekosustava na po~etku 21. stolje}a Þ Bon~ina, A., ^avlovi}, J.: Perspektive ure|ivanja {uma. Moderatori posljednjega referata i rasprave o svim izlaganjima na savjetovanju bili su izv. prof. dr. sc. Josip Margaleti} i doc. dr. sc. Tomislav Por{insky. Zadnji je referat bio: Þ U{~upli}, M., Dautba{i}, M., Tre{ti}, T., Mujezinovi}, O.: Aktualna slika zdravstvenoga stanja {uma Bosne i Hercegovine na po~etku 21. stolje}a. U raspravi su sudjelovali: akademik Slavko Mati}, prof. dr. sc. Emil Klimo s Mendelova Sveu~ili{ta u Brnu, izv. prof. dr. sc. Karl Stampfer s BOKU-a iz Be~a, doc. dr. sc. Janez Kr~ s Biotehni~koga fakulteta Sveu~ili{ta u Ljubljani, prof. dr. sc. Faruk Meki} i akademik Midhat U{~upli} sa [umarskoga fakulteta Sveu~ili{ta u Sarajevu.

1.2 Drvnotehnolo{ka sekcija U Drvnotehnolo{koj sekciji izloèno je dvanaest referata: Þ Ljuljka, B., Despot, R.: Osnivanje i razvoj drvnotehnolo{koga odsjeka na [umarskom fakultetu Sveu~ili{ta u Zagrebu Þ ûnderlik, I.: Iskustva »bolonjskoga« procesa na Fakultetu drvne tehnologije na Tehni~kom sveu~ili{tu u Zvolenu Þ Molnar, S.: Trendovi razvoja u drvnotehnolo{kom obrazovanju u Ma|arskoj Þ Guzenda, R.: Trenuta~ni problemi u poljskoj industrijskoj preradi drva – perspektive i opasnosti Þ Vlosky, R. P.: Integrirana, trì{no zasnovana metodologija za razvoj sektora drvnih proizvoda Þ Petri~, M.: Uvo|enje procesa EU VOC u proizvodnju namje{taja u Sloveniji Þ Trposki, Z.: Pobolj{anja izlaznih parametara smanjenjem tro{kova na tra~noj pili Þ Babiak, M.: Trendovi u istraìvanju drvnih svojstava

113

T. PENTEK i T. POR[INSKY

110. obljetnica visoko{kolske {umarske nastave u Republici Hrvatskoj (113–116)

Þ Horman, I.: Numeri~ka analiza pojava u drvu prouzro~enih toplinom, vlagom ili ostalim vanjskim utjecajima

Þ Rapp, O. A.: Kontrola kvalitete toplinski modificiranoga drva – nova metoda za testiranje TMT

Þ Smardzewski, J.: Antropometrijski aspekti u dizajniranju namje{taja

Þ Welzbacher, R. C., Brischke, C., Rapp, O. A.: Pona{anje toplinski modificiranoga drva (TMT) u vanjskoj primjeni – trajnost, tro{enje i izgled. Budu}i da je, kao {to je ve} istaknuto, savjetovanje uprili~eno u povodu vrlo vrijedne obljetnice [umarskoga fakulteta Sveu~ili{ta u Zagrebu, proslave njegova 110. ro|endana, treba se podsjetiti slavne pro{losti, razmotriti sada{njost te poku{ati predvidjeti doga|anja u budu}nosti ne bi li ju Fakultet do~ekao {to spremniji.

2. Pro{lost, sada{njost i budu}nost sveu~ili{ne {umarske nastave u Hrvatskoj 2.1 Povijesni pregled sveu~ili{ne {umarske nastave u Hrvatskoj Hrvatska je dio stare sredozemne civilizacije koja se na ovim prostorima razvija vi{e od tisu}u godina. Upravo su zato u sredozemnim dijelovima Hrvatske ponajprije uo~ene posljedice smanjenja {umskoga pokrova. To je potaklo ìtelje primorskih krajeva da u svojim prvim zakonskim odrednicama statuta gradova, po~ev{i od XII. stolje}a (Nin – 1103, Kor~ula – 1214, Split – 1240, Dubrovnik – 1272, Trogir – 1322, Krk – 1388 i dr.), {tite {ume. U tom su se vremenu znanja prenosila usmenom predajom, ali i pisanom rije~ju. Za~etak se naukovanja o {umarstvu u na{oj zemlji spominje u svezi sa {umarskom {kolom iz 1646. godine u Blatu na otoku Kor~uli (to je vrijeme kada su za Istru, Kvarner i Dalmaciju vrijedile zakonske odredbe Senata Venecije). Prve su {umarske {kole u Europi osnivane po~etkom XIX. st., i to: 1807. u Würtenbergu (Njema~ka), 1813. u Mariabrunnu (Austrija), 1816. u Tharandtu (Njema~ka), 1824. u Nancyju (Francuska) 1828. u Stockholmu ([vedska), 1846. u Bánskoj [tiavnici (ê{ka), 1872. u Firenzi (Italija) i u Be~u (Austrija), 1885. u Zürichu ([vicarska) itd. Prvi su se {umarski stru~njaci iz Hrvatske {kolovali u Marianbrunnu, Tharandtu i u Bánskoj [tiavnici. Njih je u tim {kolama bilo malo, ali su imali zna~ajnu intelektualnu, stru~nu i domoljubnu ulogu u ìvotu tada{nje Hrvatske. Imali su velik utjecaj na mnoga doga|anja vezana uz {umarsku struku i javni ìvot u Hrvatskoj, npr. osnivanje Hrvatsko-slavonskoga gospodarskoga dru{tva 1841. i u njem Odsjeka za {umarstvo, koji se 1846. godine osamostaljuje pod imenom Hrvatsko-slavonsko {umarsko dru{tvo i otada neprekidno djeluje, danas kao Hrvatsko {umarsko dru{tvo.

114

2.1.1 [umarska nastava na Gospodarsko-{umarskom u~ili{tu u Kri`evcima (1860 – 1898) Svestranim djelovanjem ~lanova Hrvatsko-slavonskoga {umarskoga dru{tva, a poglavito Franje [porera, Dragutina Kosa i Ante Tomi}a osniva se 1860. godine Gospodarsko-{umarsko u~ili{ta u Kri`evcima. U radu u~ili{ta tri su razvojne faze:

Þ Prvo razdoblje krièva~koga u~ili{ta (1860 – 1877) – cilj je znanstveno i prakti~no obrazovanje mladih ljudi koji }e obavljati slu`bu {umara kod vlastele ili op}ina ili se pak zaposliti u dràvnoj slu`bi. Þ Drugo razdoblje krièva~koga u~ili{ta (1877 – 1894) – 1877. godine prvi se put preustrojava Gospodarsko-{umarsko u~ili{te u Krièvcima, te se po novom zakonu ono naziva Kraljevsko gospodarsko i {umarsko u~ili{te u Krièvcih. Þ Tre}e razdoblje krièva~koga u~ili{ta (1894 – 1898) – razvojem {umarske znanosti i potrebom struke za sve kvalitetnijim {umarskim stru~nim osobljem koje }e biti priznato od svjetske stru~ne javnosti provedena je rasprava o {kolovanju {umarskih stru~njaka. [umarski odsjek Gospodarskoga i {umarskoga u~ili{ta imao je u svjetskim razmjerima razinu srednje {kole, te su {umarski stru~njaci ocijenili da bi bilo oportuno osnovati [umarsku akademiju pri Sveu~ili{tu u Zagrebu. Za prestanak je rada [umarskoga odjela na Kraljevskom gospodarskom i {umarskom u~ili{tu u Krièvcima te za otvaranje [umarske akademije pri Mudroslovnom (Filozofskom) fakultetu Sveu~ili{ta u Zagrebu zna~ajnu ulogu odigrao Zakon o {umama iz 1894. Naime, tim je zakonom o ure|enju {umsko-tehni~ke slu`be u paragrafu 6. odre|eno da se u toj slu`bi za radna mjesta vi{a od X. ~inovni~koga razreda traì akademsko obrazovanje, tj. zavr{ena visoka {umarska {kola. Stoga su sve daljnje aktivnosti bile usmjerene k osnivanju visoke {umarske {kole – [umarske akademije koja je zapo~ela raditi u listopadu 1898. godine.

2.1.2 Visoko{kolska {umarska nastava na [umarskoj akademiji pri Mudroslovnom fakultetu Sveu~ili{ta u Zagrebu (1898 – 1919) [umarska je akademija otvorena 20. listopada 1898. godine (i danas se na taj datum slavi Dan [umarskoga fakulteta Sveu~ili{ta u Zagrebu). Tim je ~inom {umarska nastava u Hrvatskoj dobila sveu~ili{ni status i utemeljena je kao ~etvrta u nizu na Zagreba~kom sveu~ili{tu. Isti se dan dogodilo, za Zagreba~ko sveu~ili{te i za {umarsku struku, nekoliko zna~ajnih doga|anja: sve~ano su otvoreni novoizgra|eni [umarski dom i [umarski muzej, odr`ana je redovita 23. glavna skup{tina Hrvatsko-slavonskoga {umarskoga dru{tva, a obavljeno je i ustoli~enje novoga rektora Sveu~ili{ta u Zagrebu za akad. god. 1898/99. Croat. j. for. eng. 29(2008)2

110. obljetnica visoko{kolske {umarske nastave u Republici Hrvatskoj (113–116)

U po~etku rada [umarske akademije, kada je studij trajao tri godine i kada je Akademija bila ~vrsto vezana uz mati~ni, Filozofski fakultet, dekan Filozofskoga fakulteta bio je istodobno i dekan [umarske akademije. Prelaskom na ~etverogodi{nji studij i otvaranjem dvogodi{njega Geodetskoga te~aja 1908. godine Akademija se osamostaljuje i bira pro~elnike (koji su imali ulogu dekana) iz redova nastavnika [umarske akademije. Nakon I. svjetskoga rata 1918. hrvatsko je {umarstvo u dosta te{kom poloàju. Raskinuta je politi~ka ovisnost o Austriji i Ugarskoj. Uprava je {uma dotad bila u rukama stranaca, a u Hrvatskoj nema dovoljno stru~noga kadra. U stru~nim su se {umarskim krugovima ~ula ~ak i razmi{ljanja da se ne moè bez stranih stru~njaka te da i dalje treba zadràti strane {umare koji su upravljali dràvnim {umama. U tom trenutku profesori [umarske akademije dr. Andrija Petra~i} i dr. \uro Nenadi} odlu~no zastupaju stajali{te kako se treba osoviti na vlastite noge te na vlastitom fakultetu odgajati {umarske stru~njake s najvi{om stru~nom spremom. Oni odmah pokre}u akciju da se [umarska akademija preustroji u samostalan odjel Filozofskoga fakulteta.

2.1.3 Visoko{kolska {umarska nastava na Gospodarsko (Poljoprivredno)-{umarskom fakultetu Sveu~ili{ta u Zagrebu (1919 – 1960) Dana 31. kolovoza 1919. potpisan je ukaz o osnivanju Gospodarsko-{umarskoga fakulteta Sveu~ili{ta u Zagrebu koji je svoj rad zapo~eo u akad. god. 1919/20. Gospodarsko-{umarski fakultet sastojao se od dva odjela: Gospodarskoga i [umarskoga, a sjedi{te je [umarskoga odjela bilo u zgradi [umarskoga dru{tva. Dana 27. sije~nja 1920. imenovana su prva dva redovita profesora za [umarski odjel: dr. Andrija Petra~i} i dr. \uro Nenadi}, a 18. i 19. oùjka prva dva redovita profesora za Gospodarski odjel. 10. travnja 1920. profesorski zbor izabrao je prvim dekanom Gospodarsko{umarskoga fakulteta prof. dr. sc. Andriju Petra~i}a. Prvi nastavni plan i program odobren je 7. srpnja 1921, a predmeti su podijeljeni kako slijedi: obvezni predmeti iz kojih se polaù diplomski (dràvni) ispiti, obvezni predmeti iz kojih se ispiti polaù pojedina~no, obvezni predmeti iz kojih se moraju polaziti samo predavanja i/ili vje`be te neobvezni predmeti koji se studentima samo preporu~uju. Prva se nau~na osnova tijekom godina mijenjala, ali su sve do 1947. godine promjene bile manjega zna~enja. Ve}e su promjene uvedene u na~in polaganja ispita tako da su svi ispiti polagani pojedina~no. Razvoj Fakulteta nije bio harmoni~an i bez problema. Gospodarsko-{umarski fakultet raspolagao je skromnim financijskim sredstvima, a ve} su se 1926. u dnevnim novinama pojavile naznake da vlast u Beogradu namjerava ukinuti Poljoprivredno-{umarski fakultet u Zagrebu. Takve su crne slutnje svoju potvrdu dobile u ~injenici da 1926. i 1927. godine nadle`no Ministarstvo prosvjete u Beogradu nije za Fakultet predCroat. j. for. eng. 29(2008)2

T. PENTEK i T. POR[INSKY

vidjelo nikakva sredstva pa je radi {tednje smanjen broj asistenata. Prili~no velike promjene u nastavnom planu dogodile su se 1947. i 1951/52. godine. 11. oùjka 1947. nastava je na [umarskom odjelu razdijeljena na dva smjera odnosno dvije grupe (to je i po~etak Drvnotehnolo{koga odsjeka): {umskouzgojni (biolo{ki – B) i {umskoindustrijski (tehni~ki – T). Biolo{ki je smjer trebao obrazovati stru~njake za uzgajanje {uma i upravljanje, a tehni~ki smjer za razli~ite tehni~ke radove u {umarstvu te stru~njake za drvnu industriju. Navedeni je nastavni plan nastao na izri~ito traènje operative koja se prva po~ela na taj plan i tuìti. Stoga su odràvani brojni sastanci unutar Fakulteta, a zatim i s predstavnicima operative. U rujnu 1951. u Sarajevu je odràna me|ufakultetska konferencija {umarskih fakulteta odnosno {umarskih odjela na podru~ju tada{nje dràve. Zaklju~eno je da {umarski inènjer mora ste}i {iroko i potpuno obrazovanje iz triju skupina nastavnih predmeta, iz biolo{kih, tehni~kih i ekonomskoorganizacijskih, a na jednom ili na dva fakulteta treba uvesti poseban odsjek za obrazovanje drvnoindustrijskih stru~njaka. Krilatica predstavnika {umarske operative bila je »do i od osovine javnoga prometa«. [umarski odjel Poljoprivredno-{umarskoga fakulteta u Zagrebu podijeljen je na dva odsjeka: [umskogospodarski odsjek i Drvnoindustrijski odsjek.

2.1.4 Visoko{kolska {umarska nastava na [umarskom fakultetu Sveu~ili{ta u Zagrebu (1960 – 2006) Zbog sve intenzivnijega razvoja svih gospodarskih grana od 1945. godine, a posebno poljoprivrede i {umarstva, sve je ja~a potreba za osamostaljenjem poljoprivredne i {umarske visoko{kolske nastave u zasebne fakultete. Oba dotada{nja odjela Poljoprivredno-{umarskoga fakulteta, Zakonom objavljenim u Narodnim novinama 8. prosinca 1959, od 1. sije~nja 1960. postaju samostalni fakulteti. [umarski fakultet Sveu~ili{ta u Zagrebu sastoji se od dva odsjeka: [umskogospodarskoga i Drvnoindustrijskoga odsjeka. Tijekom godina mijenjali su se nastavni planovi i programi (neki su predmeti nestajali, neki su se dijelili u dva nova predmeta, mijenjala se satnica predavanja, vje`bi i terenske nastave i dr.), na~in i uvjeti polaganja ispita, organizacijska struktura unutar Fakulteta (zavodi, katedre, stolice) i imena odsjeka itd.

2.2 Visoko{kolska {umarska nastava na [umarskom fakultetu Sveu~ili{ta u Zagrebu sukladno Bolonjskoj deklaraciji (od akad. god. 2005/2006) Akademske godine 2006/07. zapo~eo je nov na~in studiranja na [umarskom fakultetu Sveu~ili{ta u Zagrebu, a novi nastavni planovi i programi kreirani su u skladu s Bolonjskom deklaracijom i tada{njim nasto-

115

T. PENTEK i T. POR[INSKY

110. obljetnica visoko{kolske {umarske nastave u Republici Hrvatskoj (113–116)

janjem da se visoko{kolsko obrazovanje u Hrvatskoj preustroji unutar prili~no ~vrstih okvira kojih se trebalo pridràvati. [umarski se fakultet Sveu~ili{ta u Zagrebu danas sastoji od [umarskoga odsjeka i Drvnotehnolo{koga odsjeka. [umarski je odsjek smje{ten u novoj zgradi [umarskoga fakulteta koja je gotovo u potpunosti dovr{ena nakon punih dvanaest godina (1996. godine na Dan [umarskoga fakulteta postavljen je kamen temeljac nove zgrade [umarskoga odsjeka). Time su stvoreni preduvjeti za daljnji razvoj [umarskoga odsjeka [umarskoga fakulteta. Radi organiziranja i unapre|enja nastavnoga i znanstvenoistraìva~koga rada na [umarskom su odsjeku ustrojeni ovi zavodi i njima pripadaju}i laboratoriji: Þ Zavod za ekologiju i uzgajanje {uma (Ekolo{kopedolo{ki laboratorij i Laboratorij za {umsko sjemenarstvo i rasadni~arstvo) Þ Zavod za izmjeru i ure|ivanje {uma (Laboratorij za izmjeru {umskih resursa i Laboratorij za daljinska istraìvanja i GIS) Þ Zavod za {umarsku genetiku, dendrologiju i botaniku (Laboratorij za molekularnu biologiju i fiziologiju bilja) Þ Zavod za {umarske tehnike i tehnologije (Laboratorij za tehni~ko-tehnolo{ke izmjere u {umarstvu i Laboratorij za {umsku biomasu) Þ Zavod za za{titu {uma i lovno gospodarenje (Laboratorij za patologiju drve}a i Laboratorij za {umarsku zoologiju) Þ Zavod za nastavno-pokusne {umske objekte (NP[O Lipovljani, NP[O Velika, NP[O Zalesina, NP[O Rab, NP[O Zagreb, Dràvno lovi{te Opeke II/39 i Dràvno lovi{te Kalifront VIII/6).

2.3 Budu}nost sveu~ili{ne {umarske nastave u Hrvatskoj Kako bi visoko{kolska {umarska nastava u Republici Hrvatskoj i u budu}nosti bila na razini koja joj nesumnjivo pripada, a to je oduvijek bilo neprikosnoveno lidersko mjesto u regiji te jedno od vode}ih mjesta me|u visoko{kolskim {umarskim institucijama u Europi, potrebno je:

116

Þ objektivno, kriti~ki i odmah obaviti detaljnu

Þ Þ Þ

ra{~lambu sada{njega sustava sveu~ili{ne {umarske nastave u Republici Hrvatskoj te utvrditi probleme i manjkavosti definirati razli~ite mogu}nosti rje{enja uo~enih problema, nedostataka, propusta i manjkavosti uzimaju}i, s jedne strane, u obzir europske i svjetske smjernice u visoko{kolskom obrazovanju i u {umarstvu, a s druge strane, nastoje}i zadr`ati stoljetnu tradiciju, posebnost i prepoznatljivost zagreba~koga [umarskoga fakulteta i doma}e {umarske struke odabrati, sa svih stajali{ta, najprihvatljiviju i najbolju ina~icu rje{enja pojedinoga determiniranoga problema imenovati nositelje obveza, aktivnosti i zadataka te definirati rokove njihova izvr{enja utvrditi i osigurati stalno pra}enje provedbe pojedine aktivnosti te definirati modele i kriterije prosudbe njihove uspje{nosti.

4. Umjesto zaklju~ka 110 godina organizirane visoko{kolske {umarske nastave na Sveu~ili{tu u Zagrebu vrlo je vrijedan doseg kojim se malo koji {umarski fakultet u Europi moè podi~iti. Sveu~ili{no obrazovanje i {umarska struka u Hrvatskoj su vi{estruko, sna`no i neraskidivo povezani vi{e od stolje}a. To je jedan od klju~nih razloga dana{nje ljepote, prirodnosti i kakvo}e na{ih hrvatskih nizinskih, prigorsko-brdskih, gorskih, mediteranskih i submediteranskih {uma. Neizmjerno vrijedno prirodno bogatstvo koje smo u naslje|e dobili od svojih pradjedova, djedova i o~eva moramo, èlimo i ho}emo ostaviti svojim sinovima, unucima i praunucima u jo{ boljem stanju nego {to smo ga preuzeli jer je to na{a moralna i stru~na obveza. U tom }e nam svakako od neizmjerne pomo}i biti {umarska tradicija vrijedna strahopo{tovanja, {umarska tradicija oplemenjena suvremenim spoznajama {umarske znanosti i primijenjena u {umarskoj operativi djelovanjem inènjera, magistara struke, magistara specijalista i doktora znanosti poniklih na na{em hrvatskom, zagreba~kom [umarskom fakultetu.

Tibor Pentek i Tomislav Por{insky

Croat. j. for. eng. 29(2008)2

Original scientific paper – Izvorni znanstveni rad

Evaluation of the Feller-Buncher Moipu 400E for Energy Wood Harvesting Christian Rottensteiner, Günter Affenzeller, Karl Stampfer Abstract – Nacrtak Proper tending operations in young stands increase the quality of valuable roundwood and reduce the risk of stand damages caused by wind and snow-breaks, and infestation of bark beetles. When felling and extracting small diameter trees, costs often exceed the potential revenues. Mechanized thinning performed by using a forwarder mounted feller-buncher head could improve this cost-effectiveness. A time study was carried out in a 35–40 year old Scots Pine–Oak dominated stand. Productivity and costs were investigated of a Timberjack 1110D forwarder equipped with the felling-bunching head Moipu 400E. Further objectives were to give practical recommendations for the system in the field. The harvesting productivity was 4.11 m3/PSH0 (effective working hour) or 3.16 m3/PSH15 with an average tree volume of 0.057 m3, an average load volume of 3.71 m³, and the average forwarding distance 89 m. The supply costs from forest to plant (felling, forwarding, chipping, and transportation) were 91.60 /PSH15 or 77.84 per oven dry ton. In Austria it is possible to achieve revenues of 78.00 per oven dry ton. Therefore it is possible to gain profit. The feller-buncher head Moipu 400E is best suited to cut Pine trees up to a maximum diameter at the butt of about 30 cm, and Oak and Beech up to 25 cm. In order to keep the felling-bunching costs at a reasonable level, mechanized harvesting should be done at sites where the average volume of removed trees is over 0.05 m³ per tree. Keywords: energy wood harvesting, thinning, felling head Moipu 400E, productivity, costs

1. Introduction – Uvod Careful management and proper tending operations in young stands increase the quality of valuable roundwood and reduce the risk of stand damages caused by wind and snow-breaks, and infestation of bark beetles. According to the Austrian Forest Inventory (Österreichische Waldinventur 2004) there was a decline in tending and thinning operations. There is a potential amounting to 64 million m3 over bark available for thinning. Eberhardinger (2007) also describes a decrease in thinning utilisation in Germany. During the 1990s harvesting in Central Europe underwent considerable development by introducing the highly mechanized harvester technology. ParCroat. j. for. eng. 29(2008)2

ticularly when harvesting small-sized trees, harvesters have higher efficiencies than chainsaws used for felling and processing. Nevertheless it is difficult to achieve a positive contribution margin in first thinnings, because of higher harvesting costs and lower price for small-diameter timber (Affenzeller and Stampfer 2007). Another option to make use of the biomass that accrues in precommercial thinning is chipping to biomass fuel for power plants. Delimbing and bucking is not necessary to obtain energy wood, therefore simple feller-buncher heads without feed rollers and delimbing knives can be used instead of expensive harvester heads. Several felling heads are available for harvesting only energy wood. These heads can

117

C. ROTTENSTEINER et al.

Evaluation of the Feller-Buncher Moipu 400E for Energy Wood Harvesting (117–128)

be mounted on forwarders or tractors. The head can be constructed for single tree handling or felling and accumulating trees. Thereby it is possible to handle several trees during one crane cycle. A number of studies have been carried out on the new biomass harvesting technology. Spinelli et al. (2006) and Eberhardinger (2007) analyzed fellerbunchers for energy wood harvesting whereby felling was separated from extraction. Laitila and Asikainen (2006) examined a conventional forwarder equipped with the Moipu 400 E energy wood head that performed felling and extracting continuously. Kärhä (2006) compared the two-machine (harvester and forwarder) concept with the integrated system. Affenzeller and Stampfer (2007) examined singletree felling, loading and extraction with tractor trailer equipped with a crane, as a continuous process. As mentioned above there already exists an evaluation of the energy wood head Moipu 400E which was conducted in Finland (Laitlila and Asikainen 2006). This study was carried out on either birch or pine dominated stands. When cutting trees of an average of 0.045 m3, the productivity was about 3.5 m3/PSH0. In the study by Affenzeller and Stampfer (2007), a productivity of only 1.6 m3/PSH0 was achieved with a tractor-trailer combination in Pine stands for trees of comparable size. This is less than half of the productivity attained with the felling-bunching head Moipu 400E. At present there are no studies about Moipu 400E under Central European conditions. The influence of different tree species on productivity of the felling machine was never evaluated. For the efficiency of a tending operation not only the productivity level, but also the revenues and costs are important. Forest enterprises are focused on cost covering tending operations, and without that coverage necessary stand treatments are neglected. This investigation evaluates the influence of stand and terrain parameters on productivity and costs of a forwarder, equipped with the feller-buncher head Moipu 400E. The experiment was carried out in a Scots Pine and Oak dominated stand.

volume, tree species, number of trees in a bunch, cutting removal per hectare, forwarding distance, slope, and average tree volume of a load. The productivity model comprises 7 submodels or terms, respectively: 1. Cutting time 2. Cutting & Loading time 3. Loading time 4. Moving time 5. Forwarding time 6. Unloading time 7. Operational delay time

2.2 Machine description and harvesting system – Opis stroja i sustava pridobivanja drva The base machine of the feller-buncher was the 8-wheeled Timberjack 1110D forwarder (Fig. 1) with a weight of 14700 kg (load rating 12000 kg). The Moipu 400E head was mounted on the forwarders crane. It performs cutting, bunching, and

2. Methodology – Metodologija 2.1 Model hypothesis – Hipoteza Affenzeller and Stampfer (2007) analyzed the productivity of the felling head Naarva Grip 1500-25, Eberhardinger (2007) of Naarva Grip 1500-25E. Laitila and Asikainen studied the productivity of Moipu 400E combined with a forwarder in Finland in 2006. On the basis of these experiments the model hypothesis assumes that productivity is a function of tree

118

Fig. 1 Timberjack 1110D with Moipu 400E head Slika 1. Timberjack 1110D sa sje~nom glavom Moipu 400E Croat. j. for. eng. 29(2008)2

Evaluation of the Feller-Buncher Moipu 400E for Energy Wood Harvesting (117–128)

grappling so that one base machine performs the whole process from felling to extraction. The Moipu 400E head uses a single-action shear for cutting. The maximum cutting diameter for single trees is 30 cm and 50 cm for bunches. The opening diameter of the head is 120 cm and it weighs 540 kg (Moisio Forest 2008). The machine combination uses the following harvesting procedure: First the forwarder drives backwards into the stand and opens a strip road. Trees on the strip road are felled and piled alongside the trail. On the way back out of the stand, the machine loads the processed trees into the load space. The fully loaded forwarder drives to the landing and starts unloading. After unloading, the forwarder drives back to the stand and thins both sides of the strip road beginning at the end of the strip road. Whole-tree harvesting was carried out; trees were extracted with tops and branches. The loaded bunches of trees exceeded the forwarder’s load space. Thus the fully loaded forwarder was not capable of driving backwards; therefore a two step procedure was necessary.

2.3 Study site – Mjesto istra`ivanja Time studies were carried out in a stand close to Lockenhaus – Austria. The extent of the area was 0.96 ha with an average 7% slope. The dominant height of the 35–40 year old stand was 18 m. The major tree species is Scots Pine (Pinus sylvestris) with 50% of the stand’s volume; followed by Sessile Oak (Quercus petraea) with 40% of the stand’s volume. Beech (Fagus sylvatica) and Larch (Larix decidua) are less numerous. Oak trees originate partly from coppicing and to some extent from generative regeneration. Stand density was reduced from over 4700 stems per hectare to an average of 1800 stems per hectare. The average tree volume of removed trees was 0.057 m3. The harvested volume was 170 m3/ha.

C. ROTTENSTEINER et al.

lished of 320 m2 in size as reference unit for the observation of terrain and stand conditions (slope, cutting removal per ha). In the volume inventory the diameter at breast height (DBH) of each tree was measured with a calliper in order to use these data for calculating dry mass and volume, respectively. The diameters were grouped in DBH-classes and marked with a colour-code according to Affenzeller and Stampfer (2007). The tree volume was calculated using biomass models of Zianis et al. (2005). To achieve the volume in m3, the dry weight (kg) of the trees is divided by the oven-dry density (kg/m3). The oven-dry densities of different tree species are published in ÖNORM B 3012 (2003).

2.5 Statistical analyses – Statisti~ka analiza Variance analysis attempts to quantify the influence of nominal or ordinal-scaled variables. The statistical analysis was carried out with the computer software SPSS 15.0 for Windows, the statistical fundamentals as described in Stampfer (2002). For each part of the model, the following analysis strategy was chosen: Þ develop a linear model with all co-variables and factors, Þ evaluate non-linearity of co-variables, Þ choose a number of sub-models through removal of non-significant variables, Þ choose two-ways interactions of sub-models. Tree volume is a major part of all production functions but the relationship between productivity and tree volume is rarely linear. Therefore a power factor is used on the co-variable tree volume. Häberle (1984) recommends the estimation of this power value with an iterative procedure aimed at optimizing the coefficient of determination and the distribution of residues.

2.4 Data collection – Prikupljanje podataka

2.6 Cost analyses – Analiza tro{ka

The time study was carried out by means of the continuous time method with the use of the field computer Latschbacher EG 20. The work of the energy wood harvesting system was divided into elements with clearly recognizable starting and ending points (Table 1). The variables, covariates and the factor with two levels are shown in Table 2. A mixed stand was examined in order to figure out the effect of the factor tree species. In the stand 5 skid roads with a corridor spacing of 16 m were marked with paint. The skid roads were divided into sections of 20 m in distance. Thus plots were estab-

The calculation of the machine rates was conducted with a few modifications according to the Scheme of Food and Agriculture Organization of the United Nations (FAO 1992). The fixed costs comprise costs for interest, depreciation, and storage and insurance. The purchase price for the forwarder and the crane mounted felling head was 260,750 , and the costs for storage and insurance were 475 per year (Renner 2008. The annual interest costs were calculated at an interest rate of 4.5%. The depreciation was calculated assuming an economic life of 6 years. All calculations were done underlying 1500 scheduled system hours per year (PSH15) and an expected useful

Croat. j. for. eng. 29(2008)2

119

C. ROTTENSTEINER et al.

Evaluation of the Feller-Buncher Moipu 400E for Energy Wood Harvesting (117–128)

Table 1 Time study elements Tablica 1. Radne sastavnice Element Radna sastavnica Cutting Sje~a Cutting & Loading Sje~a s utovarom Loading Utovar Moving Premje{tanje Driving loaded Optere}eno kretanje Unloading Istovar Driving unloaded Neoptere}eno kretanje Down-time <15 Kvarovi i popravci <15 Down-time >15 Kvarovi i popravci >15 Operational delay time Povremeni radovi Miscellaneous Preostali prekidi rada

Description Opis Start: Head is in horizontal position for cutting – Po~etak: Sje~na je glava u vodoravnom poloàju za sje~u End: New cycle, or another element starts – Kraj: Novi ciklus ili po~etak druge radne sastavnice Start: Head is in horizontal position for cutting – Po~etak: Sje~na je glava u vodoravnom poloàju za sje~u End: New cycle, or another element starts – Kraj: Novi ciklus ili po~etak druge radne sastavnice Start: Head is in vertical position for loading – Po~etak: Sje~na je glava u uspravnom poloàju za utovar drva End: New cycle, or another element starts – Kraj: Novi ciklus ili po~etak druge radne sastavnice Start: Wheels are rotating after Cutting, Cutting & Loading, or Loading Po~etak: Pokretanje kota~a nakon sje~e, sje~e s utovarom ili utovara End: New element starts – Kraj: Po~etak druge radne sastavnice Start: Wheels are rotating and load space is completely loaded Po~etak: Pokretanje kota~a, a utovarni je prostor potpuno natovaren End: New element starts – Kraj: Po~etak druge radne sastavnice Start: Crane starts unloading on landing – Po~etak: Dizalica po~inje istovar na stovari{tu End: New element starts –Kraj: Po~etak druge radne sastavnice Start: Wheels are rotating after Unloading – Po~etak: Pokretanje kota~a nakon istovara End: New element starts – Kraj: Po~etak druge radne sastavnice Machine down-time shorter than 15 minutes Kvarovi i popravci kra}i od 15 minuta Machine down-time longer than 15 minutes Kvarovi i popravci dulji od 15 minuta Delays related to thinning (e.g. clearance of already cut trees) Prekidi povezani s proredom (npr. ~i{}enje ve} posje~enih stabala) Other delays Ostali prekidi rada

life of 10,000 PSH15. The operating costs comprise maintenance and repair, fuel costs, and costs for lubricants. The maintenance and repair rate was set at 0.8. The fuel consumption rate was 10 liters/hour. The lubricants costs are assumed to be 25% of the fuel costs, underlying a fuel price of 1.17 /litre. The labour costs including wages account for 25 /PSH15. All calculations are made without sales tax.

3. Results – Rezultati 3.1 Distribution of time consumption – Raspodjela utro{ka vremena Felling & loading and Loading (elements cutting, cutting & loading, and loading) represented 50% of total time consumption (Fig. 2). Time consumption for forwarding was 3% when loaded and 4% when empty. This is the case because the distance between landing and loading the first time was longer than the distance between loading the last time and the land-

120

Unit Jedinica min min min min

min min min min min min min

ing. Moving during cutting and loading and unloading at the landing were both 6% of total time consumption. 8% of the time was used for manipulation. Down-time shorter than 15 minutes represented 12%, and down-time longer than 15 minutes 11% of the recorded total time. Down-times were the result of time for breaks and repairs. 1104 cycles (bunches) for cutting and cutting & loading were recorded. The average number of trees in a bunch was 2.6 trees. The average diameter at breast height of removed trees was 9.23 cm. 44 loads with a total volume of 163 m3 were recorded. Total average forwarding distance was 89 m. The average volume of a load was 3.7 solid m3.

3.2 Productivity functions – Funkcije proizvodnosti Equation 1 shows the productivity for the entire harvesting system. Sub-models were used because of the different number of cycles for sub-models acCroat. j. for. eng. 29(2008)2

Evaluation of the Feller-Buncher Moipu 400E for Energy Wood Harvesting (117â&#x20AC;&#x201C;128)

C. ROTTENSTEINER et al.

Table 2 Variables, Factor, and Covariates of the productivity model Tablica 2. Nezavisne i zavisne varijable te faktor modela proizvodnosti Type Vrsta

Name Ime Cutting time â&#x20AC;&#x201C; tcut, min/cycle Sje~a â&#x20AC;&#x201C; tcut, min/tura

Total time for cutting trees, Productive system hour Ukupno vrijeme sje~e stabala, pogonski sat rada

Cutting & Loading time â&#x20AC;&#x201C; tcut&load, min/cycle Sje~a s utovarom â&#x20AC;&#x201C; tcut&load, min/tura

Total time for cutting and loading trees, Productive system hour Ukupno vrijeme sje~e i utovara stabala, pogonski sat rada

Forwarding time â&#x20AC;&#x201C; tfor, min/cycle Izvo`enje drva â&#x20AC;&#x201C; tfor, min/tura Dependent Variables Zavisna varijabla Moving time â&#x20AC;&#x201C; tmov, min/cycle Premje{tanje â&#x20AC;&#x201C; tmov, min/tura

Factor Faktor

Covariates Nezavisna varijabla

k1 tcut tload tcut&load tfor

Total time for driving loaded and unloaded; Productive system hour Ukupno vrijeme kretanja optere}enoga i neoptere}enoga vozila, pogonski sat rada Total time for moving between cutting & loading, Productive system hour Ukupno vrijeme premje{tanja izme|u sje~e i utovara, pogonski sat rada

Loading time â&#x20AC;&#x201C; tunload, min/cycle Utovar â&#x20AC;&#x201C; tunload, min/tura

Total time for loading, Productive system hour Ukupno vrijeme utovara, pogonski sat rada

Unloading time â&#x20AC;&#x201C; tload, min/cycle Istovar â&#x20AC;&#x201C; tload, min/tura

Total time for unloading of a load, Productive system hour Ukupno vrijeme istovara tereta, pogonski sat rada

Tree Species, 2 Levels Vrsta drve}a, 2 razreda

Scots Pine (0), Oak (1) Obi~ni bor (0), hrast (1)

Tree volume â&#x20AC;&#x201C; Vtree, m3 over bark Obujam stabla â&#x20AC;&#x201C; Vtree, m3 s korom

Average tree volume Prosje~ni obujam stabla

No. of trees in a bunch, n Broj stabala u sve`nju, n

Number of trees in a bunch Broj stabala u sve`nju (zahvatu)

Cutting removal, m3 over bark/ha Sje~na gusto}a, m3 (s korom)/ha

Timber removal quantity per ha Koli~ina posje~enoga drva po ha

Forwarding distance â&#x20AC;&#x201C; dist, m Udaljenost izvo`enja â&#x20AC;&#x201C; dist, m

Forwarding distance of a load (average of driving loaded and unloaded) Prosje~na udaljenost kretanja neoptere}enoga i optere}enoga vozila

Load volume â&#x20AC;&#x201C; load, m3 over bark Obujam tovara â&#x20AC;&#x201C; load, m3 s korom

Volume of a load Obujam utovarenoga tereta

Slope, % Nagib terena, %

Gradient of slope Kut nagiba terena

Average tree volume of load, m3 over bark Prosje~an obujam stabla u tovaru, m3 s korom

Average tree volume of a load Prosje~an obujam stabla u tovaru vozila

PROD

where: PROD

Description Opis

60 ÂŞ Âş Vcut Vcut&load k1 Â&#x2DC; ÂŤ tcut tload Â&#x2DC; tcut&load Â&#x2DC; tfor tmov tunload tod Âť V V tot tot ÂŹ Âź

Productivity of the forwarder equipped with Moipu 400E head m3/PSH15 Conversion factor from PSH0 to PSH15 Cutting time, min/m3 Loading time, min/mÂł Cutting & loading time, min/m3 Forwarding time, min/m3

Croat. j. for. eng. 29(2008)2

tmov tunload tod Vcut Vcut&load Vtot

(1)

Moving time, min/m3 Unloading time, min/m3 Operational delay time, min/m3 Volume that was first cut and then loaded, m3 Volume that was cut and loaded subsequently, m3 Total harvested volume, m3

121

C. ROTTENSTEINER et al.

Evaluation of the Feller-Buncher Moipu 400E for Energy Wood Harvesting (117â&#x20AC;&#x201C;128)

Fig. 2 Structure of time consumption Slika 2. Struktura utro{ka vremena Table 3 Means, 5% and 95% Quantile of covariates Tablica 3. Aritmeti~ke sredine te 5. i 95. percentili nezavisnih varijabli Covariate Nezavisne varijable Tree volume, m3 Obujam stabla, m3 Volume of load, m3 Obujam tovara, m3 Forwarding distance, m Udaljenost privla~enja, m Slope, % Nagib, % Number of trees in a bunch, n Broj stabala u zahvatu, n

Mean Aritmeti~ka sredina

Quantile5 Quantile95 5. percentil 95. percentil

0.057

0.012

0.212

3.71

1.13

5.61

88.8

22.0

175.6

2.249 0.362 Â&#x2DC; Vtree 0.9 R2

tcut

6.9

5.0

10.3

2.6

1.0

8.0

cording to their reference unit (bunch of trees, or load). Furthermore the detailed observation of individual working phases increases the accuracy of predictions in the entire productivity model. Table 3 shows the means, the 5th and 95th percentile of the covariates. For the entire productivity model only the covariates average tree volume, forwarding distance, and load volume were significant. The assumption that the cutting time depends on the factor tree species, as well as on the covariates number of trees in a bunch, cutting removal/ha and

122

slope could not be verified. The cutting time only depends on the tree volume (Equation 2). The number of trees in a bunch and the tree species were intercorrelated with the tree volume; therefore the number of trees in a bunch was not used in the covariance analysis. The intercorrelation between tree species and tree volume might be caused by biomass models used for different species; therefore the factor tree species was not used in the regression model. To achieve linearity the exponent -0.9 was used in the statistical model.

where: tcut Vtree

0.556

(2)

Cutting time, min/m3 Average tree volume, m3

The cutting & loading time depends on the tree volume (Equation 3). The other covariates and the factor tree species had no significant influence on the cutting & loading time per m3. tcut&load

6.774 0.267 Â&#x2DC; Vtree 0.9 R2

0.275

(3)

where: tcut&load Cutting & loading time, min/m3 Vtree Average tree volume, m3 The time consumption of forwarding (average of driving loaded and unloaded) depends on the forwarding distance (Equation 4). It accounts for 91% of Croat. j. for. eng. 29(2008)2

Evaluation of the Feller-Buncher Moipu 400E for Energy Wood Harvesting (117–128)

Table 4 Means, 5% and 95% Quantile of dependent variables Tablica 4. Aritmeti~ke sredine te 5. i 95. percentili zavisnih varijabli Dependent variable Zavisne varijable Cutting time – tcut, min/m3 Sje~a – tcut, min/m3 Cutting & Loading time – tcut&load, min/m3 Sje~a s utovarom – tcut&load, min/m3 Forwarding time – tfor, min/m3 Izvo`enje drva – tfor, min/m3 Moving time – tmov, min/m3 Premje{tanje – tmov, min/m3 Loading time – tunload, min/m3 Utovar – tunload, min/m3 Unloading time – tload, min/m3 Istovar – tload, min/m3 Operational delay time – tod, min/m3 Povremeni radovi – tod, min/m3

Mean Quantile 5 Quantile95 Aritmeti~ka sredina 5. percentil 95. percentil 10.22

2.33

25.99

13.94

4.59

24.80

0.72

0.31

1.67

1.12*

0.39

1.80

1.81*

1.00

2.69

1.14*

0.79

1.62

1.50*

0.21

3.16

* values used in the productivity model – vrijednosti kori{tene u modelu proizvodnosti

the variance of time consumption for forwarding. The influence of load volume was not significant. The division by the load volume is necessary to achieve min/m3. tfor

0.028 dist R2 load

0.91

C. ROTTENSTEINER et al.

corded time for operational delays by the total volume harvested. The constant terms for moving time, loading time, and unloading time, as well as for operational delays are summed up in Table 4. All sub-models are based upon productive system hours without down-times (PSH0). In practice down-times of up to 15 minutes are commonly included in the productive machine hours (PSH15) (Stampfer 2002). The factor k1 derived from the elemental time study was 1.15. However during elemental time collection there were almost no repairs, and also time for maintenance was not recorded. Thus for the forwarder equipped with the Moipu 400E head k1 was set at 1.3. The average productivity in this study was 4.11 m3/PSH0. Using the factor k1 the attained productivity was 3.16 m3/PSH15. This is equivalent to 8.85 cubic metre loose/PSH15 using a factor of 2.8 (ÖNORM M 7132 1998) to convert cubic metre in cubic metre loose. Fig. 3 shows the productivity of the forwarder equipped with the Moipu 400E head dependent on the tree volume. The three graphs show the range of the model’s validity in accordance with the forwarding distance (5th percentile, average, 95th percentile). Fig. 4 also shows the system productivity, but underlying the range of the model’s validity using the 5th and 95th percentile, as well as the average, of the forwarder’s load volume.

(4)

where: tfor dist

Forwarding time, min/m3 Average of driving loaded and unloaded, m load Volume of a load, m3 (over bark) The time consumption of moving, loading, unloading, and operational delays is very homogenous in each cycle. There is a relatively low variation of processes compared to the other 3 sub-models (Table 4). Therefore the means are used as constant terms. These means are calculated by dividing the respective elemental times (total time for moving, unloading and operational delays) by the total volume harvested. The time of loading was divided by the volume that was loaded. Operational delay time considers time related to thinning (e.g. clearance of already cut trees) that could not be added to another working element. For example after felling a couple of trees, the forwarder had to clear the cut trees before moving. However the clearance of already cut trees failed to appear in each cutting cycle. Thus operational delay time was collected separately from cutting. It was calculated by dividing the entire reCroat. j. for. eng. 29(2008)2

Fig. 3 Productivity model for the machine system (validity of forwarding distance) Slika 3. Pouzdanost modela proizvodnosti istra`ivanoga sustava (udaljenost izvo`enja drva) 123

C. ROTTENSTEINER et al.

Evaluation of the Feller-Buncher Moipu 400E for Energy Wood Harvesting (117–128)

sults in 587 kg/m3, the conversion factor was 1.70. Assuming a price of 78.00 per ton of oven dry chips (Österreichische Forstzeitung, 2008), results in a slightly positive contribution margin.

4. Discussion – Rasprava

Fig. 4 Productivity model for the machine system (validity of load volume) Slika 4. Pouzdanost modela proizvodnosti istra`ivanoga sustava (obujam tovara)

3.3 Costs – Tro{kovi The system costs (net value) of the harvesting system are 91.60 /PSH15. The fixed costs account for 33.20 /PSH15, and are at the same level as the operating costs of 33.40 /PSH15. The labour costs are 25.00 /PSH15. The average productivity is 3.16 m3/PSH15, and therefore the total costs for felling and forwarding are 28.99 /m3. The factor 2.8 was used to convert cubic metre in cubic metre loose (ÖNORM M 7132, 1998). The costs for chipping 3.1 /cubic metre loose found by Affenzeller and Stampfer (2007) in their study results is 8.68 /m3. According to Ganz et al. (2005) the costs for transportation range between 2.3 and 3.3 /cubic metre loose. In this study the costs for transportation were calculated with 2.9 /cubic metre loose or 8.12 /m3. The total costs of felling and forwarding, chipping, and transportation add up to 45.79 /m3 (16.35 per cubic metre loose). Overhead expenses for bookkeeping and communications, as well as office expenses are not included in the calculation. Entrepreneurial profit and moving expenses are also excluded from this calculation. The costs of 45.79 /m3 are equivalent to 77.84 per oven dry ton. Using the oven dry densities for Pine, Oak, Beech and Larch (ÖNORM B 3012, 2003) weighted by their rate on the total volume, which re-

124

The variables tree volume and forwarding distance have the major influence on productivity of the entire system. The factor tree species had no significant impact. The productivity of the Moipu 400E energy wood head mounted on a forwarder is 4.11 m3/PSH0 or 3.16 m3/PSH15. That is even higher than the productivity for comparable tree volume found by Laitila and Asikainen (2006) who obtained 3.75 m3/PSH0. In consideration of the costs of harvesting, chipping, and transportation of energy wood, compared with the prices achieved for energy wood, a positive contribution margin is possible. The costs for felling, forwarding, chipping, and transportation add up to 77.84 per oven dry ton. Assuming a price of 78.00 per oven dry ton of chips, there was a slightly positive contribution margin. However this calculation does not consider overhead expenses for bookkeeping and communications, as well as office expenses and entrepreneurial profit and moving expenses. To gain profit the timber contractor will set the price for his logging job at a higher level, which means higher energy wood harvesting costs. This would result in a negative contribution margin for the forest owner. In the economic sense of cost-accounting it must be calculated as investment costs in silvicultural tending activities with the objective of obtaining saw timber. Nevertheless, compared to an earlier study examining a similar felling device, the Moipu 400E proved to be efficient and competitive. Affenzeller and Stampfer (2007) examined single-tree felling, loading and extraction with tractor trailer equipped with a crane, as a continuous process. The felling device mounted on the crane was the Naarva Grip 1500-25, a felling head that is not capable of bunching trees. Affenzeller and Stampfer achieved a productivity of just 1.33 m3/PSH15 (1.60 m3/PSH0). They figured out that productivity of their fully mechanized system failed to cover the rate of fixed costs that arise in the mechanized system. The total costs of harvesting, extraction, chipping, and transportation to the biomass power plant were 21.20 /cubic metre loose. In this study the total costs for the supply of chips are 16.35 per cubic metre loose. When harvesting an average tree volume of 0.045 m3, as in the Naarva Grip 1500-25 study, the costs increase to 16.62 per cubic metre loose. Nevertheless this leads to a cost decrease of 4.58 per cubic metre loose. Croat. j. for. eng. 29(2008)2

Evaluation of the Feller-Buncher Moipu 400E for Energy Wood Harvesting (117–128)

Kärhä (2006) compared the whole-tree harvesting system carried out with 4 forwarders equipped with felling heads and with 5 different harvesters combined with forwarders. The productivity of the forwarder equipped with a felling head was 4.6–5 m3/PSH0 when harvesting trees with an average volume of 0.057 m3 at a forwarding distance of 250 m. This is somewhat better than in this study (4.11 m3/PSH0). The costs per productive system hour for the two machine system were approximately 70% higher than the one machine system costs. However, the difference in productivity of the two systems does not result in dramatically different harvesting costs per m3. Laitila (2008) found that the two machine system (harvester with an accumulating felling head, and forwarder) was more cost competitive than the one machine (forwarder and felling head) system. The costs of the forwarder equipped with a felling head were 3.9 /m3 higher than the costs of the logging system based on a harvester. The difference must be caused by differences in forwarding time. (The head mounted on the harvester and forwarder was the same.) With the conventional forwarder the average grapple load size in unloading was 0.6 m3 whilst the average grapple load size of the felling head used as grapple was just 0.3 m3. Laitila assumes that the explanation for this significant difference is the structure of the felling head grapple. It is designed not just for loading but also for cutting. Thus the compromise grapple is not as efficient as the purposebuilt timber grapple. A new attempt to gain profit in thinning operations when using feller-bunchers could be the integrated harvesting of energy wood for chipping and roundwood. Therefore the feller-buncher should be capable of delimbing the trees. The prototype of a modified Moipu head with delimbing capability is already available on the market. Scandinavian machine manufactures developed in the past the socalled multitree handling machine, which could process and delimb 2 trees at the same time. The problem considering roundwood production was the delimbing and bucking quality. However these requirements are not important for energy wood production. Thus this machine could get a revival.

5. References – Literatura Affenzeller, G., Stampfer, K., 2007: Energieholzbereitstellung mit Traktor und Krananhänger mit Fallbeilklingenaggregat [Energy wood supply using a guillotine-shear mounted on a tractor trailer]. Forschungsbericht für das Kooperations-abkommen Forst Holz Papier (FHP), durchCroat. j. for. eng. 29(2008)2

C. ROTTENSTEINER et al.

geführt am Institut für Forsttechnik, Universität für Bodenkultur Wien, 29. Eberhardinger, A., 2007: Schwarze Zahlen in der Schwachholzernte? [Is it possible to gain profit in first thinnings?]. Forst & Technik, 5: 2–6. FAO, 1992: Cost Control in Forest Harvesting and Road Construction. FAO Forestry Paper- 99, Food and Agriculture Organization of the United Nations, Rome. Feller, S., Remler, N., Weixler, H., 1998: Vollmechanisierte Waldhackschnitzel-Bereit-stellung. Ergebnisse einer Studie am Hackschnitzel-Harvester [Mechanized wood chip supply. Harvesting study results]. Berichte aus der Bayerischen Landesanstalt für Wald und Forstwirtschaft, Nr. 16. Freising, 76 p. Ganz, M., Holzleitner, F., Kanzian, C., 2005: Energieholzlogistik in Kärnten – Transport von Energieholz [Logistics concerning energy wood supply in Carinthia – energy wood transportation]. Institut für Forsttechnik. Universität für Bodenkultur Wien, 89 p. Häberle, S., 1984: Standardisierung zweidimensionaler Ausgleichsfunktionen über Richtgrad und Richtkonstante [Standardization of regression functions]. Forstarchiv 55(6): 220–225. Kärhä, K., 2006: Whole-tree harvesting in young stands in Finland. – Forestry Studies / Metsanduslikud Uurimused 45, 118-134. ISSN 1406-9945. Online available at: http:/ /www.hagrar-net.at/netautor/napro4/appl/na_professional/parse.php?id=2500%2C1388884%2C%2C%2CeF9 EV19IRUFERVJbMF09cHJldmlldw%3D%3D (3. 8. 2008). Laitila, J., Asikainen, A., 2006: Energy Wood Logging from early Thinnings by Harwarder Method. Baltic Forestry 12(1): 94–102. Laitila, J., 2008: Harvesting technology and the cost of fuel chips from early thinnings. Silva Fennica 42(2): 267–283. Limbeck-Lilienau, B., Stampfer, K., 2004: Sind moderne Arbeitsverfahren auch pleglich? [How careful are modern working methods?]. Arbeit im Wald 2: 1–3. Meng, W., 1978: Baumverletzungen durch Transportvorgänge bei der Holzernte: Ausmaß und Verteilung, Folgeschäden am Holz und Versuch ihrer Bewertung [Wood damages caused by transportation while harvesting: Size and location, future damages and valuation]. Schriftenreihe der Landes-forstverwaltung Baden-Württemberg, Nr. 25, 159 p. Moisioforest, 2008: Internet: www.moisioforest.com (21. 4. 2008). ÖNORM B 3012, 2003: Holzarten – Kennwerte zu den Benennungen und Kurzzeichen der ÖNORM EN 13556 [Tree species – Characteristics according to the terms and abbreviations of ÖNORM EN 13556]. Österreichisches Normungsinstitut (ed.), Wien, 12 p. ÖNORM M 7132, 1998: Energiewirtschaftliche Nutzung von Holz und Rinde als Brennstoff. Begriffsbestimmungen und Merkmale [Use of wood and bark for energetic

125

C. ROTTENSTEINER et al.

Evaluation of the Feller-Buncher Moipu 400E for Energy Wood Harvesting (117–128)

purpose. Terms and marks]. Österreichisches Normungsinstitut (ed.), Wien, 9 p. Österreichische Forstzeitung, 2008: Holzpreistabelle nach dem Holzmarktbericht der Landwirtschaftskammer Österreich [Timber prices according to the timber market report of the Austrian Chamber for Agriculture and For estry]. Österreichische Forstzeitung 119(9): 22–24. Österreichische Waldinventur, 2004: Österreichische Waldinventur 2000/02 – Hauptergebnisse [Results of the Austrian Forest Inventory]. BFW Praxis Information, Nr. 3 – 2004, Bundesamt und Forschungszentrum für Wald Wien, 23 p.

Spinelli, R., Cuchet, E., Roux, P., 2007: A new feller-buncher for harvesting energy wood: Results from European test programme. Biomass and Bioenergy 31: 205–210. Renner, H., 2008: e-mail information of Forstbedarf Renner (7. 7. 2008) Stampfer, K., 2002: Optimierung von Holzerntesystemen im Gebirge [Optimization of timber harvesting systems in mountainous regions]. Habilitationsschrift am Institut für Forsttechnik, Universität für Bodenkultur, Wien, 96 p. Zianis, D., Muukkonen, P., Mäikipaä, R., Mencuccini, M., 2005: Biomass and Stem Volume Equations for Tree Species in Europe. Silva Fennica Monographs 4, 63 p.

Sa`etak

Ocjena vi{ezahvatne sje~ne glave Moipu 400E pri pridobivanju drva za energiju Pravilna njega mladih sastojina pove}ava kakvo}u vrijednoga drva i smanjuje opasnost od o{te}ivanja sastojina vjetrolomima, snjegolomima te gradacijom potkornjaka. Pri pridobivanju drva iz pretkomercijalnih proreda tro{kovi ~esto nadilaze mogu}i prihod zbog niske vrijednosti drva malih dimenzija. Jedna je od mogu}nosti upotrebe sitnoga drva iz pretkomercijalnih proreda usitnjavanje stabala u iverje, koje se rabi kao gorivo u bioenerganama. Kresanje grana i trupljenje debla nije nu`no pri pridobivanju {umske biomase te se stoga umjesto skupih harvesterskih glava mogu rabiti jednostavne vi{ezahvatne sje~ne glave bez posmi~nih valjaka i noèva za kresanje grana. Strojno prorje|ivanje uporabom forvardera s ugra|enom vi{ezahvatnom sje~nom glavom moè pobolj{ati djelotvornost u pretkomercijalnim proredama zbog zahvatanja i sje~e vi{e tanjih stabala u jednom radnom hodu dizalice. Cilj je ovoga istraìvanja vrednovanje utjecaja sastojinskih i terenskih ~imbenika na proizvodnost i tro{kove rada osmokota~noga forvardera Timberjack 1110D (mase 14,7 t te dopu{tene nosivosti 12 t) na ~iju je dizalicu (umjesto hvatala) ugra|ena vi{ezahvatna sje~na glava Moipu 400E pri pridobivanju energentskoga drva (slika 1). Sje~na glava sije~e, skuplja i zahva}a stabla te tako jedno vozilo sije~e, ali i privla~i drvo. Sje~na glava Moipu 400E te{ka je 540 kg, ima {kare za sje~u, najve}i radni promjer pri sje~i pojedina~nih stabala je 30 cm, a 50 cm pri sje~i sve`nja stabala. Pri izra~unu tro{kova strojnoga rada primijenjena je modificirana metodologija FAO-a (1992). Fiksni tro{kovi obuhva}aju amortizaciju stroja, kamate za investiciju te tro{kove osiguranja i garaìranja. Nabavna cijena forvardera s dizalicom te ugra|enom sje~nom glavom iznosi 260 750 EUR, a godi{nji tro{ak garaìranja i osiguranja 475 EUR. Godi{nji tro{ak vezan uz nabavu stroja zasnovan je na kamati od 4,5 %. Amortizacija je temeljena na vremenu zastarijevanja stroja od 6 godina. Svi su izra~uni zasnovani na 1500 pogonskih sati rada godi{nje, odnosno 10 000 pogonskih sati rada u normalnom uporabnom razdoblju. Varijabilni tro{kovi obuhva}aju odràvanje i popravke stroja te tro{kove goriva i maziva. Tro{ak odràvanja i popravaka iznosi 80 % amortizacije, a potro{nja goriva 10 L po satu rada. Tro{ak maziva pretpostavlja 25 % tro{ka goriva, ~ija je jedini~na cijena 1,17 EUR/L. Tro{ak radnika obuhva}a njegovu pla}u u iznosu od 25 EUR po pogonskom satu rada stroja. Svi izra~uni su bez PDV-a. Na osnovi navedenih ulaznih podataka fiksni tro{kovi iznose 33,2 EUR, varijabilni 33,4 EUR, a tro{ak radnika 25 EUR, {to ukupno daje tro{ak sje~e i izvoènja drva u iznosu od 91,6 EUR po pogonskom satu rada. Za istraìvanje je odabrana mje{ovita sastojina obi~noga bora i hrasta kitnjaka zbog utvr|ivanja utjecaja vrste drve}a na razinu proizvodnosti opisanoga stroja. Istraìvanje je provedeno u odjelu povr{ine 0,96 ha, prosje~na nagiba terena od 7 %. Sastojina je u dobi izme|u 35 i 40 godina, s dominantnom visinom od 18 m. U drvnoj zalihi bijeli bor sudjeluje s 50 %, hrast kitnjak s 40 %, a obi~na bukva i europski ari{ pridolaze u primjesi. Kitnjakova stabla djelomi~no potje~u iz panja, odnosno iz sjemena. Proredom je gusto}a sastojine smanjena s 4700 stabala/ha na 1800 stabala/ha. Sje~na je gusto}a iznosila 170 m3/ha, a prosje~ni obujam posje~enoga stabla 0,057 m3. Tijekom rada istraìvani je stroj radio na sljede}i na~in. Na po~etku, kre}u}i se unazad, vozilo ulazi u sje~inu, pri ~emu si otvara vlaku. Stabla na trasi vlake sije~e i uhrpava uz njezin rub. Pri povratku iz sje~ine vozilo tovari posje~ena i uhrpana stabla u svoj utovarni prostor, te zavr{etkom utovara nastavlja se kretati prema pomo}nomu

126

Croat. j. for. eng. 29(2008)2

Evaluation of the Feller-Buncher Moipu 400E for Energy Wood Harvesting (117–128)

C. ROTTENSTEINER et al.

stovari{tu, gdje zapo~inje s istovarom. Pri ponovnom vra}anju u sje~inu po~inje prorje|ivati sastojinu s obje strane, i to s kraja prethodno prosje~ene vlake. Pri radu se rabi stablovna metoda izradbe drva, tako da se posje~ena stabla privla~e zajedno s ovr{inama i granama. Utovareni sve`njevi stabala, obujamno (ali ne maseno) nadilaze mogu}nosti utovarnoga prostora vozila. Tako natovareno vozilo nije se u mogu}nosti kretati unazad prema pomo}nomu stovari{tu, te je stoga prijeko potreban opisani na~in rada, tj. u dva koraka. Za potrebe istraìvanja bojom je ozna~eno 5 traktorskih vlaka me|usobna razmaka 16 m. Traktorske su vlake razdijeljene po duljini na segmente od 20 m te su tako oblikovane plohe povr{ine 320 m2, na kojima su prikupljani podaci o terenskim i sastojinskim ~imbenicima (nagib terena, sje~na gusto}a i dr.). Prsni su promjeri stabala mjereni s promjerkom, a za utvr|ivanje biomase stabala kori{ten je model koji su dali Zianis i dr. (2005). Obujam i masa suhe tvari utvr|eni su pomo}u pretvorbenih koeficijenata za razli~ite vrste drve}a (ÖNORM B 3012, 2003). Studij rada i vremena proveden je proto~nom metodom kronometrije i primjenom terenskoga ra~unala Latschbacher EG 20. Radni proces pridobivanja drva za energiju razdijeljen je u radne sastavnice s jasno odre|enim fiksa`nim to~kama (tablica 1). Hipoteza istraìvanja pretpostavlja da je proizvodnost opisanoga stroja funkcija obujma stabla, vrste drve}a, broja stabala u zahvatu sje~ne glave, sje~ne gusto}e, udaljenosti privla~enja drva, nagiba terena i prosje~noga obujma stabla u tovaru vozila. Model se proizvodnosti sastoji od sedam podmodela utro{aka vremena, i to: 1) sje~e, 2) sje~e i utovara, 3) utovara, 4) premje{tanja, 5) izvoènja, 6) istovara, 7) povremenih radova. Nezavisne i zavisne varijable te faktor u modelu proizvodnosti prikazani su u tablici 2. Analizom varijance nastojao se kvantificirati utjecaj nominalnih i ordinalnih varijabli. Statisti~ka je analiza provedena uz pomo} ra~unalnoga programa SPSS 15.0. Analiza svakoga pojedinoga podmodela provedena je po sljede}oj strategiji: Þ razvoj linearnoga modela sa svim nezavisnim varijablama i faktorima Þ ocjena nelinearnosti nezavisnih varijabli Þ izbor broja podmodela izbacivanjem statisti~ki nezna~ajnih varijabli Þ izbor dvostruke interakcije podmodela. Tijekom istraìvanja ostvarena su 1104 ciklusa zahvatanja sve`nja stabala pri sje~i, odnosno pri sje~i i utovaru. Prosje~an broj stabala u sve`nju iznosio je 2,6, a prosje~an prsni promjer posje~enih stabala 9,2 cm. U 44 forvarderska turnusa ukupno je izvezeno 163 m3 energentskoga drva, s prosje~nim obujmom tovara od 3,7 m3/turi. Prosje~na je udaljenost izvoènja drva iznosila 89 m. Radne sastavnice sje~a te sje~a i utovar zastupljene su s 50 % ukupnoga vremena rada (slika 2). Utro{ci vremena optere}enoga kretanja vozila zauzimaju 3 %, odnosno neoptere}enoga kretanja 4 %, {to je posljedica ve}e udaljenosti izme|u pomo}noga stovari{ta i mjesta prvoga utovara u odnosu na udaljenost izme|u mjesta zadnjega utovara i pomo}noga stovari{ta. Na premje{tanje vozila tijekom sje~e i utovara te istovara drva na pomo}nom stovari{tu otpada 6 %, a na manipulaciju 8 % ukupnoga vremena rada. Na kvarove i popravke kra}e od 15 minuta otpada 12 %, a na dulje od 15 minuta 11 % ukupnoga vremena rada. Izraz 1 predstavlja proizvodnost sustava koji obuhva}a sje~u i privla~enje drva za energiju. Razlog je kori{tenju podmodela jedini~nih utro{aka vremena (min/m3) razli~it broj ciklusa s razli~itim jedinicama izrade (sveànj stabala, tovar forvardera) koji se javljaju u jednom proizvodnom ciklusu sje~e i privla~enja drva opisanim vozilom. [tovi{e, op{irnije opaànje pojedinih sastavnica rada pove}ava to~nost predvi|anja cijeloga modela proizvodnosti. Tablica 3 prikazuje aritmeti~ke sredine te 5. i 95. percentile nezavisnih varijabli. Za cijeli model proizvodnosti statisti~ki su zna~ajne samo ove nezavisne varijable: prosje~ni obujam stabla, udaljenost privla~enja drva te obujam tovara. Nije potvr|ena pretpostavka da vrijeme sje~e ovisi o faktoru vrste drve}a, o nagibu terena, broju stabala u sve`nju i sje~noj gusto}i kao nezavisnim varijablama. Utro{ak je vremena sje~e ovisan samo o obujmu stabla (izraz 2). Tako|er samo o obujmu stabla ovisi i utro{ak vremena sje~e i utovara (izraz 3). Utro{ak vremena izvoènja drva ovisi o udaljenosti privla~enja, koja obja{njava 91 % varijabilnosti podataka (izraz 4). Me|utim, iako statisti~ki nezna~ajan, u izraz 4 je uvr{ten i obujam tovara koji je nuàn za izra~unavanje jedini~noga utro{ka vremena (min/m3). Utro{ci vremena premje{tanja, utovara, istovara te povremenih radova vrlo su homogeni za svaki pojedini turnus te je utvr|ena njihova mala varijabilnost (tablica 4) u odnosu na utro{ke vremena opisane podmodelima izraz 2, 3 i 4. Stoga su njihovi prosjeci kori{teni kao konstantne vrijednosti (tablica 4) u modelu proizvodnosti. Svi podmodeli jedini~nih utro{aka vremena (min/m3) ne obuhva}aju prekide rada te su zasnovani na efektivnom satu rada (PSH0). Praksa je da su prekidi rada, kra}i od 15 minuta, obi~no ve} uklju~eni u pogonski sat rada (PSH15). Poveznicu izme|u proizvodnosti iskazanoj po efektivnom satu, odnosno pogonskom satu rada, predstavlja faktor dodatnoga vremena (k1), koji je u ovom istraìvanju utvr|en u iznosu od 1,15. Tijekom

Croat. j. for. eng. 29(2008)2

127

C. ROTTENSTEINER et al.

Evaluation of the Feller-Buncher Moipu 400E for Energy Wood Harvesting (117–128)

istraìvanja nije bilo kvarova niti se stroj morao odràvati, te je u model proizvodnosti uklju~en faktor dodatnoga vremena u iznosu od 1,3. Pouzdanost modela proizvodnosti istraìvanoga vozila za udaljenost izvoènja drva prikazuje slika 3, a za obujam tovara slika 4. Prosje~na ostvarena proizvodnost sje~e i privla~enja drva za bioenergane opisanim strojem iznosi 4,11 m3/PSH0, odnosno 3,16 m3/PSH15, s jedini~nim tro{kom od 28,99 EUR/m3. Istraìvanjem je utvr|eno optimalno podru~je rada vi{ezahvatne sje~ne glave Moipu 400E koje iznosi do 30 cm promjera u panju za bor, odnosno do 25 cm za hrast i bukvu. Za djelotvornu primjenu mehaniziranoga pridobivanja drva istraìvanim strojem pogodne su sje~ne jedinice u kojima obujam srednjega sje~noga stabla nadilazi 0,05 m3. Osim navedenoga tro{ka dobava drvnoga iverja optere}ena je i tro{kom iveranja te tro{kom njegova prijevoza. Za pretvorbu kubnoga metra (m3) u nasipni kubni metar (mn3) kori{ten je pretvorbeni faktor u iznosu od 2,8 (ÖNORM M 7132, 1998). Affenzeller i Stampfer (2007) utvr|uju tro{ak iveranja u iznosu 8,68 EUR/m3 (3,1 EUR/mn3), a Ganz i dr. (2005) tro{ak prijevoza iverja u rasponu od 2,3 do 3,3 EUR/mn3. Za potrebe ovoga istraìvanja tro{ak prijevoza iverja zasnovan je na vrijednosti od 2,9 EUR/mn3 ili 8,12 EUR/m3 tako da ukupan tro{ak dobave drvnoga iverja (sje~a, privla~enje, iveranje, prijevoz) na energanu doseè 45,79 EUR/m3 (16,35 EUR/mn3). Op}i tro{kovi knjigovodstva i komuniciranja, kao i uredski tro{kovi te tro{kovi premje{tanja, ali i poduzetni~ka dobit nisu uklju~eni u kalkulaciju. Tro{ak dobave iverja od 45,79 EUR/m3 odgovara tro{ku od 77,84 EUR/t (suhe tvari) koji je izra~unat kori{tenjem koeficijenata gusto}e suhe tvari bora, hrasta, bukve i ari{a (ÖNORM B 3012, 2003) koji su ponderirani s udjelom pojedine vrste u posje~enom drvu. Prosje~na gusto}a suhe tvari iznosila je 587 kg/m3 (pretvorbeni faktor 1,7). Uz cijenu iverja od 78 EUR/t suhe tvari (Österreichische Forstzeitung, 2008) ukupni se tro{kovi pribliàvaju granici isplativosti. S obzirom na to da ukupni tro{ak dobave iverja ne obuhva}a op}e tro{kove i dobit, {umarski }e poduzetnik, da bi ostvario dobit, povisiti cijenu sje~e, {to pove}ava tro{kove pridobivanja drva za energiju te nepovoljno utje~e na {umovlasnika. U ekonomskom smislu izra~un tro{kova treba se promatrati kao ulaganje u njegu sastojina s budu}nosnim ciljem proizvodnje pilanske oblovine. Klju~ne rije~i: pridobivanje drva za energiju, proreda, vi{ezahvatna sje~na glava Moipu 400E, proizvodnost, tro{kovi

Authors’ address – Adresa autorâ:

Received (Primljeno): July 18, 2008 Accepted (Prihva}eno): November 11, 2008

128

Christian Rottensteiner, MSc. e-mail: christian.rottensteiner@boku.ac.at Günter Affenzeller, MSc. e-mail: guenter.affenzeller@boku.ac.at Assoc. Prof. Karl Stampfer, PhD. e-mail: karl.stampfer@boku.ac.at University of Natural Resources and Applied Life Sciences Vienna Department of Forest and Soil Sciences Institute of Forest Engineering Peter Jordan Straße 82 1190 Wien AUSTRIA Croat. j. for. eng. 29(2008)2

Orginal scientific paper – Izvorni znanstveni rad

Harvesting Short-Rotation Poplar Plantations for Biomass Production Raffaele Spinelli, Carla Nati, Natascia Magagnotti Abstract – Nacrtak In Italy, short rotation forest has become very popular in recent years, with over 4,000 hectares already planted – almost exclusively with clone poplar. The study models the performance of modified forage harvesters on a range of short-rotation poplar plantations, identifies technical obstacles to the deployment of these machines and suggests solutions that may expand the capability of modified forage harvesters when treating short-rotation poplar. Data were collected from 16 operations, covering a total of 50 hectares and producing over 1000 green tonnes of wood chips. The average yield of the fields harvested during the trials was about 20 green tonnes/ha year, equivalent to 8 oven-dry tonnes/ha for a 60% average moisture content, measured in the laboratory. Gross machine productivity ranged from 9 to 44 green tonnes/scheduled machine hour (gt/SMH), with an average value of 25 gt/SMH. Of course, this result is affected by other factors than just forager performance, which is potentially much higher. A model was developed to predict harvesting performance and cost, showing that harvesting cost can be maintained below the 15 Euro/green tonne (2 Euro/GJ) ceiling only if field stocking exceeds 40 or 50 gt/ha when rows are long 300 and 100 m, respectively. The study also shows the need to optimize operations. Over a quarter of the total worksite time is occupied by unproductive delays, which may be reduced with improved planning and maintenance. Keywords: short-rotation forest, biomass production, harvesting, forager, Italy

1. Introduction – Uvod European farmers are increasingly attracted to energy crops, following the most recent changes in the Common Agricultural Policy and the rapid development of the bioenergy sector. Among potential sources of energy biomass, dedicated crops from surplus agricultural land have the highest potential contribution, and in the medium term they could account for three quarters of the total supply of energy biomass (Hoogwijk et al. 2003). Compared to other sources, dedicated crops offer the advantage of a highly intensive management that assures a strong impact relative to the land area involved (Alig et al. 2000). Of course, management intensity does not exclude multiple land-use, where the production of biomass is integrated with groundwater protection, ecological planning, etc. (Londo et al. 2004). This is especially the case with woody crops, including short rotation coppice (Heller et al. 2003). Among various cropping modules, short rotation coppice (SRC) seems to best reflect the expectations of farmCroat. j. for. eng. 29(2008)2

ers, who are used to short return times and generally show little enthusiasm for traditional wood plantations, harvested at 10–30 years intervals. However, SRC is an industrial crop, designed to produce large quantities of low-priced raw materials and to be successful all operations must be conducted with the utmost efficiency. Harvesting cost is estimated to be above 50% of the total cost of biomass produced from wood plantations (Moiseyev and Ince 2000), which underscores the special needs for optimizing these operations. In Europe, Sweden has opened the way by launching an extensive plantation programme based on willow coppice: to date over 15,000 ha of short rotation willow coppices have been planted in this country (Larsson et al. 1998). Willow plantations are established at a very high density, and harvested every third or fourth year using modified forage harvesters, which have proved very effective (Danfors et al. 1998). More to the South – in Germany and in Italy, for example – poplar is considered better suited to

129

R. SPINELLI et al.

Harvesting Short-Rotation Poplar Plantations for Biomass Production (129–139)

the local environment than willow, and plantation programmes are largely based on this species. Modern poplar hybrids are highly suitable for ex-arable land. Short rotation coppice established with poplar resembles the Swedish model, but it also has some peculiar characteristics that may affect harvesting technology – among other things. As compared to willow, poplar wood is lighter and more brittle than willow wood, and poplar stools tend to generate fewer and larger sprouts when coppiced (Tharakan et al. 2003), which may have a considerable impact on harvesting performance. In Italy, short rotation forest (SRF) is very popular in the North, along the Po Valley, where the Italian agricultural industry is concentrated. The Regional Government of Lombardy has been the first one to release grants for the establishment and the management of SRF crops: after that, several other regions have followed, but none is yet offering the same level of subsidies (Table 1). The result is evident: in less than four years, 3000 ha of SRF crops have been established in Lombardy, representing three quarters of the total SRF surface established in Italy. Plantations in Northern Italy are established almost exclusively with poplar: sites are fresh enough, farmers are already familiar with the species and specific clones are available for biomass production (Frison et al. 1990). In fact, several nurseries have got into developing new clones, and have obtained a remarkable success. Hence the interest in finding the most effective harvesting system, exploring first the capacity of modified foragers, which have already proved the best option in the Nordic Countries.

Table 1 Grants released for the establishment and maintenance of SRF in Italy Tablica 1. Odobrene potpore za osnivanje i gospodarenje {umama kratkih ophodnji u Italiji Establishment Maintenance Compensation Max. cycle Osnivanje Gospodarenje Naknada Najdu`i period Euro/ha year Euro/ha year years Euro/ha Euro/ha god. Euro/ha god. godine Lombardy 3,150 620* 105–725 15 Friuli 4,000 – – 15 Veneto 4,000 – – 15 Piemonte 4,000 – – 15 Emilia 3,840 – – 15 Tuscany 2,000 – – 15 Umbria 2,320 – – 15 Lazio 2,000 – – 15 Region Regija

* first 2 years – prve 2 godine

130

The goal of this study is: 1) to document the performance of modified forage harvesters on a range of short-rotation poplar plantations; 2) to identify technical obstacles to the deployment of these machines and 3) to suggest solutions that may expand the capability of modified forage harvesters when treating short-rotation poplar. In the process, a costing model will be developed and applied, for providing unambiguous reference figures.

2. Materials and methods – Materijal i metode The study tested four different Claas foragers of the series Jaguar Mega, and namely the 840, 850, 860 and 880 models, with engines capable of delivering 254, 286, 306 and 340 kW, respectively. The Claas SRF harvesting system is based on a standard forage harvester, fitted with a special SRF header. Claas have produced two header versions, specifically designed for the Scandinavian market: the HS-1 and the HS-2 headers. In 2004 two Italian contractors purchased three of the newer HS-2 models and mounted them on already available Jaguar Mega foragers. Compared to the older HS-1, the HS-2 is a completely new design (Fig. 1). It is a purpose-built tool and not a modified sugar cane header. The circular saws placed at the bottom have a larger diameter, and the two vertical feed rollers placed above them on the old model have been replaced by crop-collectors with solid steel fingers: these move cut stems to a couple of horizontal in-feed rollers and eventually to the chopper of the forager – the same used for chopping maize, but with every other blade removed in order to produce larger wood chips. The new header looks more compact than the old one and offers better visibility to the driver. The machine was developed in Germany, but the Swedish users have added a few modifications, as suggested by practical experience. Besides the original Claas HS-2 header, the study also tested an Italian-made header – the GBE-1 model – very similar in design to the German unit but heavier and stronger, possibly better suited to handling of large stems. In both cases, the system is based on a forager and 2 to 4 tractor-trailer units which receive the chips from the forager and move them to a collection point: once there, the chips are loaded on transportation vehicles and moved to the plant (Fig. 2). The machines were studied while carrying out their scheduled commercial activity, on 16 different sites representative of the two main cropping modules used for SRF poplar in Italy: the annual and the biannual system. Plantations managed on the annual system are harvested at 1-year intervals and Croat. j. for. eng. 29(2008)2

Harvesting Short-Rotation Poplar Plantations for Biomass Production (129â&#x20AC;&#x201C;139)

R. SPINELLI et al.

Fig. 1 HS-2 header on Class forager Jaguar Mega 840 Slika 1. @etvena glava HS-2 na Classovu sila`nom kombajnu Jaguar Mega 840

Fig. 2 Harvesting system based on a forager and tractor-trailer unit Slika 2. Sustav pridobivanja drvnoga iverja pomo}u sila`noga kombajna i traktora s prikolicom Croat. j. for. eng. 29(2008)2

131

R. SPINELLI et al.

Harvesting Short-Rotation Poplar Plantations for Biomass Production (129–139)

Fig. 3 Planting systems Slika 3. Raspored sadnje adopt a planting density of about 10,000 cuttings/ha. Cuttings are planted in twin-rows, with a spacing of 1.8–2.7 m between twin-rows, 75 cm between the rows forming a pair and 45 cm along the rows (Fig. 3). Stem size at harvest reaches 2–3 cm (cut level), with peaks of 6–8 cm. Seeking a better fibre-to-bark ratio, many farmers resort to the biannual system, where the plantation is harvested at 2–3-years intervals, and is accordingly less thick. Cuttings are planted in single rows, with a spacing of 2.8–3.0 m between the rows and 0.5–0.7 m along the rows, which results in a planting density of 6,000–7,000 cuttings/ha. Stem size at harvest reaches 10–12 cm at cut level. Overall, the tests spread over 50 hectares of experimental plots. The study was designed to evaluate machine productivity and to identify the most significant variables affecting it. The data collection procedure consisted of a set of detailed time-motion studies conducted at the cycle level. In general, detailed time studies are more discriminating than shift-level studies and can detect smaller differences between treatments than shift-level studies can detect (Olsen et al. 1998). Cycle times for each machine were defined and split into time elements considered to be typical of the functional process analyzed. This was done with the intent of isolating those parts of a routine that are dependent on one or more external factors in order to enhance the accuracy of the productivity models (Bergstrand 1991). The criteria considered for such

132

subdivisions were: 1) isolating significant cycle elements, 2) reflecting as much as possible other similar existing protocols (Bjorheden et al. 1995) and, 3) avoiding unnecessary detail. All time elements and the related time-motion data were recorded with Husky Hunter® hand-held field computers running Siwork3® time-study software (Kofman 1995). Output was estimated by measuring the volume of all chip containers produced during each test, and by taking all of the containers to a certified weight bridge. Moisture content determination was conducted on samples, collected in sealed bags and weighted fresh and after drying for 48 hours at a temperature of 103°C in a ventilated oven. Row spacing was measured with a tape, and the length of row harvested for each load with a laser range-finder. This way, it was also possible to calculate the surface actually harvested at each site. Tests were conducted from 2004 to 2006 on 4 different machines (all the units currently used in Italian commercial operations) that harvested a total of 1036 green tonnes on 49 hectares at 16 different sites. The valid time study sessions lasted 54 hours. Data were statistically analyzed with both ANOVA and regression techniques to detect and formalize significant relationships (SAS 1999). Operating costs were calculated using the procedures described by Miyata (1980), on an estimated annual utilization of 1200 hours for the forager and 500 hours for the header. The corresponding investment costs are 250,000 and 140,000 Euro, respectively, and in both Croat. j. for. eng. 29(2008)2

Harvesting Short-Rotation Poplar Plantations for Biomass Production (129–139)

R. SPINELLI et al.

Table 2 Description of 16 test sites Tablica 2. Opis 16 pokusnih ploha Surface Povr{ina

Diameter Promjer

Harvesting density Sje~na gusto}a

gt/ha

gt/SMH*

gt/ha year gt/ha god.

first prva

9.2

2.05

9.0

9.2

9.0

twin dvostruki

second druga

3.2

2.34

25.5

19.5

25.5

twin dvostruki

third tre}a

8.4

2.13

10.9

10.5

10.9

HS-2

twin dvostruki

second druga

0.6

1.89

40.8

44.2

40.8

Mega 850

HS-2

twin dvostruki

first prva

0.6

4.31

43.1

13.2

21.6

Cura

Mega 850

HS-2

twin dvostruki

first prva

2.5

3.91

28.2

23.1

14.1

Albuzzano

Mega 840

HS-2

twin dvostruki

first prva

2.7

1.76

7.2

12.1

7.2

Linarolo

Mega 840

HS-2

twin dvostruki

third tre}a

1.2

1.58

27.6

28.7

27.6

Torbole

Mega 860

HS-2

twin dvostruki

first prva

2.0

2.23

13.5

20.2

13.5

Travagliato

Mega 860

HS-2

twin dvostruki

first prva

4.3

1.71

8.4

11.7

8.4

Pudiano

Mega 860

HS-2

twin dvostruki

second druga

3.1

2.29

37.9

38.0

37.9

Carpignano Mega 850

HS-2

twin dvostruki

second druga

1.3

3.37

71.9

33.3

36.0

Eraclea

Mega 880

GBE-1

single jednostruki

second druga

1.0

4.64

24.8

35.0

12.4

Caorle

Mega 880

GBE-1

single jednostruki

second druga

4.5

4.43

31.4

41.7

15.7

Arre

Mega 880

GBE-1

single jednostruki

second druga

1.2

4.20

53.5

33.7

26.8

Conselve

Mega 880

GBE-1

single jednostruki

second druga

3.6

3.70

27.8

24.3

13.9

Mean – Srednja vrijednost

3.1

2.9

28.8

24.9

20.1

Minimum – Najmanja vrijednost

0.6

1.6

7.2

9.2

7.2

Maximum – Najve}a vrijednost

9.2

4.6

71.9

44.2

40.8

Forager Sila`ni kombajn

Header @etvena glava

Age Dob

type tip

years godine

Sforzesca

Mega 840

HS-2

twin dvostruki

Bigli

Mega 840

HS-2

Frascarolo

Mega 840

HS-2

Alperolo

Mega 840

Calignano

Place Mjesto

Rows Redovi

Rotation Ophodnja

Productivity Annual yield U~inak Prinos

* green tonne per scheduled machine hour – tona svje`e tvari po ukupnom radnom satu

Croat. j. for. eng. 29(2008)2

133

Harvesting Short-Rotation Poplar Plantations for Biomass Production (129–139)

134

Diameter, cm Promjer, cm

Utilization, % Iskori{tenost, %

Availability, % Ispravnost, %

Blockage, % on harvesting time Kvarovi, % od vremena sje~e

Table 2 contains a description of test sites and some preliminary results. The fields present sample variations, with individual sizes ranging from 0.6 to 9 hectares and stocking from 7 to 70 green tonnes/ha. They have all been established using the new Alasia clones »Pegaso« and »AF2«: measured under operational conditions, yields varied between 7 and 40 green tonnes/ha year depending on site fertility and tending care. Hence the halved values refer to 2 year rotations (annual yield is half the actual yield at harvest age), while full values refer to 1 year rotations. Underscored values denote top performance obtained with good planting material on good soils, and may be indicative of future crops, once sufficient experience is gathered. Gross machine productivity ranges from 9 to 44 green tonnes/scheduled machine hour (gt/SMH), with an average value of 25 gt/SMH. Of course, this result is affected by other factors than just forager performance, which is potentially much higher. Harvesting progress is slowed down by a number of delays, caused by mechanical breakdowns, operator fatigue and machine interference within the support fleet. In fact, actual harvesting represents about 70% of the total worksite time, whereas machine maintenance and waiting for the transport units account respectively for 13% and 12% of the total time (Fig. 4). Defined as the percent ratio between maintenance-free worksite time and total worksite time, machine availability can give a measure of how the harvester copes with the strain of handling wood, rather than softer forage or maize. The data in Table 3 may suggest that machine maintenance becomes more intense when harvesting 2-year-old plantations, whose bigger stems may cause higher mechanical stress on the harvester. This observation may be corroborated by the higher occurrence of minor blockages during the harvesting of older plantations.

Place Mjesto

Rotation Ophodnja

3. Research results – Rezultati istra`ivanja

Table 3 Machine availability and utilization rates Tablica 3. Odnosi iskori{tenosti i ispravnosti stroja

Rows Redovi

cases the depreciation period was assumed to be 8 years. Repair and maintenance was estimated to 70% of depreciation, while labour cost was set at 16 Euro/hour. Fuel cost was assumed to be 0.90 Euro/L (subsidized fuel for agricultural use). The total costs are inclusive of 20% profit and overheads. Similar assumptions were used for the trailer and the tractor-trailer units. The resulting operating costs are 234 Euro/h and 71 Euro/h, respectively for the forager and the tractor-trailer unit. Conversion into energy figures was obtained on the assumption that the energy content of dry hardwood equals 18.5 GJ/t (Hartmann et al. 2000).

Age, years Dob, godine

R. SPINELLI et al.

Sforzesca

twin dvostruki

first prva

2.05

78.2

92.8

0.0

Bigli

twin second 2.34 dvostruki druga

55.6

76.9

0.0

Frascarolo

twin dvostruki

2.13

82.0

93.7

3.5

Alperolo

twin second 1.89 dvostruki druga

97.7

100.0

6.8

Calignano

twin dvostruki

first prva

4.31

54.3

70.7

9.6

Cura

twin dvostruki

first prva

3.91

90.9

94.7

3.4

Albuzzano

twin dvostruki

first prva

1.76

97.7

100.0

5.6

Linarolo

twin dvostruki

third tre}a

1.58

95.7

100.0

3.8

Carpignano

twin second 2.23 dvostruki druga

54.5

77.0

5.6

Torbole

twin dvostruki

first prva

1.71

93.5

100.0

0.0

Travagliato

twin dvostruki

first prva

2.29

76.6

79.8

0.0

Pudiano

single second 3.37 jednostruki druga

89.9

99.3

1.7

Caorle

single first jednostruki prva

4.64

78.8

100.0

2.6

Eraclea

single first jednostruki prva

4.43

48.8

100.0

7.0

Arre

single second 4.20 jednostruki druga

52.0

79.9

5.4

Conselve

single second 3.70 jednostruki druga

40.7

65.2

3.2

Mean – Srednja vrijednost

74.2

89.4

3.6

Minimum – Najmanja vrijednost

40.7

65.2

0.0

Maximum – Najve}a vrijednost

97.7

100.0

9.6

third tre}a

However, ANOVA testing did not confirm the inference, because the 10% difference in machine availability between the treatments (1-year-old and 2year-old) did not prove statistically significant at the 0.05 level (Table 4). Yet, the near significance of the test at the 0.10 level raises some questions and demands for further investigations in future. Croat. j. for. eng. 29(2008)2

Harvesting Short-Rotation Poplar Plantations for Biomass Production (129–139)

R. SPINELLI et al.

Table 4 Results of t-test for machine availability vs. rotation age Tablica 4. Rezultati t-testa odnosa ispravnosti stroja i godina ophodnje Machine availability – Ispravnost stroja

Rotation age, years Ophodnja, godine

Count Broj opa`anja

Mean Sredina

Variance Varijanca

Std. dev. Standardna devijacija

Std. error Standardna pogre{ka

93.611

83.534

9.140

3.047

83.933

203.847

14.265

5.392

t-test

Degree of freedom Stupanj slobode

Mean difference Srednja razlika

t value t vrijednost

p-value p vrijednost

1, 2

9.678

1.653

0.1205

Fig. 4 Breakdown of worksite time (excl. preparation and relocation) Slika 4. Ra{~lamba radnoga vremena (bez vremena pripreme i premje{tanja)

–0.450

0.154

–0.450 –2.931 0.0041

Crop density Gusto}a stabala

10.736

0.656

0.776

Forager model –0.430 Tip sila`noga kombajna

0.090

–0.226 –4.757 <0.0001

R2 = 0.783, n = 115 Croat. j. for. eng. 29(2008)2

p-value p vrijednost

Std. error Stand. pogre{ka

Intercept Slobodni ~lan

Variables Varijable

t-value t vrijednost

Coefficient Koeficijent

Std. Coeff. Stand. devijacija

Table 5 Regression of harvesting time vs. crop density and forager model Tablica 5. Regresijska analiza vremena sje~e u ovisnosti o gusto}i stabala i tipu sila`noga kombajna

16.374 <0.0001

Like with most agricultural machinery, net harvesting time (machine progression through the crop) is related to crop density, and this relation can be calculated by statistical analysis. The equation in Table 5 is the best fit to the experimental data points obtained from 115 observations, each equal to one loaded trailer. It shows that productivity increases with crop density, and also with the use of the most powerful forager. Before calculating the regression, ANOVA post-hoc testing allowed detecting that the net productivity of the 880 forager model is significantly different from that of the other models, which show no significant differences among themselves. Hence the decision of including this effect as an indicator variable in the regression. End-row turn times can also be modelled, simply by adopting the median value of 0.57 minutes per occurrence. The average operation consists of one forager and two transport units, for an hourly cost of 376 Euro. For the productivities recorded in the study, the harvesting cost of chips delivered to the farm centre varies between 8 and 40 Euro/green tonne, with an average value of 15 Euro/green tonne (2 Euro/GJ).

4. Discussion and conclusions – Rasprava sa zaklju~cima The average yield of the fields harvested during the trials is about 20 green tonnes/ha year, equivalent to 8 oven-dry tonnes/ha for a 60% average moisture content, measured in the laboratory. This value increases to 9.2 odt/ha if the figures from the three evidently failed plantations are removed. It is interesting to notice that the average yield of the Italian short-rotation stands approaches the top yield measured further north for the best poplar (Karacic et al. 2003, Pellis et al. 2004) and willow (Nordh and Verwijst 2004) clones. Moreover, the best fields encountered in the study reached annual yields in the range of 15 odt/ha, proving good quality of the Ital-

135

R. SPINELLI et al.

Harvesting Short-Rotation Poplar Plantations for Biomass Production (129â&#x20AC;&#x201C;139)

Fig. 5 Harvesting cost as a function of crop density, row spacing and machine type Slika 5. Tro{kovi pridobivanja u odnosu na gusto}u stabala, razmak redova i tip stroja ian clones and climate. These values are net of the harvesting losses, which averaged 0.6 odt/ha. The study also shows the need to optimize operations. Over a quarter of the total worksite time is occupied by unproductive delays, which may be reduced with improved planning and maintenance. In particular, waiting time can be curtailed by a better balance of the operation, which should be designed so that the capacity of the support fleet matches that of the forager. This entails predicting both productivities, which vary according to several factors, and namely: payload capacity and forwarding distance for the shuttles, crop density and forager model. The data contained in this study allow predicting the productivity of the forager, whereas that of the tractor is simpler to estimate. Maintenance is also an important source of delays, and seems to take more time when harvesting 2-year-old plantations, due to higher mechanical stress on the harvester. Although this observation lacks statistical confirmation, it must be remembered that the observation of a distinct pattern in breakdowns would require longer term studies, and that the somewhat blurred picture (p = 0.12) transpiring from this study might just be the outline of a real phenomenon. Modified foragers can reach a very high productivity, with peak values up to 80 green tonnes per hour, excluding turns and delays. However, top performance is only obtained when several factors concur, and namely: good terrain conditions, adequate machine choice, high crop density and appropriate row spacing. The forager is a heavy machine that cannot traffic wet or sloping soils, and should only

136

be applied to flat and solid terrain. The most powerful version in the tested range seems to have a remarkable edge on the other units, especially when harvesting single-row two-year-old plantations: fitted with the heavier GBE-1 head, the Mega 880 can reach twice the productivity of the other models combined (84 gt/h vs. 41 gt/h) and the difference is statistically significant to p <0.0001. Of course, such a high productivity might also be the result of a different crop structure, as the 880 model operated on fields established according to the new single-row plantation module. The new plantations are designed to produce fewer and bigger stems, with a higher fibre content: this might be paid with a lower annual yield, due to the less intense exploitation of available space, as Table 2 seems to indicate. Modified foragers cannot harvest stems that are too big and too close: cut stems have to be placed horizontally to enter the chopper, and if they are too long and too near to each other, they often get entangled with the uncut stems ahead, jamming into the header. This problem does not occur with small stems, which are shorter and more flexible, so that their tops bend and the butts can be fed horizontally to the chopper. Therefore, effective harvesting of large-size stems requires an accordingly large spacing, so that the tops of cut stems can sneak between the standing crop ahead and the stems can be laid horizontally. Similarly, row distance must follow strict rules, because the forager-based harvesting system is quite rigid with respect to crop spacing. Both the Claas and the GBE SRF headers have been built for harvesting twin rows placed 75â&#x20AC;&#x201C;80 cm apart: any significant variation in row spacing makes harvesting difficult or even imCroat. j. for. eng. 29(2008)2

Harvesting Short-Rotation Poplar Plantations for Biomass Production (129–139)

possible. The distance between twin rows must also be adequate to allow machine traffic, and generally between 2.4 and 2.8 m. These same headers can also harvest single-row plantations, by working slightly offset to the row alignment, but in this case row spacing must be 3 m: typically these plantations are managed on two-year-rotations and produce larger stems. Harvesting cost vary with the same factors listed above. The relationships calculated from this study allowed building a simple deterministic model to predict harvesting productivity and cost as a function of: crop density, row length, machine type and expected level of delays. The model also calculates transport fleet balance and operation relocating time, using the measured forager road speed of 22 km/h. Figure 5 shows the result of a simple simulation, calculated for a forwarding distance of 2 km, a 20% incidence of delays, a relocation distance of 10 km and an average lot size of 5 ha. Depending on crop density, row spacing and machine type, the total harvesting cost including forwarding to a collection site, preparation and relocation ranges from 10 to 40 Euro/green tonne (1.3 to 5.4 Euro/GJ). Harvesting cost is restrained below the 15 Euro/green tonne (2 Euro/GJ) ceiling only if field stocking exceeds 40 or 50 gt/ha when rows are long 300 and 100 m, respectively. The most powerful forager is fast enough to compensate for slightly lower stocking levels, and can harvest fields with only 30 gt/ha within the 15 Euro/gt limit.

5. References – Literatura Alig, R., Adams, D., McCarl, B., Ince, P., 2000: Economic potential of short-rotation woody crops on agricultural land for pulp fiber production in the United States. Forest Products Journal 50(5): 67–74. Bergstrand, K. G., 1991: Planning and analysis of forestry operation studies. Skogsarbeten Bulletin 17: 1–63. Björheden, R., Apel, K., Shiba, M., Thompson, M. A., 1995: IUFRO Forest work study nomenclature. Swedish University of Agricultural Science, Dept. of Operational Efficiency, Garpenberg, 16 p. Danfors, B., Ledin, S., Rosenqvist, H., 1998: Short-rotation willow coppice grower manual. Swedish Institute of Agricultural Engineering, Uppsala, Sweden. Facciotto, G., Schenone, G., 1998: Poplar as a source of renewable energy (Il pioppo fonte di energia rinnovabile) Sherwood 35: 19–26. Frison, G., Bisoffi, S., Allegro, G., Borelli, M., Giorcelli, A., 1990: Short rotation forestry in Italy: past experience and present situation. IEA/BA Task V Energy Production System Workshop. ISP, Casale Monferrato, Italy, 42 pp.

Croat. j. for. eng. 29(2008)2

R. SPINELLI et al.

Hartmann, H., Böhn, T., Maier, L., 2000: Naturlebene biogene festbrennstoffe – Umweltrelewante Eigenschaften und Einflussmöglichkeiten. Umwelt & Entwicklung Nr. 154. Bayerisches Staatsministerium für Landesentwicklung und Umweltfragen, München, Germany. Heller, M., Keoleian, G., Volk, T., 2003: Life cycle assessment of a willow bioenergy cropping system. Biomass and Bioenergy 25(2): 147–165. Hoogwijk, M., Faaij, A., Van der Broek, R., Berndes, G., Gielen, D., Turkenburg, W., 2003: Exploration of the ranges of the global potential of biomass for energy. Biomass and Bioenergy 25(2): 119–133. Karacic, A., Verwjist, T., Weih, M., 2003: Above-ground woody biomass production of short rotation populus plantations on agricultural land in Sweden. Scandinavian Journal of Forest Research 18(5): 427–437. Kofman, P., 1995: Siwork 3: User Guide. Danish Forest and Landscape Research Institute, Vejle, Denmark, 37 pp. Larsson, S., Melin, G., Rosenqvist, H., 1998: Commercial harvest of willow wood chips in Sweden. Proceedings of International Conference Biomass for Energy and Industry, June 8–11, 1998, Würzburg. Editor: CARMEN, Rimpar, Germany, 200–203. Londo, M., Roose, M., Dekker, J., De Graaf, H., 2004: Willow short-rotation in multiple land-use systems: evaluation of four combination options in the Dutch context. Biomass and Bioenergy 27(3): 205–221. Miyata, E. S., 1980: Determining fixed and operating costs of logging equipment. General Technical Report NC-55. Forest Service North Central Forest Experiment Station, St. Paul, MN, 14 pp. Nordh, N. E., Verwjist, T., 2004: Above-ground biomass assessments and first cutting cycle production in willow coppice. Biomass and Bioenergy 27(1): 1–8. Olsen, E., Hossain, M., Miller, M., 1998:. Statistical Comparison of Methods Used in Harvesting Work Studies. Oregon State University, Forest Research Laboratory, Corvallis, OR. Research Contribution 23: 1–31. Pellis, A., Laureysens, I., Ceulemans, R., 2004: Growth and production of a short rotation coppice culture of poplar I. Biomass and Bioenergy 27(1): 9–19. SAS Institute Inc. 1999. StatView Reference. SAS Publishing, Cary, NC. ISBN-1-58025-162-5, pp. 84–93. Tharakan, P., Volk, T., Abrahamson, L., White, E., 2003: Energy feedstock characteristics of willow and hybrid poplar clones at harvest age. Biomass and Bioenergy 25(6): 571– 580.

Note – Bilje{ka This model is available for free, and can be requested to the authors at: spinelli@ivalsa.cnr.it

137

R. SPINELLI et al.

Harvesting Short-Rotation Poplar Plantations for Biomass Production (129–139)

Sa`etak

Pridobivanje biomase sje~om {umskih plantaà topola u kratkim ophodnjama [ume kratkih ophodnji na poljoprivrednim zemlji{tima smatraju se industrijskim usjevima. One su podignute radi proizvodnje velikih koli~ina drvnoga materijala niske cijene te su velik potencijalni izvor biomase za pridobivanje energije. Na sjeveru Italije poljoprivrednici sve ve}u pa`nju usmjeravaju prema energetskim usjevima i {umama kratkih ophodnji. Slijede}i iskustva poljoprivrednika skandinavskih zemalja, osnivaju se {umske plantaè velikoga broja stabala po povr{ini koje se sijeku prilago|enim sila`nim kombajnima svakih nekoliko godina. Pri tom se za podizanje {umskih plantaà koriste ponajprije klonovi topola, za razliku od {umskih vrbovih plantaà u Skandinaviji. Regionalna vlada Lombardije me|u prvima je odobrila nov~ane potpore za osnivanje i gospodarenje {umama kratkih ophodnji. Slijedile su je ostale regionalne vlade, ali s manjom razinom financiranja (tablica 1). Rezultat svega je osnivanje 3000 ha {uma kratkih ophodnji u manje od ~etiri godine u Lombardiji, {to predstavlja ¾ ukupne povr{ine {uma kratkih ophodnji u Italiji. Cilj je rada istraìti u~inkovitost prilago|enih sila`nih kombajna pri sje~i {uma kratkih ophodnji, ustanoviti tehni~ke pote{ko}e pri njihovu radu, predloìti rje{enja za ve}u u~inkovitost sila`nih kombajna pri sje~i stabala iz {uma kratkih ophodnji te odrediti utjecajne ~imbenike na tro{kove rada. Istraìvanje je provedeno na 4 razli~ita tipa sila`nih kombajna Claas opremljena specijaliziranim ètvenim glavama za sje~u drvenastih vrsta (slika 1). Claas je razvio dva tipa takvih ètvenih glava za skandinavsko trì{te (HS-1 i HS-2). One su opremljene kru`nim pilama ve}ih promjera na podno`ju, dok su dva vertikalna uvla~na valjka zamijenjena ~vrstim ~eli~nim hvata~ima koja dovode posje~eni drvni materijal do usitnjiva~a. U Italiji je razvijen novi tip teè i ja~e ètvene glave (GBE-1) koja je pogodnija pri sje~i ve}ih stabalaca. Sustav pridobivanja drvne biomase iz {umskih plantaà uklju~uje uz sila`ni kombajn i najmanje dva traktora s prikolicama za prihvat drvnoga iverja i prijevoz do stovari{ta ili mjesta utovara na transportna vozila (slika 2). Istraìvanje je provedeno u 16 topolovih {umskih plantaà kojima se gospodari u jednogodi{njoj ili dvogodi{njoj ophodnji. Mnogi poljoprivrednici danas prelaze na dvogodi{nju ophodnju {umskih plantaà jer se njihovom sje~om dobiva ve}i udio drvnih vlakana u pridobivenom drvnom iverju. [umske plantaè kojima se gospodari u jednogodi{njoj ophodnji osnovane su s gusto}om sadnje od 10 000 stabala po hektaru, u dvostrukim redovima, s razmakom od 1,8 do 2,7 m izme|u dvostrukih redova, 75 cm izme|u dva reda te 45 cm unutar jednoga reda (slika 3). Srednji sje~ni promjer iznosi 2 – 3 cm, odnosno najvi{e 6 – 8 cm. [umske se plantaè dvogodi{nje ophodnje osnivaju s gusto}om sadnje od 6000 do 7000 stabala po hektaru, u jednostrukim redovima, s razmakom sadnje 2,8 – 3,0 m izme|u redova i 0,5 – 0,7 m unutar reda. Srednji sje~ni promjer iznosi 10 – 12 cm. Razmak i duljina redova izmjereni su laserskim daljinomjerom. Proveden je studij rada i vremena pri sje~i topolovih {umskih plantaà na svim ispitnim plohama. Utro{ak je vremena radnih sastavnica mjeren terenskim ra~unalom Husky Hunter pomo}u ra~unalnoga programa Siwork 3. Podaci su statisti~ki obra|eni u ra~unalnom programu ANOVA. Mjerenjem dimenzija kontejnera na prikolici i mase svakoga punoga tovara prikolice na mosnoj vagi ustanovljeni su obujam i masa pridobivenoga drvnoga iverja. Sadràj je vlage odre|en su{enjem uzoraka drvnoga iverja tijekom 48 sati na temperaturi od 103°C. Tablica 2 daje opis 16 ispitnih ploha te ostvareni u~inak radnih strojeva. U~inci se kre}u u rasponu od 9 tona svjeè tvari na sat do 44 tone svjeè tvari na sat. Velik je raspon rezultata uzrokovan razli~itim prekidima rada zbog mehani~kih kvarova, umora radnika i ~ekanja traktora s prikolicom za rad sila`noga kombajna. U ra{~lambi radnoga vremena (slika 4) efektivno vrijeme ~ini oko 70 % ukupnoga utro{ka vremena. Ispravnost je stroja odre|ena kao odnos utro{enoga radnoga vremena (bez vremena odràvanja stroja) i ukupnoga utro{enoga vremena rada te se govori o mogu}nosti stroja za rad s drvenastim biljkama u odnosu na krmna bilja i kukuruz. Rezultati u tablici 3 ukazuju na ve}i udio vremena odràvanja stroja i ~e{}e prekide pri radu u {umskim plantaàma dvogodi{njih ophodnji. Zaklju~ak je da ve}i promjeri stabala uzrokuju ve}i mehani~ki stres stroja. Provedena regresijska analiza na 115 tovara prikolice pokazuje da se vrijeme utro{eno na pridobivanje punoga tovara prikolice drvnim iverjem smanjuje s gusto}om sadnje stabala i uporabom sila`noga kombajna ve}e snage

138

Croat. j. for. eng. 29(2008)2

Harvesting Short-Rotation Poplar Plantations for Biomass Production (129–139)

R. SPINELLI et al.

motora (tablica 5). Testom uz pomo} programa ANOVA utvr|eno je zna~ajno ve}a u~inkovitost sila`noga kombajna Claas Mega 880 od ostalih tipova. Tro{kovi su pridobivanja drvnoga iverja odre|eni na osnovi procijenjenih 1200 pogonskih sati sila`noga kombajna i 500 pogonskih sati ètvene glave. Investicijski tro{kovi iznose 250 000 EUR, odnosno 140 000 EUR s vremenom amortizacije od 8 godina. Tro{kovi popravka i odràvanja procijenjeni su na iznos od 70 % od nabavne cijene. Tro{ak radnika iznosi 16 EUR/h, a tro{ak goriva 0,90 EUR/L (subvencionirana cijena za poljoprivrednike). Na osnovi navedene kalkulacije ukupni tro{kovi iznose 234 EUR/h za sila`ni kombajn, odnosno 71 EUR/h za traktor s prikolicom. Kako rad pridobivanja drvnoga iverja iz {umskih plantaà podrazumijeva uporabu sila`noga kombajna i najmanje 2 traktora s prikolicom, ukupni tro{ak iznosi 376 EUR/h. Za ostvarene u~inke tro{kovi pridobivanja drvnoga iverja iznose od 8 EUR/t svjeè tvari do 40 EUR/t svjeè tvari, odnosno prosje~no 15 EUR/t svjeè tvari. Na osnovi energetske vrijednosti drvnoga iverja od 18,5 GJ/t (Hartmann i dr. 2000) najve}i tro{ak pridobivanja drvnoga iverja moè iznositi pribli`no 2 EUR/GJ. Istraìvanje ukazuje na potrebu optimizacije sustava pridobivanja drvnoga iverja sila`nim kombajnima iz {uma kratkih ophodnji. Vi{e od ~etvrtine ukupnoga radnoga vremena otpada na neproizvodne prekide, {to se mora umanjiti pobolj{anim planiranjem rada. Zastoje zbog ~ekanja traktora s prikolicom treba izbje}i boljom organizacijom rada koja }e se temeljiti na utjecajnim ~imbenicima, kao {to su veli~ina tovara prikolice, udaljenost prijevoza, gusto}a sadnje i tip sila`noga kombajna. Prilago|eni sila`ni kombajni mogu posti}i vrlo veliku u~inkovitost pri povoljnim terenskim uvjetima, velikoj sje~noj gusto}i i prikladnim razmacima redova. Sila`ni su kombajni te{ki strojevi koji ne mogu raditi na vla`nim tlima i nagnutim terenima, ve} se preporu~uje njihov rad na ravnim terenima dobre nosivosti tla. Tako|er se javljaju pote{ko}e pri radu sila`noga kombajna ako su stabla ve}ih promjera i/ili su na malom me|usobnom razmaku. Pri malom razmaku stabala ~esto dolazi do nagnje~enja i nakupljanja neposje~enih stabala u ètvenoj glavi, {to uzrokuje prekid rada. Stoga u~inkovita sje~a stabala ve}ih promjera iz {uma kratkih ophodnji zahtijeva odre|eni razmak izme|u i unutar redova. Pri dvostrukim redovima sadnje specijalizirane ètvene glave mogu ispravno raditi pri razmaku 75 – 80 cm unutar dvostrukih redova i 2,4 – 2,8 m izme|u dvostrukih redova. Potreban razmak jednostrukih redova iznosi 3 m za pravilan rad ètvene glave. Jednostruki se redovi primjenjuju pri gospodarenju {umama kratkih ophodnji u dvogodi{njim ophodnjama, {to pretpostavlja i ve}e promjere stabala te se stoga zahtijeva i ve}i razmak izme|u redova. Sila`ni kombajn najve}e snage motora Class Mega 880 pri pridobivanju drvnoga iverja iz dvogodi{njih {uma kratke ophodnje zasa|enih u jednostrukim redovima mogu posti}i dvostruko ve}i u~inak u odnosu na ostale tipove. Rezultati se istraìvanja mogu koristiti za odre|ivanje u~inkovitosti sustava pridobivanja drvnoga iverja iz {uma kratkih ophodnji na osnovi sje~ne gusto}e, duljine redova, tipa sila`noga kombajna i o~ekivanih prekida rada. Slika 5 prikazuje tro{kove pridobivanja drvnoga iverja u odnosu na gusto}u stabala, razmak redova i tip sila`noga kombajna. Pri tome su pretpostavljeni ovi ulazni podaci: udaljenost prijevoza traktora s prikolicom od 2 km, udaljenost premje{tanja od 10 km, povr{ina {umske plantaè od 5 ha i udio prekida od 20 % u ukupnom vremenu rada. Ukupni se tro{ak pridobivanja drvnoga iverja kre}e od 10 do 40 EUR/t svjeè tvari (1,3 do 5,4 EUR/GJ). Tro{ak manji od 15 EUR/t svjeè tvari (2 EUR/GJ) mogu}e je posti}i jedino pri sje~noj gusto}i od 40 gt/ha pri duljini redova od 300 m, odnosno 50 gt/ha pri duljini redova od 100 m. Jedino sila`ni kombajn ve}e snage motora moè ostvariti tro{kove manje od 15 EUR/t svjeè tvari pri sje~noj gusto}i od samo 30 t (svjeè tvari)/ha. Klju~ne rije~i: {uma kratke ophodnje, pridobivanje biomase, sje~a, sila`ni kombajn, Italija

Authors’ address – Adresa autorâ:

Received (Primljeno): July 28, 2008 Accepted (Prihva}eno): November 11, 2008 Croat. j. for. eng. 29(2008)2

Raffaele Spinelli, PhD. e-mail: spinelli@ivalsa.cnr.it Carla Nati, PhD. e-mail: nati@ivalsa.cnr.it Natascia Magagnotti, BSc. e-mail: magagnotti@ivalsa.cnr.it CNR – Ivalsa via Madonna del Piano Pal. F I-50019 Sesto Fiorentino ITALY

139

108

Croat. j. for. eng. 29(2008)1

Orginal scientific paper â&#x20AC;&#x201C; Izvorni znanstveni rad

Damage to Young Forest Due to Harvesting in Shelterwood Systems Bo{tjan Ko{ir Abstract â&#x20AC;&#x201C; Nacrtak Different types of shelterwood system and group selection forests were studied to discover the extent of damage caused by logging. Motor-manual cutting and mainly tractor skidding were included. Sampling transects were used to estimate the damage to young forest and remaining stands. The whole research area was regenerated on average 31%, of which 21% was damaged. We found a higher density of designated and undesignated skid trails on larger regeneration areas. Damage to young forest and damage to remaining productive stands were compared. In this respect the whole rotation period was divided into three time intervals, the first of which designated a mixed pattern of young forest area and younger phases prior to commercial thinning, in which the last of the old mature trees are removed. The second phase is a mix of currently productive stands and some young forest, in which the first and second commercial thinnings begin, until the final stage, in which young forest becomes increasingly abundant and perspective. In the last period, damage to productive stands is high (around 70%), since they have accumulated over a long time period. The fact that better forest stand opening with skid trails means less damage to young forest, but slightly more damage to mature stands suggests the conclusion that the abundance and position of young forest patches should dictate the density and position of skid trails. Keywords: shelterwood, young forest, damage, harvesting, motor-manual, skid trail

1. Introduction â&#x20AC;&#x201C; Uvod The share of close-to-nature stands in Slovenia is fairly high, on average. Clear felling with artificial regeneration has been forbidden since 1949 (Perko 2005). The idea of co-natural practice has been slowly developing and this development can be followed through many forest acts since then. In recent decades, the actual practice has mainly been inspired by silvicultural practice (selective thinning) in other countries (Mlin{ek 1968). There have, however, been some modifications (Kotar 2005), which were necessary because the terrain, sites, forest stand conditions, ownership and other factors of forest management, including technical level, differ from those in other countries. Natural regeneration is the premise on which the philosophy of silviculture rests. The way of regeneration follows the goals of each owner, be they uniform shelterwood forest on large areas, shelterwood forest with group regeneration or regeneration more like selection forest (Matthews 1999). Adaptation to the natural conditions is thus very good and flexible. Cutting starts with the first comCroat. j. for. eng. 29(2008)2

mercial thinning (often too late from the silvicultural point of view) and ends when the entire area under the remains of the old stand is covered with young forest. Private forests are managed according to the wishes and needs of the owner. There are, of course, many exemptions (Kr~ 1999), where the actual situation requires artificial intervention in the regeneration process. The prevailing technology is motor-manual cutting and extraction with adapted agricultural tractors or cable skidders, some of them made in Slovenia. The cut-to-length method has already been used in many places, but the technology is still developing. Modern all-terrain cable cranes are used in some more mountainous parts of the country. The shelterwood system, just like selection forest, requires very good accessibility of forest with forest roads, skid roads and skid trails. In state forests there are more than 25 m/ha of forest roads and around 100 m/ha of permanent skid roads (Ko{ir 2003). Accessibility in private forests is at least half that in state forests (Medved 2003).

141

B. KO[IR

Damage to Young Forest Due to Harvesting in Shelterwood Systems (141–153)

In this article, skid roads and skid trails will be combined in the same category of designated skid trails. Designated skid trails can be understood as potentially or even predominantly permanent forest thoroughfares (depending on terrain features) – skid roads. Designated skid trails on difficult terrain are regularly built before forest operations commence, on easy terrain they are marked in the stand and on the map, and trees are removed. On the other hand, undesignated skid trails are not permanent thoroughfares, since they serve the actual needs of thinning and they are chosen by the tractor operator. Undesignated skid trails are not built, but can be sometimes a necessity if the pattern of permanent skid trails (designed through operational planning) is inadequate. Silvicultural planning is a valuable tool that has been used in state forests for at least three decades. It can be seen through prescriptions in the silvicultural plan whether a specific young forest area is important enough to be part of future stand development or not. It is essential to know in which direction felled trees can be moved without damaging the young forest and remaining trees. In some cases, this simply cannot be done. Timber harvesting is performed many times during the rotation period and, consequently, damage to the remaining stand accumulates and tends to reach as much as the total number of trees (Ko{ir and Cedilnik 1996, Ko{ir 1996, Ko{ir 2001). The practice of too frequent thinning has already been criticised, and possibilities of improvement discussed (Ko{ir 1998a, Ko{ir 1998b, Ko{ir 2001). Many stands, mostly in small patches, have a proportion of young forest in addition to more or less adult trees. According to silvicultural goals, this young forest is a potential candidate for a new stand in the next rotation period. Such young forest is also subject to damage, which is less visible, but equally important. We therefore measured damage to the young stand under the cover of the old stand and combined this damage with the damage to the productive stand. It is equally important to make studies of damage due to traditional motor-manual and tractor skidding, in order to have a reliable basis for future comparisons with more advanced mechanized cutting and timber forwarding techniques. Damages are connected with moving – falling or trasport of heavy loads. When felling trees, the direction of fall can be somehow chosen so as to work safely, enable further transport and minimize damage. Directional felling is therefore dependent upon tree chacteristics and position in relation to skid trail network. Damages are directly related to greater or smaller probability of contact between the load and remaining trees or regenerated area. In this context

142

different degrees and extent of damages could be expected. Our principal goal has been however to reveal the degree of probability for damage to appear in mature (Ko{ir 2001) and in regenerated stands.

2. Methods – Metode Damaged young forest has been considered as a regenerated area on the transect on which healthy new forest cannot develope – in the future. This happens for two reasons: 1) damaged young plants (uprooted, broken, etc.) or 2) displaced or damaged upper layers of soils together with plants. In both cases previous studies indicated that area (in m2) is a better measurement criteria than the number of damaged plants. This paper only deals with studies that were conducted in state forests. Research took place at four sites, in order to balance terrain and stand structure differences: Þ NW site with mostly Alpine conditions, predominantly spruce stands; Þ NE site with mountainous terrain, spruce is the predominant species; Þ SE site with hilly and ravine terrain, stands of beech and other broadleaf species; Þ SW site on High Karst, silver fir – beech forest predominates. Among many cutting units, those in which normal thinning had taken place (no salvage cuttings etc.) were selected. We studied the working units for which good silvicultural and operational plans were available. Most of the stands had the characteristics of different shelterwood systems, but in some areas of silver fir – beech forest on High Karst, certain peculiarities of group selection forest distinguished these stands from the others. Damage caused by logging was measured immediately after operations had been finished. In some places time studies took place before damage was measured. In each working unit we first surveyed the area and analysed the density of the designated skid trails, together with skid trails used by the tractor driver without previous planning (undesignated or »wild« skid trails). The maps of working units with skid trails position were the basis for making a sampling plan along the used skid trails. We used a systematic sampling method of 20–40 m long and 4 m wide transects (20 m in young forest only, 30 m in polewood only, 40 m in mature stand possibly with young forest), which lay perpendicular to skid trails at a distance of 50 m (Robek and Ko{ir 1996, comparable to: Han and Kellog 2000). Croat. j. for. eng. 29(2008)2

Damage to Young Forest Due to Harvesting in Shelterwood Systems (141–153)

B. KO[IR

Table 1 Number of transects according to age class Tablica 1. Broj primjernih ploha po dobnim razredima Age class, years Dobni razredi, godine <20 20–39 40–59 60–79 80–99 100–119 120–139 >140 Total – Ukupno

Share of transects with Number of transects with young forest in number of of transects with damaged young forest Number of all transects Number all strips, % young forest Ukupan broj primjernih Broj primjernih ploha s Broj primjernih ploha s Udjel primjernih ploha s ploha o{te}enim mladim mladim sastojinama u mladim sastojinama sastojinama odnosu na ukupan broj primjernih ploha, % 1 1 – 100 11 1 – 9 112 5 4 4 422 53 42 13 385 100 66 26 90 41 26 46 26 7 5 27 121 74 56 61 1168 281 199 24

The aim of the method was to measure the damage in the remaining stand (count of injuries larger than 10 cm2) and also to estimate the damage to young forest area, which was present in many cases under the cover of adult trees. For each logging unit a string of variables was gathered, calculated, estimated or measured. Young stand damage included: broken, severely bent, compressed, injured or in any other way disturbed young trees, with a poor hope of recovery. The following assumptions were made when choosing the logging unit in which measurements were taken: Þ Measurements were carried out in state owned stands in which silvicultural planning has been updated and operational planning is obligatory; Þ Operations were conducted with professional machinery, equipment and skilled workers employed by forest enterprises; Þ Minimum cut per hectare in the working unit was 15 m3/ha; Þ Young forest damage was recorded when such young forest was described in silvicultural plans as potentially perspective or important. This article presents an analysis of young stand damage measured in the same transects as damage to productive stand. The age class was computed on the basis of the number of trees in each transect, on premises taken from common growth tables for the purpose of classifying each transect in the stand, since the majority of Slovenian forests are more or less mixed in terms of age and tree species. Croat. j. for. eng. 29(2008)2

Share of transects with damaged young forest in number of transects with young forest, % Udjel primjernih ploha s o{te}enim mladim sastojinama u odnosu na broj primjernih ploha s mladim sastojinama, % – – 80 79 66 63 71 76 71

3. Results – Rezultati 3.1 Young forest damage – O{te}enje mlade sastojine The research covered 51 working units with an average 23 transects per unit. The average working unit had 11.45 ha, which makes a total area of 584 ha covered by sampling in different parts of the country. Within this area, around 140 ha of young forest was found and measured for damage (Table 1). The assumption that young forest is not important during the »thinning« phase is simply not true. The applied mosaic structure gives importance to almost any presence of young forest. This raises the question of whether this silvicultural practice is really co-natural and whether it is rational. The often discussed question is whether we still need a trained forester for forest management, if the real situation is a randomized (managed by nature) pattern of development phases. Each working unit was classified into a forest management type such as: uniform shelterwood system, group shelterwood system or group selection forest. This classification is approximate, since in reality many stands do not follow a strict pattern, bearing in mind the size and natural regeneration method. A much larger sample would be required for investigating distinctions among these tree forest management systems. In the analysis we, therefore, used an average value for all three systems, since they are really all shelterwood-systems. The age classes in the following tables and figures are approximate, but the analysis showed that age classes give us a good general picture.

143

B. KO[IR

Damage to Young Forest Due to Harvesting in Shelterwood Systems (141â&#x20AC;&#x201C;153)

Fig. 1 Number of transects according to age class and share of transects with young forest Slika 1. Broj primjernih ploha s obzirom na dobni razred i udjel primjernih ploha s mladim sastojinama

At the beginning of the new rotation period (age class less than 39 years) the share of regenerated area should normally be larger, but in older classes this dependency becomes weaker (Fig. 1). In the very late stage of the stand, the share of regeneration area reaches only one third of the stand area, but more than 60% of all observed strips. This is presumably due to random dispersion of young forest patches under the canopy of older stands. Such a random pattern makes silvicultural and operational planning difficult, not to speak of environmentally friendly logging. Optimal access to felled timber and directional felling can hardly be planned in detail, but forest operations must be performed without unnecessary damage. In all transects where young forest was found (Tables 1 and 2, Fig. 1) some damage to young forest was recorded. The main source of damage is skidding, in which heavy loads are moved through the stand. Felling and processing trees causes some damage, but this damage can be understood as a constant when comparing different skidding means (Table 3). Tractor skidding is made entirely on the ground, which is why this method results in higher damage to young forest. Cable cranes normally lift the front end of the timber, which reduces the contact between the load and the ground surface. We

144

have paid more attention to tractor skidding, since this is the predominant skidding form in the research area. On average, the regenerated area in stands with more than 50% conifers was 34% and in stands where broadleaf prevail 27%. Analysis showed that damaged young forest in coniferous stands is 19% and in broadleaf stands 21%, which is not a significant difference. Analysis also showed that in coniferous stands the average cut was 31 m3/ha and in predominantly broadleaf stands 41 m3/ha per thinning. These cut concentrations are low, but we must take into account that thinning is done repeatedly, often more than once in ten years of forest management plan period. The impact of the logging technique on remaining stand has to be understood as complex (Leinss 1991, Frohm 1993, Spinelli 1999). Skidding, as a part of logging, means transport through the forest. This means also taking into consideration, in addition to machines and techniques, the way the machine approaches the tree or timber. A closer look at the true interrelation between silvicultural and operational planning in respect of young forest damage can be achieved by examining tractor skidding only. Designated skid trails in coniferous stands were 161 m/ha and in broadleaf stands 139 m/ha. Undesignated Croat. j. for. eng. 29(2008)2

Damage to Young Forest Due to Harvesting in Shelterwood Systems (141–153)

B. KO[IR

Table 2 Distribution of young forest damage according to age class and forest management system Tablica 2. Raspodjela o{te}enja mladih sastojina po dobnim razredima i na~inu gospodarenja Group selection forest system Group shelterwood system Uniform shelterwood system Average Grupimi~no preborno Oplodne sje~e na malim Oplodne sje~e na velikim Prosje~no gospodarenje povr{inama povr{inama 2 Young forest area – Povr{ina mladih sastojina, m /ha <20 – – – – 20–39 – – 500 500 40–59 1400 – 750 1140 60–79 2455 2158 4321 2842 80–99 2747 2862 3186 2934 100–119 2279 5077 3278 3385 120–139 1500 2417 – 2286 >140 2097 4467 4200 3430 Average – Prosje~no 2387 3311 3621 3056 Damaged young forest – O{te}ene mlade sastojine, m2/ha <20 – – – – 20–39 – – – – 40–59 228 – 750 359 60–79 492 475 658 538 80–99 557 718 642 635 100–119 416 1025 564 667 120–139 750 194 – 305 >140 327 1030 882 713 Average – Prosje~no 448 735 722 627 Damaged young forest, % of regenerated area – O{te}ene mlade sastojine, % od pomladne povr{ine <20 – – – – 20–39 – – – – 40–59 16 – 100 31 60–79 20 22 15 19 80–99 20 25 20 22 100–119 18 20 17 20 120–139 50 8 13 >140 16 23 21 21 Average – Prosje~no 19 22 20 21 Share of stripes with young forest, % of all stripes number – Udjel primjernih ploha s mladim sastojinama, % od ukupnoga broja primjernih ploha <20 – – – – 20–39 – – 13 9 40–59 14 – 6 4 60–79 12 17 10 13 80–99 24 32 24 26 100–119 48 59 32 46 120–139 14 50 0 27 >140 59 60 64 61 Average – Prosje~no 25 26 22 24 Age class, years Dobni razredi, godine

Croat. j. for. eng. 29(2008)2

145

B. KO[IR

Damage to Young Forest Due to Harvesting in Shelterwood Systems (141–153)

Table 3 Average of damaged regeneration area according to skidding means Tablica 3. Prosje~no o{te}enje pomladne povr{ine s obzirom na sredstvo privla~enja drva Age class, years Dobni razredi, godine 40–59 60–79 80–99 100–119 120–139 >140 Average – Prosje~no

Skidding means – Sredstvo privla~enja drva Tractor Cable crane Average Traktor @i~ara Prosje~no % of damaged regeneration area % o{te}ene pomladne povr{ine 31 – 31 20 9 19 21 19 20 24 13 20 45 7 13 21 21 21 21 16 20

skid trails were also used in coniferous stands to a greater extent (55 m/ha) than in predominantly broadleaf stands (44 m/ha). On average, the relative proportion between designated and undesignated skid trails shows that a greater share of undesignated skid trails can be expected in the middle of the rotation period (Fig. 2)

and at the very end of stand life. Young forest expressed as a regeneration area percentage has an increasing tendency toward the end of the rotation period, and the same is true of the share of damaged young forest area. The principal goal of operational planning is to designate enough skid trails for normal work, and no undesignated skid trails should be tolerated. This is, of course, more theory than practice. During our measurements it was often difficult to assess whether a particular undesignated skid trail had been used because of inadequate operational planning, or was a »wild« skid trail, used by a tractor driver for higher efficiency or comfort without respect to stand and young forest damage. The impact of secondary forest opening on young forest damage is shown in Fig. 3. The dependencies are fairly reliable, and the conclusions that follow are: Þ The larger the area of regeneration, the larger the area of damaged young forest. Þ The larger the area of regeneration, the higher must be the density of designated skid trails. Þ The larger the area of regeneration, the more skid trails were recorded. In addition to a higher density of designated skid trails, a higher density of undesignated skid trails also occurs. The two trends are almost parallel.

Fig. 2 Planning and actual use of skid trails as a potential cause of young forest damage Slika 2. Planirane i stvarno kori{tene traktorske vlake kao mogu}i uzrok o{te}enja sastojine 146

Croat. j. for. eng. 29(2008)2

Damage to Young Forest Due to Harvesting in Shelterwood Systems (141–153)

Þ Relative damage to young forest rises with increased regeneration area, and reaches its maximum at 1700 m2/ha (= 17% of the stand area) and thereafter decreases again. The relation between access to felled timber and the possibility of damage to young forest is influenced by the position and abundance of existing regeneration areas and possible positions of skid trails. Although designated skid trails try to avoid regenerated areas, damage to young forest tends to increase from designated skid trail densities ranging between approximately 100 m/ha and 140 m/ha. Higher densities obviously arrange timber transport more successfully, in such a way that relative damage to young forest decreases to about 20% of the regenerated area. Designated skid trails have a stronger impact on relative damage to young forest: too low densities require higher densities of unplanned passes of tractors and cause more damage to the regenerated area, while higher densities are reflected in lower relative damage to young forest. It is a controversial issue but the results show that when the density of undesignated skid trails increases, damage to young forest decreases, on average. The problem can also be understood in this way: tractor skidding has two main operations that risk causing damage – timber bunching with a winch and traveling empty or loaded with timber towards the forest

B. KO[IR

road. Damage from dragging timber is concentrated along skid trails, while bunching is the main cause of damage between skid trails, where damage to young forest has been assessed. When skid trail density is higher, there is less bunching (shorter distances) and, consequently, less damage. A higher skid trail density, on the other hand, means greater temporary loss of productive forest area.

3.2 All stand damages – O{te}enje cijele sastojine In this section, the connection between productive stand damages and young forest damages is discussed, since these damages occur at the same time and should not be treated separately. The boundary between young forest and productive stands was 10 cm DBH. Damage to the remaining stand tends to increase through the rotation period. This has already been proven by model (Ko{ir and Cedilnik 1996, Ko{ir 1996) and field observations (Ko{ir 2000). The main reason for this is the accumulating nature of damage. A comparison of total stand damage (damage to trees that accumulates during the rotation period) due to harvesting with damage to young forest of the same age shows that total stand damage clearly increases with the forest age, while young stand damage does not show such a tendency.

Fig. 3 Relations between designated and undesignated skid trails and damage to young forest Slika 3. Ovisnost planiranih i neplaniranih traktorskih vlaka o o{te}enosti mladih sastojina Croat. j. for. eng. 29(2008)2

147

B. KO[IR

Damage to Young Forest Due to Harvesting in Shelterwood Systems (141–153)

The results show that the share of damage to young forest (new generation) decreases towards the end of the regeneration period and, at the same time, the share of damage to pole-wood starts to increase. When a new regeneration period starts, the share of damage to new young forest starts to climb again. It seems that different factors govern damage to young forest and in young stands where more or less normal thinning has begun. The answer might be simple: damage to remaining stand is expressed as the number of damaged trees (or as the share of remaining trees), while damage to young forest is expressed as damaged area (or as the share of damaged area). The number of trees per hectare and area of canopy projection (cover) are connected, since the cover is close to 1 in dense stands during the rotation period, while at the end of the rotation period, when regeneration cutting begins, the cover is much less (tends to reach 0 at the end of regeneration felling, observing only the old stand). These trends are logical, but have to be confirmed with more measurements, so as to make clearer the distinction between young forest and the early stages of a new developing stand. In this respect, the whole rotation period can be divided into three time intervals (Fig. 4 and Fig. 5): Þ In the first phase, when the young forest develops and slowly becomes a future stand, damage goes from zero (when no cutting takes place), increasing momentarily as the

last of the old trees are removed, and decreasing because there is a time when little or no thinning is done. Þ The second phase, when young forest is present but is not perspective, is when the existing stand is of interest. The first commercial thinning begins. Damage rises according to cutting intensity and skid trail planning and use. The first cutting causes the first stand damage to trees (DBH>10 cm), which are mingled in groups with young forest stages. Thinning continues and stand damage accumulates correspondingly. Þ In the final stage, when the young forest becomes perspective, damage decreases as ever fewer adult trees are present in the regenerated area. The majority of adult trees have already been damaged, many of them more than once. The analysis showed that there is no significant difference in young forest distribution among age classes based on forest management systems. This is due to the mosaic distribution of development phases, whereby the distinction between different forest management types is not always transparent. It is of major importance for practical foresters to know which part of the forest is to be regenerated in the current and next cutting, and where the best routes for timber transport are today and in the future. The question related with the previous chapter is whether there is a connection between forest skid

Fig. 4 Development of damage in a forest stand according to approximate age Slika 4. Razvoj o{te}enja sastojine s obzirom na njezinu pribli`nu dob 148

Croat. j. for. eng. 29(2008)2

Damage to Young Forest Due to Harvesting in Shelterwood Systems (141â&#x20AC;&#x201C;153)

B. KO[IR

Fig. 5 Development of relative damage in a forest stand according to approximate age Slika 5. Razvoj relativne o{te}enosti sastojine s obzirom na njezinu pribli`nu dob

Fig. 6 Damages to a productive stand (DBH>10 cm) and young forest in relation to skid trail density Slika 6. Ovisnost o{te}enja proizvodnoga dijela sastojine (DBH>10 cm) i mlade sastojine o gusto}i traktorskih vlaka trail density and damage to stands â&#x20AC;&#x201C; on young and on mature trees. The impact of skid trail density on damage to young forest has already been discussed above. Here we compare this relation to the correlation between skid trail density and new damage to the remaining stand. All new damage corresponds Croat. j. for. eng. 29(2008)2

to all injured remaining trees, regardless of the fact that some of them have already been injured during previous thinning. Fig. 6 shows damage to a productive stand in relation to skid trail density. The conclusions are as follows: better forest stand accessibility with tractor

149

B. KO[IR

Damage to Young Forest Due to Harvesting in Shelterwood Systems (141–153)

skid trails means less damage to young forest, but slightly more damage to the mature stand. The two curves match at a skid trail density around or a little more than 200 m/ha.

4. Discussion and conclusions – Rasprava i zaklju~ci Further analysis should be focused on damages due to new harvesting methods. Damages of mature stands have been already studied. A more detailed study should be carried out of damages of young forest under the cover in selection cutting. The transect method should be replaced with other types of sampling like circular areas. Mechanized cutting brings several significant changes into our expectation of probable young forest appearance: 1) easier directional felling (this fact is in favor of smaller damages), 2) shortwood production and extraction and 3) heavy machines with greater ground pressure in combination with 4) greater share of area necessary for machine movement (this fact increases the probability of damaging young forest). We can also expect that the types of damaged young forest will change (use of broad tires, track chains for rear drive and chains for front drive). The used sampling method is suitable for motor-manual felling and tractor or cable skidding, where the density of secondary forest opening is relatively low or, in other words, where the average distance between skid trails is relatively large. From our field measurements and from other sources (Butora and Schwager 1986) we know that the possibility of greater damage is higher close to skid trails and damage is lower midway between skid trails. This method of sampling would not be suitable for the cut-to-length method, since skid trails are very dense, and each strip would cross several skid trails. It would be better to use the Frödig method, which can be easily adapted to our needs (Frödig 1992, Ko{ir and Robek 2000). The sampling method should change in future research, when different harvesting and silvicultural systems will be compared. Young forest is important as the foundation of the next stand and carries information to the future. It will be necessary to organize measurements in such a way as to enable us to make a reliable distinction between different types of forest management, but also to include different types of damages. This will be possible in practice, but also difficult, because of the mosaic mixture of tree composition and development phases. Young forest appears under the cover of more or less mature trees in almost every development phase of stands, regardless of the forest manage-

150

ment system. When first thinning begins, damage to the remaining stand together with young forest starts to accumulate. New damage to the remaining stand is divided into injuries to previously undamaged trees, and injuries to those that have already been injured more than once. This damage accumulation is more evident and easier to measure on trees that compose the actual productive stand, while old damage to young forest is less visible, which means that we can only assess current damage to young forest. The density of young forest is high, as well as the vitality. In general it is impossible to determine the influence of a certain share of damaged young trees on the future stand development, since new seedlings fill the gaps and replace missing or severely damaged young trees. Tree felling, bunching and skidding are sensitive operations in terms of damage to young forest, especially in a mature stand. Proper forest area opening is therefore essential. Silvicultural and operational planning have precisely this task – to make timber transport not only technically possible, but also environmentally friendly – causing the least possible damage. A comparison of the relative area of damaged young forest to total regenerated area (m2/ha) shows that a larger area of damaged young forest can be expected on a larger regeneration area. The trend is not linear since relative damage to young forest rises with increased regeneration areas, and after a maximum at around 17% of the regenerated area, decreases again. With a higher share of regeneration area, the need for more detailed forest opening (higher density of designated skid trails) is more important. This is also reflected in the fact that we found a higher density of undesignated skid trails in such cases. Links between previous research on remaining stand damage (DBH>10 cm) and young forest damage is also interesting since it is connected with numerous questions of reasonable and close-to-nature forest management. We divided the whole rotation period into three time intervals, the first of which is a mixed pattern of young forest area and younger phases, mostly prior to commercial thinning, when the last of the old mature trees are removed. The second phase represents a mix of present productive stand and some young forest, which is normally less important as a productive or protection function carrier, when the first and second commercial thinning take place (and productive stand damages start to increase) until the final stage, when young forest becomes more and more abundant and perspective. Productive stand damage is high since it has been accumulating over a long period. The fact that better forest stand opening with skid trails means less Croat. j. for. eng. 29(2008)2

Damage to Young Forest Due to Harvesting in Shelterwood Systems (141–153)

damage to young forest but slightly more damage to the mature stand suggests that the abundance and position of young forest patches actually dictate the density and position of skid trails.

5. References – Literatura Butora, A., Schwager, G., 1986: Holzernteschaden in Durforstungbestanden. Berichte, 288, Birmensdorf, Eidgenossische Anstalt fur das forstliche Verzuchswezen, 51 p. Diaci, J., McConnell, S., 1996: Close-to-nature forestry and ecosystem management. Zb. gozd. lesar. 49: 105–127. Dvoøak, J., 2005: Variability of Tree Damage with Respect to Felling – Technological Factors that can be Changed in Short Term. Proceedings: FORMEC 2005, Ljubljana, 139–146. Eriksson, L., 1981: Strip roads and Damages Caused by Machines when Thinning Stands. The Swedish University of Agricultural Sciences, Dep. of Operational Efficiency, Garpenberg 1992, Rep. No. 193: 1–44. Frödig, A., 1992:. Thinning damage – A study of 403 stands in Sweden in 1988. The Swedish University of Agricultural Sciences, Dep. of Operational Efficiency, Garpenberg 1992, Rep. No. 193: 1–45. Frohm, S., 1993: Efficient and Safe Thinning. In: Efficient, Sustainable and Ecologically Sound Forestry, Skogforsk Report 5: 43–49. Halaj, J., Grék, J., Pánek, F., Petrá{, R., Øehák, J., 1987: Rastové tabul’ky hlavnýh drevín ^SSR. Príroda, Bratislava, p. 362. Han, H. S., Kellog, L. D., 2000: A Comparison of Sampling Methods for Measuring Residual Stand Damage from Commercial Thinning. J. of For. Eng. 11(1): 63–71. Harstela, P., 1995: Environmental impacts of wood harvesting in Nordic countries. Environmental impacts of Forestry and Forest Industry. EFI Proc. 3: 37–44. Ivanek, F., 1976: Vrednotenje po{kodb pri spravilu lesa v gozdovih na Pohorju. IGLG, Strokovna in znanstvena dela 51, Ljubljana, 142–147. Ko{ir, B., 1996: How to manage thinning with low damages of standing trees – experience from the model. Proceedings »Planning and implementing forest operations to achive sustainable forests« 19th Annual Meeting of COFE & IUFRO SG S3.04–00, July 29–August 1, 1996, Marquette, Michigan USA, 82–91.

B. KO[IR

Ko{ir, B, Robek, R., 2000: Characteristics of the stand and soil damage in cut-to-lenght thinning on the @ekanc working site (SW Slovenia). Zb. gozd. lesar. 62: 87–115. Ko{ir, B., 2000: Primerjava rezultatov modela po{kodb drevja v sestoju zaradi pridobivanja lesa in terenskih opazovanj. – Research Reports, University of Ljubljana, Biotechnical Fac., Dep. of Forestry and Forest Resources, 62: 135–151. Ko{ir, B., 2001: Frequent thinning – impact on stand quality. In: Thinnings: a valuable forest management tool., Montreal, Quebec, Canada, September 2001, Canadian forest service, 2003. Ko{ir, B., 2003: Wood harvesting technologies in regional forest management plans for the period from 2001 to 2010. In: Obmo~ni gozdnogospodarski na~rti in razvojne perspektive slovenskega gozdarstva: conference proceedings. Ljubljana: Biotechnical Faculty, Department of Forestry and Renewable Forest Resources, 153–165. Ko{ir, B., 2004: Factors affecting technological changes. Gozd. vestn. 62(1): 3–11. Kotar, M., 2005: Zgradba, rast in donos gozda na ekolo{kih in fiziolo{kih osnovah. Zveza gozdarskih dru{tev Slovenije in Zavod za gozdove Slovenije, Ljubljana, 500 p. Kr~, J., 1999: Analysis of Forest Quantities and Species Composition Alteration Using Two Different Methods with Comparisons. Zb. gozd. lesar. 60: 211–236. Krivec, A., 1975: Racionalizacija delovnih procesov v se~nji in izdelavi ter spravilu lesa glede na delovne razmere in po{kodbe. Research Reports, University of Ljubljana, Biotechnical Fac., Dep. of Forestry, 13, 2, Ljubljana, 145–193. Leinss, C., 1991: Unterzuchungen zur Frage der nutzungstechnischen Folgen nach Fall-und Ruckeshaden bei Fichte (Picea abies /L./ Karst.). Mitteilungen der Forstlichen Versuchs- und Forschungsanstalt Baden-Wurttemberg, Freiburg im Breisgau, Heft 157, 172 S. Matthews, J. D., 1999: Sylvicultural systems. Oxford Univ. Press, Oxford, 98–137. Medved, M., 2003: Property conditions and forest operations in private forests. Gozd. vestn. 61(9): 347–359. Mlin{ek, D., 1968. Spro{~ena tehnika gojenja gozdov na osnovi nege. Poslovno zdru`enje gozdnogospodarskih organizacij, Ljubljana, Jugoslavenski poljoprivredni {umarski centar, Beograd, 117 p.

Ko{ir, B., Cedilnik, A., 1996: The model of number increasing of tree damages at thinnings. Zb. gozd. lesar. 48: 135–151.

Mlin{ek, D., 1977: Übertragbarkeit und die Bedeutung des Prinzipes der Nachhaltigkeit und der Theorie der Waldpflege für die Naturgerechte Bewirtschaftung von erneuerbaren Naturgutern. Die Waldpflege in der Mehrweckforstwirtschaft. Österreichischer Agrarverlag, Wien, 45–57.

Ko{ir, B., 1998a: Damage to mountain spruce stands due to harvesting. Conference proceedings »Gorski gozd«, Ljubljana: Biotechnical Faculty, Department of Forestry and Renewable Forest Resources, 95–107.

Mlin{ek, D., 1994: Was ist naturnahe Waldwirtschaft?. In: HATZFELDT, Hermann Graf (Ed.). Ökologische waldwirtschaft : Grundlagen – Aspekte – Beispiele (Alternative Konzepte, 88). Heidelberg: C. F. Müller, 67–76.

Ko{ir, B., 1998b: Critical evaluation of frequent thinnings from the aspect of energy consumption and damage in the stands. Zb. gozd. lesar. 56: 55–71.

Nicholls, A., Bren, L., Humphreys, N., 2004: Harvester Productivity and Operator Fatique: Working Extended Hours. Int. J. of For. Eng. 15(2): 57–65.

Croat. j. for. eng. 29(2008)2

151

B. KO[IR

Damage to Young Forest Due to Harvesting in Shelterwood Systems (141–153)

Perko, F., 2005: Trpeli so na{i gozdovi. Ljubljana, Zalo`ba Jutro, 327 p.

Biotechnical Fac., Dep. of Forestry and Forest Resources, 52: 119–136.

Robek, R., Ko{ir, B., 1996: Razvoj metode vzor~nega ocenjevanja motenj pri izkori{~anju gozdov. Izzivi gozdne tehnike, Zbornik mednarodnega posveta, UL, GIS,SZ, Ljubljana, 73–81.

Siren, M., 1999: One-Grip Harvesting Operations, Silvicultural Results and Possibilities to predict Tree Damage. In: Proc. IUFRO 3.09.00 Harvesting and Economic of Thinnings, Ennis, Ireland, 152–167.

Robek, R., Medved, M., 1997: Po{kodbe drevja zaradi izvajanja gozdarskih del po podatkih popisov propadanja gozdov Sloveniji. Research Reports, University of Ljubljana,

Spinelli, R., 1999: The Environmental Impact of Thinning: More Good than Bad? In: Proc. IUFRO 3.09.00 Harvesting and Economic of Thinnings, Ennis, Ireland, 136–143.

Sa`etak

O{te}enja mladih sastojina nakon pomla|ivanja metodama pod zastorom kro{anja Istraìvanje se bavi utvr|ivanjem razine o{te}enja sastojina po zavr{etku pridobivanja drva pri razli~itim na~inima gospodarenja {umama (oplodne sje~e na malim, odnosno velikim povr{inama te grupimi~ne preborne sje~e). Istraìvanje je provedeno na ~etiri podru~ja radi obuhvata terenskih i sastojinskih razli~itosti: Þ SZ dio Slovenije, ve}inom alpski predjeli, prevladavaju smrekove sastojine Þ SI dio Slovenije, planinski predjeli, prevladavaju}a je vrsta smreka Þ JI dio Slovenije, razvedeni brdski tereni, sastojine bukve i tvrdih lista~a Þ JZ dio Slovenije, podru~je visokoga kr{a, jelovo-bukove sastojine. Obuhva}eni sustav pridobivanja drva zasnovan je na ru~no-strojnoj sje~i i izradbi drva motornom pilom te na privla~enju drva skiderom. U ovom radu traktorske vlake i traktorski putovi shva}aju se kao jedinstvena kategorija – planirane traktorske vlake, koje predstavljaju mreù mogu}ih ili ~ak preteìto trajnih (izgra|enih) sekundarnih {umskih prometnica u skladu s operativnim planovima. Neplanirane se traktorske vlake ne grade, sluè trenuta~nim zahtjevima proreda, a posljedica su odluka voza~a skidera koje on donosi zbog nezadovoljavaju}ega rasporeda planiranih traktorskih vlaka u sje~noj jedinici. Za procjenu o{te}enosti mlade sastojine (mladik, letvik) i preostalih dube}ih stabala nakon sje~e kori{tena je metoda uzorkovanja primjernim plohama. Mladom o{te}enom sastojinom na primjernim plohama smatra se sastojina koja se u budu}nosti ne moè razviti u novu zdravu {umu zbog dvaju razloga: 1) o{te}enosti mladih biljaka, 2) premje{tanja i o{te}ivanja gornjih slojeva tla zajedno sa samim biljkama. U oba slu~aja studije su pokazale da je za mjeru o{te}enosti bolje koristiti jedinicu povr{ine (m2) nego broj o{te}enih biljaka (stabala). Nastala su o{te}enja mjerena odmah po zavr{etku radova. U svakoj je istraìvanoj sje~ini analizirana gusto}a planiranih traktorskih vlaka, ali i gusto}a neplaniranih (tzv. »divljih«) traktorskih vlaka koje su napravili sami voza~i skidera, te njihov utjecaj na o{te}enost sastojine. Sistematskim uzorkom primjerne su plohe {irine 4 m postavljane okomito na traktorske vlake s me|usobnim razmakom od 50 m. Duljina je primjernih ploha u mladim sastojinama (letvik) iznosila 20 m, u srednjodobnim sastojinama (stupovlje) 30 m te u zrelim sastojinama 40 m. Pri izboru sje~ina prikladnih za istraìvanje vodilo se ra~una da: Þ sastojine pripadaju dràvnim {umama koje podlijeù uzgojnim radovima i u kojima je operativno planiranje obvezno Þ radovi budu izvo|eni s profesionalnim strojevima i opremom te sa {kolovanim radnicima zaposlenima u {umarskom poduze}u Þ najmanja sje~na gusto}a bude 15 m3/ha Þ budu zabiljeèna o{te}enja mladih sastojina koje su s uzgojnoga stajali{ta smatrane va`nim i perspektivnim. O{te}enja su mjerena u 51 sje~noj jedinci, s prosjekom od 23 primjerne plohe po jedinici. Ukupno su uzorkom obuhva}ena 584 ha (prosje~no 11,45 ha po sje~ini). Od ukupne povr{ine uzorka mlade sastojine zauzimaju 140 ha ili 31 % (tablica 1). Pretpostavka da su o{te}enja stadija mladih sastojina pri izvo|enju proreda neva`na, nije to~na jer upravo mozai~na struktura istraìvanih sastojina postavlja pitanje da li su ti uzgojni radovi opravdani. Na svim

152

Croat. j. for. eng. 29(2008)2

Damage to Young Forest Due to Harvesting in Shelterwood Systems (141–153)

B. KO[IR

primjernim plohama s razvojnim stadijima mlade sastojine (tablica 1 i 2, slika 1) prona|ena su o{te}enja ~iji je glavni uzrok bio privla~enje drva, dok sje~a i izradba uzrokuju manje o{te}enje koje je stalno kada se usporede razli~ita sredstva privla~enja drva (tablica 3). O{te}enje je mlade sastojine kod ~etinja~a iznosilo 19 %, a kod lista~a 21 % povr{ine. Gusto}a mreè planiranih traktorskih vlaka u sastojinama ~etinja~a iznosila je 161 m/ha, a u sastojinama lista~a 139 m/ha. Gusto}a mreè neplaniranih traktorskih vlaka u sastojinama ~etinja~a je 55 m/ha te 44 m/ha u sastojinama lista~a. Radi lak{ega poimanja o{te}enja razvojnih stadija mladih sastojina (udjel od povr{ine) koje se razvijaju pod zastorom kro{anja zrelih stabala, koja su tako|er predmet o{te}enja (broja stabala), ophodnja je podijeljena u tri faze (slike 4 i 5): Þ Mlada sastojina raste i polako po~inje predstavljati budu}u sastojinu, o{te}enja se kre}u od 0 (u slu~aju izostanka sje~e) uz porast do trenutka sje~e zadnjega zreloga stabla te se opet smanjuju zbog rijetkih proreda. Þ Mlada je sastojina formirana i zapo~inju komercijalne prorede, o{te}enja rastu s porastom intenziteta sje~e i gusto}e mreè traktorskih vlaka. Prva sje~a uzrokuje i prva o{te}enja na stablima (prsni promjer >10 cm) koja su grupimi~no uklopljena s mladim stadijima budu}e {ume. Prorede se nastavljaju i o{te}enja sukladno tomu rastu. Þ U posljednjoj fazi o{te}enja se smanjuju jer je sve manje zrelih stabala za sje~u na pomladnoj povr{ini. Ve}ina zrelih stabala ima vi{e od jednoga o{te}enja. Ve}a gusto}a traktorskih vlaka omogu}uje bolju kretnost i mogu}nost prihvata drva sredstvima sekundarnoga transporta tako da se o{te}enost mlade sastojine smanji na 20 % pomladne povr{ine. Istraìvanje je pokazalo da se kod pove}ane gusto}a neplaniranih traktorskih vlaka o{te}enja u sastojini u prosjeku smanjuju. U slu~ajevima kada je gusto}a planiranih traktorskih vlaka velika, manje su udaljenosti skupljanja drva vitlom, a time i manje o{te}enja, no ipak se s ve}om gusto}om gubi i vi{e proizvodne povr{ine {umskoga tla. U prosjeku se o~ekuje vi{e neplaniranih traktorskih vlaka u sredini i na kraju ophodnje (slika 2). Rezultati pokazuju da se udio o{te}enja u mladoj sastojini smanjuje kroz ophodnju, a u isto vrijeme se pove}ava u srednjodobnoj sastojini. Razlog tomu je {to se o{te}enja u mladoj sastojini prikazuju udjelom o{te}ene povr{ine (m2), dok se o{te}enja dube}ih stabala prikazuju brojem o{te}enih stabala. Budu}a bi se istraìvanja trebala usmjeriti na mehaniziranu sje~u i izradbu drva. O{te}enja dube}ih, zrelih stabala za sje~u ve} su istraìvana te treba istraìti o{te}enja mladih sastojina (pod zastorom kro{anja stabala). Strojna sje~a i izradba drva sustavom harvester – forvarder donosi nekoliko va`nih promjena u predvi|anjima strukture o{te}enja budu}ih mladih sastojina: Þ lak{e usmjereno ru{enje stabala, {to dovodi do manjih o{te}ivanja Þ obvezna primjena sortimentne metode izradbe drva (forvarder) Þ te{ka vozila s velikim dodirnim tlakovima Þ ve}i udjeli povr{ine pod mreòm sekundarnih {umskih prometnica nu`nih za kretanje vozila. Tako|er se moè o~ekivati da }e se vrste o{te}enja mladih sastojina promijeniti zbog upotrebe {irih guma i lanaca na prednjim i stra`njim pogonskim kota~ima. Bolja otvorenost {uma zna~i i manje o{te}enja mladih sastojina, ali i vi{e o{te}enja zrelih stabala, {to name}e zaklju~ak da poloàj i zastupljenost pomladne povr{ine i te kako trebaju utjecati na gusto}u i poloàj traktorskih vlaka. Klju~ne rije~i: oplodne sje~e, mlade sastojine, ru~no-strojno pridobivanje drva, traktorske vlake

Author’s address – Autorova adresa:

Received (Primljeno): April 7, 2008 Accepted (Prihva}eno): November 11, 2008 Croat. j. for. eng. 29(2008)2

Prof. Bo{tjan Ko{ir, PhD. e-mail: bostjan.kosir@bf.uni-lj.si University of Ljubljana, Biotechnical Faculty Department of Forestry and Forest Resources Ve~na pot 83 1000 Ljubljana SLOVENIA

153

108

Croat. j. for. eng. 29(2008)1

Orginal scientific paper – Izvorni znanstveni rad

Growth of Pedunculate Oak Seedlings under Soil Contamination by Mineral and Biodegradable Oils Milan Or{ani}, Dubravko Horvat, Nikola Pernar, Marijan [u{njar, Darko Bak{i}, Damir Drvodeli} Abstract – Nacrtak The purpose of this study is to research the effects of different concentrations of mineral and biodegradable oils for chainsaws on seedling growth of pedunculate oak (Quercus robur L.) during two-year vegetation period. The paper also examines the influence of the above mentioned oils and their concentrations on the development of the seedling root system at the end of the second vegetation. Tests were carried out in forest nurseries on seven 1 m2 plots. Three plots were treated, after the planting of the pedunculate oak acorns, with biodegradable oil in the concentrations of 0.1 L/m2, 0.2 L/m2, and 0.5 L/m2. The other three plots were treated with mineral oil of the same concentrations. The control plot received only seed sowing. The analysis of the root system of the two-year-old seedlings of pedunculate oak was carried out with WinRHIZO ProLA 2400 software. The SAS and Statistica 7 program packages were used for the statistic data processing and for determination of statistically significant differences among variables. Results showed that the lower oil concentration the larger root lengths. The lowest heights of the two-year-old pedunculate oak seedlings were measured from the test plots treated with mineral oil and the highest on those from the plots treated with bio-oil, although the heights of seedlings from the control plots were very close to heights of those treated with bio-oil. Keywords: pedunculate oak seedlings, mineral oil, biodegradable oil, root system, growth

1. Introduction – Uvod Oils for lubricating power chainsaw chains are called »Total Loss Oils«, because they end completely and irreversibly in sawdust on timber surface and the soil, or arrive on the surface of leaves of the surrounding plants. Skoupý (2004) established that 75–77% of the oil was absorbed by sawdust, 7–13% adhered to the surface of the cut timber, while 12–16% ended up in the surface. Biodegradable chainsaw oils appeared in 1986. They can be made either of base-fluids of artificial origin (saturated and unsaturated esters), or of natural vegetable oils and animal fats. The most common base-fluid is rape oil, though other base-fluids can also be used. At present, pine oil, also named tall oil, Croat. j. for. eng. 29(2008)2

is being investigated in Finland (Takalo and Lauhanen 1994). It has been assessed that the total quantity of chainsaw lubricants discharged into the environment in Finland amounts to 2 million litres whereas in Croatia this quantity is about 420,000 litres per year (Anon 1996). According to the data obtained from the company »Hrvatske {ume«, which manages 80% of the total forest areas, the unit consumption of chainsaw oil amounts to 0.168 L/m3, which is far more than in Finland where due to highly mechanised felling and processing the unit consumption ranges from 0.015 L/m3 to 0.027 L/m3. Horvat and [u{njar (2003) established a considerably lower unit consumption of biodegradable and mineral oil for

155

M. OR[ANI] et al.

Growth of Pedunculate Oak Seedlings under Soil Contamination by Mineral ... (155â&#x20AC;&#x201C;162)

lubricating chainsaw chains. Thus, in the final cut of pedunculate oak they used 0.07 L/m3, in thinning of beech stands the respective amount was 0.04 L/m3, while 0.035 L/m3 was required for the selective cut of fir stands. Takalo and Lauhanen (1994) established a considerably lower wear of the lower parts of chain teeth when the chain was lubricated with biodegradable oil, than when it was lubricated with mineral oil. However, their low oxidation stability requires low operating temperatures. Oil viscosity is inversely proportionate to the temperature, which can lead to problems with the lubricant flow quality during winter. In the summer, the incidence of too abundant lubrication is possible due to decreased operating viscosity of oil. The use of additives with biodegradable oils can improve their viscosity index (change of viscosity with the change of temperature) which in that case can be compared to the one of mineral oils if not even better (Augu{tin et al. 2000). Lauhanen et al. (2000) treated plants with biodegradable oils in the laboratory and found no harmful effects. However, the interviewed operators, who used mineral and biodegradable oils, reported a much higher rate of skin and allergic diseases when biodegradable oils were used. While investigating the impacts of spilling two concentrations (0.002 L/m2 and 4.0 L/m2) upon the germination of Scots pine (Pinus sylvestris L.) seeds in glasshouse over three weeks, Lauhanen and Kolppanen (2003) established that lower concentrations of both oil types decreased the germination and that mineral oil was less suitable. A higher concentration (4 L/m2), which corresponds to an incidental oil spilling, was lethal for all seeds.

Or{ani} et al. (2008) investigated the influence of 3 concentrations (0.1 L/m2, 0.2 L/m2, 0.5 L/m2) of chainsaw oils (biodegradable and mineral) on germination of oak acorns in two nurseries. Results showed that biodegradable oil, regardless of the concentration, does not significantly reduce nursery germination of oak acorns, and in one nursery it lead to an increase of germination. Considering the harmful impacts of oil upon the soil and plants, it should be pointed out that a film is formed upon the surface composed of firm particles. The film prevents the contact of the particles with water and air. Impeded breathing of oots, metabolic disorders, and even root dieback were the established consequences (Ba{i} et al. 1999). Accordingly, the purpose of this study is the research on the impacts of different concentrations of both mineral and biodegradable oils on the growth of pedunculate oak seedlings (Quercus robur L.) during the two years vegetation in the nursery. The impact of different oil concentration of mineral and biodegradable oils on certain morphological characteristics of the root system of two-year-old seedlings would also be examined.

2. Materials and methods â&#x20AC;&#x201C; Materijal i metode The experiment was set in forest nursery. The soil in the nursery is classified as lowland pseudogley (stagnosol). In the upper 30 cm, the soil is a clayey loam by its texture. Deeper down, it acquires a slightly heavier texture, and turns into light clay. The soil reaction is between neutral and slightly acid. The upper 10 cm of the soil is moderately supplied

Fig. 1 Test plots Slika 1. Pokusne plohe 156

Croat. j. for. eng. 29(2008)2

Growth of Pedunculate Oak Seedlings under Soil Contamination by Mineral ... (155–162)

M. OR[ANI] et al.

Table 1 Mean values of root characteristics Tablica 1. Srednje vrijednosti zna~ajki korijena Root characteristic Zna~ajka korijena

Oil type – Vrsta ulja

Oil concentracion – Koncentracija ulja

Biodegradable Biorazgradivo

Mineral Mineralno

Control Kontrola

0 L/m2

0.1 L/m2

0.2 L/m2

0.5 L/m2

Root lenght, mm Duljina korijena, mm

626.67

501.53

632.33

572.60

560.81

558.87

Root diameter, mm Promjer korijena, mm

1.22

1.24

1.17

1.30

1.17

1.22

Root volume, mm Obujam korijena, mm

7.37

5.99

7.14

7.39

6.05

6.60

No. of seedlings Broj sadnica

with humus. The contents of the organic matter decrease in relation to the soil depth. The forest nursery contained seven 1 m2 test plots (Fig. 1). First, the planting of pedunculate oak acorns was carried out. The acorns were covered with the soil from the nursery, with the layer thickness of two acorn diameters. Upon covering the seeds, the plots were treated with different concentrations of biodegradable and mineral oils. Using a compressor and a sprinkler, the oil was evenly sprayed over the test plots. Three plots were treated with biodegradable oil in the concentrations of 0.1 L/m2, 0.2 L/m2, and 0.5 L/m2. The other three plots were treated with mineral oil of the same concentrations. The control plot received only seed sowing. The height and the root collar diameter of twoyear-old seedlings were measured. The root collar diameter was measured with a digital movable measurer of an accuracy of 0.01 mm, while the heights were measured with a measuring rod, the accuracy of which was 1 mm. During vegetation, the usual nursery tending routine was carried out with the exception of root cutting. After two vegetations of growth 10 seedlings were taken out from each test pot (the total of 70 seedlings). After washing the particles the root system was scanned by Epson Expression 10000XL and then analyzed using WinRHIZO ProLA2400 software for analyzing washed roots. The SAS and Statistica 7 program packages were used for the statistic data processing.

control plot (not treated with oil) had the best average root length of 632.33 cm (Table 1). The root length of seedlings from the bio treated plots was on average by 125.14 cm longer compared to the seedlings from the mineral oil treated plots (501.53 cm). The total root length increased with lower oil concentration. With the oil concentration of 0.5 L/m2 it was 558.87 cm, with the oil concentration of 0.2 L/m2 it was 560.81 cm and with 0.1 L/m2 the root length was 572.60 cm (Fig. 2). The univariant analysis revealed a significant difference (p = 0.0324) in the total root lengths with respect to the treatment (bio and mineral oils) while

3. Research results – Rezultati istra`ivanja The analysis of the root system obtained using WinRHIZO ProLA2400 software revealed that the two-year-old seedlings of pedunculate oak from the Croat. j. for. eng. 29(2008)2

Fig 2 Impact of different oil types and concentrations on root lengt Slika 2. Utjecaj razli~itih vrsta i koncentracija ulja na duljinu korijena 157

M. OR[ANI] et al.

Growth of Pedunculate Oak Seedlings under Soil Contamination by Mineral ... (155–162)

Table 2 Univariant analysis of root characteristics Tablica 2. Univarijantna analiza zna~ajki korijena Root characteristic Zna~ajka korijena

Effect Djelovanje Treatment – Tretiranje Root length Concentration – Koncentracija Duljina korijena Error – Gre{ka Treatment – Tretiranje Root diameter Concentration – Koncentracija Promjer korijena Error – Gre{ka Treatment – Tretiranje Root volume Concentration – Koncentracija Obujam korijena Error – Gre{ka

Sum of squares Suma kvadrata 234898 2208 3197000 0.001002 0.180857 2.016676 28.4516 17.9345 624.6202

the oil concentration itself did not significantly affect this property (Table 2). The seedlings grown on mineral oil treated plots had the largest average root diameter of 1.24 mm. An average root diameter of seedlings from bio oil treated plots was on average by 0.05 mm larger than that of seedlings from control plots (1.17 mm). Seedlings from the plots treated with 0.1 L/m2 had the largest average root diameter, while those from the plots treated with 0.5 L/m2 had the average root diameter of 1.22 mm, followed by root diameter of 1.17 mm on plots treated with 0.2 L/m2 (Fig. 3). The univariant analysis did not show any significant differences in the average seedling root diameter with respect to the treatment (p = 0.5718) and oil concentration (p = 0.0613) The largest average root volume (7.37 cm3) was measured on seedlings from the plots treated with bio oil and the smallest (5.99 cm3) on those from mineral oil treated plots. The root volume of seedlings from the control plot was 7.14 cm3. Certain regularities between an average root volume and average root diameter have been proved related to oil concentration. The largest root volume of 7.39 cm3 was measured with seedlings from the plots with oil concentration of 0.1 L/m2, followed by the average root volume of 6.60 cm3 from the plots treated with 0.5 L/m2 and 6.05 cm3 for the plots treated with 0.2 L/m2 (Fig. 4). The univariant analysis did not show any significant differences in the average root volume of seedlings with respect to the treatment (p = 0.0901) and oil concentration (p = 0.3985). The results of the descriptive statistics show that seedlings from mineral oil treated plots had the poorest average heights (Table 3). Considering the concentration of spilling oil on test plots, the best heights (698.11 mm) were obtained with the 0.2 L/m2

158

Degree of freedom Stupanj slobode 1 2 65 1 2 65 1 2 65

Variance Varijanca 234898 1104 49184.6 0.001002 0.090428 0.031026 28.4516 8.9672 9.6095

F value F vrijednost 4.7758 0.0224

p-value p vrijednost 0.0324 0.9778

0.3229 2.9146

0.5718 0.0613

2.9607 0.9332

0.0901 0.3985

Fig 3 Impact of different oil types and concentrations on root diameter Slika 3. Utjecaj razli~itih vrsta i koncentracija ulja na promjer korijena

of biodegradable oil, followed by seedlings from the control plot (653.03 mm) and the plot treated with 0.5 L/m2 of biodegradable oil (630.69 mm). The poorest heights (476.83 mm) were measured on seedlings from the test plot treated with mineral oil in concentration of 0.5 L/m2 (Fig. 5). At the end of the second growth vegetation, the largest average root collar diameter was measured on the control plot seedlings (7.90 mm), followed by those from the bio-oil treated plots (7.75 mm at the 0.1 L/m2 concentration, 7.17 mm at the 0.2 L/m2 concentration, 7.19 mm at the 0.5 L/m2 concentraCroat. j. for. eng. 29(2008)2

Growth of Pedunculate Oak Seedlings under Soil Contamination by Mineral ... (155â&#x20AC;&#x201C;162)

M. OR[ANI] et al.

The univariant analysis of the data revealed significant differences in the root collar diameter with respect to the type of oil while oil concentration did not have a major impact on this morphological feature. The Tukey HSD test (Table 6) showed significant differences in the seedling root collar diameter

Fig 4 Impact of different oil types and concentrations on root volume Slika 4. Utjecaj razli~itih vrsta i koncentracija ulja na obujam korijena tion). The lowest values were measured on seedlings from plots treated with mineral oil (Fig. 6). The univariant analysis (Table 4) established significant differences in seedling heights with respect to oil type and oil concentration (p = 0.00008). The Tukey HSD test (Table 5) revealed significant differences in seedling heights with respect to oil concentration in all cases except between the oil concentration of 0.2 L/m2 and 0.5 L/m2 (p = 0.7916).

Fig 5 Influence of different oil types and concentrations on height growth of seedlings Slika 5. Utjecaj razli~itih vrsta i koncentracija ulja na visinski rast sadnica

Biodegradable Biorazgradivo Mineral Mineralno Control Kontrola

No. of seedlings Broj sadnica

Oil type Vrsta ulja

Oil concentracion Koncentracija ulja

Table 3 Mean values of seedling dimensions Tablica 3. Srednje vrijednosti veli~ina sadnica

L/m2 0.1 0.2 0.5 0.1 0.2 0.5

100 100 100 80 80 80

Seedling dimension Veli~ina sadnice Root collar Height diameter Visina Promjer vrata korijena mm 622.74 7.75 698.11 7.17 630.69 7.19 484.79 6.33 479.28 6.70 476.83 6.66

653.03

Croat. j. for. eng. 29(2008)2

7.90

Fig 6 Influence of different oil types and concentrations on root collar diameter of seedlings Slika 6. Utjecaj razli~itih vrsta i koncentracija ulja na debljinski rast sadnica 159

M. OR[ANI] et al.

Growth of Pedunculate Oak Seedlings under Soil Contamination by Mineral ... (155–162)

Table 4 Univariant analysis of seedling dimensions Tablica 4. Univarijantna analiza veli~ina sadnica Seedling dimension Veli~ina sadnice

Effect Djelovanje Treatment – Tretiranje Height Concentration – Koncentracija Visina Error – Gre{ka Treatment – Tretiranje Root collar diameter Concentration – Koncentracija Promjer vrata korijena Error – Gre{ka

Sum of squares Suma kvadrata 1443741 94925 3246080 102.088 19.849 5194.78

Table 5 Influence of oil concentration on heights of seedlings – Tukey HSD test Tablica 5. Utjecaj koncentracija ulja na visinski rast sadnica – Tukey HSD test Concentration Koncentracija 0 L/m

0 L/m

0.0000

0.1 L/m

0.5 L/m

0.0000

0.2 L/m

0.1 L/m

2 2

0.2 L/m

0.5 L/m

0.0296

0.0025

0.0001

0.0296

0.0001

0.0025

0.0036

0.0036 0.7916

0.7916

Table 6 Influence of oil type on root collar diameter of seedlings – Tukey HSD test Tablica 6. Utjecaj koncentracija ulja na debljinski rast sadnica – Tukey HSD test Treatment Tretiranje

Control Kontrola

Control Kontrola Biodegradable oil Biorazgradivo ulje

0.9984

Mineral oil Mineralno ulje

0.0282

Biodegradable oil Biorazgradivo ulje

Mineral oil Mineralno ulje

0.9984

0.0282 0.0009

0.0009

between the control test plot and plots treated with mineral oils (p = 0.0282) as well as between plots treated with bio-oil and mineral oil (p = 0.0009). No significant differences in this morphological feature were established only between the control plot and plots treated with bio-oil (p = 0.9984).

4. Conclusions – Zaklju~ci The analysis of the root system of the twoyear-old seedlings of pedunculate oak carried out with WinRHIZO ProLA 2400 showed that seedlings

160

Degree of freedom Stupanj slobode 1 2 655 1 2 655

Variance Varijanca 1443741 47462 4956 102.088 9.9246 7.931

F value F vrijednost 291.321 9.577

p-value p vrijednost 0 0.00008

12.872 1.251

0.00036 0.28679

from the control plot (no oil spilling) had the best average total root length and that the lower the oil concentration the bigger the root lengths. The univariant analysis revealed significant differences in the total length of the seedling root system with respect to the type of oil while no significant differences were found for the other morphological features of the root (average diameter, volume). The same test showed that oil concentration had no influence on the total root length. The lowest heights of the two-year-old pedunculate oak seedlings were measured from the test plots treated with mineral oil and the highest on those from the plots treated with bio-oil, although the heights of seedlings from the control plot were very close to heights of those treated with bio-oil. The univariant analysis showed significant differences in seedling heights with respect to type of oil and oil concentration. Significant differences in the root collar diameter were established only between the types of oil while the oil concentrations had no major impact on this feature. Additional research should try to explain physiological condition of seedlings treated with mineral and biodegradable oil.

5. References – Literatura Anon., 1996: Razvoj i organizacija hrvatskog energetskog sektora. Knjiga 6, »Gospodarenje {umama u Hrvatskoj«, Energetski institut »Hrvoje Po`ar«, Zagreb, 1–76. Augu{tin, H., Dekani}, S., Martini}, I., Sever, S., 2000: Enviromentally friendly hydraulic fluids for forestry machines – conditions and prospects. Meh. {umar. 25(1–2): 41–58. Ba{i}, F., Kisi}, I., Mesi}, M., Butorac, A., 1999: Stanje tala i djelotvornost sanacije na mjestu puknu}a naftovoda OS Stru`ec – Rafinerija Sisak. Studija, Agronomski fakultet Sveu~ili{ta u Zagrebu, 1–42. Croat. j. for. eng. 29(2008)2

Growth of Pedunculate Oak Seedlings under Soil Contamination by Mineral ... (155–162)

Horvat, D., [u{njar, M., 2003: Application of biodegradable chain-saw lubricants. Proceedings of the 2nd International Scientific Conference »Forest and wood-processing technology vs. environment«, Brno, Czech Republic, Mendel University of Agriculture and Forestry Brno, 109–121. Lauhanen, R., Kolppanen, R., Takalo, S., Koukkanen, T., Kola, H., Valimaki, I., 2000: Effects of biodegradeble oils on forest enviroment and forest machines. Proceedings »Forest and wood technology vs. enviroment« Mendel University of agriculture and forestry Brno, Czech Republic, 203– 206. Lauhanen, R., Kolppanen, R., 2003: Effects of spent hydralic oil on the germination of scots pine seed, Proceedings of the 2nd International Scientific Conference »Forest and wood-processing technology vs. environment«, Brno, Czech

M. OR[ANI] et al.

Republic, Mendel University of Agriculture and Forestry Brno, 221–223. Or{ani}, M., Horvat, D., Pernar, N., [u{njar, M., Bak{i}, D., Drvodeli}, D., 2008: Utjecaj mineralnog i biorazgradivog ulja na rasadni~ku klijavost i rast sadnica hrasta lu`njaka (Quercus robur L.). (Influence of mineral and bio oil on the germination of acorn and the growth of pedunculate oak /Quercus robur L./ seedlings). [umarski list 132 (1–2): 3–9. Skoupý, A., 2004: Biodegradable oils used for saw chain lubrications. Ukrainskij der`avnij lisotehni~eskij universitet, Naukovij visnik 14(3): 41–49. Takalo, S., Lauhanen, R., 1994: Tall oil as a new lubricant in forestry – preliminary results and experiences. Forsitrisk, 4 – 8 July 1994, Feldafing, Germany, 1–4.

Sa`etak

Rast sadnica hrasta lu`njaka pri one~i{}enju tla mineralnim i biorazgradivim uljima Ovim se radom istraùje utjecaj razli~itih koncentracija mineralnoga i biolo{ki razgradivoga ulja za podmazivanje lanaca motornih pila na rast sadnica hrasta lu`njaka (Quercus robur L.) i razvoj korijenskoga sustava sadnica na kraju druge vegetacije. Ulja za podmazivanje lanaca motornih pila nazivaju se »total loss oils«, jer sve ulje nepovratno odlazi na povr{inu prepiljenoga drva, u piljevinu, u tlo ili zavr{ava na lisnim povr{inama okolnoga bilja. Skoupý je (2004) ustanovio da se 75 – 77 % ulja apsorbira u piljevinu, 7 – 13 % ostaje na povr{ini prepiljenoga drva i 12 – 16 % odlazi u tlo. Ukupna koli~ina ulja za podmazivanje lanaca motornih pila koja je ispu{tena u okoli{ u Finskoj je procijenjena na 2 mil. litara, dok se ta koli~ina u Hrvatskoj kre}e oko 420 000 litara godi{nje (Anon. 1996). No, prema podacima poduze}a »Hrvatske {ume«, koje gospodari s 80 % ukupne povr{ine {uma, jedini~na potro{nja ulja za podmazivanje lanca iznosi 0,168 L/m3, {to je mnogo vi{e od finskoga primjera, gdje se zbog visoko mehanizirane sje~e i izrade jedini~na potro{nja kre}e od 0,015 L/m3 do 0,027 L/m3. Horvat i [u{njar (2003) zamijetili su mnogo manju jedini~nu potro{nju biorazgradivoga i mineralnoga ulja za podmazivanje lanaca motornih pila koja se kretala od 0,07 L/m3 kod dovr{noga sijeka hrasta lu`njaka, preko 0,04 L/m3 u proredi bukove sastojine, do 0,035 L/m3 kod preborne sje~e u jelovoj sastojini. U {umskom je rasadniku osnovano 7 pokusnih ploha povr{ine 1 m2. Na pokusnim je plohama posijano sjeme. @ir je prekrivan mati~nom zemljom iz rasadnika u debljini od dva promjera sjemena. Nakon prekrivanja sjemena pokusne su plohe tretirane razli~itim koncentracijama biorazgradivoga i mineralnoga ulja. Pomo}u kompresora i brizgaljke ulje se pod tlakom ravnomjerno raspr{ivalo po pokusnim plohama. Tri su pokusne plohe tretirane biorazgradivim uljem u koncentracijama od 0,1 L/m2, 0,2 L/m2 i 0,5 L/m2 (slika 1). Sljede}e su tri pokusne plohe tretirane mineralnim uljem u istim koncentracijama. Na kontrolnoj je plohi samo posijano sjeme. U rasadniku tlo taksonomski pripada pseudogleju ravni~nomu. U gornjih 30 cm tlo je prema teksturi glinasta ilova~a, a dublje poprima neznatno teù teksturu i prelazi u laku glinu. Tlo je neutralne do slabo kisele reakcije te osrednje opskrbljeno humusom u povr{inskih 10 cm. Sadràj organske tvari opada s obzirom na dubinu tla. Nakon dva vegetacijska razdoblja na pokusnim su plohama mjereni visina i promjer vrata korijena sadnica. Promjer vrata korijena sadnica mjeren je pomo}u digitalne pomi~ne mjerke s to~no{}u 0,01 mm, dok su visine mjerene mjernom letvom s to~no{}u od 1 mm. Tijekom vegetacije obavljena je standardna njega sadnica koja se ina~e provodi u rasadnicima. Od tih radova nije obavljeno jedino podrezivanje korijena sadnica. Nakon dviju vegetacija uzgoja, sa svake pokusne plohe, izva|eno je po 10 komada sadnica, a nakon ispiranja od ~estica tla korijenski je sustav digitaliziran uz pomo} skenera Epson Expression 10000XL, dok je njegova analiza obavljena pomo}u softvera za analizu opranoga korijenja WinRHIZO ProLA2400. Za statisti~ku obradu podataka kori{teni su programski paketi SAS i Statistica 7. Analiza korijenskoga sustava sadnica dobivena uz pomo} softvera WinRHIZO ProLA2400 pokazala je kako su dvogodi{nje sadnice hrasta lu`njaka s kontrolne plohe (bez tretiranja uljima) imale prosje~no najve}u ukupnu

Croat. j. for. eng. 29(2008)2

161

M. OR[ANI] et al.

Growth of Pedunculate Oak Seedlings under Soil Contamination by Mineral ... (155–162)

duljinu korijena od 632,33 cm (tablica 1). Duljina korijena sadnica s ploha tretiranih biouljima bila je prosje~no za 125,14 cm ve}a od sadnica s ploha tretiranih mineralnim uljima (501,53 cm). Ukupna duljina korijena pove}avala se sa smanjenjem koncentracije ulja (slika 2). Univarijantnom analizom dobivena je signifikantno zna~ajna razlika u ukupnoj duljini korijena s obzirom na tretiranje (biolo{ka i mineralna ulja), dok koncentracija ulja nije zna~ajno utjecala na to svojstvo (tablica 2). Najve}i prosje~ni promjer korijena od 1,24 mm imale su sadnice uzgojene na pokusnim plohama tretiranim mineralnim uljima. Prosje~ni je promjer korijena sadnica s ploha tretiranih biouljima za 0,05 mm ve}i od sadnica s kontrolne plohe (1,17 mm). Najve}i prosje~ni promjer korijena od 1,30 mm imale su sadnice s ploha doziranih s 0,1 L/m2 (slika 3). Univarijantna analiza nije pokazala postojanje signifikantno zna~ajnih razlika u prosje~nom promjeru korijena sadnica s obzirom na tretiranja (p = 0,5718) i razli~ite koncentracije ulja (p = 0,0613). Prosje~no najve}i obujam korijena (7,37 cm3) imale su sadnice s ploha tretiranih biouljem, a najmanji (5,99 cm3) s ploha tretiranim mineralnim uljem. Obujam korijena sadnica s kontrolne plohe iznosio je 7,14 cm3 (slika 4). Univarijantna analiza nije pokazala postojanje signifikantno zna~ajnih razlika u prosje~nom obujmu korijena sadnica s obzirom na tretiranje (p = 0,0901) i koncentraciju ulja (p = 0,3985). Rezultati deskriptivne statistike pokazuju kako su najmanju visinu imale sadnice s ploha tretiranih mineralnim uljima (tablica 3). Najve}e visine sadnica (698,11 mm) dobivene su pri koncentraciji ulja od 0,2 L/m2, slijede sadnice s kontrolnih pokusnih ploha (653,03 mm) i ploha tretiranih koncentracijom ulja od 0,5 L/m2 (630,69 mm). Najmanje su visine sadnica (476,83 mm) izmjerene na pokusnoj plohi tretiranoj mineralnim uljem u koncentraciji 0,5 L/m2 (slika 5). Prosje~ne najve}e vrijednosti promjera vrata korijena sadnica hrasta lu`njaka na kraju druge vegetacije uzgoja imale su sadnice s kontrolne pokusne plohe (7,90 mm), slijede promjeri sadnica s ploha tretiranih biouljem. Najmanje su vrijednosti imale sadnice uzgajane na plohama tretiranim mineralnim uljem (slika 6). Univarijantnom analizom (tablica 4) dobivena je signifikantno zna~ajna razlika u visinama sadnica s obzirom na vrstu ulja (p = 0,000000) i koncentraciju (p = 0,000125). Parametrijski Tukey HSD test (tablica 5) pokazao je signifikantno zna~ajnu razliku u visini sadnica s obzirom na koncentraciju u svim slu~ajevima osim izme|u doze 0,2 L/m2 i 0,5 L/m2 (p = 0,791636). Univarijantnom analizom dobivena je signifikantno zna~ajna razlika u promjeru vrata korijena sadnica s obzirom na vrstu ulja, dok razli~ite koncentracije nisu zna~ajno utjecale na ovo morfolo{ko obilje`je. Tukey HSD test (tablica 6) pokazao je signifikantno zna~ajnu razliku u promjeru vrata korijena sadnica izme|u kontrolne pokusne plohe i ploha tretiranih mineralnim uljima (p = 0,0282) te izme|u ploha tretiranih biouljem i mineralnim uljem (p = 0,0009). Nisu dobivene zna~ajne razlike u ovom morfolo{kom svojstvu jedino izme|u kontrolne plohe i ploha tretiranih biouljem (p = 0,9984). Klju~ne rije~i: sadnice hrasta lu`njaka, mineralno ulje, biorazgradivo ulje, korijenski sustav, rast

Authors’ address – Adresa autorâ:

Received (Primljeno): July 25, 2008 Accepted (Prihva}eno): November 11, 2008

162

Assoc. Prof. Milan Or{ani}, PhD. e-mail: milan.orsanic3@zg.htnet.hr Prof. Dubravko Horvat, PhD. e-mail: horvat@sumfak.hr Assoc. Prof. Nikola Pernar, PhD. e-mail: npernar@sumfak.hr Assist. Prof. Marijan [u{njar, PhD. e-mail: susnjar@sumfak.hr Assist. Prof. Darko Bak{i}, PhD. e-mail: baksa@hazu.hr Damir Drvodeli}, BSc. e-mail: ddrvodelic@inet.hr Forestry Faculty of Zagreb University Sveto{imunska 25 HR–10000 Zagreb CROATIA Croat. j. for. eng. 29(2008)2

Original scientific paper â&#x20AC;&#x201C; Izvorni znanstveni rad

GIS-Based Decision-Support Program for Planning and Analyzing Short-Wood Transport in Russia Yuri Gerasimov, Anton Sokolov, Timo Karjalainen Abstract â&#x20AC;&#x201C; Nacrtak Extraction of short-wood from harvesting operations is becoming common practice in Russia. Logging companies are faced with a large number of options for short-wood transport, but they have limited knowledge of logistics potential. Developed GIS-based decision support program is a unique tool assisting logging companies in making comprehensive decisions on organizational options for the most suitable short-wood transport. Application of the program allows to increase efficiency when introducing cut-to-length technology in Northwest Russia, decrease wood transport costs and improve utilization of short-wood truck fleet. Testing of the program and comparison of alternative delivery plans show that the efficiency of short-wood transport can be increased by 40%. This program could also be used for other applications, such as road planning, fuel supply or logistics in silviculture, and also provides an excellent opportunity to convey knowledge gained through research to the companies in a clear and practical way. Keywords: Russia, logging, cut-to-length, logistics, delivery plan, truck fleet, optimization, GIS, MapInfo

1. Introduction â&#x20AC;&#x201C; Uvod In Russia, logging operations are traditionally divided into three stages: harvesting, transport and work at the central processing yard. Wood harvesting is conducted according to full-tree, tree-length or cut-to-length methods. These methods are different regarding the applied technology and namely delimbing and cross cutting taking place at the stump, road-side or central processing yard (Karvinen et al. 2006). Although the lack of appropriate domestic machinery hinders the implementation of cut-tolength method, it is becoming increasingly common in Russia due to technology transfer from the Nordic countries. The reasons for increasing popularity of cut-to-length method have been as follows: better suitability of this method to other fellings than just clear-cutting, smaller environmental impacts, cleaner wood, less requirements for road-side landings compared to full-tree and tree-length methods (Gerasimov 2004). For example, in the Republic of Karelia and Leningrad region, approximately 70% of harvested wood is already logged with cut-to-length Croat. j. for. eng. 29(2008)2

method, whereas its share in other regions of Russia is considerably smaller, at the level of the whole Russian Federation it is approximately 30% (Gerasimov et al. 2005). Methods of wood transport depend on the used harvesting methods: wood is transported either directly to the end user from the road side storage or via intermediate storages or central processing yard. It is quite easy to manage logistic issues related to traditional tree-length method as all tree-length wood from cutting areas is transported to one central processing yard. Application of the cut-to-length harvesting method or using of the processor at a road-side storage require more attention on wood transport logistics as different timber assortments or short-wood from cutting areas should be delivered directly to several customers: pulp mills, sawmills, wood-based boards mills, wood terminals, and railway stations. The short-wood logistics is complicated and can not be realized by current tree-length approaches effectively (Sikainen et al. 2005). Logistical approaches for short-wood transport are not yet well developed in Russia. Software and tools developed in countries having long experi-

163

Y. GERASIMOV et al.

GIS-Based Decision-Support Program for Planning and Analyzing Short-Wood Transport ... (163–175)

ence of cut-to-length method and short-wood logistics, namely Finland and Sweden (Andersson et al. 2007, Forsberg et al. 2005, Fjeld and Hedlinger 2005, Uusitalo 2005, Hedlinger et al. 2005, Helstad 2006), are not necessarily applicable in Russian conditions. This is due to the specific organizational structure of Russian logging companies, which include a transport department with own vehicle fleet, garages and repair workshops. Russia also has specific requirements for axle load of trucks, own standard of round-wood, category of roads, poor state and maintenance of roads, seasonality of road availability, uneven distribution of logging during the year, etc. Moreover solutions are usually company specific, so that tailored programming tools need to be developed for improving planning and optimization of wood transport in operational and tactical tasks.

2. Objectives of the program – Smjernice programa The objective is to develop a GIS-based decision support program for planning and analyzing shortwood transport for a logging company level in Russian conditions. The program should give the logging company comprehensive information about the benefits and limitations of different short-wood transport options. The logging company should get sufficient information to make sound short-term and long-term decisions. Development of the program has been supported by the latest research results that have been produced as part of the »Intensification of forest management and improvement of wood harvesting in Northwest Russia« project (web-address of the project is http://www.metla.fi/ hanke/3384/subproject-2.htm#background) as well as some other projects. The economic feasibility of logging operations that provide short-wood is a critical element for the development of forestry and wood harvesting in Russia (Karjalainen et al. 2005). The decision support program also acts as a set of guidelines for logging companies since it takes economical aspect into consideration, draws attention to the lack of short-wood trucks and gives recommendations for organizational management of logistics (i.e. delivery planning, locations of garages and temporary wood terminals) when required.

3. Problem set – Problematika The problem in the short-wood transport is to define delivery plans, which allows maximizing wood removals and rationalizing the usage of short-wood truck fleet in a logging company. The term delivery

164

plan means an output schedule for truck fleet for a given time period, including for example places and time for loading and unloading, and type of transporting assortments. Let us formalize the short-wood transport problem. The logging company has several operation units: cutting areas, customers, railway stations and garages (Fig. 1). The following data are known: allowable and actual short-wood storages at roadsides, daily productions in cutting areas by assortments, and their accessibility for wood transport in winter or all-seasons. The company has valid wood trade contracts with some customers and monthly delivery volumes by assortment are known for each customer. The type of assortment depends on tree species, use (sawlog, pulpwood, energy wood), size or dimensions (diameter and length), and quality of wood (domestic or export requirements). The size of an assortment can be specified by limiting values (minimum, maximum), tree species can be specified directly (pine, spruce, birch, aspen and other) or given as a general information (coniferous, deciduous, any). Moreover, a customer may accept unsorted roundwood. In such a case, two different assortments in the cutting area can be equal raw material in the mill and vice versa. Therefore the procedure of assortment identification has to distinguish between assortments nomenclatures in cutting areas and at customer. All cutting areas and customers are connected by road and/or railway. Trans-shipment from trucks to railway wagons is organized in terminals at railway stations. Wood from cutting areas to mills or terminals is delivered by short-wood trucks. The number of trucks and their characteristics (model, carrying capacity, etc) are established. Each truck registers in a concrete garage. There can be several garages. Geographical information system (GIS) should be used to locate and connect cutting areas, terminals, customers and garages.

4. Program structure – Struktura programa 4.1 Overall structure of the program – Sveobuhvatna struktura programa Decision support program has been constructed in MapInfo environment using Map Basics for coding and Microsoft Excel for reporting, i.e. with very common software. MapInfo environment provides the possibility to build a program with user interfaces and custom dialog boxes with MS Excel. An overview of the proCroat. j. for. eng. 29(2008)2

GIS-Based Decision-Support Program for Planning and Analyzing Short-Wood Transport ... (163â&#x20AC;&#x201C;175)

Y. GERASIMOV et al.

Fig. 1 Example of logistic management units of a logging company Slika 1. Primjer logistike organizacijskih jednica poduze}a za pridobivanje drva gram structure and its most important components is presented in Fig. 2. Data module includes information about roads and their quality, locations of logistic management units (i.e. cutting areas, customers, truck garages, and railway stations) and their characteristics. The user can easily manage data with a user friendly interface. The second part of the program is Graph module. In this module the user can generate a layer of roads including logistic management units. Several subCroat. j. for. eng. 29(2008)2

modules have been created for the managing graph (construction, editing, deleting, and adding). The module of Optimal Routes helps the user to search with heuristic optimization method better variant of short-wood transporting route. The module of Optimal Delivery Plan helps the user to optimize by dynamic programming daily tasks for each truck. The Reporting module contains reports of optimal routes and delivery for short-wood transport for the logging company.

165

Y. GERASIMOV et al.

GIS-Based Decision-Support Program for Planning and Analyzing Short-Wood Transport ... (163–175)

Fig. 2 An overview of the program structure Slika 2. Pregled strukture programa

4.2 Data – Podatci

Þ Road maps in MapInfo format

Data required for planning and analyzing shortwood transport include:

Þ Location of logistic management units (cutting aresas, customers, railway stations, garages)

166

Croat. j. for. eng. 29(2008)2

GIS-Based Decision-Support Program for Planning and Analyzing Short-Wood Transport ... (163–175)

Y. GERASIMOV et al.

Fig. 3 Screenshots for a cutting area Slika 3. Zaslon ra~unala pri odabiru zna~ajki sje~ine Þ Characteristics of logistic management units Cutting areas (Fig. 3): starting date of logging; type of cutting site (winter, summer, the whole year round); type of producible assortments and their characteristics: tree species, size, and quality class; average production of daily logging; growing stock by assortments: actual cut and allowable cut; possibility to use heavy trucks with trailer; possible customers for each assortment. Customers (Fig. 4): type of customer (local customer means that direct delivery by truck is possible, remote customer means that transshipment from trucks to railway wagons is needed); distance from railway station to remote customer; type of used assortments and their characteristics: tree species, size, quality Croat. j. for. eng. 29(2008)2

class; monthly contracted deliveries by assortment. Garages (Fig. 5): number of registered trucks; characteristics of each truck: model, trailer or semi-trailer availability, registration number, carrying capacity, average time for loading and unloading. Railway station: name, code; costs of trans-shipment from trucks to wagons via terminal per m Þ Wood transporting costs and trans-shipment costs at terminals are taken into account when searching optimal routes.

4.3 Graph – Grafi~ki prikazi Before searching optimal routes, the initial layer of roads has to be transferred into the graph. The first step is the creation of the layer of nodes. Nodes

167

Y. GERASIMOV et al.

GIS-Based Decision-Support Program for Planning and Analyzing Short-Wood Transport ... (163–175)

Fig. 4 Screenshots for a customer Slika 4. Zaslon ra~unala pri pode{avanju podataka o kupcu

are numbered and saved in the database. The next step is the creation of the layer of arcs – every road is transferred into several independent segments. The starting and the ending points of segments coincide with dotty objects of the layer of nodes. Type of the road, number of starting and final dots, arc length and computed time of moving are

168

entered into database for each arc. The user has to put down the average speeds of all types of roads for the calculation of moving time. If the user knows the specific properties of the road sections – their state, complicated turns, and other factors affecting speed – the program has special tools for specifying them. Fig. 1 shows an examCroat. j. for. eng. 29(2008)2

GIS-Based Decision-Support Program for Planning and Analyzing Short-Wood Transport ... (163–175)

ple of the graph including logistic management units of a logging company.

4.4 Search of optimal routes – Utvr|ivanje optimalnih ruta The search of optimal routes helps to find the route with the lowest transport costs. Relative or absolute wood transport costs per 1 m3 by different types of roads and trans-shipment costs at the terminals have to be established. Estimation of moving time and costs between the logistic management units are important elements for optimization. Moving time depends on the distance and the average speed of moving along the road, under different condition. Usually several paths can be used for moving. There are several approaches to optimal route searching (Dijkstra 1956, Hart et al. 1968, Stefankis and Kavouras 1995, Jonsson 2003, Huurinainen and Ikonen 2007). The Dijkstra algorithm is the most simple and precise one. Moreover, an absolute optimum can always be reached. Application of the Dijkstra algorithm for this task showed that the algorithm does not work properly when there is a huge number of nodes in the graph. Therefore an original heuristic method based on the Dijkstra algorithm was applied, allowing taking into account all nodes of the graph for every step of the algorithm (Appendix). All routes and their characteristics are saved in the database and downloaded from there when que-

Y. GERASIMOV et al.

ries are repeated. This saves time significantly during calculation of new alternatives for the delivery plan of the same graph.

4.5 Search of optimal delivery plan – Odre|ivanje optimalnoga plana isporuke The synthesis of the delivery plan can not be solved by classical approaches (Andreev and Gerasimov 1999). This problem may be classified as »open« and »without end«. The process of the delivery plan calculation for every truck stops and the procedure for return to the garage starts because shift ends; lack of short-wood in cutting areas, obligations of wood trade contracts already performed. The original algorithm based on dynamic programming was developed for these tasks (Sokolov and Gerasimov 2004). The criterion for optimization is wood transport per shift for every truck. Total time of the truck moving is minimized during limited shift without nontechnological stops. The established optimal decision directly corresponds to maximum wood transport per shift, i.e. number of runs. During conditional optimization on every step of the dynamic programming for every current cutting area in turn sets customers with minimum total moving time. Moving time is calculated from the beginning of the shift to the arrival to the current cutting area. During unconditional optimization (from the end to the beginning) the plan with maximum runs is defined. If several alternative plans with the same number of runs are defined then the plan where the

Fig. 5 Screenshots for a trucks garage Slika 5. Zaslon ra~unala pri pode{avanju podataka vozila Croat. j. for. eng. 29(2008)2

169

Y. GERASIMOV et al.

GIS-Based Decision-Support Program for Planning and Analyzing Short-Wood Transport ... (163–175)

Fig. 6 An example of a delivery plan Slika 6. Primjer plana isporuke truck is back to garage as late as possible is selected (use of truck is maximized). The assortment with the highest priority is selected if alternative types of assortments are allocated for transport from the optimal cutting area to optimal customer. The assortment priority is moved in corresponding user’s dialog (characteristics of cutting area or customer). All trucks are included in the total list by garage according to user’s priority. Trucks priority is set in corresponding user’s dialog (characteristics of garage). The first plan is calculated for the first truck in the list, then for the second one (for undelivered wood) and so on. In case of several garages, the first plans are calculated for the first trucks of all garages. Next plans are calculated for the second trucks of all garages and so on as long as there is wood to be delivered. The results are saved as Microsoft Excel file, every sheet in the file is a delivery plan for all trucks of a single garage.

5. Efficiency of delivery plans – U~inkovitost planova isporuke 5.1 Testing – Testiranje The efficiency of the developed program was tested in the actual logging process. Three delivery plans were compared for a logging company operating in the Republic of Karelia. The company provided forest inventory and infrastructure information and thus the following map layers were created: roads (5 types of quality), forest stands, and cutting areas. The »basic« delivery plan (Plan 1) was made in a traditional way without program support. Two delivery plans (Plan 2 and Plan 3) were made with

170

the program. The difference between the second and third delivery plan is that in the third plan (Plan 3) the trucks change the drivers on the route without retuning to the garage in every shift. The delivery plans were created for four consecutive working days using two shifts per day for the same conditions of logistic management units (cutting areas, customers, routes, fleet, etc). There were five trucks based in one garage, four cutting areas, and four customers (three sawmills and one wood terminal). Capacities for short-wood trucks were 50–52 m3 depending on the model (Volvo, Scania). Daily outputs in cutting areas were 140–420 m3 depending on the site, the actual cut per cutting area was 5,000–15,000 m3. A half of the actual cut is coniferous sawlogs including 9% of small size spruce sawlogs, 18% – coniferous pulpwood, 22% – birch pulpwood, 10% – energy wood (Gerasimov et al. 2005).

5.2 Performance indexes – Indeksi u~inkovitosti Delivery plans were compared using the following performance indexes: total work time (hours), total run (kilometers); total number of runs, total volume of wood transport (m3), total cargo run (kilometers), required number of trucks, fleet utilization rate per shift, index of loaded distance; index of operation work (m3/km). Fleet utilization rate per shift is calculated as follows: tp ku = (1) ts n where: tp total work time per day, hours ts total length of shift, hours n number of working trucks, units Croat. j. for. eng. 29(2008)2

GIS-Based Decision-Support Program for Planning and Analyzing Short-Wood Transport ... (163–175)

The fleet utilization rate per shift has somewhat different meaning than standard fleet utilization rate. This rate shows truck utilization within a shift, i.e. how effectively trucks are utilized in the delivery plan. If the truck was standing idle during a day, it was excluded from the calculation. The most efficient delivery plan means the least working trucks for the same daily short-wood transport or, vice versa, the biggest short-wood transport for the same number of working trucks. The index of loaded distance means the ratio between the total cargo run and the total run. The operation work shows how much short-wood is delivered per 1 km of the total truck’s run.

5.3 Results – Rezultati Comparison of the results between delivery plans when applying the basic method (Plan 1) and the program (Plans 2 and 3) are presented in Table 1. The change in the indexes (in percents compared to the basic Plan 1) is shown in parentheses. Optimization of the schedule using the program with Plan 2 shows that the total delivered wood volume increases from 2740 m3 to 2997 m3 (+9%). The total run is the same, but the total working time decreases by 17%. The required fleet is the same, 5 short-wood trucks. The fleet utilization rate decreases slightly (-4%), the index of loaded distance increases by 22%, the total volume of transporting roundwood per km increases by 9%.

Y. GERASIMOV et al.

Optimization of the schedule using the program with Plan 3 shows that the total delivered wood volume increases from 2740 m3 to 3000 m3 (+10%). The total run decreases from 7382 km to 5743 km (-22%), the total working time decreases from 307 h to 234 h (-22%). It reduces the required fleet from 5 to 4 trucks. The fleet utilization rate increases by 19%, the index of loaded distance increases by 30%, the total volume of transporting roundwood per km increases by 42%.

6. Discussion and conclusion – Rasprava i zaklju~ci Developed decision support program can be used for planning and analysis of short-wood transport. One logging company was asked to provide the actual data for testing the program. Different transport options were then presented to the logging company, and feedback was received for further development of the program. Testing of the program and comparison of alternative delivery plans show that the efficiency of shortwood transport can be increased by 40%. Application of the program allows computer based processing of delivery plans and thus provides possibilities for producing several alternatives and taking into account possible changes both inside and outside the organization. And most importantly, the program makes it possible to optimize transportation operations. The program may not be able to find global optimum in some cases. Testing shows, however, that

Total volume, m3 Ukupni drvni obujam, m3

Total cargo run, km Ukupna udaljenost vo`nje optere}enoga vozila, km

Required number of trucks Potreban broj kamiona

Fleet utilization rate Stupanj iskori{tenosti flote vozila

Index of loaded distance Indeks udaljenosti optere}enoga vozila

Operation work, m3/km Operativan rad, m3/km

Number of runs Broj turnusa

Total run, km Ukupna udaljenost, km

Total working time, h Ukupno vrijeme rada, h

Plan – Plan

Table 1 Comparison between the basic delivery Plan 1 and delivery Plan 2 and Plan 3 made with the decision support program Tablica 1. Poredba osnovnog plana (plan 1) i preostala dva plana isporuke (plan 2, plan 3) nastala kao rezultat programa za pomo} pri dono{enju odluka

307 255 (–17%) 239 (–22%)

7382 7382 (0%) 5743 (–22%)

53 58 (+9%) 58 (+9%)

2740 2996 (+9%) 3000 (+10%)

2212 2697 (+22%) 2872 (+30%)

5 5 (0%) 4 (–20%)

0.754 0.728 (–4%) 0.895 (+19%)

0.300 0.365 (+22%) 0.499 (+66%)

0.371 0.406 (+9%) 0.526 (+42%)

Plan 1 was made in a traditional way without program support, the trucks return to the garage in every shift. Plan 2 was made using the program, the trucks return to the garage in every shift. Plan 3 was made using the program, the trucks change the drivers on the route without retuning to the garage in every shift. Plan 1 izra|en je tradicionalnim na~inom bez pomo}i ra~unala, vozila (kamioni) vra}aju se u spremi{te na kraju svake radne smjene. Plan 2 izra|en je uz pomo} programa, vozila se vra}aju u spremi{te na kraju svake radne smjene. Plan 3 izra|en je uz pomo} programa, promjene voza~a na trasi puta bez vra}anja u spremi{te na kraju svake radne smjene.

Croat. j. for. eng. 29(2008)2

171

Y. GERASIMOV et al.

GIS-Based Decision-Support Program for Planning and Analyzing Short-Wood Transport ... (163–175)

problems appear only in the case of complicated graphs with chaotic structure. In reality forest road networks are not located randomly. They have certain directions, and therefore the developed algorithm for searching optimal routes can be considered reliable. Mathematical programming has been used as the main tool for optimization. Coding of algorithms has been done in a simple Map Basic environment. Obviously such universal algorithmic languages as C++ or Visual Basic would provide better processing speed and flexibility of the program. Extraction of short-wood from harvesting processes is becoming more common practice in Northwest Russia, particularly in Karelia, Leningrad, Pskov and Novgorod regions. Short-wood transport is also expected to increase in other parts of Russia. Application of cut-to-length harvesting method would allow to increase productivity of wood harvesting and thus to improve the economics of logging operations. At the same time, harvesting of forest resources by cut-to-length method causes less environmental impacts than traditional methods and improves the ecological state of forest sites both in the short and long term. Review of the existing logistic methods and approaches applied in Russia show that logging companies are using different approaches. These approaches do not provide the basis for economic analysis. Moreover decision making is strongly based on the experience of logistic manager without software support. Approaches are suitable for companies which utilize traditional tree-length technology and one central processing yard. Introduction of the Nordic cut-to-length technology requires more attention to wood transport logistics as roundwood from cutting areas has to be delivered directly to several customers, terminals, and railway stations. GIS-based decision support program has been developed to assist logging companies in decision making related to planning, utilization and optimization of vehicle fleet. Searching of optimal routes could also be used for other applications, i.e. forest road planning, fuel supply, seedling transportation, etc.

7. References – Literatura Andersson, G., Flisberg, P., Liden, B., Rönnqvist, M., 2007: RuttOpt – A decision support system for routing of logging trucks. Discussion Papers, Department of Finance and Management Science, Norwegian School of Economics and Business Administration (NHH), No 2007/16, 34 p. Huurinainen, S., Ikonen, O., 2007: Watt Specialists Software for Bionergy Networks. More and Better Bioenergy with Less Work from Smaller Areas. In: Savolainen, M. (ed.), Bioenergy 2007. 3rd International Bioenergy Confer-

172

ence and Exhibition from 3rd to 6th of September 2007, Jyväskylä, Finland, Proceedings, FINBIO Publications 36. Benkert, M., Wolff, A., Widmann, F., Shirabe, T., 2006: The minimum Manhattan network problem: Approximations and exact solutions. Computational Geometry: Theory and Applications 35(3):188–208. Helstad, K., 2006: Managing timber procurement in Nordic purchasing sawmills. Acta Wexionensia 93/2006, Växjö University press, 56 p. Karvinen, S., Välkky, E., Torniainen, T., Gerasimov, Y., 2006: Northwest Russian forestry in a nutshell. Working Papers of the Finnish Forest Research Institute 30, 98 p. Fjeld, D., Hedlinger, C., 2005: The Transport Game – A Tool for Teaching Basics of Transport Decision Proficiency. International Journal of Forest Engineering 16(2): 57–64. Forsberg, M., Frisk, M., Rönnqvisty, M., 2005: FlowOpt – A Decision Support Tool for Strategic and Tactical Transportation Planning in Forestry. International Journal of Forest Engineering 16(2): 101–114. Hedlinger, C., Nilsson, B., Fjeld, D., 2005: Service Divergence In Swedish Round Wood Transport. International Journal of Forest Engineering 16(2): 153–166. Sikanen L., Asikainen, A., Lehikoinen, M., 2005: Transport control of forest fuels by fleet manager, mobile terminals and GPS. Biomass and Bionergy 28: 183–191. Uusitalo, J., 2005: A Framework for CTL Method-Based Wood Procurement Logistics. International Journal of Forest Engineering 16(2): 37–46. Gerasimov, Y., 2004: Cut-to-length method in the wood procurement of Russia: SWOT analysis. Silva Carelica 45: 338–344. Sokolov, A., Gerasimov, Y., 2004: Corporative information systems for wood procurement development in Karelia. Silva Carelica 45: 166–172. Jonsson, M., 2003: An Optimal Pathfinder for Vehicles in Real-World Terrain Maps. The Royal Institute of Science, School of Engineering Physics, Stockholm, Sweden. Andreev, V., Gerasimov, Y., 1999: Optimal Decision Making: Theory and Application in Forest Engineering and Forestry. Joensuu University Press. 200 p. (In Russian. Summary in English). Stefanakis, E., Kavouras, M., 1995: On the Determination of the Optimum Path in Space. Proceedings of the European Conference on Spatial Information Theory COSIT 95. Hart, P., Nilsson, N., Raphael, B., 1968: A Formal Basis for the Heuristic Determination of Minimum Cost Paths. IEEE Transactions of Systems Science and Cybernetics SSC 4(2): 100–107. Dijkstra, E., 1959: A Note on Two Problems in Connexion With Graphs. Numerische Mathematik 1: 269–271. Karjalainen, T., Mutanen, A., Torniainen, T., Viitanen, J., 2005: Changes and Challenges in the Russian Forest Sector. Finnish Forest Sector Economic Outlook 2005–2005, 50–53. Gerasimov, Y., Siounev, V., Chikulaev, P., Pechorin, V., Dyakonov, V., Komkov, V., Sikanen, L. Karjalainen, T., 2005: An analysis of logging companies in the Republic of Karelia. Working Papers of the Finnish Forest Research Institute 16, 39 p. Croat. j. for. eng. 29(2008)2

GIS-Based Decision-Support Program for Planning and Analyzing Short-Wood Transport ... (163–175)

Karanta, I., Jokinen, O., Mikkola, T., Savola, J., Bounsaythip, C., 2000: Requirements for a vehicle routing and scheduling system in timber transport. In: Sjöström, K. (ed), Logistics in the forest sector, Timber Logistics Club, Helsinki, 235–251 p.

7. Define the list of nodes of the graph which are connected with active node by arcs and value of elements array S less than 2. 8. Give S(k) = 1, if B(k) > B(j)+A(j,k) than B(k) = B(j)+A(j,k) and C(k) = j for all nodes found on previous step list. Where k – number of the node from the concerned list; j – number of the active node; A(j,k) – transport costs per 1 m3 by the arc from node j to node k. 9. Among nodes from the list obtained in step 7 find the node with minimal sum of values of elements of arrays B and H. Make this node active. 10. If the active node is the final point of the route go to step 12. 11. Go to step 6. 12. Insert the final node into the route. 13. Calculate the value of array C for the final node. Let value is z. 14. Insert the node with number z into the route. 15. If C(z) = z, then go to 18. 16. z = C(z) 17. Go to 14. 18. Calculate the route in reverse order, from the last inserted node to the first inserted node.

Appendix – Dodatak Heuristic algorithm uses four arrays which have N numbers. N means number of nodes of the graph. Array B keeps the current shortest distances from the initial point to the corresponding node. Array C keeps the number of the next to the last point on the current shortest way from the initial point to the node. Array S keeps the marks of node consideration. The mark can have three meanings: 0 – the node is not considered, 1 – the node is included in consideration list, and 2 – if the node has already been considered. Array H keeps the heuristic estimation of transport costs per m3 from the tested node to the final point of the route. The Manhattan method was used for the calculation of transport costs on the hypothetical road consisting of perpendicular segments: the first segment is located along a parallel and the second segment along a meridian (Benkert et al. 2006). The algorithm consists of the following steps: 1. Let i to go from i = 1 to N. Give value 0 to S(i) and value i to C(i). 2. Give for B transport costs per 1 m3 from initial point to every node of the graph. If the arc from the initial point to the considered node is absent, B is infinite. 3. Give for H heuristic transport costs per m3 from the tested node to the final point of the route. 4. Give meaning 0 to element array C which corresponds to the initial point of the route. 5. Make active status for the initial node of the route. 6. Give meaning 2 to element array S which corresponds to the active point of the route.

Y. GERASIMOV et al.

Due to the fact that only several nearest nodes are checked in every cycle of the algorithm, the processing time does not depend on the number of nodes. The processing time depends on the distance between the initial and final points of the route, measured both in arcs and degrees of graph branching. In comparison with the Dijkstra algorithm the processing time has decreased a hundred times. The double-ply search provides for the decrease of probability of errors. The search goes from the initial point to the final point and after that back from the final to the initial point. The next best route is selected if the searching results are not similar. This option increases the processing time by two times, but eliminates the possibility of mistakes.

Sa`etak

Ra~unalni program temeljen na GIS-u kao potpora odlu~ivanju pri planiranju i ra{~lambi transporta kratkoga drva u Rusiji U Rusiji su {umski radovi tradicionalno podijeljeni u tri vrste: sje~a, transport te radovi na stovari{tu. Sje~a i izradba stabala provodi se deblovnom, stablovnom i sortimentnom metodom. Te se metode razlikuju s obzirom na primijenjene postupke te mjesto kresanja grana i trupljenja debla, koje mo`e biti u sje~ini, na {umskoj cesti ili na stovari{tu. Nedostatak doma}e mehanizacije prije~i primjenu sortimentne metode, koja postaje sve uobi~ajenija u Rusiji zbog transfera tehnologije iz nordijskih zemalja. Ta je tehnologija pogodnija ne samo kod ~istih ve} i kod ostalih vrsta sje~a zbog manjega utjecaja na okoli{, smanjenoga udjela ne~isto}e na oblovini te su, me|u ostalim, i zahtjevi za pomo}na stovari{ta manji

Croat. j. for. eng. 29(2008)2

173

Y. GERASIMOV et al.

GIS-Based Decision-Support Program for Planning and Analyzing Short-Wood Transport ... (163–175)

nego kod stablovne metode, {to je pove}alo popularnost sortimentne metode izradbe drva. Na primjer, u Republici Kareliji i u lenjingradskoj regiji pribli`no 70 % drva posje~eno je i izra|eno primjenom sortimentne metode, dok je udjel te metode u ostalim regijama manji, a u cijeloj Ruskoj Federaciji iznosi oko 30 %. Jednostavna logistika prati tradicionalnu deblovnu metodu jer se sve drvo transportira na glavno mehanizirano stovari{te. Primjena sortimentne metode ili uporaba procesora na pomo}nom stovari{tu zahtjevnija je za logistiku zbog proizvodnje razli~itih sortimenata ili zbog potrebe neposredne dostave brojnim korisnicima: tvornicama celuloze, pilanama, proizvo|a~ima drvnih plo~a te na glavna stovari{ta. Logisti~ki sustavi u proizvodnji kratkoga drva sloèni su i nemogu}a je u~inkovita neposredna primjena postoje}e logistike deblovne metode. Logisti~ki sustavi za transport kratkoga drva u Rusiji su u povojima. Ra~unalni softver i alati razvijeni u zemljama s dugogodi{njim iskustvom u proizvodnji kratkoga drva sortimentnom metodom, kao {to su Finska i [vedska, nisu primjenjivi za ruske uvjete zbog specifi~ne organizacijske strukture ruskoga {umarstva koja uklju~uje transportni sektor s vlastitim voznim parkom, spremi{tima i radionicama, zahtjeve za osovinskim optere}enjima vozila, vlastiti standard za oblo drvo, posebnost kategorizacije prometnica, nezadovoljavaju}e stanje cesta, sezonsku dostupnost prometnica, neravnomjerni prostorno-vremenski raspored sje~ina itd. Osim toga poslovna su rje{enja obi~no posebnost pojedinoga poduze}a, tako da programski alati moraju biti razvijani i/ili prilago|avani za kvalitetnije planiranje i optimizaciju na razini operativnih i takti~kih zadataka. Cilj je bio razviti GIS, sustav za potporu odlu~ivanju pri planiranju i ra{~lambi transporta kratkoga drva u ruskim uvjetima za poduze}a ~ija je djelatnost pridobivanje drva. Poduze}a bi trebala dobiti zadovoljavaju}e informacije potrebne za kratkoro~no i dugoro~no dono{enje prihvatljivih odluka. Ekonomska izvedivost {umarskih operacija pri proizvodnji kratkoga drva kriti~na je sastavnica razvoja {umarstva u Rusiji. Sustav potpore odlu~ivanju djeluje kao skup smjernica za poduze}a jer u obzir uzima i ekonomski aspekt, te na zahtjev upozorava na manjak kamionskih skupova i daje preporuke za organizaciju logistike (planiranje dostave, lokacije spremi{ta i pomo}nih stovari{ta). Za pojednostavljenje daljinskoga transporta kratkoga drva pri modeliranju daljinskoga prijevoza potrebni su ovi podaci: sje~na podru~ja, kupci, èljezni~ki kolodvori i spremi{ta/garaè (slika 1). Potrebno je poznavati najve}i dopu{teni obujam sje~a i stvarno stanje na pomo}nim stovari{tima uz cestu, dnevnu proizvodnju po vrstama sortimenata i otvorenost {uma mreòm prometnica (zimi i tijekom cijele godine). Na osnovi valjanih ugovora o kupoprodaji drva poznate su traène koli~ine po sortimentima za isporuku na mjese~noj razini za svakoga kupca. Veli~ina sorimenta moè biti odre|ena i ograni~avaju}im vrijednostima (minimum i maksimum). Vrsta se drva moè specificirati neposredno ili moè biti zadana op}enito (~etinja~e, lista~e). Osim toga postoji mogu}nosti da kupac prihva}a nesortirano oblo drvo. U takvu slu~aju dva razli~ita sortimenta iz sje~ine mogu biti jednaka po obujmu na skladi{tu kupca, ne znaju}i to~no koji je koji. Zbog toga postupak identifikacije sortimenata mora omogu}iti prepoznavanje i na pomo}nom stovari{tu i kod kupca. Sve sje~ine i kupci me|usobno su povezani javnim prometnicama – cestama i/ili èljeznicom. Prekrcaj je drva s kamiona u èljezni~ke vagone organiziran na me|ustovari{tima uz èljezni~e stanice. Drvo se iz sje~ina prevozi kamionskim skupovima. Stoga je potrebno imati podatke o koli~ini (broju) kamionskih skupova i njihovim tehni~kim zna~ajkama – model, nosivost itd. Svaki kamion vezan je uz pripadaju}e spremi{te (garaù). Geografski informacijski sustav (GIS) koristi se za lociranje i povezivanje sje~ina, stovari{ta, kupaca i spremi{ta. Program potpore odlu~ivanju izra|en je u okruènju MapInfo kori{tenjem Map Basica za kodiranje i Microsoft Excela za izvje{tavanje. U okruènju MapInfo izra|eno je korisni~ko su~elje s uobi~ajenim MS Excel dijalo{kim prozorima (dialog box). Pregled programske strukture i njegovih najva`nijih sastavnica prikazan je na slici 2. Podatkovni modul (Data modul) uklju~uje informacije o cestama i njihovoj kakvo}i, lokacijama logisti~kih jedinica (sje~ine, kupci, kamionske garaè i èljezni~ki kolodvori) te njihove zna~ajke. Korisnik moè lako upravljati podacima preko grafi~koga korisni~koga su~elja. Drugi je dio programa grafi~ki modul (Graph module). U tom modulu korisnik moè generirati sloj cesta uklju~uju}i logisti~ke jedinice. Nekoliko podmodula stvoreno je za upravljanje grafi~kim prikazima (izrada, ure|ivanje, brisanje i dodavanje). Modul optimalnih putova (Optimal routes) pomaè korisniku da metodom heuristi~ke optimizacije odabere bolju ina~icu transportnoga puta, a modul optimalne isporuke (Optimal delivery plan) pomaè korisniku da uz pomo} dinami~noga programiranja svakodnevno odredi zadatak za svaki kamion. Izvje{tajni modul (Reporting module) sadràva izvje{taje optimalnih putova i planova isporuke za transport drva. Prije utvr|ivanja optimalnih putova po~etni sloj prometnica mora biti grafi~ki prikazan. Prvi je korak stvaranje sloja ~vori{ta. ^vori{ta su pobrojena i pohranjena u bazi podataka. Sljede}i je korak stvaranje sloja krivina – svaka je cesta prevedena u nekoliko neovisnih sastavnica. Po~etne i zavr{ne to~ke sastavnica podudaraju se s to~kastim objektima sloja ~vori{ta. Vrsta ceste, velik broj po~etnih i zavr{nih to~aka, duljine lukova i izra~unato vrijeme prijevoza upisano je u bazu podataka za svaku krivinu. Korisnik mora unijeti prosjek brzina svih tipova cesta za izra~un vremena kretanja. Ako korisnik poznaje specifi~na svojstva cestovnih sastavnica (dijelova cesta) – njihovo stanje, sloèna skretanja i druge ~imbenike koji imaju utjecaja na brzinu, program ima dodatne alate za njihovu specifikaciju. Na slici 1 nalazi se grafi~ki prikaz na kojem su uklju~ene logisti~ke jedinice poduze}a. Potraga za optimalnim putovima pomaè pri traènju puta s najniìm transportnim cijenama. Potrebno je zadati relativni ili apsolutini tro{ak prijevoza po kubnom metru koji razlikuju razli~ite tipove cesta te cijene prekrcaja drva na stovari{tima. Va`ne sastavnice za optimizaciju je prora~un vremena vo`nje i cijena prijevoza. Vrijeme ovisi o udaljenosti i prosje~noj brzini kretanja prometnicama. Obi~no se nekoliko razli~itih putova moè koristiti za prijevoz. Poznato je nekoliko pristupa pri optimalnom utvr|ivanju puta. Dijkstrin algoritam je najjednostavniji i jedan od preciznijih. Osim toga uvijek se moè utvrditi apsolutan optimum. Primjena Dijkstrina algoritma za ovaj zadatak

174

Croat. j. for. eng. 29(2008)2

GIS-Based Decision-Support Program for Planning and Analyzing Short-Wood Transport ... (163–175)

Y. GERASIMOV et al.

pokazala je da algoritam ne daje valjane rezultate u slu~aju postojanja ve}ega broja ~vori{ta. Zato su primijenjenoj izvornoj heuristi~koj metodi temeljenoj na Dijkstrinu algoritmu pridruèna i uzeta u razmatranje sva ~vori{ta grafi~koga prikaza za svaki korak algoritma (v. Dodatak). Svi putovi i njihove zna~ajke pohranjeni su u bazi podataka i kori{teni pri ponovljenim upitima. To zna~ajno smanjuje vrijeme tijekom prora~una alternativnih planova isporuke. Uop}avanje plana isporuke ne moè se rije{iti klasi~nim pristupima. Ovaj se problem moè klasificirati kao »vje~no otvoren«. Proces izra~una plana isporuke svakoga zaustavljanja kamionskoga skupa i postupak povratka u spremi{te po~inje zbog kraja smjene, zbog nedostatka drva u sje~inama, a obveze iz trgovinskih ugovora uzete su u razmatranje. Izvoran algoritam temeljen na dinami~nom programiranju ve} je bio razvijen za ovakve zada}e. Ukupno vrijeme kamionskoga prijevoza smanjeno je tijekom smjene ograni~enoga trajanja i bez prekida rada. Prona|ena optimalna odluka o putu utvr|uje maksimalni transport po smjeni te broj turnusa. Tijekom uvjetne optimizacije na svakom koraku dinami~noga programiranja za svaku sje~inu odre|uju se kupci za koje je utvr|eno najmanje ukupno vrijeme vo`nje. Ukupno je vrijeme ra~unato od po~etka smjene pa do dolaska u sje~inu. Prilikom bezuvjetne optimizacije (od kraja prema po~etku) odre|en je plan koji ima najve}i broj turnusa. Ako postoji nekoliko alternativnih planova s jednakim brojem turnusa, odabire se onaj kod kojega se kamion najkasnije vra}a u spremi{te (upotreba kamiona je maksimizirana). Sortimenti najvi{ega prioriteta izabrani su u slu~aju vi{e ina~ica prijevoza od optimalne sje~ine do optimalnoga kupca. Izbor prioritetne vrste sortimenata je u odgovaraju}em korisni~kom dijalogu (zna~ajke sje~ina ili kupaca). Rezultati su sa~uvani kao Microsoft Excel dokument, svaki list (sheet) dokumenta je plan isporuke svih kamiona jednoga spremi{ta. Razvijeni program moè biti kori{ten za planiranje i ra{~lambu transporta kratkoga drva. Testirani program i usporedba alternativnih planova isporuke pokazuju da u~inkovitost transporta moè rasti i do 40 %. Primjena programa osigurava ra~unalnu obradu plana isporuke i tako pruà mogu}nost za izradu nekoliko zamjenskih ina~ica uz promjenu unutar i izvan organizacije poduze}a. Najzna~ajnije je to {to program moè optimizirati sve prometne sastavnice. Program u nekim slu~ajevima ne nalazi op}eprihva}eni optimum. Testiranje pokazuje da se te situacije pojavljuju u slu~aju sloènih problema s kaoti~nim rasporedom logisti~kih jedinica. U stvarnosti se {umske prometnice ne postavljaju nasumi~no. One imaju stru~no odre|en raspored pa se razvijeni alogoritam moè smatrati pouzdanim. Matemati~ko je programiranje kori{teno kao glavni alat za optimizaciju. Za kodiranje algoritama upotrijebljen je MapBasic. O~ito je da bi univerzalni programski jezici kao {to su C++ ili Visual Basic pruìli brù obradu i ve}u prilagodljivost programa. Dono{enje odluka dosada se temelji na iskustvu {umarskih stru~njaka bez pomo}noga softvera. Ovaj je pristup pogodan za poduze}a koja koriste tradicionalno stablovnu ili deblovnu metodu i jedno glavno mehanizirano stovari{te. Uvod sortimentne metode zahtijeva vi{e pa`nje prema transportnoj logistici jer se oblo drvo iz sje~ina dostavlja neposredno nekolicini korisnika te na glavna stovari{ta. Potpora GIS-a va`na je pri planiranju, iskori{tavanju i optimizaciji rada voznoga parka. Utvr|ivanje optimalnih putova moè se koristiti i za druge namjene, npr. planiranje mreè {umskih prometnica, pri opskrbi pogonskim gorivom, transportu sadnica itd. Klju~ne rije~i: Rusija, pridobivanje drva, sortimentna metoda, logistika, plan isporuke, vozni park, optimizacija, GIS, MapInfo

Authors’ addresses – Adresa autorâ: Yuri Gerasimov, PhD. e-mail: yuri.gerasimov@metla.fi Prof. Timo Karjalainen, PhD. e-mail: timo.karjalainen@metla.fi Finnish Forest Research Institute Joensuu Research Unit Yliopistokatu 6 80101 Joensuu FINLAND

Received (Primljeno): March 8, 2008 Accepted (Prihva}eno): September 11, 2008 Croat. j. for. eng. 29(2008)2

Assoc. Prof. Anton Sokolov, PhD. e-mail: a_sokolov@psu.karelia.ru Petrozavodsk State University Forest Engineering Faculty A. Nevsky av. 58 185030 Petrozavodsk Republic of Karelia RUSSIA

175

108

Croat. j. for. eng. 29(2008)1

Original scientific paper – Izvorni znanstveni rad

Predicting Wood Skidding Direction on Steep Terrain by DEM and Forest Road Network Extension Janez Kr~, Bo{tjan Ko{ir Abstract – Nacrtak The study presents the procedure and functioning of the model for large-scale determination of wood skidding direction on steep slopes. The determination of wood skidding direction significantly depends on the position of the forest stand and the adjacent forest road, the characteristics of skid trail /ground/ and forest operation technology with special attention to the applied skidding means. The model evaluation of wood skidding direction is determined on the basis of forest road layout on the slope (slope, valley and ridge). The data of digital terrain model and digitalized forest road network were used as source data. The software can be modified according to specific technology needs by increasing the range and/or the level of suitability of individual wood skidding directions between two roads. The classification is presented for case study Forest Management Unit comprising 3000 ha. The application of the model discussed, as wll as its limitation and adaptation to changing technological condition and its development in the future. Keywords: forest operation, wood skidding direction, model, forest road network, steep terrain

1. Introduction – Uvod Ecology and technology are interconnected forest management issues. It is a matter of dispute whether economy belongs to the former or the latter and whether social and political issues should be discussed in connection with the basic two issues or separately. In this study, the terrain and stand conditions are origins of assessments and decision making, where forest owners and foresters seek help from modern techniques. Terrain classification is a procedure where the area is divided into homogenous units based on criteria set in advance. Different procedures and factors are applied. Forestry has accepted terrain classification as an important and logical tool for forest management, operational and silviculture planning (Löffler 1984). In this respect alpine region is a very demanding region where tractor skidding prevails, but with ever increasing alternative of cable skidding or forwarding. Operational planning provides specific guidelines for logging options in specific conditions on the basis of technological models. An essential part is the thematic map, i.e. a map showing the location of terrain Croat. j. for. eng. 29(2008)2

characteristics which are important for forest operation (Kr~ 1999) and economics of managing private woodlots. The criteria for determination of technological models are related to technical, environmental, economical and social criteria, and limitations which enable and ensure the co-natural and sustainable forest resource management. In the procedure of determination of technological models, as a base for terrain classification, the following data are used: data of topography, soil type, ground water content, stream lines layout, forest roads and skid roads layout, etc. The selection of the technological model is highly affected by solving skidding problems (Ko{ir 1982, Löffler 1984, Saarilahti 2002). The use of mechanized cutting has changed the relative impact of terrain characteristics on terrain classification (Berg 1992). Both parts of forest operation (cutting and skidding) are influenced by the skidding direction. There are three classes of skidding direction, depending on the terrain gradient (uphill, downhill and even). On steep slopes, the network of forest roads greatly influences the skidding direction as well as the execution of cutting (Heralt 2002, Saarilahti 2002). Skidding forms

177

J. KR^ and B. KO[IR

Predicting Wood Skidding Direction on Steep Terrain by DEM ... (177–188)

such as tractor skidding, cable skidding or forwarding add confusion to procedures of skidding direction design, where the whole design of the forest openness with skid roads and cable corridors is closely connected with environmental considerations – danger of erosion processes (Hay 1998, Pentek et al. 2004). While the skidding forms on wheels are technically no different when skidding uphill or downhill, it is a different matter when using cable cranes. Cable cranes can be gravity or all terrain; they can be classic or mobile with towers. The methods for assessing skidding direction in combination with the skidding means should, therefore, be simplified (Rowan 1996). The intention of this article was to use the existing databases but original algorithms so as to improve and to economise the determination of skidding direction as a basis for operational planning. The accuracy of such a model is of secondary importance at present and depends on the accuracy of input variables. The stress is therefore on the procedure, which should be simple and cheap.

2. Existing research methods – Postoje}e metode istra`ivanja The assessment of the specific model suitability depends mostly on skidding variants from stump to the roadside. Information technology provides many possibilities in the assessment of specific technological solutions in forest operation. There are many studies related to the optimization of the layout of new forest roads (Chung 2001, Aruga et al. 2004), and the feasibility assessment and impacts of changing the forest road density (Ko{ir and Kr~ 2000). The most frequent database used for designing the models is Digital Terrain Model (DTM), which serves as a powerful means for acquiring solutions of forest openness on steep terrains. Similar issues are also addressed by research projects dealing with forest road network density, forest roads layout in relation to the environment (Newnham 1995, Tucek 1995, Yoshimura 1997) and the choice of skidding mainly means comparison between the tractor and cable skidding (Tucek 1999, Stuckelberger 2006). Literature offers several solutions of a similar problem of how to identify watershed areas. Numerous authors have presented and described algorithms and software tools for extracting topographic structures and water catchment areas from DTM data (Mark 1983, Jenson and Dominguem 1988). The problem of identifying the water catchment is similar, but not the same as defining the wood skidding direction. The main difference between the two procedures is in the flow direction – generally, water

178

flows only downhill while wood can be yarded in basically two (three) directions (uphill, downhill and even). Evidently so far forest practice has placed little confidence in the results acquired by using computer simulation or taken from forest inventory databases. The data of functional classification based on past experiences have mostly been used (Rowan 1996). The consequence of this distrust is low efficiency and low rate of application of forest inventory databases, a practice which has to be changed.

3. Objectives of the study – Cilj istra`ivanja The main goal of this research has been defining the model for terrain classification on the basis of skidding direction. In the process of classification the two main data layers (Digital Terrain Model and Digital data of forest road network) together with GIS software modules (Eastman 1997) have been used. The classification of forests into expected categories of wood skidding direction would serve as additional information source for future building, verification, validation and assessment of the models dealing with logging feasibility evaluation. The wood skidding direction influences the efficiency and costs of logging operation. The accuracy of skidding performance assessment for certain areas can be improved, and the cost calculation can be made more reliable if the share and the location of specific wood skidding direction are known. With wood flow to a specific forest road the need for road maintenance activities in the region can be better estimated (Kr~ 2006). There are additional benefits connected with the assessment of forest accessibility as well as with the selection of technological models in logging operations. Newly built forest roads can significantly influence the proportion between uphill and downhill skidding area in typical Alpine conditions. The proposed algorithm can therefore also be useful as an evaluation tool, in case several choices of newly planned forest road are possible. Secondly, there is also the question of proportion between uphill and downhill area between two roads on the slope. So far it has been assumed that ? of the space between roads should account for downhill skidding and the rest for uphill skidding. This can be true on average for moderate inclination and for tractor skidding, while in case of cable skidding just the opposite situation can be expected. The challenge of developing this procedure was also to distinguish between these two skidding options. In Alpine region, uphill skidding with cable cranes (gravity or all-terrain) is still favoured against downhill Croat. j. for. eng. 29(2008)2

Predicting Wood Skidding Direction on Steep Terrain by DEM ... (177â&#x20AC;&#x201C;188)

J. KR^ and B. KO[IR

Fig. 1 Procedure of forest road strip zone classification into cut and fill bank Slika 1. Postupak razredbe ome|enih povr{ina oko {umskih cesta na stranu nasipa i stranu iskopa Croat. j. for. eng. 29(2008)2

179

J. KR^ and B. KO[IR

Predicting Wood Skidding Direction on Steep Terrain by DEM ... (177–188)

direction, which means that one slope road on cable terrains covers more area downwards and less upwards. This also means that in case of tractor skidding or forwarding the situation is probably just the opposite. The main objective of this article was to find a simple solution to the question what to dot, where and how.

4. Method – Metoda rada The assumptions of wood skidding direction from the junction of the secondary forest communication (tractor skidding or cable skidding corridor) with forest road have been used. The following assumptions have been used for the determination of wood skidding direction: Þ The connections of skid trails and cable corridors on forest roadcut bank provide for downhill wood skidding direction in the catchments area; Þ The connections of skid trails and cable corridors on forest roadfill bank provide for uphill skidding direction. The even wood skidding category was not included into the model for mathematical and practical reasons because even skidding on steep terrain is hardly probable. The large-scale (depending on data and software availability) classification of wood skidding direction can be made in the following six steps (Fig. 1): Þ Determination of buffer zone around forest roads which can be of different width depending on specific skidding means. The determination procedure of buffer zone around forest road was conducted using GIS software (Eastman 1997). Buffer zones were created by running DISTANCE module and then by running RECLASS module on the output distance image. Þ Working out raster files with DTM of forest roads and DTM of adjacent buffer zones. Þ The assignment of forest road DTM value parallel to buffer zone. The result is a broad buffer strip with altitude equal to the adjacent forest road altitude. Þ Overlay the actual altitude (DTM) forest road buffer zone raster files with the altitudes parallel to road body altitude in buffer zone (overlay file from step 3 with file from step 2). Þ Reclassification of the result of overlay operation (step 4). The quotient of cut bank is greater than one while the fill bank of forest road has values (quotient) lower than one.

180

Þ The roadfill and roadcut banks are used as destination areas in the process of allocation of the remaining forest area (i.e. forest stand, which is not included in forest road buffer zones). The allocation (module ALLOCATE, Eastman 1997) of the remaining area to the destination (fill and cut banks) classifies uphill skidding to roadfill bank and downhill skidding to roadcut bank.

5. Results of the case study – Rezultati izra|ene studije The case study object is a part of the Jelendol Forest Management Unit (FMU). The unit area covers 3,653 ha and is located on prevailingly steep terrain in the vicinity of Slovenian border with Austria. The difference of accessible forest stands had a strong influence on cutting and skidding costs and consequently on expected forest rent value. Furthermore, the skidding models frequently change due to terrain conditions on short distances. On the prevailingly steep terrain, there is a large area suitable for cable crane skidding. The altitude of FMU ranges from 750 to 2,065 m above sea level. The forest roads density is 23 m/ha; average stand growing stock is 311 m3/ha with prevailingly coniferous trees with 80% of growing stock (Skuber 1990). The northern part of FMU is determined as protected area taken out of wood production. All forest roads in FMU have the same exit point from the area which is located in the south west part and is at the same time the point with the lowest elevation of FMU (Fig. 2). The three-dimensional presentation of FMU Jelendol study area has been used for verifying the skidding direction classification. The position of the individual forest road section has been evaluated with respect to the position on the slope. Three dif-

Fig. 2 Digital Terrain Model of FMU Jelendol with forest roads Slika 2. DMR gospodarske jedinice Jelendol s ucrtanom mre`om {umskih cesta Croat. j. for. eng. 29(2008)2

Predicting Wood Skidding Direction on Steep Terrain by DEM ... (177–188)

J. KR^ and B. KO[IR

Fig. 3 Extraction of roadcut and roadfill banks used as target features for classification of wood skidding direction Slika 3. Razredba smjera privla~enja drva temeljem polo`aja pomo}nih stovari{ta uz {umsku cestu (na strani nasipa ili na strani iskopa) ferent positions of forest road section on the slope have been determined: Þ Valley (forest road section only has a roadcut bank), Þ Slope located (forest road section has – both roadcut and roadfill banks) and Þ Ridge located (forest road section has roadfill banks on both sides, Fig. 3). In the procedure of allocating stands to roadcut and roadfill banks, specific friction influencing the skidding direction has been applied. Frictions by definition (Eastman 1997) means inhibit movement through space. The following important influential friction factors, affecting wood skidding operation, have been considered: existence of skid trails, ridge points (lines), terrain slope and other topographical, traffic and ground features with significant influCroat. j. for. eng. 29(2008)2

ence on wood skidding. Same factors act in positive (skid rails), other in negative friction (ridge points) regarding wood skidding. Determination of specific topographic elements using DTM has already been worked out by many studies and algorithms (Jenson et al. 1988, Kr~ 2006). It is possible to modify (i.e. increase or decrease) the share of specific skidding direction on the slope terrain between two roads. The modification is done by scaling the buffer zone of roadcut or roadfill banks. A broader bank zone promotes corresponding skidding direction, while a narrow bank zone influences in the opposite direction. The result of classification of FMU Jelendol into predicted classes of skidding direction is shown in Fig. 4. The evaluation of classification has been made by analyzing the share of raster cells in the file repre-

181

J. KR^ and B. KO[IR

Predicting Wood Skidding Direction on Steep Terrain by DEM ... (177–188)

Fig. 4 Map of wood skidding direction acquired from DTM for FMU Jelendol Slika 4. Prikaz smjera privla~enja drva preuzet s DMR-a za GJ Jelendol senting the prediction of wood skidding direction. Knowing ground cell equivalent of raster cell permits deriving data per hectare. In the case study, the 0.25 ha (50 x 50 m) resolution of raster point (cell) has

been used. For the case study area of FMU Jelendol the share of 65% (2,235 ha) is predicted as suitable for downhill skidding, and the remaining 35% (1,220 ha) for uphill skidding.

Table 1 Cross tabulation indicating terrain classification into wood skidding direction classes Tablica 1. Zajedni~ka distribucija kao pokazatelj razredbe terena u razli~ite kategorije smjera privla~enja drva Inventory database Model Inventurni podatci Model Down – Niz nagib Down – Niz nagib Up – Uz nagib Down – Niz nagib Down – Niz nagib Up – Uz nagib Up – Uz nagib Up – Uz nagib Sum – Ukupno

182

Ground cell equivalent (raster cells) Ekvivalent rasterskih polja terena 5,882 2,033 2,775 2,914 13,604

Share, % Udjel, % 43 15 20 21 100

Comment Napomena Compliance – Sukladno Difference – Razli~ito Difference – Razli~ito Compliance – Sukladno

Croat. j. for. eng. 29(2008)2

Predicting Wood Skidding Direction on Steep Terrain by DEM ... (177–188)

6. Model validation – Pouzdanost modela Validity of the model has been tested by comparison of model results with terrain classification in forest inventory data maintained by Public Forest Service. In forest inventory database the predominant skidding direction is defined as downhill, uphill and even for each basic unit – forest compartment. Classification of skidding directions provided by Slovenian Public Forest Service (field inventoried) for Forest Management Unit Jelendol is shown in Fig. 5. It has been proven that the accuracy of determination of skidding direction in forest inventory database is quite poor, speaking about predominant (average) skidding direction in the forest compartment. Table 1 shows the results of cross tabulation – error matrix. Error Matrix showed a 35% of different classified area based on which »ground truth« cells have

J. KR^ and B. KO[IR

been assigned differently as in the case of terrain inventory. The analysis of different classified areas showed that the majority of misclassified areas in inventory data base are caused by treating the forest compartment as an undivided unit with one average assigned skidding direction. There are many forest compartments positions which are placed between two forest roads on slopes. The unique classifications of wood skidding direction are a correct decision only in cases where there is no forest road on borders or if one of the »border« forest roads is unproductive for wood skidding. Determination of prevailing wood skidding direction for forest compartment has also the effect of compensation misclassified areas in both classes (Up, Down). The phenomena can be stated in our

Fig. 5 Classification of skidding direction provided by Slovenian Public Forest Service for Forest Management Unit Jelendol Slika 5. Razredbu smjera privla~enja drva za gospodarsku jedinicu Jelendol provela je slovenska [umarska agencija Croat. j. for. eng. 29(2008)2

183

J. KR^ and B. KO[IR

Predicting Wood Skidding Direction on Steep Terrain by DEM ... (177–188)

Fig. 6 Location of areas for additional uphill skidding after model modification (advantage is given to uphill skidding on steep terrain) Slika 6. Prikaz dodatnih podru~ja za privla~enje drva uzbrdo (nakon prilagodbe modela) case study, too. The difference between total Up and Down classified areas is only –9% (Down, GCE: (79158657)/8657) and +15% (Up, GCE: (5689-4947)/4947), which is much less than individually misclassified cells (35.34%). We have also tested a modified model where the uphill skidding on the steepest terrain was given advantage. The reason was that in this way we estimated possible changes in the classification with a broader use of full tree method and cable cranes with a processor head. The width of buffer zones of cut bank was cut by 2/3, which has resulted in a relative approach of the steep terrain to fill banks in order to promote uphill skidding direction. The result of such terrain allocation is an increase in the share of uphill wood skidding class by 18%. At the same time there is a 10% drop in areas classified

184

as suitable for downhill wood skidding direction. The differences between the basic and changed model are shown in Fig. 6. All additional areas can be further analysed for wood stock, allowable cut, tree species, etc.

7. Discussion and conclusions – Rasprava i zaklju~ci The presented solution for the prediction of wood skidding direction is simple and not demanding in terms of specific software solution and required data availability. The solution can be derived by using only the procedure of combining standard modules in GIS software packages (Eastman 1997). In this respect we can expect that the procedure will be used for analytical and operational purposes in Croat. j. for. eng. 29(2008)2

Predicting Wood Skidding Direction on Steep Terrain by DEM ... (177–188)

developing optimal forest roads and second forest opening with skid roads and cable corridors in state forest and in case of large scale private property. The procedure will be further developed and adapted to new possibilities in forest operation technology. The changing of influential factors is evident through the introduction of: Þ a) mechanized cutting in combination with cable skidding (Visser and Stampfer 1998) and Þ b) combination of full tree method, cable skidding and processing on the forest road (Stampfer 2001). On mountainous terrains the share of uphill skidding will probably increase. The reason is very efficient cable skidding in combination with mechanized processing. This method is less expensive than downhill skidding – especially for slopes with gradients between 40 and 65% (Streif 2001) – as it needs fewer roads and skid trails construction. The development of different cable systems is permanent. Cable systems are suitable, more than before, for downhill skidding, and the setting up and down times are shorter. The new cable systems are environmentally friendly in comparison to other skidding means – especially ground-based skidding (Owende et al. 2003). Cable crane layout (parallel, fan-shaped) and skidding distances (= cable corridor lengths) influence the cost of forest operation. On the basis of costs, optimal cable length can be calculated (Ko{ir 2003). The cost calculation is not included in the present model development, while the indirect influence of costs using friction system allows various modifications. The described model is also not intended to optimise the situation in the described case study, but merely to make such an optimisation possible with further analysis. The problem of model accuracy is connected with the possibility of alternative skidding means on the same area, but the share of the so called »Gray« area in Alpine conditions has not been sufficiently analysed. There is, therefore, the problem of »what is true« or better »what is optimal«. The proposed problem solution may be a step away from the relative accuracy of field determination of skidding direction, but it also means a step toward a simpler and faster problem solution.

8. References – Literatura Ahamed, N. T. R., Rao, G. K., Murthy J. S. R., 2002: Automatic extraction of tank outlets in a sub-watershed using digital elevation models. Agricultural Water Management 57(1): 1–10. Croat. j. for. eng. 29(2008)2

J. KR^ and B. KO[IR

Aruga, K., Sessions, J., Akay, A., Chung, W., 2004: Optimizing horizontal and vertical alignments of forest roads using a high resolution DEM. Proceedings of the 12th International Mountain Logging Conference, June 13–16, 2004, Vancouver, BC. Berg, S., 1992: Terrain Classification System for forestry work. Skogsarbeten, Forestry Operations Institute, Uppsala, Sweden, 1–28. Chung, W., Sessions, J., 2001: Designing a forest road network using heuristic optimization techniques. Proceedings of the 24th Meeting of the Council of Forest Engineering, July 15–19, 2001, Snowshoe, West Virginia. Eastman, J. R., 1997: IDRISI for Windows: A grid-based geographic analysis, Version 2.0. Graduate School of Geography, Clark University, Worcestor, MA. Hay, R., 1998: Forest road design. Proceedings of the Seminar on Environmentally Sound Forest Roads and Wood Transport. June 17–22, 1996, Sinaia, Romania, FAO Rome, Italy, 44–49. Heralt, L., 2002: Using the ROADENG system to design an optimum forest road variant aimed at the minimization of negative impacts on natural environment. J. of For. Sci. 48(8): 361–365. Jenson, S. K., Dominguem, J. O., 1988: Extracting topographic structure from digital elevation data for geographic information system analysis. Photogramm. Engg. Remote Sens. 54(11): 1593–1600. Ko{ir, B., 1982: Informacija o klasifikaciji terena za organizacijsko tehnolo{ke potrebe procesa {umarstva. Meh. {umar. 7(5–6): 146–148. Ko{ir, B., Kr~, J., 2000: Where to Place and Built Forest Roads – Experience From the Model. Journal of Forest Engineering 11(1): 7–19. Ko{ir, B., 2003: Optimal line lengths when skidding wood with the Syncrofalke Cable Crane in Slovenian conditions. Proceedings of Workshop »New Trends in Wood Harvesting with Cable Systems for Sustainable Forest Management in the Mountains«, Joint FAO/ECE/ILO & IUFRO, June 18–24, 2001, Ossiach (Austria), FAO Rome, 81–90. Kraj~i~, D., 1996: Transport rules of forest wood sortiments in the forestry Nazarje. Research reports Forestry and Wood Science and Technology 48, Ljubljana, 53–75. Kr~, J., 1999: The influence of a road on increasing the forest yield according to a computer model calculation. Research reports Forestry and Wood Science and Technology 59, Ljubljana, 121–139. Kr~, J., 2000: Selecting a wood transportation route with GIS technology. Research reports Forestry and Wood Science and Technology 61, Ljubljana, 49–73. Kr~, J., 2006: A model for evaluating forest road load by forest operations. Proceedings of the International Precision Forestry Symposium »Precision Foretry in plantations, semi-natural and natural forests«, Stellenbosch University, March 5–10, 2006, South Africa, 1–14. Mark, D. M., 1983: Automated detection of drainage networks for digital elevation models. Proceedings of Auto-Carto 6, Vol. 2, Ottawa, Ontario, Canada, 288–298.

185

J. KR^ and B. KO[IR

Predicting Wood Skidding Direction on Steep Terrain by DEM ... (177–188)

Newnham, R. M., 1995: ROADPLAN: A tool for designing forest road networks. Journal of Forest Engineering 6(2): 17–26. Owende, P. M. O., Tiernan, D., Ward, S. M., Lyons, J., 2003: Is there a role for cable extraction on low gradient sensitive sites? Proceedings of Workshop »New Trends in Wood Harvesting with Cable Systems for Sustainable Forest Management in the Mountains«, Joint FAO/ECE/ILO & IUFRO, June 18–24, 2001, Ossiach (Austria), FAO Rome, 227–234.

Streif, A., 2001: View of the contractor for forest companies. Proceedings of Workshop »New Trends in Wood Harvesting with Cable Systems for Sustainable Forest Management in the Mountains«, Joint FAO/ECE/ILO & IUFRO, June 18–24, 2001, Ossiach (Austria), FAO Rome. Stuckelberger, J., Heinimann, H. R., Chung, W., Ulber, M., 2006: Automatic road-network planning for multiple objectives. Proceedings of the 29th Meeting of Council on Forest Engineering, July 30 – August 2, 2006, Coeur d’Alene, ID.

Pentek, T., Pi~man, D., Neve~erel, H., 2004: Environmental – ecological component of forest road planning and designing. Proceedings of International scientific conference »Forest constructions and ameliorations in relation to the natural environment«, September 16–17, 2004, Zvolen, Slovakia, 94–102.

Tucek, J., 1995: Computer aided forest roads (localization) and forest operations planning. XX. IUFRO Congress, S3.06 Meeting, Tampere, IUFRO XX. World Forestry Congress.

Rowan, A., A. 1996. Terrain Classification. Forestry Commission, Technical Development Branch, Technical Note 16/95, Edinburgh, 1–6.

Visser, R., Stampfer, K., 1998: Cable Extraction of Harvester-Felled Thinnings: An Austrian Case Study. Journal of Forest Engineering 9(1): 39–46.

Saarilahti, M., 2002: Soil interaction model, Dynamic terrain classification – Modelling of the seasonal variation of the trafficability on forest sites. Project deliverable D2 (Work package No. 1) of the Development of a Protocol for Ecoefficient Wood Harvesting on Sensitive Sites (ECOWOOD). EU 5th Framework Project (Quality of Life and Management of Living Resources) Contract No. QLK5-1999-00991 (1999-2002), Appendix Report No 1, 1–22.

Voinov, A., Voinov, H., Costanza, R., 1999: Surface water flow in landscape models: 2. Patuxent watershed case study. Ecol. Model. 119(2): 211–230.

Seppelt, R., Voinov, A., 2002: Optimization methodology for land use patterns using spatially explicit landscape models. Ecol. Model. 151(2–3): 125–142. Skuber, J., 1990: Forest management Plan Jelendol. Tr`i~, Forest Enterprise Kranj, 160 p. Stampfer, K. 2001: Harvester, Leistungs Daten, MHT Robin, Neuson 11002HV, Impex Konigstiger. FPP, Wien, 15 p.

Tucek, J., Pacola, E., 1999: Algorithms for skidding distance modelling on a raster digital terrain model. Journal of Forest Engineering 10(1): 67–79.

Wattenbach, M., Hattermann, F., Weng, R., Wechsung, F., Krysanova, V., Badeck F., 2005: A simplified approach to implement forest eco-hydrological properties in regional hydrological modeling. Ecol. Model. 187(1): 40–59. Yoshimura, T., 1997: Development of an expert system planning a forest road based on the risk assessment. PhD Thesis, Kyoto University, Faculty of Agriculture, pp. 82. Zhu, Z., Arp, P. A., Meng, F., Bourque, C. P. A., Foster, N. W., 2003: A forest nutrient cycling and biomass model (ForNBM) based on year-round, monthly weather conditions. Part I: Assumption, structure and processing. Ecol. Model. 169(2–3): 347–360.

Sa`etak

Odre|ivanje smjera privla~enja drva prema postoje}oj mreì {umskih cesta na strmim terenima U radu je predstavljen model odre|ivanja smjera privla~enja drva na strmim terenima. Model se koristi za razredbu ve}ih {umskih povr{ina, a prikazana je njegova primjenjivost u gospodarskoj jedinici Jelendol u Sloveniji. Pri gospodarenju {umskim resursima ekolo{ka i tehnolo{ka sastavnica sna`no su povezane. Razredbom su terena obuhva}ene odre|ene {umske povr{ine koje su razvrstane u homogene jedinice temeljem prethodno vrlo precizno definiranih kriterija. Pri tome se primjenjuju razli~ite metode te razlikuju mnogi utjecajni ~imbenici. [umarstvo je, prema Löffleru (1984), prihvatilo razredbu terena kao vaàn logi~ki alat za gospodarenje {umskim resursima, planiranje radova u {umarstvu te provedbu radova u uzgajanju {uma i u pridobivanju drva. Alpsko je podru~je sa stajali{ta operativne razredbe terena vrlo zanimljivo. Naime, pri redovitom pridobivanju drva prevladavaju zglobni traktori (vu~a drva), me|utim vrlo su ~este {umske ì~are kojima se iznosi drvo i razli~iti oblici strojeva za izvoènje drva – forvarderi. Operativno planiranje nudi smjernice za radove u pridobivanju drva u specifi~nim uvjetima uz preporuku najpovoljnijega na~ina rada. Va`na je sastavnica operativnoga planiranja, kako zbog samoga obavljanja {umskih radova tako i zbog ekonomski isplativijega gospodarenja, tematska karta koja prikazuje obilje`ja terena (Kr~ 1999). Definiranje se najpogodnijega na~ina naj~e{}e temelji na tehni~ko-tehnolo{kim, okoli{no-ekolo{kim i sociolo{kim kriterijima uz ograni~enja koja name}u smjernice za potrajnim i odrìvim gospodarenjem {umskim ekosustavima.

186

Croat. j. for. eng. 29(2008)2

Predicting Wood Skidding Direction on Steep Terrain by DEM ... (177–188)

J. KR^ and B. KO[IR

Pri izradi se modela primjenjivih postupaka za pridobivanje drva, uz odabir najpovoljnijega s obzirom na razredbu terena, koriste: topografski podaci, podaci o tipu tla, nosivost tla i sadràj vode u tlu, mreà vodotokâ, mreà primarnih {umskih prometnica i mreà sekundarnih {umskih prometnica. Na sje~u i izradu te na privla~enje drva, posebno na nagnutom terenu, utje~e odabir smjera privla~enja drva do pomo}noga stovari{ta (do mreè {umskih cesta). Smjer se privla~enja drva dijeli u tri kategorije: privla~enje uzbrdo, privla~enje nizbrdo i privla~enje po ravnom terenu. Na strmom terenu prostorni poloàj mreè {umskih cesta ima snaàn utjecaj na smjer privla~enja drva i na provedbu sje~e stabala (usmjereno ru{enje). Mogu}i su oblici privla~enja drva na nagnutim terenima: vu~a drva po tlu zglobnim traktorima, izno{enje drva {umskim ì~arama i izvoènje drva forvarderima, uz njih povezani sustavi sekundarnih {umskih prometnica (traktorski putovi i ì~ne linije) koji su povezani s opasno{}u od erozije. Sve to ~ini dizajniranje modela zahtjevnim. U dosada{njim se sli~nim istraìvanjima, pri optimizaciji mreè {umskih cesta na strmim terenima, naj~e{}e koristio digitalni model terena (DMR). Mnoge su se studije bavile najboljim mogu}im rje{enjem pri izboru poloàja novih {umskih cesta (Chung 2001, Aruga i dr. 2004), ali i mogu}no{}u izvedivosti pojedinih ina~ica te njihovim utjecajem na gusto}u {umskih cesta (Ko{ir i Kr~ 2000). Dosta se istraìvao odnos prostornoga poloàja {umskih cesta, njihove gusto}e i okoli{a (Newnham 1995, Tucek 1995, Yoshimura 1997) te izbor najpogodnijega sredstva za privla~enje drva uspore|uju}i naj~e{}e zglobni traktor i {umsku ì~aru (Tucek 1999, Stuckelberg 2006). Razvidno je da danas {umarska operativa ne pridaje dovoljno va`nosti rezultatima dobivenim razli~itim ra~unalnim simulacijama temeljenim na provjerenim ra~unalnim modelima ve}, radije i ~e{}e, koristi rezultate funkcionalnih razredbi temeljenih na dosada{njim iskustvima. Osnovni je cilj ovoga istraìvanja kreiranje modela za razredbu terena na osnovi smjera privla~enja drva. Pri tome su kori{tena dva glavna sloja podataka: digitalni model terena i digitalni podaci o mreì {umskih cesta te uz njih GIS-ov programski modul (Eastman 1993). Smjer privla~enja drva utje~e na u~inkovitost i tro{ak pridobivanja drva. Model }e omogu}iti realniji izra~un tro{kova privla~enja drva ako se zna udio i prostorni poloàj pojedine kategorije smjera privla~enja. Poznavanje toka drva do pojedine {umske ceste pruà mogu}nost boljega planiranja tro{kova njezina odràvanja (Kr~ 2006). Nadalje, model pruà dodatne koristi za lak{u i uspje{niju provedbu postupka daljnjega primarnoga i sekundarnoga otvaranja {umskoga podru~ja, ali i postupka izbora najboljega na~ina pridobivanja drva. Postavlja se pitanje gdje se nalazi granica privla~enja drva do pomo}nih stovari{ta uz rub dviju {umskih cesta izgra|enih na padini kontinuiranoga nagiba i ravnomjerno raspore|ena drva. Do sada se pretpostavljalo kako drvo sa {umske povr{ine izme|u dviju {umskih cesta treba s 2/3 {irine povr{ine privla~iti na donju cestu (privla~enje nizbrdo), a s 1/3 {irine povr{ine privla~iti na gornju cestu (privla~enje uzbrdo). Navedena tvrdnja moè vrijediti za vu~u drva zglobnim traktorom, ali se pri izno{enju drva ì~arama moè o~ekivati obrnuta situacija. Prona}i jednostavno rje{enje i odgovor na pitanje {to i kako raditi na kojem podru~ju osnovna je ideja ovoga rada. Za podru~je istraìvanja odabrana je gospodarska jedinica Jelendol povr{ine 3653 ha. Smje{tena je na nagnutom terenu u Sloveniji, u alpskom podru~ju, blizu granice s Austrijom. Nadmorska joj je visina izme|u 750 i 2065 m. Gusto}a {umskih cesta iznosi 23 m/ha, prosje~na drvna zaliha je 311 m3/ha uz omjer smjese ~etinja~a i lista~a 80 : 20 (Skuber 1990). Sjeverni je dio gospodarske jedinice Jelendol izlu~en kao za{ti}eno podru~je bez propisanoga etata. Mreà je {umskih prometnica na mreù javnih prometnica povezana samo na jednom mjestu (jugozapadni dio gospodarske jedinice), a to je ujedno i to~ka najniè nadmorske visine u GJ Jelendol. Radi odre|ivanja smjera privla~enja drva spoj sekundarnih {umskih prometnica (traktorskih putova i ì~nih linija) s rubom {umskih cesta (pomo}nim stovari{tem) razdijeljen je u dvije kategorije: Þ sekundarna se {umska prometnica sa {umskom cestom spaja na strani iskopa zasjeka – privla~enje se drva obavlja nizbrdo Þ sekundarna se {umska prometnica sa {umskom cestom spaja na strani nasipa zasjeka – privla~enje se drva obavlja uzbrdo. Privla~enje po ravnom nije uklju~eno u model jer je malo vjerojatno da na nagnutom terenu postoji ova kategorija privla~enja drva. Na slici 1 prikazan je postupak razredbe ome|enih povr{ina oko {umskih cesta na iskopanu i nasipanu stranu zasjeka kroz {est koraka. 3 D prikaz GJ Jelendol kori{ten je u postupku provjere kategorije smjera privla~enja drva. Procijenjen je poloàj svake dionice (segmenta) {umske ceste u odnosu na padinu te su formirane tri kategorije: Þ dolinske dionice (dionica {umske ceste ima samo iskope; rije~ je o usjeku) Þ padinske dionice (dionica {umske ceste ima i iskopanu i nasipanu stranu zasjeka; rije~ je o zasjeku) Þ grebenske dionice (dionica {umske ceste ima samo nasipe; rije~ je o nasipu). Rezultati su navedene kategorizacije prikazani na slici 3.

Croat. j. for. eng. 29(2008)2

187

J. KR^ and B. KO[IR

Predicting Wood Skidding Direction on Steep Terrain by DEM ... (177â&#x20AC;&#x201C;188)

Prepoznavanje specifi~nih topografskih elemenata uz primjenu DMR-a tako|er je obra|eno putem razli~itih studija i algoritama (Jenson i dr. 1988, Kr~ 2006). Mogu}e je modificirati udio pojedine kategorije privla~enja drva na strmom terenu, a na povr{ini izme|u dviju {umskih cesta. Modifikacija se obavlja promjenom {irine ome|ene povr{ine na strani nasipa ili iskopa. Rezultat razredbe GJ Jelendol u definirane kategorije smjera privla~enja drva prikazan je na slici 4. Udio podru~ja pogodnih za privla~enje nizbrdo iznosi 65 % (2235 ha), a za privla~enje uzbrdo 35 % (1220 ha). Primjenjivost modela potvr|ena je usporedbom rezultata dobivenih modelom i razredbom terena koju je obavila Slovenska savjetodavna slu`ba tijekom inventure {uma ove gospodarske jedinice. Za svaki je odsjek smjer privla~enja drva definiran kao: privla~enje uzbrdo, privla~enje nizbrdo i privla~enje po ravnom terenu. Rezultati ove, prili~no jednostavne, ali i neprecizne razredbe terena (jer su uop}eni podaci za cijeli odsjek), prikazani su na slici 5. Iz tablice 1 vidi se da je na 35 % povr{ine utvr|ena druga~ija kategorija smjera privla~enja drva primjenom modela i procjenom na terenu. U ve}ini je slu~ajeva do pogre{ne kategorizacije do{lo zbog procjene prosje~ne kategorije smjera privla~enja na razini odsjeka (ima puno odsjeka koji se nalaze smje{teni na padinama izme|u dviju {umskih cesta).Tako|er je pri procjeni smjera privla~enja po odsjecima prisutan u~inak kompenzacije pogre{no razvrstanih povr{ina u oba razreda smjera nagiba (razlika izme|u ukupnih razreda smjera nagiba nizbrdo i uzbrdo mnogo je manja od razlike individualno krivo procijenjenih razreda nagiba po poljima dimenzija 50 x 50 m). Obavljena je i modifikacija modela radi pove}anja udjela podru~ja pogodnih za privla~enje drva uzbrdo na najstrmijim terenima. Pri tome bi se koristila stablovna metoda i {umska ì~ara opremljena procesorskom glavom. [irina ome|enih povr{ina oko {umske ceste na strani iskopa zasjeka smanjena je za 2/3, {to je utjecalo na pove}anje udjela privla~enja drva uzbrdo za 18 % uz istodobno smanjen udio privla~enja drva nizbrdo za 10 % (sva podru~ja koja su dodatno uvr{tena u razred privla~enja drva uzbrdo zahtjevaju dodatne analize drvne zalihe, etata, vrste drve}a itd.). Razlike izme|u baznoga i modificiranoga modela prikazane su na slici 6. Model se moè iskoristiti za analizu postoje}e te planiranje i razvoj budu}e mreè primarnih i sekundarnih (traktorski putovi i ì~ne linije) {umskih prometnica u dràvnim {umama i ve}im kompleksima privatnih {uma. Postupak izrade modela trebat }e unaprijediti, dalje razvijati i prilagoditi novim postupcima u pridobivanju drva: mehanizirana sje~a u kombinaciji sa {umskom ì~arom (Visser i Stampfer 1998), stablovna metoda uz primjenu {umske ì~are opremljene procesorskom glavom i izradu na {umskoj cesti (Stampfer 2001). U planinskim }e podru~jima udio privla~enja drva uzbrdo rasti. Razlog je vrlo u~inkovito izno{enje drva {umskom ì~arom u kombinaciji s izradom na {umskoj cesti (procesorska glava). Ta je metoda jeftinija od privla~enja drva zglobnim traktorima nizbrdo, posebno za nagibe terena izme|u 40 i 65 %, a manja je i potreba izgradnje {umskih cesta i traktorskih putova (Streif 2001). Stalno se razvijaju razli~ite {umske ì~are. One postaju sve pogodnije za izno{enje drva nizbrdo i okoli{no su najprihvatljivije za pridobivanje drva, posebno u usporedbi sa strojevima kretnim po zemlji (Owende i dr. 2003). Primijenjeni model ne uklju~uje izra~un tro{kova niti ima namjeru optimizirati postupke pridobivanja drva, iako je odre|ena optimizacija, uz daljnje analize, mogu}a. Postoji problem primjene alternativnih sredstava za privla~enje drva u alpskim uvjetima jer jo{ uvijek nemamo odgovor na pitanje {to je optimalno. Klju~ne rije~i: pridobivanje drva, smjer privla~enja drva, model, mreà {umskih cesta, strmi teren

Authors address â&#x20AC;&#x201C; Adresa autora:

Received (Primljeno): August 21, 2008 Accepted (Prihva}eno): November 11, 2008

188

Assist. Prof. Janez Kr~, PhD. e-mail: janez.krc@bf.uni-lj.si Prof. Bo{tjan Ko{ir, PhD. e-mail: bostjan.kosir@bf.uni-lj.si University of Ljubljana, Biotechnical Faculty Department of Forestry and Forest Resources Ve~na pot 83 1000 Ljubljana SLOVENIA Croat. j. for. eng. 29(2008)2

Orginal scientific paper – Izvorni znanstveni rad

Accuracy Analysis of GPS Positioning Near the Forest Environment Atinç Pirti Abstract – Nacrtak GPS has become an essential tool for georeferencing. In some cases, GPS is used for unfavorable conditions although it was developed for open field studies. This paper analyzes the achievable accuracy and performance of GPS near the forest. Three surveying marks have been established with the distance seperation five meter in length. Two GPS campaigns were conducted for the selected marks in the forest. The same campaign was repeated once again after the forest was cut off. The experiments demonstrate the degradation of the GPS accuracy due to the forest. As a result, the largest horizontal accuracy errors were found to be in the forest. Horizontal accuracy errors were the smallest in the area without obstacles. Large vertical accuracy errors were produced in the forest; however vertical accuracy errors were also relatively small after the forest was cut off. The standard deviations improved by about 50–70% for both baselines and height differences when the forest was cut off. In conclusion, tree canopies greatly affected both horizontal and vertical accuracy. Keywords: GPS, forest, accuracy, terrestrial measurements

1. Introduction – Uvod Global Positioning System (GPS) has been applied successfully in many areas of forest industry. Typical applications include fire prevention and control, harvesting operations, insect infestation, boundary determination, and aerial spraying. The past scientific literature found that the equivalent accuracies could be obtained under a canopy compared with the open field. These results are not supported by recent studies. While the topic seems to be somewhat avoided in the relevant scientific literature (GPS not being for use in the non-open environment in the first place), it is still recurrent in many discussion lists on the web. Deckert and Bolstadt (1996) studied the effects of terrain, forest canopy, number of consecutive position fixes and Position Dilution of Precision (PDOP) on GPS accuracy. They found that the positional accuracy was higher for open sites compared to sub-canopy and higher for deciduous sites versus coniferous. Sigrist et al. (1999) discussed the impact of the forest canopy on quality and accuracy assessment of GPS measurements. Hasegawa and Yoshimura (2003) studied the performance of dualfrequency GPS receivers for static surveying under tree canopies. Croat. j. for. eng. 29(2008)2

Forest and natural resource applications can be achieved efficiently employing GPS data collection technologies. However, there are limiting factors in environments, such as forest canopy, that cause adverse effects on the reception of GPS signals. Steep terrain and heavy forest cover make GPS data capture slow due to reception of acceptable satellite coverage. So, position accuracy is often degraded in difficult terrain conditions, and in some cases it may not meet accuracy standards and requires resurveying. In the forests, canopy cover may interfere with satellite signal reception and make it difficult to make reliable measurements. The combined effects of forest cover and terrain will degrade the performance of all GPS receivers. The GPS signals are affected by the surrounding trees and earth and that affects adversely both accuracy (how close the lines and points are to their true location) and productivity (how much of the time the receiver is tracking enough satellites). The users are limited by the view of the sky in a tree canopy environment resulting in the GPS receiver to be locked to only high elevation satellites. Satellite constellation has a large effect on the quality of the data collected in forested environments such as data bias. Constantly changing constellations re-

189

A. PIRTI

Accuracy Analysis of GPS Positioning Near the Forest Environment (189–199)

sult in inconsistent and poor relative data accuracy. Forest canopy affects the GPS signals because of obstruction, attenuation, and reflection. So, line-of-sight GPS signals are obstructed by solid objects. The signal is blocked by tree trunks, larger branches, and terrain features such as mountains. The signals are weakened and attenuated by leaves and small branches. This attenuation can make it very difficult for a GPS receiver to track the signals. At some point, the receiver will not be able to track the signal at all and the effect will be the same as if the signal was obstructed. Even if the signal can be tracked, some receivers will have difficulty in measuring the pseudoranges accurately. The phenomenon of a satellite signal reaching an antenna by more than one path (direct and some reflected paths) is called multipath. This multipath can cause large variations in position estimates in a variety of environment, e.g., under forest canopy. The main effect of signal obstruction is to convey an increase in PDOP. As PDOP is related to the satellite geometry and number of satellites logged, a lower PDOP is expected when more satellites are observed. PDOP is a unitless measure indicating the quality of satellite geometry. When the satellites are spread around the sky, the PDOP value is low and the computed position is more accurate. In the case when satellites are grouped closely, the PDOP is high and positions are less accurate. As the PDOP is directly related to the position accuracy, more satellites and a lower PDOP will usually mean better accuracy under forest canopy. Modern GPS systems have been improved for the satellite tracking technology so that weaker signals can be observed under trees with foliage (Note that dense foliage will still cause cycle slips). In spite of this advanced tracking capability, the signals are noisier, weaker and more likely to be subject to multipath and diffraction. The surveyor should be aware that positions may not be

accurate despite the quality indicators showing good solutions. To overcome this situation, the surveyors are required to check out the GPS results using a total station. In such cases, terrestrial survey can help productivity in difficult terrain conditions and be carried out to obtain an independent result of the position for assessing the accuracy of the GPS results in the forest environment (Parkinson and Spilker 1996, Hoffmann et al. 2000, Pirti 2005, Rabbany 2006). The aim of this study is to assess the achievable accuracy of GPS surveys under forest environment.

2. Material and Methods – Materijal i metode In this paper two experiments are designed to show the effect of increasing relative distance to a tree canopy as well as quantifying the magnitude of multipath effect. The two experiments were performed in the Samandýra area of Istanbul, Turkey (Fig. 1). The GPS (static) measurements were taken both in the presence of a forest (July 30, 2003 – Day of year – DOY 211) and after the forest was cut off (September 16, 2003 – DOY 259), Fig. 2. To study signal multipath and diffraction effects on static GPS baselines due to forest, three stations (P1, P2 and P3) were located at a distance of about 0 m, 5 m and 10 m from the forest environment consisting of around 8–10 m tall pine trees (Fig. 2a). Starting with point P1 (the border of the forest), two other points with distances of about 5 m (P2) and 10 m (P3) from P1 were marked. Three stations were observed both using GPS and terrestrial measurement methods. In the first step of this study, GPS measurements were carried out for both situations either when the forest exists (Fig. 1a) or is cut off in the project area (Fig. 1b). The data were recorded at three stations by static GPS measurements at time intervals of 6 hours.

Fig. 1 Project area and GPS network Slika 1. Podru~je istra`ivanja i polo`aj GPS prijamnika 190

Croat. j. for. eng. 29(2008)2

Accuracy Analysis of GPS Positioning Near the Forest Environment (189–199)

A. PIRTI

Fig. 2 Stations in the project area Slika 2. Stajali{ta istra`ivanoga podru~ja Three Ashtech Z Surveyor receivers and Ashtech geodetic antennas were used for the static GPS measurements. The coordinates of the three points were determined in two GPS sessions (with and without forest) with 5 seconds sampling interval, 10 degrees elevation cut-off for two days. The RINEX data obtained were processed with Bernese 5.0 GPS software in order to compute the coordinates of P3 for both sessions in ITRF 2000 on both days (see Tables 1 and 2). IGS reference station ISTA – (Fig. 1b) was fixed in the processing.

The forest cause severe obstruction of almost 50% of the sky for P1 on DOY 211 (Fig. 1b and Fig. 2a). The coordinates of P1 were much affected by the forest environment on DOY 211, and however, the coordinates of P3 were less affected on both days by forest environment than the other points. Therefore, the coordinates of P3 were fixed in the static and kinematic (epoch-by-epoch) processing by using Ashtech Solution 2.60 GPS Software to compute the coordinates of the other two points (P1 and P2) because of its far distance from the forest environment (Tables 1 and 2).

Table 1 Coordinates and standard deviations of three points in the project area on DOY 211 Tablica 1. Koordinate i standardne devijacije za tri stajali{ta na istra`ivanom podru~ju prvoga dana mjerenja Point Stajali{te

Coordinate Koordinata jITRF

Standard deviation Standardna devijacija sj, mm

Coordinate Koordinata lITRF

Standard deviation Standardna devijacija sl, mm

Coordinate Koordinata HITRF, m

Standard deviation Standardna devijacija sH, mm

ISTA P1 P2 P3

41° 06' 16''.01024 40° 58' 17''.48034 40° 58' 17''.54318 40° 58' 17''.59966

0 27 27 27

29° 01' 09''.62368 29° 12' 55''.88682 29° 12' 56''.08298 29° 12' 56''.28366

0 27 27 27

147.246 180.450 179.738 180.218

0 36 36 36

Table 2 Coordinates and standard deviations of three points in the project area on DOY 259 Tablica 2. Koordinate i standardne devijacije za tri stajali{ta na istra`ivanom podru~ju drugoga dana mjerenja Point Stajali{te

Coordinate Koordinata jITRF

Standard deviation Standardna devijacija sj, mm

Coordinate Koordinata lITRF

Standard deviation Standardna devijacija sl,, mm

Coordinate Koordinata HITRF, m

Standard deviation Standardna devijacija sH, mm

ISTA P1 P2 P3

41° 06' 16''.01024 40° 58' 17''.48028 40° 58' 17''.54320 40° 58' 17''.59962

0 10 10 10

29° 01' 09''.62368 29° 12' 55''.88546 29° 12' 56''.08319 29° 12' 56''.28360

0 13 13 13

147.246 180.416 179.731 180.218

0 13 13 13

Croat. j. for. eng. 29(2008)2

191

A. PIRTI

Accuracy Analysis of GPS Positioning Near the Forest Environment (189–199)

3. Results and discussion – Rezultati s diskusijom P1–P3 and P2–P3 baselines were processed by Ashtech Solution 2.60 GPS Software in kinematic mode to investigate epoch-by-epoch variations of the coordinates (P3 fixed). The aim is to examine multipath and diffraction effects on the coordinate

results. In data processing, x, y and h coordinate component residuals (Dx, Dy, Dh) of the two stations were obtained for every epoch. The residuals were referred to as the difference between the estimated coordinates and the reference ones for each epoch. Fig. 3a and Fig. 3b show epoch-by-epoch coordinate residuals of P1 on DOY 211 and 259, respectively. It was shown in Fig. 3a and 3b that the standard devia-

Fig. 3 Epoch-by-epoch processing results for P1 Slika 3. Rezulati izmjere po epohama za stajali{te P1 192

Croat. j. for. eng. 29(2008)2

Accuracy Analysis of GPS Positioning Near the Forest Environment (189â&#x20AC;&#x201C;199)

A. PIRTI

Fig. 4 Epoch-by-epoch processing results for P2 Slika 4. Rezultati izmjere po epohama za stajali{te P2 tions and mean values of P1 on DOY 211 are considered to be a factor approximately three or four times larger than that of P1 on DOY 259. The standard deviations improved by about 70 to 80 percent for P1 when the forest was cut off. It is clear that the forest caused a significant bias in the coordinate residuals on DOY 211 (Fig. 3a). The time series of coordinate residuals of the GPS session confirm that there is a Croat. j. for. eng. 29(2008)2

strong bias of about 10 cm in horizontal components and about 25 cm in height components on DOY 211 indicating that the forest causes significant multipath effect. At certain times, enough satellites were not tracked to fix the ambiguity value for P1, e.g., 9:30â&#x20AC;&#x201C;10:30 UT (Fig. 3a). The three components are presented for P2 on both days in Fig. 4. Fig. 4a shows epoch-by-epoch co-

193

A. PIRTI

Accuracy Analysis of GPS Positioning Near the Forest Environment (189–199)

Fig. 5 PDOP and number of satellites for P1–P3 and P2–P3 baselines Slika 5. PDOP i broj satelita za osnovne linije P1–P3 i P2–P3 194

Croat. j. for. eng. 29(2008)2

Accuracy Analysis of GPS Positioning Near the Forest Environment (189–199)

A. PIRTI

Table 3 Mean values of the distances and height differences between two points by using the terrestrial and static GPS measurements Tablica 3. Srednje vrijednosti udaljenosti i visinske razlike izme|u stajali{ta klasi~nom metodom izmjere totalnom stanicom i izmjera stati~nim GPS prijamnikom Distance – Udaljenost Baseline Osnovna linija

Terrestrial Terenska

GPS DOY 211 Dan 1.

Height difference – Visinska razlika GPS DOY 259 Dan 2.

Geo. levelling Nivelacija

GPS DOY 211 Dan 1.

GPS DOY 259 Dan 2.

Dh, m

s, m P1–P3

10.020

9.985

10.013

±0.192

±0.232

±0.198

P2–P3

5.026

5.028

5.022

±0.491

±0.480

±0.487

ordinate results of P2 on DOY 211. P2 was mounted at a distance of about 5 m from the forest environment. These Dx and Dy components change between a few millimetres up to 10 cm on that day. The height component is, however, less consistent and sometimes shows differences up to 15 cm. Figure 4b shows coordinate residuals for P2 on DOY 259. It is also

shown in Figures 4a and 4b that standard deviations and mean values of P2 on DOY 211 are considered to be a factor approximately two or four times larger than the following day. Again, the standard deviations improved by about 50 to 70 percent for P2 when the forest was cut off. It is clear that forest caused a significant bias in the coordinate results on

Fig. 6 The epoch by epoch changes of P1–P3 baseline (comparison of GPS and terrestrial survey) Slika 6. Promjene po epohama za osnovnu liniju P1–P3 (usporedba izmjere GPS-om i totalnom stanicom) Croat. j. for. eng. 29(2008)2

195

A. PIRTI

Accuracy Analysis of GPS Positioning Near the Forest Environment (189–199)

Fig. 7 The epoch by epoch changes of P2-P3 baseline (comparison of GPS and terrestrial survey) Slika 7. Promjene po epohama za osnovnu liniju P2-P3 (usporedba izmjere GPS-om i totalnom stanicom) DOY 211 (Fig. 4a). The epoch-by-epoch coordinates of the GPS session confirm that there is a strong bias of about 10 cm in horizontal components and 15 cm in height components indicating that the forest causes significant multipath. The difference in precision between good and poor satellite configurations P1–P3 and P2–P3 baselines on both days can be seen in Fig. 5a and 5b. The PDOP changes according to the number and distribution of the satellites tracked. The good satellite configuration on DOY 259 results in small coordinate variations compared to the coordinate variations on DOY 211 for P1 and P2 (Fig. 3, 4 and 5). It is obvious that the epoch-by-epoch horizontal components are not affected by high multipath effects on DOY 259. In order to compare the GPS results with those obtained by using an independent measurement method, distances between points were measured with a total station. Terrestrial surveys were used to check the static and kinematic (epoch by epoch) GPS

196

results, especially for the spatial distances and height differences. Topcon DL-102 digital level surveying instrument (with a measurement accuracy of 1.5 mm/km) and a barcode rod were used to determine the height differences and Nikon DTM 330 Total Station (measurement accuracy for angles and distances ±1.5 mgon and distances 3 mm + 2 ppm, respectively) was used to measure spatial distances between all points. Distance and height measurements were made (10 series) and then the mean value of all measurements computed as shown in Table 3. In this test, the variation of the geoid was neglected since the distances are very close. The quality of the static GPS results was checked out against spatial distances and height differences determined by the terrestrial measurements. GPS and terrestrial methods show differences up to 4 cm for horizontal and vertical coordinates in the presence of multipath due to the forest environment, whereas about 1 cm for horizontal and vertical coordinates without forest environment (Table 3). As expected, P1 was Croat. j. for. eng. 29(2008)2

Accuracy Analysis of GPS Positioning Near the Forest Environment (189–199)

mostly affected by multipath due to forest. This effect can be seen in the solutions of P1–P3. P2–P3 is less affected by the multipath since P2 and P3 points are installed further from the forest. Fig. 6 shows epoch by epoch changes of the distance differences (DS) of P1–P3 differencing distances from total station and GPS. It is found that the variations were greater in height differences and smaller in distances at the project area. Figures 6a and 6b show the standard deviation of DS for P1–P3 as ±1.2 cm on DOY 211 and ±0.3 cm with and without forest, respectively. The mean value of distance difference for P1–P3 was 3.0 cm and 0.8 cm for the days with and without forest, respectively. The standard deviation of height difference (DH) variations for P1–P3 was ±3.2 cm on DOY 211 and ±1.0 cm on DOY 259 (Figures 6a and 6b). The mean value of the height difference for P1–P3 was 3.6 cm on DOY 211 and 0.8 cm on DOY 259. So, signal multipath due to forest environment affected the horizontal and vertical components. Positioning accuracy under forest canopy was considerably lower than the other case for horizontal and vertical components. As for P2–P3, the standard deviation of DS for P2–P3 is ±1.1 cm on DOY 211 and ±0.3 cm on DOY 259 (Fig. 7a and 7b). The mean value of DS for P2–P3 is 1.2 cm in multipath environment, whereas 0.2 cm for the other case. The standard deviation of DH variations for P2–P3 is ±2.1 and ±1.0 cm; the mean values are 1.5 and 0.7 cm with and without forest, respectively. As a result, the signal blockage due to tree canopies could be considered as the main problem affecting the use of GPS near the forest environment despite the presence of good satellite windows. It is clear that the multipath effect disappears from the solutions as the forest is cut off. Improvements can be observed of up to 4 cm in both distance and height solutions. Naesset et. al (2000) demonstrate that accuracy can be achieved for static measurement under the forest canopy only within 1–9 cm. The obtained results in this paper are consistent with this study.

4. Conclusions – Zaklju~ci This study indicates that the extent of forest obstruction has a significant effect on the accuracy, pre-

Croat. j. for. eng. 29(2008)2

A. PIRTI

cision and performance of GPS positions. The observation points should be carefully installed, i.e. distances to the forest border should be incremented so that the point is less affected by the multipath. Comparison of individual session solutions shows that the accuracy of GPS results was degraded for both horizontal and vertical components. As expected, the accuracy of the height component is about 2 or 4 times lower. The standard deviation of coordinate solutions gradually improves as the surveying station is moved away from the border of the forest. High multipath environment (forest) causes the standard deviations and the mean values of coordinate estimates to become lower by about 50–70%. Comparisons of GPS results with terrestrial surveys also reveal that the effect of tree canopy resulting in multipath effect is obvious. Spatial distances and height differences were degraded by about 4 centimeters.

5. References – Literatura Deckert, C., Bolstadt, P. V., 1996: Forest Canopy, terrain and distance effects on global positioning system point accuracy. Photogrammetric Engineering and Remote Sensing 62: 317–321. Hasegawa, H., Yoshimura T., 2003: Application of dual frequency GPS receivers for static surveying under tree canopies. Japan Journal Forest Society 8: 103–110. Hoffmann-Wellenhof, B., Lichtenegger, H., Collins, J., 2000: GPS Theory and Practice. Fifth Revised Edition, Wien New York, Springer-Verlag, 1–382. Næsset, E., Bjerke, T., Øvstedal, O., Ryan, L. H., 2000: Contributions of differential GPS and GLONASS observations to point accuracy under forest canopies. Photogrammetric Engineering & Remote Sensing 66: 403–407. Parkinson, B. W., Spilker, J. J., 1996: Global Positioning System, Theory and Applications. Stanford University and Telecom, Stanford, America Institute of Aeronautics & Ast, California, 1–793. Pirti, A., 2005: Using GPS near the forest and quality control. Survey Review 38: 286–298. Rabbany, A., 2006: Introduction to GPS. Second Edition, New York, USA, Artech House Publishers, 1–230. Sigrist, P., Coppin, P., Hermy, M., 1999: Impact of forest canopy on quality and accuracy of GPS measurements. Journal of Remote sensing 20: 3595–3610.

197

A. PIRTI

Accuracy Analysis of GPS Positioning Near the Forest Environment (189–199)

Sa`etak

Analiza to~nosti pozicioniranja GPS-a uz {umski okoli{ U dana{njem je svijetu GPS postao prijeko potreban alat za odre|ivanje pozicije, tzv. georeferenciranje. GPS je tako|er na{ao primjenu u razli~itim podru~jima {umarstva, kao {to su ure|ivanje {uma, pridobivanje drva, za{tita {uma od poàra i biolo{kih {tetnika. Kori{tenje GPS-a za prikupljanje podataka na {umskim podru~jima pokazalo se zahtjevnim, ali i u~inkovitim. Me|utim, postoje ograni~avaju}i ~imbenici, kao {to su sastojinski sklop i konfiguracija terena, koji uzrokuju smetnje u prijmu GPS signala. Strmi tereni i gust sklop kro{anja smanjuju kakvo}u signala, zbog ~ega je preciznost odre|ivanja pozicije nepouzdana i ~esto ne zadovoljava propisane norme. Stalna promjena poloàja satelita zna~ajno utje~e na kakvo}u prikupljenih podataka u {umskim predjelima, {to se o~ituje kao odstupanje u to~nosti podataka. Debla i kro{nje stabala slabe i ometaju signal, a to rezultira slabijom kakvo}om prikupljenih podataka, jer signal ne uspijeva sti}i do GPS prijamnika. Pojava pri kojoj satelitski signal dolazi do antene prijamnika razli~itim putovima naziva se vi{epu}e. Upravo se zbog te pojave u podru~ju obraslom {umom doga|aju velika odstupanja u odre|ivanju pozicije. Kakvo}a je signala tako|er povezana s PDOP veli~inom koja je u svezi s poloàjem satelita i brojem satelita ~iji signal GPS prijamnici primaju. Kada su sateliti ravnomjerno raspore|eni, PDOP vrijednost je mala i izra~un pozicije je to~niji. U slu~aju kada su sateliti grupirani, PDOP vrijednost raste, a to~nost se odre|ivanja pozicije smanjuje. Cilj je ovoga istraìvanja procijeniti mogu}u to~nost u {umskom okoli{u. U ovom su radu oblikovana dva pokusa koji prikazuju utjecaj pove}anja relativne udaljenosti u odnosu na kro{nju stabla te kvantificiranje veli~ine vi{eputnoga efekta. Pokusi su obavljeni u okolici Istanbula, u podru~ju Samandira (slika 1). Zbog potreba istraìvanja vi{eputnoga signala i efekta distrakcije na stati~ni GPS ure|aj, tri su stajali{ta postavljena (P1, P2 i P3) na udaljenosti 0, 5 i 10 metara od sastojine (slika 2). Na stajali{tima je primijenjena klasi~na metoda izmjere totalnom stanicom i izmjere GPS prijamnicima. Podaci su prikupljeni pomo}u prijamnika Ashtech Z Surveyor i pripadaju}im geodetskim antenama Ashtech na svakom stajali{tu. GPS ure|ajem mjereno je na istom {umskom podru~ju u dva navrata: u neposrednoj blizini {ume i na {umskoj povr{ini nakon sje~e. Dobiveni su podaci (format RINEX) obra|eni pomo}u aplikacije Bernese 5.0 GPS kako bi koordinate bile izra~unate prema referentnomu koordinatnomu sustavu ITRF 2000 (tablice 1 i 2). Sklop je sastojine prouzro~io zna~ajne smetnje pri prvom snimanju na stajali{tu P1 jer je 50 % vidljivoga horizonta bilo zaklonjeno {umom. Za odre|ivanje to~nih koordinata kao referentno je uzeto stajali{te P3 prema kojem su izra~unate to~ne koordinate za preostala stajali{ta iz razloga {to je to stajali{te najmanje bilo pod utjecajem blizine {ume. Osnovne su linije P1–P3 i P2–P3 obra|ene pomo}u Ashtech Solution 2.60 GPS aplikacije kinemati~kom metodom radi ispitivanja tzv. epoch-by-epoch oscilacije koordinata. Prilikom obrade podataka razlike su x, y i h varijabli dvaju GPS prijamnika dobivene za svaku epohu (za svaki podatak snimljen u pojedinom intervalu), a odnose se na razliku izme|u referentnih i izmjerenih koordinata. Slike 3a i 3 prikazuju razlike koordinatnih varijabli po epohama na stajali{tu P1. Standardna je devijacija smanjena 70 – 80 % na stajali{tu P1 nakon {to je {uma posje~ena. O~ito je kako {uma uzrokuje zna~ajnu sustavsku pogre{ku u razlici koordinatnih varijabli (slika 3a). Vremenski niz koordinatnih razlika pokazuje veliku sustavsku pogre{ku od pribli`no 10 cm za horizontalnu sastavnicu i 25 cm za visinsku sastavnicu zbog utjecaja blizine {ume na pojavu vi{epu}a. Slike 4a i 4b prikazuju koordinatne varijable po epohama na stajali{tu P2 za prvi i drugi dan mjerenja. Vodoravna se i okomita sastavnica mijenjaju u opsegu od nekoliko milimetara do 10 cm tijekom prvoga dana mjerenja. Standardna se devijacija smanjila za 50 % do 70 % drugoga dana mjerenja na {umskoj povr{ini nakon sje~e.

198

Croat. j. for. eng. 29(2008)2

Accuracy Analysis of GPS Positioning Near the Forest Environment (189–199)

A. PIRTI

Razlika u preciznosti izme|u povoljnoga i nepovoljnoga razmje{taja satelita vidljiva je na slikama 5a i 5b. Vrijednosti PDOP-a mijenjaju se s obzirom na broj i vidljivost satelita. Radi usporedbe rezultata GPS izmjere obavljena je tako|er klasi~na izmjera totalnom stanicom Nikkon DTM 330. Razlike u izmjerenim rezultatima prikazane su u tablici 3. Vidljivo je zna~ajnije odstupanje izmjerenih vrijednosti na pravcu P1–P3. Uzrok tomu je poloàj stajali{ta P1 uz rub {ume pri ~emu sklop sastojine utje~e na ve}e rasipanje podataka. Srednje vrijednosti razlike udaljenosti izme|u stajali{ta P1 i P3 iznose 3 cm tijekom prvoga dana mjerenja u blizini {ume, odnosno 0,8 cm nakon sje~e {ume (slika 6a i 6b). Srednje vrijednosti visinske razlike izme|u stajali{ta P1 i P3 iznose 3,6 cm tijekom prvoga dana mjerenja u blizini {ume, odnosno 0,8 cm nakon sje~e {ume. Zaklju~uje se da je to~nost izmjere zna~ajno manja zbog utjecaja sklopa sastojine. Na slikama 7a i 7b prikazane su srednje vrijednosti razlike udaljenosti i visinske razlike izme|u stajali{ta P2 i P3. Istraìvanje je pokazalo da opseg smetnji prouzro~en {umom ima zna~ajan utjecaj na to~nost, preciznost i rezultat odre|ivanja GPS pozicije. Mjesta opaànja trebaju biti pa`ljivo postavljena, odnosno udaljenost od granice {ume trebala bi biti pove}ana kako bi mjesta opaànja bila manje izloèna pojavi vi{epu}a. Usporedba rje{enja pojedinih mjerenja pokazuje da je to~nost smanjena i za horizontalnu i za vertikalnu sastavnicu. Preciznost visinske sastavnice je 2 – 4 puta manja od o~ekivane. Standardna devijacija stupnjevito se pobolj{ava {to je ve}a udaljenost snimanja od granice {ume. Utjecaj okoli{a, odnosno visok stupanj vi{epu}a uzrokuje ve}u standardnu devijaciju i rasipanje srednjih vrijednosti rezultata za 50 – 70 %. Usporedba podataka dobivenih pomo}u GPS prijamnika i klasi~na izmjera totalnom stanicom pokazuju kako je o~it utjecaj sklopa sastojine na to~nost izmjerenih podataka. Klju~ne rije~i: GPS, {uma, to~nost, izmjera terena

Author’s address – Autorova adresa:

Received (Primljeno): March 17, 2008 Accepted (Prihva}eno): September 11, 2008 Croat. j. for. eng. 29(2008)2

Assist. Prof. Atinç Pirti, PhD e-mail: atinc@yildiz.edu.tr Yildiz Technical University Faculty of Civil Engineering Department of Geodesy and Photogrammetry Engineering 34349 Beºiktaº–Istanbul TURKEY

199

108

Croat. j. for. eng. 29(2008)1

Preliminary note – Prethodno priop}enje

Artificial Neural Networks in the Assessment of Stand Parameters from an IKONOS Satellite Image Damir Klobu~ar, Renata Pernar, Sven Lon~ari}, Marko Suba{i} Abstract – Nacrtak The paper explores the possibilities of assessing five stand parameters (tree number, volume, stocking, basal area and stand age) with the application of a multi-layer perceptron artificial neural network. An IKONOS satellite image (PAN 1 m x 1 m) was used to asses parts of stands in the sixth (121–140 yrs) and seventh (141–160 yrs) age class of pedunculate oak management class in the »Slavir« Management Unit of Otok Forest Office. Six features extracted from the first order histogram and five texture features extracted from the second order histogram were used as input data for neural network training. Data from the Management Plan were used as outputs of the neural network. An early stopping method and scaled conjugate gradient algorithm with error back propagation were used to improve generalization property of the neural network. Two neural network models were applied to assess the required stand parameters. The first model has one neuron in the output layer, where separate neuron network training was conducted for each stand parameter. The second model has five neurons in the output layer related to five assessed stand parameters. Both networks were trained and tested simultaneously. The conducted research showed that both of these neuron network models have good generalization properties. However, further analysis gave precedence to the second neural network model. Assessment of five quantitative stand parameters did not show any statistically significant differences between the Management Plan data and the neuron network model in terms of tree number, volume, stocking, basal area and stand age analysis. Keywords: artificial neuron networks, IKONOS – 2, stand parameter assessment, texture

1. Introduction – Uvod At present, remote sensing information is generally gathered using digital procedures (De Jong et al. 2006). Image analysis and scene interpretation are complex problems that require knowledge of the objects contained in a scene and spatial distribution of objects. Image analysis and scene interpretation are one of the most difficult issues in the sphere of intelligent systems (Gonzalez and Woods 2002). In addition to statistical methods and operational research methods, based on the theory of learning, artificial intelligence has advanced the possibility of using previous knowledge (e.g. expert systems or neural networks) to foster more effective decisionmaking processes (Haykin 1999). Croat. j. for. eng. 29(2008)2

For a number of years, empirical statistical methods or complex mathematical models have been applied in forest research and management to complement valid decision making processes. These models are expressed as mathematical equations. However, some decision making procedures contain qualitative components, which do not allow integration into mathematical equations. The technology of artificial intelligence makes it possible to process knowledge that will be used in decision making as an additional tool. The application of artificial neural networks in predictions of non-linear systems behaviour has become an alternative to traditional statistical methods (Peng and Wen 1999). In the years to come an increasing number of research teams will be dealing with artificial neural

201

D. KLOBU^AR et al.

Artificial Neural Networks in the Assessment of Stand Parameters ... (201–211)

networks and artificial intelligence in general. An interdisciplinary approach to this issue has become the imperative of our time. The degree of interdisciplinarity is expected to rise. At present, artificial neural networks have such broad applications that we can safely say that this is the period of transition to artificial neural network technology. Extensive research has been conducted in the applicability of satellite images to the study of the Earth’s surface. Satellite images used in forest research have proven their applicability in a number of issues: determining methods of land use, identifying tree species and monitoring the condition of forest stands, making forest inventories, assessing biomass, monitoring and identifying changes in a forest, detecting fires, assessing the conditions immediately after natural disasters (floods, volcano eruptions, earthquakes, etc.), hunting, etc. The advent of the new era in remote sensing (late 1990s) and the launching of the new generation of high resolution satellites (IKONOS) have enabled scientists to investigate their applications in natural resource monitoring. Scientific research predominantly focused on radiometric and geometric accuracy of IKONOS satellite images (Helder et al. 2003, Pagnutti et al. 2003, Zanoni et al. 2003), and on automatic detection of forms (features), recognition and regeneration. Their applicability to interpretation, mapping and photogrammetry was also investigated (Kristof et al. 2002, Dial et al. 2003). Some authors also used IKONOS satellite images in forestry to evaluate structural variables: age, height, number of trees, volume and basal area (Astola et al. 2000, Kayitakire et al. 2006). In their research, Shresta and Zinck (2001), Hagner (2002), classify and compare stand volumes from satellite images with different spatial resolutions (IKONOS, IRS, LANDSAT ETM+, SPOT). Katoh (2004) studies and describes tree species classification in mixed stands and their spectral characteristics. Kawamura et al. (2004) describe a method of parameter recognition necessary for the discrimination and identification of forest types on IKONOS satellite images, as well as spectral and textural features of these species. Chubey et al. (2006) describe object-based classification for the assessment and acquisition of soil cover, stand height and stand age parameters.

1.1 Artificial neural networks – Umjetne neuronske mre`e The application of artificial intelligence in forestry and natural resource management began with the development of an expert system for problem solving and decision making (Coulson et al. 1987).

202

Initial experiments in the application of neural networks to forestry began in the USA and Canada in the late 1980s. In order to understand an artificial neural network model, it is necessary to have some basic knowledge of biological neuron structure. There are a number of criteria for discriminating the architecture of neural networks. The basic discrimination factors are: number of layers, type of learning, direction in which a signal travels through the network, type of connection between neurons, input and transfer functions. According to Peng and Wen (1999) and Liu et al. (2003), the advantages of artificial neural networks stem from: Þ the possibility of learning complex patterns and monitoring data trends, Þ significant tolerance of imperfect data (absence of values), Þ robustness towards highly interconnected data. Artificial neural networks have been developed as an alternative approach to modeling non-linear and complex phenomena in forestry science (Gimblett and Ball 1995, Lek et al. 1996, Peng and Wen 1999, Liu et al. 2003). They are generally used for segmentation and classification purposes and are recommended for solving problems with highly diverse data. Such a use is forest inventory. Sui (1994) groups the application of artificial neural networks in spatial data handling into two main categories: the application of neural networks in remote sensing and integration of neural network with GIS for purposes of spatial modeling. In general, the application of artificial neural networks in remote sensing began in the early 1990s (Benediktsson et al. 1990, Civco 1993, Paola and Schowengerdt 1995). The most commonly used neural network model in remote sensing is the multi-layer perceptron (Atkinson and Tatnall 1997, Kanellopoulos and Wilkinson 1997, Foody 2001, Ashis 2002, Cetin et al. 2004, Shah and Gandhi 2004, Berberoglu and Curran 2006), whereas neural networks with radial basis functions (RBF) and probabilistic neural networks (PNN) (Foody 2001) are used less frequently. Neural networks with unsupervised learning, such as a self-organizing neural network (Beamish 2001) are also used. In their research authors use different types of images, but the focus is on satellite images with varied spectral, sensor and temporal characteristics. Skidmore et al. (1997) apply the error back propagation algorithm to forest mapping using GIS data and data obtained with Landsat TM images. Croat. j. for. eng. 29(2008)2

Artificial Neural Networks in the Assessment of Stand Parameters ... (201–211)

Wang and Dong (1997) apply the multi layer perceptron to determine stand parameters using radar scenes. Ardö et al. (1997) use the error back propagation algorithm to classify conifer damage with multitemporal satellite images (Landat TM) and topographic data. They compare this algorithm with multinominal logistic regression and do not favor any of these approaches. Moisen and Frescino (2002) compare five techniques of forest features predictions and confirm that the neural network technique is equally valuable as statistical methods. Ingram et al. (2005) map the structure of tropical forests using the multi layer perceptron with error back propagation on the basis of Landsat ETM+ scenes. Kuplich (2006) uses artificial neural networks in the analysis of satellite scenes (SAR, Landsat TM) to discriminate between forests and pastures, or to classify the age structure of forests. Joshi et al. (2006) use the feed forward multi layer network with feedback error propagation and Landsat ETM+ scene to determine (the density of) forest canopy. They compare the neural network with three methods: multiple linear regression, forest canopy density mapper and the highest probability classification. Their research confirms higher accuracy of the neural network model in relation to the three methods.

D. KLOBU^AR et al.

Verbeke et al. (2006) also use the feed forward multi layer network with feedback error propagation and CIR aerial photographs to assess the number of trees. Research in the field of remote sensing use in forestry has shown the merits of artificial neural networks as an alternative approach to classical statistical methods.

2. Research aim– Cilj istra`ivanja The basic goal is to investigate the simplest and the most acceptable procedure for operational application of artificial neural networks in the assessment of five stand parameters: volume, tree number, basal area, stocking and stand age from an IKONOS satellite image (PAN 1 m x 1 m). In order to achieve the set goal, the research was carried out as follows: Þ assessment of stand parameters in the images using the neural network method, Þ experimental validation of the obtained results, Þ analysis and comparison of the obtained results, Þ analysis of the strengths and weaknesses of artificial neural networks in remote sensing as support in forest management.

Fig. 1 IKONOS satellite image of a part of the study area (Spa~va basin – Croatia) Slika 1. IKONOS-ov satelitski snimak dijela istra`ivanoga podru~ja (Spa~va – Hrvatska) Croat. j. for. eng. 29(2008)2

203

D. KLOBU^AR et al.

Artificial Neural Networks in the Assessment of Stand Parameters ... (201–211)

3. Materials and Methods – Materijal i metode 3.1 Research Area – Podru~je istra`ivanja This research was conducted in a part of the Spa~va forest basin. The forests of the Spa~va Basin cover about 40,000 ha. They are predominantly developed in the floodplain area of the Sava River and its tributaries. The part of the Spa~va basin, where the research was carried out, relates to the age class VI (121–140 yrs) and VII (141–160 yrs) of pedunculate oak management class in the »Slavir« Management Unit (MU). An IKONOS satellite image of the Spa~va basin covering an area of 132 km2 was obtained in 5 spectral channels: PAN (1 x 1 m) and 4 MS Bundle. The satellite image was processed (Fig. 1) with ERDAS IMAGINE 9.2 software.

3.2 Extracting textural features of the stand scene – Ekstrakcija teksturnih zna~ajki sastojinske scene In order to determine the textural features, a sample of the satellite image was cut out for each stand scene. A total of 120 compartments/subcompartments (stand scenes) in the sixth and seventh age class of pedunculate oak management class were cut out. The reason for choosing these two age classes lies in the fact that this management unit has an irregular age structure and that in terms of surface area these are the two best represented age classes (64% of the management class area, or 76% without the first age class). A typical procedure in texture analysis relates to statistical intensity features of the first order histogram. The MATLAB statxture function (Gonzales and Woods 2004) was used, and it yielded six statistical values: arithmetic mean, standard deviation, smoothness, third moment, uniformity and entropy. Texture measures calculated only from the first order histogram data have a drawback, because they do not provide information on the relative relationship between the pixels themselves (Gonzales and Woods 2002). According to Coburn and Roberts (2004), remote sensing researchers commonly use data obtained with second order histograms to make texture analyses and classifications, while first order histograms are used less frequently. Kayitakire et al. (2006) state that features obtained with second order histograms were often used in texture classification or segmentation (Hay et. al. 1996, Franklin et al. 2000, 2001, Coburn and Roberts 2004), and however they were very rarely

204

used in stand parameter assessments (Kayitakire et al. 2006). Berberoglu and Curran (2006), Kayitakire et al. (2006) report that only six out of 14 defined textural features of the second order (Haralick et al. 1973), (energy, contrast, variance, homogeneity, correlation and entrophy) are used in remote sensing more frequently. To determine the features of the second order histogram, the MATLAB imtexfeat function was used on blocks sized [M N] with a certain vector shift [Dx Dy]. In this case the block was represented by the cut stand scene, while the shift was [1 1]. Five textural features were calculated: the absolute difference value, inertia, covariance, entropy and energy. The following three facts were taken into account to select the sample size: Þ image features are extracted for stand scenes (compartments/subcompartments) that were already stratified according to forest management criteria, Þ selection of matrix size (window) is not important only for calculation reasons, but also for purposes of defining a representative sample (Hodgson 1994, Franklin et al. 2000), Þ MATLAB is the interpreter language, and the implementation of the function imtexfeat, which was used to calculate features of the second order histogram, may take some time. Accordingly, stand scene features were extracted using spatial features of the first and second order histograms. A total of eleven texture features were extracted for each stand scene (compartment/subcompartment) using the described procedure. A data set (vectors) was formed as an input to the neural network model. Data from Management Plans were used as output values.

3.3 Construction of an Optimal Structure of Multi Layer Perceptron – Izrada optimalne strukture vi{eslojnoga perceptrona After extracting textural features of stand scenes for 120 compartments/subcompartments, the optimal neural network architecture was produced in MATLAB 6.5 software program. According to Davies (2005), there are generally two approaches to optimizing the network architecture. The first approach involves gradual upgrade of the network by adding one by one neuron. The second approach involves the construction of a complex network structure, which is gradually reduced until the optimal network structure is obtained. The same author also states that research so far favors the first approach, and that the universal optimization Croat. j. for. eng. 29(2008)2

Artificial Neural Networks in the Assessment of Stand Parameters ... (201–211)

Table 1 Results of repeated-measures analysis of variance for the MU »Slavir« Tablica 1. Rezultati analize varijance ponovljenih mjerenja za GJ »Slavir« Volume, m3/ha – Obujam, m3/ha

Intercept Error

44501936

33458.07

0.000000

78475

1330 0.43

0.649579

Model

694

347

Error

94512

118

801

Trees number per ha – Broj stabala po ha Intercept

14216695

Error

199449

3380

Model

629

315

Error

119777

118

1015

Intercept

2745109

Error

4360

Model

121

Error

3329

118

4205.503

0.000000

0.310

0.734021

Stand age, years – Dob sastojine, godine 37145.96

0.000000

2.14

0.122463

94398.69

0.000000

3.16

0.045852

Stocking – Obrast Intercept

162.8612

Error

0.1018

0.0017

Model

0.0080

0.0040

Error

0.1485

118

0.0013

Basal area, m2/ha – Temeljnica, m2/ha Intercept

163557.1

Error

243.9

4.1

Model

9.5

4.7

Error

348.9

118

3.0

39572.03

0.000000

1.60

0.205497

process has not yet been formulated. Generalization is the property of the network to work »well« with vectors, which were not contained in the set of examples used for network training. An early stopping method and scaled conjugate gradient algorithm with error back propagation were used to improve generalization property of the neural network. The early stopping method actually involves a statistical cross-validation method in which the total data set is divided into three sets: for training, validation and testing. Out of the total data set, 50% or 60 compartments/subcompartments were allocated to the training set, while the two remaining sets were divided in equal amounts: 25% (30 compartments/subcompartments) accounted for the vaCroat. j. for. eng. 29(2008)2

D. KLOBU^AR et al.

lidation set and 25% (30 compartments/subcompartments) accounted for the testing set. Prior to neural network training, the data were preprocessed. In this sense two operations were performed using MATLAB functions: input-out value normalization and analysis of the main input value components. Normalization yielded the mean value equal to zero, or standard deviation equal to one of input and output data. The main components analysis reduced the input vectors dimension. In this case, all those components participating with less than 1% in the total input data variance were eliminated, thus reducing the number of extracted texture features from 11 to 5. After the data were trained, generalized and normalized, they were converted into standard units. To construct a neural network model, we used an algorithm with one hidden layer and the logarithmic sigmoidal function at its output and the hyperbolic-tangential-sigmoidal function in the output layer. A total of 14 architectures were trained (the number of neurons in the hidden layer: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 15, 20, 25, 30), starting with one neuron in the hidden layer, and adding one more neuron with each new iteration until 10 neurons were obtained in the hidden layer, after which the number of neurons increased by 5 until there were 30 neurons in the hidden layer. Two models were used to determine stand parameters. In model 1, there was one neuron in the output layer. For each stand parameter, separate neural network training was performed. A neural network with five neurons in the output layer was also applied (Model 2), while the number of input and hidden neurons, as well as the applied activation functions were identical to Model 1. The five neurons in the output layer relate to the five listed stand parameters, which were trained, i.e. tested simultaneously in this case. The minimal value of the mean square error in the testing set was applied for both models so as to select the optimal architecture. As mentioned before, this set consists of 30 compartments/subcompartments. In the operative stage of neural network application, 30 additional compartments (15 compartments from each age class) were subsequently tested. Consequently, generalization for both models was conducted in 60 compartments/subcompartments.

4. Results – Rezultati The minimal value of the mean square error of the testing set was used to select the optimal archi-

205

D. KLOBU^AR et al.

Artificial Neural Networks in the Assessment of Stand Parameters ... (201–211)

Fig. 2 Means and 95% confidence intervals of stand age for MU »Slavir«, Model 1, Model 2 Slika 2. Aritmeti~ke sredine i 95 %-ni intervali pouzdanosti dobi sastojina za GJ »Slavir«, model 1, model 2

Fig. 3 Means and 95% confidence intervals of tree number per ha for the MU »Slavir«, Model 1, Model 2 Slika 3. Aritmeti~ke sredine i 95 %-ni intervali pouzdanosti broja stabala po hektaru za GJ »Slavir«, model 1, model 2

tecture of the multi layer perceptron. In Model 1, to determine basal area/ha, stocking and number of trees/ha the optimal architecture had two neurons in the hidden layer (5 – 2 – 1), to determine volume/ha the optimal architecture had three neurons in the hidden layer (5 – 3 – 1), and to determine stand age 20 neurons in the hidden layer were used (5 – 20 – 1). The following values of the mean square testing set error were determined: basal area/ha (0.0473), stocking (0.0633), number of trees/ha (0.1092), volume/ha (0.0468) and age (0.1631). In Model 2, the optimal architecture contained five neurons in the hidden layer (5 – 5 – 5), and the obtained mean square error of the testing set was 0.1804. To test the differences in the values of five quantitative stand parameters of the MU »Slavir« Management Plan and the artificial neural networks model, repeated-measures analysis of variance was used (Table 1) as well as Tukey’s HSD test for multiple comparisons in STATISTICA 7.1 software. The results obtained with Model 1 and Model 2 were compared with those of the Management Plan data. According to Table 1, the statistically significant difference between the Management Plan data and the neural network model related only to the assessment of the stocking. The application of Tukey’s HSD test revealed the difference related to Model 1 and Model 2.

Fig. 2 shows the relationship of stand ages from the Management Plan and their assessment with neural network models. It is clear that Model 2 shows slightly higher correspondence with the Man-

206

Fig. 4 Means and 95% confidence intervals of basal area per ha for the MU »Slavir«, Model 1, Model 2 Slika 4. Aritmeti~ke sredine i 95 %-ni intervali pouzdanosti temeljnice po hektaru za GJ »Slavir«, model 1, model 2 Croat. j. for. eng. 29(2008)2

Artificial Neural Networks in the Assessment of Stand Parameters ... (201–211)

D. KLOBU^AR et al.

Fig. 5 Means and 95% confidence intervals of volume per ha for the MU »Slavir«, Model 1, Model 2 Slika 5. Aritmeti~ke sredine i 95 %-ni intervali pouzdanosti volumena po hektaru za GJ »Slavir«, model 1, model 2

Fig. 6 Means and 95% confidence intervals of stocking for the MU »Slavir«, Model 1, Model 2 Slika 6. Aritmeti~ke sredine i 95 %-ni intervali pouzdanosti obrasta za GJ »Slavir«, model 1, model 2

agement Plan value range, but also that assessment did not include the bottom part of confidence interval in both models; however, the difference is not significant. In tree number assessment (Fig. 3), the arithmetic means of neural network models showed high correspondence with the mean value of tree number (279 trees/ha) in the Management Plan. It is also clear that both models for the most part encompass the value range contained in the Management Plan. In the assessment of basal area (Fig. 4), values from the upper part of confidence interval from the Management Plan were assessed in both models. Model 2 showed a slightly higher value range. In terms of stand volume assessment, both neural network models showed good generalization properties, which is reflected in the values of arithmetic means, as well as in correspondence with the value range of interval confidence in the Management Plan (Fig. 5). In the assessment of stocking, Model 2 has better generalization properties in relation to Model 1, where the assessed values show a small value range (Fig. 6).

erties. In the assessment of five quantitative parameters there was no statistically significant difference between the Management Plan data and the results obtained by neural network in the analyses of tree number, volume, stocking, basal area and stand age. Data obtained for basal area correspond to the research by Kayitakire et al. (2006) according to which it is not recommended to determine basal area per ha with remote sensing methods for intensively managed stands without additional (ancillary) information. The present research did not show high accuracy of basal area assessment despite the application of additional information (Model 2). Further, it is clear that Model 2 responds better to the value range of stand parameters contained in the Management Plan. The conclusion is that in order to assess stand parameters with remote sensing methods it is better to use the architecture of a neural network which has a higher number of neurons in the output layer; in other words, the network is trained for simultaneous assessment of a higher number of stand parameters, as is the case with Model 2. Compared to terrestrial measurements, the applied procedure is much more acceptable from material and temporal aspects. To assess five quantitative stand parameters for 60 compartments/subcompartments in the generalization procedure, the total time needed to extract textural features of stand scenes and simulate (assess) them with a trained

5. Discussion – Rasprava Table 1 and graphs (Fig. 2–6) clearly show that Model 1 and Model 2 have good generalization propCroat. j. for. eng. 29(2008)2

207

D. KLOBU^AR et al.

Artificial Neural Networks in the Assessment of Stand Parameters ... (201–211)

(optimal architecture) neural network was one working day. The area in question was 1326.96 ha. Image feature extraction, i.e. preparation of a data set for the neural network, is faced with the problem of heterogeneity and complexity of data from natural surroundings. The problem, described by Cherakassky et al. (2006), relates to temporal, dynamic, spatial, biometric and other components of data collection, whether quantitative and qualitative variables are determined terrestrially or with remote sensing. Another problem occurring in remote sensing research relates to the difficulty of having at our disposal aerial and satellite images, as well as terrestrial data that are gathered in the same time period (Foody and Curran 1994, Ingram et al. 2005). From the aspect of forestry profession, the results obtained from remotely-sensed five quantitative stand parameters using artificial neural network models from a high-resolution IKONOS panchromatic satellite image (PAN 1 m x 1 m) for stands in the management class of pedunculate oak in the sixth and seventh age class can be considered acceptable. Forestry is a field of economy in which multitudinous and varied measurements are performed almost every day. The artificial neural network model based on the theory of learning could considerably improve the handling of such a large number of data. Until now, the problem has been dealt exclusively with statistical methods and methods of operational research. Artificial neural networks are more accurate than statistical methods, especially when the problem is poorly defined or incomprehensible, or when the solution is not a priori know by the user.

6. Conclusions – Zaklju~ci The conclusions can be summarized into the following statements: Þ Artificial neural networks have proved to be a robust remote sensing tool in forest management, Þ The multi layer perceptron has good generalization properties in the assessment of quantitative stand parameters (volume per ha, number of trees per ha, age, stocking, basal area per ha) from an IKONOS (PAN 1 m x 1 m) satellite image, Þ Based on the conducted research and experience with the use of the multi layer perceptron with error back-propagation, one hidden layer has proved sufficient for solving problems in the field of using artificial neural networks in remote sensing for the needs of forest management,

208

Þ This research has also confirmed the strengths (no need to know the data model, possibility of application in the analysis of new conditions, tolerance of data imperfections) and weaknesses (determination of the optimal architecture, impossibility of assessment outside the value range of learning data) of artificial neural networks.

7. References – Literatura Atkinson, P. M., Tatnall, A. R. L., 1997: Neural networks in remote sensing. International Journal of Remote Sensing 18(4): 699–709. Ärdo, J., Pilesjo, P., Skidmore, A., 1997: Neural networks, multitemporal Landsat Thematic Mapper data and topographic data to classify forest damage in the Czech Republic. Canadian Journal of Remote Sensing 23(3): 217–219. Ashish, D., 2002: Land – use classification of aerial images using artificial neural networks. Master’s Thesis, The University of the Georgia, 65 p. Astola, H., Bounsaythip, C., Ahola, J., Häme, T., Laura Sirro, E., Veikkanen, B., 2000: Highforest – forest parameter estimation from high resolution remote sensing data. ISPRS 2003 Annual Conference, Istanbul. Beamish, D., 2001: A Review of Neural Networks in Remote Sensing, 1–45. Benediktsson, J. A., Swain, P. H., Evsoy, O. K., 1990: Neural network approach versus statistical methods in classification of multi-source remote sensing data. IEEE Transactions on Geoscience and Remote Sensing 28(4): 540–551. Berberoglu, S., Curran, P. J., 2004: Merging spectral and textural information for classifying remotely sensed images. In: De Jong, S. M. and Van der Meer, F. D. (eds.), Remote Sensing Image Analysis: Including the Spatial Domain. Dordrecht, The Netherlands, Kluwer Academic, 113–136. Cetin, M., Kavzoglu, T., Musaoglu, N., 2004: Classification of multi – spectral, multi – temporal and multi – sensor images using principal components analysis and artificial neural networks: Beykoz case. ISPRS’ 2004, Istanbul. Cherkassky, V., Krasnopolsky, V., Solomatine, D. P., 2006: Computational intelligence in earth sciences and environmental applications: Issues and challenges. Neural Networks 19(2): 113–121. Chubey, M. S., Franklin, S. E., Wulder, M. A., 2006: Object-based Analysis of Ikonos-2 Imagery for Extraction of Forest Inventory Parameters. Photogrammetric Engineering & Remote Sensing 72(4): 383–394. Civco, D. L., 1993: Artificial neural networks for land cover classification and mapping. International Journal of Geographical Information Systems 7(2): 173–186. Coburn, C. A., Roberts, A. C. B., 2004: A multiscale texture analysis procedure for improved forest stand classificaCroat. j. for. eng. 29(2008)2

Artificial Neural Networks in the Assessment of Stand Parameters ... (201–211)

tion. International Journal of Remote Sensing 25(2): 4287– 4308. Coulson, R. N., Folse, L. J., Loh, D. K., 1987: Artificial intelligence and natural resource management. Science 237: 262–267. Davies, E. R., 2005: Biologically Inspired Recognition Schemes, 724 – 755., Texture, 756 – 779. Machine Vision, Third edition. De Jong, S. M., van der Meer, F. D., Clevers, J., G. P. W., 2006: Basics of Remote Sensing. Remote sensing image analysis, 1 – 15, Springer, Netherlands. Dial, G., Bowen, H., Gerlach, F., Grodecki, J., Oleszczuk, R., 2003: IKONOS satellite imagery, and products. Remote Sensing of Environment 88(1): 23–36. Foody, G. M., Curran, P. J., 1994: Estimation of tropical forest extent and regenerative stage using remotely sensed data. Journal of Biogeography 21(3): 223–244. Foody, G. M., 2001: Thematic mapping from remotly sensed data with neural networks: MLP, RBF and PNN based appraches. Journal of Geographical Systems 3(3): 217–232. Franklin, S. E., Hall, R. J., Moskal, L. M., Maudie, A. J., Lavigne, M. B., 2000: Incorporating texture into classification of forest species composition from airborne multispectral images. International Journal of Remote Sensing 21(1): 61–79. Franklin, S. E., Maudie, A. J., Lavigne, M. B., 2001: Using spatial cooccurrence texture to increase forest structure and species composition classification accuracy. Photogrammetric Engineering and Remote Sensing 67(7): 849–855. Gimblett, R. H., Ball, G. L., 1995: Neural network architectures for monitoring and simulating changes in forest resources management. AI Applications 9(2): 103–123. Gonzales, R. C., Woods, R. E., 2002: Digital Image Proceessing. Second edition, Prentice – Hall, Inc, New Jersey, 793. Gonzales, R. C., Woods, R. E., Eddins, S. L., 2004: Digital Image Proceessing using MATLAB. Prentice – Hall, Inc, New Jersey. Hagner, O., 2002: Combined estimation of forest parameters from high and medium resolution satellite imagery and forest inventory data. In: Proceedings of the ForestSAT conference in Edinburg Scotland, 5–9. Haralick, R. M., Shanmugam, K., Dinstein, I., 1973: Textural features for image classification. IEEE Transactions on Systems, Man and Cybernetics 3(6): 610–621. Hay, G. J., Niemann, K. O., McLean, G. F., 1996: An object – specific image – texture analaysis of H – resolution forest imagery. Remote Sensing of Environment 55(2): 108–122. Haykin, S., 1999: Neural Networks: A Comprehensive Foundation. 2nd Ed. Prentice-Hall, Inc., New Jersey Helder, D., Coan, M., Patrick, K., Gaska, P., 2003: IKONOS geometric characterization. Remote Sensing of Environment 88(1–2): 68–78. Croat. j. for. eng. 29(2008)2

D. KLOBU^AR et al.

Hodgson, M. E., 1994: Window size and visual image classification accuracy: An experimental approach. ASPRS Technical Papers, 1994 ASPRS – ACSM Ann. Conv., V. 2, 209– 218. Ingram, J. C., Dawson, T. P., Whittaker, R. J., 2005: Mapping tropical forest structure in southeastern Madagascar using remote sensing and artificial neural networks. Remote Sensing of Environment 94(4): 491–507. Joshi, C., De Leeuw, Jan, Skidmore, A. K., van Duren, I. C., van Oosten, H., 2006: Remotely sensed estimation of forest canopy density: A comparison of the performance of four methods. International Journal of Applied Earth Observation and Geoinformation 8(2): 84–95. Kanellopoulos, I., Wilkinson, G. G., 1997: Strategies and best practise for neural network image classification. International Journal of Remote Sensing 18(4): 711–725. Katoh, M., 2004: Classifying tree species in a northern mixed forest using high resolution IKONOS data. Journal of For Res. 9(1): 7–14. Kayitakire, F., Hamel, C., Defourny, P., 2006: Retrieving forest structure variables based on image texture analysis and IKONOS – 2 imagery. Remote Sensing of Environment 102(3–4): 390–401. Kawamura, M., Tsujiko, Y., Tsujino, K., Sakai, T., 2004: Time – series fire – induced forest hazard mapping usin Landsat and IKONOS imageries. Geoscience and Remote Sensing Symposium, 2004. IGARSS apos., 04. Proceedings. 2004 IEEE International Volume 4, September 20–24, 2004, 256– 259. Kristof, D., Csato, E., Ritter, D., 2002: Application of highresolution satellite images in forestry and habitat mapping-evaluation of ikonos images through a Hungarian case study. Symposium on Geospatial Theory, Processing and Applications, ISPRS, Ottawa. Kuplich, T. M., 2006: Classifying regenerating forest stages in Amazonia using remotely sensed images and a neural network. Forest Ecology and Management 234(1–3): 1–9. Lek, S., Delacoste, M., Baran, P., Dimopoulos, I., Lauques, J., Aulagnier, S., 1996: Application of neural networks to modelling nonlinear relation-ships in ecology. Ecol. Modell. 90(1): 39–52. Liu, C., Zhang, L., Davis, C. J., Solomon, D. S., Brann, T. B., Caldwell, D. S., 2003: Comparision Of Neural Networks and Statistical Methods in Classification of Ecological Habitats Using FIA Data. Forest Science 49 (4): 619–631. Moisen, G. G., Frescino, T. S., 2002: Comparing five modelling techniques for predicting forest characteristics. Ecological Modelling 157(2): 209–225. Pagnutti M., Ryan R. E., Kelly M., Holekamp K., Zanoni V., Thome K., Schiller S., 2003: Radiometric characterization of IKONOS multispectral imagery. Remote Sensing of Environment 88(1): 53–68. Paola, J. D., Schowengerdet, R. A., 1995: A review and analysis of backpropagation neural networks for classifi-

209

D. KLOBU^AR et al.

Artificial Neural Networks in the Assessment of Stand Parameters ... (201–211)

cation of remotley – sensed multi spectral imagery. International Journal of Remote Sensing 16(16): 3033–3058.

using GIS and remotely sensed data. Photogrammetric Engineering and Remote Sensing 63(5): 501–514.

Peng, C., Wen, X., 1999: Recent Application of Artificial Neural Networks in Forest Resource Management: An Overview. In: Environmental Decision Support Systems and Artificial Intelligence, Ulises Corté and Miquel Srnchez – Marrc, Cochairs (eds.). pp. 15–22. Tech. Rep. WS-99-07, AAAI Press, Menlo Park, California.

Sui, D. Z., 1994: Recent applications of neural networks for spatial data handling. Canadian journal of remote sensing 20(4): 368–380.

Shah, S. K., Gandhi, V., 2004: Image Classification Based on Textural Features using Artificial Neural Network (ANN). Electronics and Telecom Engineering 84:72–77. Shrestha, D. P., Zinck, J. A., 2001: Land use classification in mountainous areas: Integration of image processing, digital elevation data and field knowledge (application to Nepal). International Journal of Applied Earth Observation and Geoinformation 3(1): 78–85. Skidmore, A. K., Turner, B. J., Brinkhof, W., Knowles, W., 1997: Performance of a neural network: mapping forests

Verbeke, L. P. C., Van Coillie, F. M. B., DeWulf, R. R., 2006: Object-based forest stand density estimation from very high resolution optical imagery using wavelet-based texture measures. In: 1st International Conference on Object-based Image Analysis (OBIA 2006). Zanoni, V., Stanley, T., Ryan, R., Pagnuti, M., Baldridge, B., Roylance, S., Snyder, G., Lee, G., 2003: The Joint Agency Commercial Imagery Evaluation (JACIE) team: Overview and IKONOS joint characterization approach. Remote Sensing of Environment 88(1–2): 17–22. Wang, Y., Dong, D., 1997: Retrieving forest stand parameters from SAR backscatter data using a neural network trained by a canopy backscatter model. International Journal of Remote Sensing 18(4): 981–989.

Sa`etak

Umjetne neuronske mreè u procjeni sastojinskih parametara s IKONOS-ova satelitskoga snimka Danas se pridobivanje informacija u daljinskim istraìvanjima uglavnom provodi digitalnim postupkom (De Jong i dr. 2006). Polazi{ne osnove analiza slike i interpretacija scena su kompleksni problemi koji zahtijevaju znanje o objektima sadrànim na sceni te o me|usobnom prostornom rasporedu objekata. Navedene analize slike interpretacije scena jedne su od najteìh u podru~ju inteligentnih sustava (Gonzalez i Woods 2002). Ve} mnogo godina u istraìvanju i rukovo|enju u {umarstvu koriste se empiri~ke statisti~ke metode ili sloèni matemati~ki modeli, koji upotpunjuju dono{enje pravovaljanih odluka. Ti su modeli izraèni kao matemati~ke jednad`be. Me|utim, neki postupci dono{enja odluka sadrè kvalitativne komponente koje ne dopu{taju integraciju u matemati~ke jednad`be. Tehnologija umjetne inteligencije omogu}uje procesiranje znanja koje }e biti uklju~eno kao dodatni alat u odlu~ivanju. Primjena umjetnih neuronskih mreà u predikciji pona{anja nelinearnih sustava postaje alternativa tradicionalnim statisti~kim metodama (Peng i Wen 1999). Naime, umjetna inteligencija temeljena na teoriji u~enja unaprijedila je mogu}nost kori{tenja prethodnoga znanja (npr. ekspertni sustavi ili neuronske mreè) i podataka radi dono{enja u~inkovitih odluka (Haykin 1999). Stoga se u radu istraùju mogu}nosti procjene pet sastojinskih parametara (broja stabala, obujma, obrasta, temeljnice i dobi sastojina) primjenom vi{eslojnoga perceptrona, kao najkorisnijega modela umjetnih neuronskih mreà u daljinskim istraìvanjima. Za tu je potrebu kori{ten IKONOS-ov satelitski snimak (PAN 1 m x 1 m) dijela sastojina VI. (121 – 140 god.) i VII. (141 – 160 god.) dobnoga razreda, ure|ajnoga razreda hrasta lu`njaka, gospodarske jedinice »Slavir«, [umarije Otok. Za ulaz u neuronsku mreù uzeto je {est vrijednosti teksturnih zna~ajki histograma prvoga reda i pet vrijednosti teksturnih zna~ajki histograma drugoga reda. Za izlazne vrijednosti kori{teni su podaci Osnove gospodarenja. Radi unapre|enja generalizacije primijenjena je metoda ranijega zaustavljanja (engl. early stopping), te scaled conjugate gradient algoritam s povratnom propagacijom pogre{ke. Prije treniranja neuronske mreè provedeno je preprocesuiranje podataka, dok je za izradu optimalnoga modela neuronske mreè kori{ten jedan skriveni sloj s razli~itim brojem neurona. Primijenjene aktivacijske funkcije su logaritamsko-sigmoidna u skrivenom, odnosno hiperboli~ko-tangentno-sigmoidna funkcija u izlaznom sloju.

210

Croat. j. for. eng. 29(2008)2

Artificial Neural Networks in the Assessment of Stand Parameters ... (201–211)

D. KLOBU^AR et al.

Da bi se procijenili sastojinski parametri primijenjena su dva modela: model 1 s jednim neuronom u izlaznom sloju, gdje je za svaki sastojinski parametar provedeno zasebno treniranje neuronske mreè, i model 2 s pet neurona u izlaznom sloju koji se odnose na pet procjenjivanih sastojinskih parametara, koji su u ovom slu~aju trenirani, odnosno testirani istodobno. Kod oba modela u odabiru optimalne arhitekture primijenjena je najmanja vrijednost srednje kvadratne pogre{ke na setu za testiranje. Za testiranje razlike u vrijednostima pet kvantitativnih parametara sastojine podataka Osnove gospodarenja GJ »Slavir« i modela umjetnih neuronskih mreà primijenjena je analiza varijance ponovljenih mjerenja te Tukey HSD test za me|usobne vi{estruke usporedbe u programu STATISTICA 7. 1. Svi postupci koji se odnose na tehnologiju umjetnih neuronskih mreà odra|eni su u programu MATLAB 6.5. Satelitski je snimak obra|en pomo}u programskoga paketa ERDAS IMAGINE 9.2. Provedenim istraìvanjem utvr|eno je da oba modela neuronskih mreà imaju dobra generalizacijska svojstva, s tim da je daljnjom analizom prednost dana modelu 2. Naime, u procjeni pet kvantitativnih parametara nema statisti~ki zna~ajne razlike izme|u podataka Osnove gospodarenja i modela neuronske mreè u analizi broja stabala, obujma, obrasta, temeljnice i dobi sastojina. [umarstvo je podru~je gospodarstva u kojem se gotovo svakodnevno provodi velik broj razli~itih mjerenja, a upravo umjetne neuronske mreè predstavljaju model temeljen na teoriji u~enja, kojim bi se zna~ajnije moglo unaprijediti kori{tenje tako velikoga broja podataka, gdje su se problemi do sada rje{avali isklju~ivo statisti~kim metodama i metodama operacijskih istraìvanja. Umjetne su neuronske mreè to~nije od statisti~kih metoda, pogotovo kada je problem slabo definiran ili nerazumljiv, odnosno kada korisnik a priori ne zna rje{enje. Klju~ne rije~i: umjetne neuronske mreè, IKONOS – 2, procjena sastojinskih parametara, tekstura

Authors’ address – Adresa autorâ: Damir Klobu~ar, PhD. e-mail: damir.klobucar@hrsume.hr »Hrvatske {ume« d.o.o. Zagreb Farka{a Vukotinovi}a 2 HR–10000 Zagreb CROATIA Assoc. Prof. Renata Pernar, PhD. e-mail: rpernar@sumfak.hr Forestry Faculty of Zagreb University Sveto{imunska 25 HR–10000 Zagreb CROATIA

Received (Primljeno): August 20, 2008 Accepted (Prihva}eno): November 29, 2008 Croat. j. for. eng. 29(2008)2

Prof. Sven Lon~ari}, PhD. e-mail: sven.loncaric@fer.hr Marko Suba{i}, PhD. e-mail: marko.subasic@fer.hr University of Zagreb Faculty of Electrical Engineering and Computing Unska 3 HR–10000 Zagreb CROATIA

211

Editorial – Uvodnik TIBOR PENTEK, TOMISLAV POR[INSKY 110 Years of Universty Forestry Education in the Republic of Croatia ......................... 109 110. obljetnica visoko{kolske {umarske nastave u Republici Hrvatskoj

Orginal scientific papers – Izvorni znanstveni radovi CHRISTIAN ROTTENSTEINER, GÜNTER AFFENZELLER, KARL STAMPFER Evaluation of the Feller-Buncher Moipu 400E for Energy Wood Harvesting .................... 117 Ocjena vi{ezahvatne sje~ne glave Moipu 400E pri pridobivanju drva za energiju RAFFAELE SPINELLI, CARLA NATI, NATASCIA MAGAGNOTTI Harvesting Short-Rotation Poplar Plantations for Biomass Production ........................ 129 Pridobivanje biomase sje~om {umskih planta`a topola u kratkim ophodnjama BO[TJAN KO[IR Damage to Young Forest Due to Harvesting in Shelterwood Systems ......................... 141 O{te}enja mladih sastojina nakon pomla|ivanja metodama pod zastorom kro{anja MILAN OR[ANI], DUBRAVKO HORVAT, NIKOLA PERNAR, MARIJAN [U[NJAR, DARKO BAK[I], DAMIR DRVODELI] Growth of Pedunculate Oak Seedlings under Soil Contamination by Mineral and Biodegradable Oils .. 155 Rast sadnica hrasta lu`njaka pri one~i{}enju tla mineralnim i biorazgradivim uljima YURI GERASIMOV, ANTON SOKOLOV, TIMO KARJALAINEN GIS-Based Decision-Support Program for Planning and Analyzing Short-Wood Transport in Russia ... 163 Ra~unalni program temeljen na GIS-u kao potpora odlu~ivanju pri planiranju i ra{~lambi transporta kratkoga drva u Rusiji JANEZ KR^, BO[TJAN KO[IR Predicting Wood Skidding Direction on Steep Terrain by DEM and Forest Road Network Extension ... 177 Odre|ivanje smjera privla~enja drva prema postoje}oj mre`i {umskih cesta na strmim terenima ATINÇ PIRTI Accuracy Analysis of GPS Positioning Near the Forest Environment ......................... 189 Analiza to~nosti pozicioniranja GPS-a uz {umski okoli{

Preliminary note – Prethodno priop}enje DAMIR KLOBUÂR, RENATA PERNAR, SVEN LONÂRI], MARKO SUBA[I] Artificial Neural Networks in the Assessment of Stand Parameters from an IKONOS Satellite Image .. 201 Umjetne neuronske mreè u procjeni sastojinskih parametara s IKONOS-ova satelitskoga snimka

ISSN 1845-5719

9 771845 571000