P. ZorogastĂşa, International Potato Center (CIP), Lima, Peru R. Quiroz, International Potato Center (CIP), Lima, Peru M. Potts, International Potato Center (CIP), SSA1 S. Namanda, International Potato Center (CIP), SSA V. Mares, International Potato Center (CIP), Lima, Peru L. Claessens, International Potato Center (CIP), SSA 1
2007- 5 Working Paper
Utilization of high-resolution satellite images to improve statistics for the sweetpotato cultivated area of Kumi district, Uganda
Now at Catholic Relief Services (CRS), Nairobi.
ISBN 978-92-9060-330-6 Natural Resources Management Division Working Paper No. 2007-5
Table of Contents Utilization of high-resolution satellite images to improve statistics for the sweetpotato cultivated area of Kumi district, Uganda INTRODUCTION Agricultural statistics and remote sensing Physical basis of the discrimination between crops MATERIALS AND METHODS Material and equipment Image Processing RESULTS AND DISCUSSION CONCLUSIONS REFERENCES
1 1 2 3 6 7 7 9 13 14
iii
Preface This working paper has been written by scientists from different research divisions of the International Potato Center (CIP). It describes a methodology based on the use of high-resolution satellite images and spectroradiometry to gather data to construct reliable statistics on sweetpotato cropping areas. The pilot study was conducted in a region located in eastern Uganda, where sweetpotato is cultivated in two cropping seasons. The results suggest that the tested method substantially improves estimation of cropping areas compared to traditional statistics based on field surveys and censuses.
Acknowledgements The authors appreciate the technical support of fellow members of the NRM Division.
v
Utilization of high-resolution satellite images to improve statistics for the sweetpotato cultivated area of Kumi district, Uganda INTRODUCTION Crop statistics information is important for the implementation of research and development projects. Agricultural statistics are utilized for decision making with regard to planning development activities to benefit the populations occupying a target region. FAO (1999, 2003) points out that in order to plan agricultural research and rural development strategies, it is absolutely necessary to have an information system that provides reliable data on the total area cultivated under a specific crop, the size of production units, the agroecological areas where the crops are found, soil and climate characteristics, the socioeconomic condition of the producers and the cropping schedule, among others. In recent years, CIP has introduced several orange-fleshed sweetpotato clones in Africa as a strategy to solve the severe vitamin A deficiency affecting the human population. This introduction requires precise information on the current area of sweetpotato cultivation to assess potential dissemination of new germplasm and its beneficiaries on both spatial and time scales. This knowledge is necessary for estimating production, marketable volumes, per capita consumption, required inputs and to orient research. There are reasons to believe that public statistics on cultivated extensions in Africa do not bear much relation to the actual cultivated area. This reasonable doubt is based on the fact that crop statistics related to small producers are obtained through field sampling and survey techniques that have some limitations. For this reason, CIP’s Natural Resources Management Division is developing and validating a methodology based on spectroradiometry and the use of high-resolution remote sensing images that provides reliable, accurate, and dynamic information for estimating cropping areas. This methodology has been tested in Uganda where CIP has recently introduced orange-fleshed sweetpotato clones and where sweetpotato is perceived as a replacement for cassava as one of the principal staple crops. According to recent data, sweetpotato is the fifth most important crop in Uganda, with respect to cultivated area, after banana, beans, corn, and millet and is third in terms of fresh weight produced (Odongo et al., 2004). In eastern Uganda, sweetpotato is cultivated in two cropping seasons, the first between March and June, and the second in August through November. These cropping seasons are determined by the occurrence of precipitation
U T I L I Z A T I O N
O F
H I G H
R E S O L U T I O N
S A T E L L I T E
I M A G E S
1
C I P
•
N A T U R A L
R E S O U R C E S
M A N A G E M E N T
W O R K I N G
P A P E R
2 0 0 7 - 5
and the availability of moisture in the soil. Most of the sweetpotato is produced by poor farmers, in plots of less than 0.1 ha. Agricultural statistics and remote sensing The National Agricultural Statistics Service (NASS) of the United States Department of Agriculture (USDA) has been involved since 1972 in applying information-gathering techniques for statistical purposes and agricultural monitoring, having utilized LANDSAT scenes since 1976 to establish a sampling frame as primary auxiliary variable. However, in this case the images were available only at the end of the growing season (Hanuschak and Delince, 2004). One of the largest projects to obtain crop statistics from images of LANDSAT satellite, agricultural and climate data and advanced data processing, was the Large Area Crop Inventory Experiment – LACIE, that was carried out between 1974 and 1978 by three US agencies, in order to predict in real time the production of wheat of the Soviet Union (MacDonald and Hall, 1980). Estimation of the extent of crop areas based on the integration of field surveys and satellite images of high spatial resolution has been carried out operationally in the United States and Europe since 1980 (Allen et al., 2002). However, due to limitations imposed by the spatial resolution of satellite products, until the end of the 1990s many institutions utilized the images only as a sampling frame. In the European Union, progress was made by the expansion of the project Monitoring Agriculture through Remote Sensing (MARS) to more countries (Hanuschak and Delince, 2004). An example of evolution in the utilization of remote sensing techniques for the acquisition of data on agriculture is the case of Hungary, where changes in the economic model brought about changes in land tenure, size of farms and number of farmers, as well as in agricultural technology and investment. This created the need for an efficient information system, which led to the establishment in 1980 of the Hungarian Agricultural Remote Sensing Program – HARSP (Csornai et al., 2006), whose objective was to implement a monitoring system for the gathering of reliable, accurate information about major crops, including the extent of cropping areas, their growth and problems (particularly the evaluation of drought risks), as well as to predict yields. A key step in producing agricultural statistics is the stratification of the target region by dividing it into sub-units (Roller and Colwell, 1986), which have less variability than the region as a whole, allowing for more precise target crops assessment. In some cases the administrative limits of a region are used for stratification, such as the instance described by Fang (1998) concerning the estimation of rice yields through remote sensing data. 2
U T I L I Z A T I O N
O F
H I G H
R E S O L U T I O N
S A T E L L I T E
I M A G E S
C I P
•
N A T U R A L
R E S O U R C E S
M A N A G E M E N T
W O R K I N G
P A P E R
2 0 0 7 - 5
In April 1986 the program SPOT Image led by France and other European countries was launched. CNESS SPOT has put in orbit several satellites for observing terrestrial resources. Currently, the SPOT 5 satellite is operational. The characteristics of this satellite are presented in Table 1. Table 1. Characteristics of the products generated by the SPOT Satellite.
Source:
http://www.spotimage.fr/automne_modules_files/standard/public/p229_3a1cd2cb59b76fc75e20286a6abb7efesatSpot_E.pdf
Physical basis of the discrimination between crops When radiant energy strikes any material an interaction occurs. Thus, the crops interact with solar radiation. A fraction of incoming radiation is absorbed by the plant pigments chlorophyll, carotenoids, xanthophylls and others (Figure 1), another fraction is transmitted toward the bottom of the canopy and the soil, and a further fraction is reflected back to space (Colwell, 1983: 2310). The property that is utilized to obtain information on the interaction between radiant energy and objects on the earth’s surface is reflectance, which is the ratio of the energy reflected in a given wavelength to the incoming energy (Lillesand and Kiefer, 1994). Vegetation, soil, and water bodies have distinct spectral patterns of reflectance (Figure 2), which makes it possible to differentiate them. Furthermore, different kinds of vegetation coverage have distinct spectral U T I L I Z A T I O N
O F
H I G H
R E S O L U T I O N
S A T E L L I T E
I M A G E S
3
C I P
•
N A T U R A L
R E S O U R C E S
M A N A G E M E N T
W O R K I N G
P A P E R
2 0 0 7 - 5
patterns or spectral signatures (Jensen, 1996: 316; Richards, 1993) that are utilized to distinguish them from other plant coverage and land uses, making it possible to quantify their respective areas. This differentiation between plant covers is based on the radiometric differentiation of the classes being considered (Lemoine and Kidd, 1998).
Figure 1. Spectra of absorption of chlorophyll and carotenoids.
Source: http://generalhorticulture.tamu.edu/lectsupl/Physiol/physiol.html
To accommodate the diversity of plot sizes, crop species, varieties, clones and cropping systems found in a region, satellite products with the appropriate level of spatial, spectral and time resolution are needed. Spatial resolution is the property that makes it possible to distinguish between two objects when they are very close. Spectral resolution, with a given number of bands that covers not only the visible region of the electromagnetic spectrum, but also the nearinfrared, makes it possible to discriminate different spectral signatures. In turn, time resolution permits registry of information from the same place with a given time sequence. The spectral bands that the SPOT 5 satellite registers are presented in the Table 2
4
U T I L I Z A T I O N
O F
H I G H
R E S O L U T I O N
S A T E L L I T E
I M A G E S
C I P
•
N A T U R A L
R E S O U R C E S
M A N A G E M E N T
W O R K I N G
P A P E R
2 0 0 7 - 5
Figure 2. Spectral patterns of vegetation, soils and water.
Source: http://landsat.usgs.gov/resources/remote_sensing/remote_sensing_applications.php
Sensors
Electromagnetic Spectrum
Pixels Size
Spectral Bands
SPOT 5
Panchromatic B1 : green B2 : red B3 : near-infra-red B4 : short-wave infrared (SWIR)
2.5 m OR 5 m 10 m 10 m 10 m 20 m
0.48 – 0.71 µm 0.50 – 0.59 µm 0.61 – 0.68 µm 0.78 – 0.89 µm 1.58 – 1.75 µm
Table 2. Characteristics of SPOT 5 satellite sensors and resolution.
Source: www.spotimage.com
U T I L I Z A T I O N
O F
H I G H
R E S O L U T I O N
S A T E L L I T E
I M A G E S
5
C I P
•
N A T U R A L
R E S O U R C E S
M A N A G E M E N T
W O R K I N G
P A P E R
2 0 0 7 - 5
MATERIALS AND METHODS Kumi district (Figure 3), located in northeast Uganda at latitude 1º 29’15 N and longitude 33º 55’ 58 E, was chosen as a pilot site for the presence of sweetpotato cropping fields. This district has an area of 2,848 km2, a population of 388,015 inhabitants and a population density of 136.24 inhab/km2, which is higher than the national average. Most of the population (97.8%) lives in the rural area. Two areas were stratified as a function of annual precipitation. In the dry area there is limited cultivated land, mostly used for drought tolerant crops such as millet and sorghum and some sweetpotato. The moist area has much more cultivated land with crops such as sweetpotato, peanut, corn, manioc, Chinese beans, and other species. Figure 3. Kumi district –Uganda.
Source: Google Maps
6
U T I L I Z A T I O N
O F
H I G H
R E S O L U T I O N
S A T E L L I T E
I M A G E S
C I P
•
N A T U R A L
R E S O U R C E S
M A N A G E M E N T
W O R K I N G
P A P E R
2 0 0 7 - 5
Material and equipment Cartographic material. Multispectral (XS) and panchromatic (P) images from the SPOT satellite with a spatial resolution of 10 and 5 meters, respectively, captured on May 6, 2006 (Figure 4) and on October 15, 2006. The initial spatial resolution in the multispectral scene was 10 m and the panchromatic band with a 5 m resolution was used for resampling the scene to this resolution. Since the photogrammetric chart for the geometric correction was not available, the requested scenes were corrected with the parameters of navigation of the SPOT 5 satellite. Handheld ASD FieldSpec pectroradiometer. Magellan Platinum Global Positioning System. Laptop Dell Inspiron 600 m. Canon Digital Camera. Environment for Visualizing Images (ENVI) v. 4.1 Software.
Image Processing The methodology utilized for sweetpotato cropping area data acquisition is based on the use of high-resolution images captured by the SPOT satellite, complemented by radiometric evaluation of the cultivated fields. Transects from Mbale to Soroti were marked by the access roads, following the distribution of the areas with the highest density of sweetpotato fields, where radiometric measurements were made with a handheld ASD FieldSpec spectroradiometer. The coordinates of the cultivated fields were recorded in order to locate them in the SPOT satellite scenes. ENVI software was used for processing the digital satellite data. An unsupervised classification was initially carried out, using clustering algorithms for ordering the high variability of land use and plant coverage. A supervised classification was subsequently carried out based on the spectral patterns recorded in the field, which was used for obtaining the spatial distribution of the sweetpotato plots, as well as the cultivated area through the maximum likelihood classifier. The digital process consisted of the sampling of the scene areas covered with sweetpotato, verified in the field in May and November of 2006, which were then graphically shown in the corresponding scene in order to carry out a supervised classification.
U T I L I Z A T I O N
O F
H I G H
R E S O L U T I O N
S A T E L L I T E
I M A G E S
7
C I P
•
N A T U R A L
R E S O U R C E S
M A N A G E M E N T
W O R K I N G
P A P E R
2 0 0 7 - 5
Figure 4. Multispectral Spot Image of the study area.
Grass Mango tree
House roof
Source: SPOT Image
Road
Subsequently, the geographical area of the district was delimited by means of a mask (image bitmap), which defined the cropping areas of sweetpotato and of other crops (Figure 5). In order to establish the percentage of success of the selected classes, the contingency matrix was calculated. It showed that the misclassification error for sweetpotato was 6%, which is satisfactorily low.
8
U T I L I Z A T I O N
O F
H I G H
R E S O L U T I O N
S A T E L L I T E
I M A G E S
C I P
•
N A T U R A L
R E S O U R C E S
M A N A G E M E N T
W O R K I N G
P A P E R
2 0 0 7 - 5
Figure 5. SPOT Image of the Kumi district.
House roof Road Dry grassTree Source: SPOT Image
RESULTS AND DISCUSSION Figure 6 shows the different spectral patterns of the main crops and land uses in Kumi. Sweetpotato can be differentiated both in the range of the visible spectrum, where it shows low reflectance attributed to the foliage pigmentation and total soil coverage, and in the nearinfrared where the reflectance is very high. This high reflectance of sweetpotato fields in the nearinfrared spectrum has been instrumental in distinguishing it from other crops. Figure 7 is the supervised classification carried out with the May 2006 image. It shows that most of the sweetpotato fields are located in moist areas while the drier areas are primarily covered by grasses. For ease of reading, the legend of Figure 7 is represented in Table 3. The table shows that sweetpotato covered 24,556 ha on the date the satellite image was captured. This area, representing 8.6% of the area of the district, corresponds to the sweetpotato crop in the first growing season only. On the other hand, the satellite image captured in October 2006 (Figure 8) processed and ground truthed with the same methods as the other image, shows that sweetpotato covers 20,064 ha, or 7% of the area of the district. It appears that this decreased U T I L I Z A T I O N
O F
H I G H
R E S O L U T I O N
S A T E L L I T E
I M A G E S
9
C I P
•
N A T U R A L
R E S O U R C E S
M A N A G E M E N T
W O R K I N G
P A P E R
2 0 0 7 - 5
cropping area observed in the second season was due to abnormally delayed rains. Adding up the sweetpotato cropping areas observed in the images from the two cropping seasons, the yearly total for 2006 amounts to 44,620 ha. However, the more recent official data from the Department of Statistics of Uganda (Figure 9), point out that in the whole year of 2003 the area cultivated with sweetpotato in Kumi was 28,000 ha. Assuming that this estimation remains approximately constant for the next years, including 2006, our remote sensing data suggests that the official statistics record just the 63% of the sweetpotato cropping area in Kumi district. In years with normal precipitation, the shortage of the official data would be even greater. If the proportion holds true for other areas, the total area of sweetpotato in Uganda might well exceed 1 M ha.
Figure 6. Spectral patterns of the main crops and land uses in Kumi. Reflectance
CROPS REFLECTANCES 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 300
400
500
600
700
800
900
1000
1100
1200
Wavelength Maize
10
U T I L I Z A T I O N
Bare Soil
O F
H I G H
Millet
R E S O L U T I O N
Sweet potato
S A T E L L I T E
Cassava
I M A G E S
Banana
C I P
•
N A T U R A L
R E S O U R C E S
M A N A G E M E N T
W O R K I N G
P A P E R
2 0 0 7 - 5
Figure 7. Classified SPOT Image of Kumi district, May 2006.
Source: CIP, NRM Geospatial Analysis Laboratory, based on a SPOT scene of May 6, 2006
U T I L I Z A T I O N
O F
H I G H
R E S O L U T I O N
S A T E L L I T E
I M A G E S
11
C I P
•
N A T U R A L
R E S O U R C E S
M A N A G E M E N T
W O R K I N G
P A P E R
2 0 0 7 - 5
Figure 8. Classified SPOT Image of Kumi district, October 2006.
Source: CIP, NRM Geospatial Analysis Laboratory, based on a SPOT scene of Oct. 15, 2006
Table 3. Area covered by different land use and plant cover categories, May and October 2006.
Category Water bodies Wetland Grassland/other crops Sweetpotato Forest/Mangoes Urban areas Fallow land Main/Secondary road Clouds No data Total
12
U T I L I Z A T I O N
O F
.--/
H I G H
LEGEND May-06 Oct-06 Ha % Ha % 39310 13.8 37823 13.3 23858 8.4 22932 8.1 118324 41.5 104218 36.6 24556 8.6 20064 7.0 17710 6.2 17698 6.2 11 0.0 11 0.0 30840 10.8 70031 24.6 242 0.1 242 0.1 4872 1.7 0 0.0 25078 8.8 11782 4.1 284800 100.0 284800 100.0
R E S O L U T I O N
S A T E L L I T E
I M A G E S
C I P
•
N A T U R A L
R E S O U R C E S
M A N A G E M E N T
W O R K I N G
P A P E R
2 0 0 7 - 5
Figure 9. Sweetpotato cropping area in Kumi district according to official statistics of Uganda.
Area planted to Sweetpotato in Kumi 30,000 25,000
Ha.
20,000 15,000 10,000 5,000 0 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2,003
Year
Source: Uganda Bureau of Statistics (UBOS)
CONCLUSIONS Radiometric evaluation and processing of SPOT scenes of Kumi district, Uganda, have made it possible to determine that sweetpotato foliage has a distinct spectral pattern defined by low reflectance in the visible range of the spectrum and high reflectance in the near-infrared range. This spectral pattern makes it possible to identify a sweetpotato crop with a high degree of certainty, which allows defining the cultivated area and spatial distribution of the crop with precision through the utilization of high-resolution SPOT images. The results suggest that traditional statistics underestimate the total annual sweetpotato cropping area in Kumi. Using the tested method substantially improves estimation of cropping areas.
U T I L I Z A T I O N
O F
H I G H
R E S O L U T I O N
S A T E L L I T E
I M A G E S
13
C I P
•
N A T U R A L
R E S O U R C E S
M A N A G E M E N T
W O R K I N G
P A P E R
2 0 0 7 - 5
REFERENCES Allen, R., G. Hanuschak, and M. Craig. 2002. Future vision for the utilization of remotely sensed data and Geographic Information Systems (GIS) in the National Agricultural Statistics Service. From: http://www.usda.gov/nass/nassinfo/remotevision.htm Colwell, R.N., ed. 1983. Manual of Remote Sensing. 2d Ed. 2 vols. Falls Church, Virginia: American Society of Photogrammetry. Csornai, G., Cs. Wirnhardt, Zs. Suba, P. Somogyi, G. Nádor, L. Martinovich, L. Tikász, A. Kocsis, B. Tarcsai, Gy. Zelei. 2006. Remote Sensing based crop monitoring in Hungary. From: http://www.fomi.hu/Internet/magyar/Projektek/remsensmonit.htm and abstract at: http://www.igik.edu.pl/earsel2006/abstracts// Fang, H. 1998. Rice crop area estimation of an administrative division in China using remote sensing data. International Journal of Remote Sensing, Vol. 19(17): 3411–3419. FAO. 1999. Crop Information Systems. From: http://www.fao.org/sd/EIdirect/EIre0078.htm FAO. 2003. Remote Sensing for Decision-makers. From: http://www.fao.org/sd/EIdirect/EIre0072.htm Hanuschak, G., and J. Delince. 2004. Utilization of data captured through remote sensing and of geographic information systems for the livestock statistics in the United States and the European Union. Third International Conference on Agricultural Statistics 2–4 November, 2004, Cancun, Mexico. Jensen, J. R. 1996. Introductory Digital Image Processing. 2d ed. Saddle River NJ: Prentice Hall. Lemoine, G., R. Kidd. 1998. An alternative sampling technique for deriving crop area statistics from hybrid remote sensing data. In: Proceedings: Retrieval of Bio- and Geo-Physical Parameters from SAR Data for Land Applications Workshop. ESTEC, 21–23 October. Lillesand, T.M., and R.W. Kiefer. 1994. Remote Sensing and Image Interpretation. 3d ed. New York: John Wiley & Sons. MacDonald, R. B., and F. G. Hall. 1980. Global Crop Forecasting. Science, Vol. 208 (4445): 670–679. Odongo, B., R.O.M. Mwanga, and C. Niringiye, 2004. Promoting improved sweetpotato varieties and processing technologies in Lira, Pallisa and Soroti Districts. Final Technical Report. Namulonge Agricultural and Animal Production Research Institute (NAARI), Kampala, Uganda. Richards, J.A. 1993. Remote Sensing Digital Image Analysis: An Introduction. 2d ed. N.Y.: Springer Verlag. Roller, N. E. G., and J. E. Colwell. 1986. Coarse-resolution satellite data for ecological surveys.
Bioscience Vol. 36(7): 468–75.
14
U T I L I Z A T I O N
O F
H I G H
R E S O L U T I O N
S A T E L L I T E
I M A G E S
CIP’s Mission The International Potato Center (CIP) seeks to reduce poverty and achieve food security on a sustained basis in developing countries through scientific research and related activities on potato, sweetpotato, and other root and tuber crops, and on the improved management of natural resources in potato and sweetpotato-based systems. The CIP Vision The International Potato Center (CIP) will contribute to reducing poverty and hunger; improving human health; developing resilient, sustainable rural and urban livelihood systems; and improving access to the benefits of new and appropriate knowledge and technologies. CIP will address these challenges by convening and conducting research and supporting partnerships on root and tuber crops and on natural resources management in mountain systems and other less-favored areas where CIP can contribute to the achievement of healthy and sustainable human development. www.cipotato.org CIP is supported by a group of governments, private foundations, and international and regional organizations known as the Consultative Group on International Agricultural Research (CGIAR). www.cgiar.org
International Potato Center Apartado 1558 Lima 12, Perú • Tel 51 1 349 6017 • Fax 51 1 349 5326 • email cip@cgiar.org