Measuring Sustainability: Operationalization and Interpretation of the performance of urban form by Diellza Elshani

Measuring Sustainability Operationalization and interpretation of the performance of the urban form using generative design methods Case study: Hernals, Vienna

Master Thesis | 2020 | Diellza Elshani

Measuring Sustainability: Operationalization and interpretation of the performance of the urban form using generative design methods. Case study: Hernals, Vienna

Master Thesis in Integrated Urban Development and Design IUDD October 2020

Author: Diellza Elshani 120499

Supervisors: Prof. Dr. Reinhard König Prof. Dr. Sven Schneider M.Sc. Serjoscha Düring

Faculty of Architecture and Urbanism Chair of Computational Architecture | Bauhaus-University Weimar | Germany City Intelligence Lab CIL | Austrian Institute of Technology AIT | Austria

Acknowledgments I want to express a high appreciation and endless gratitude to my supervisor Prof. Dr. Reinhard König for his academic guidance and lessons throughout my master's studies and primarily through my master thesis period. I am also thankful to my supervisor Prof. Dr. Sven Sneider for his valuable support. It's a pleasure to extend my sincere thanks to the Austrian Institute of Technology, and its City Intelligence Lab CIL, for allowing me to write my thesis with them. I want to thank the head of the lab, Angelos Chronis, and all my colleagues for their support. Especially Serjocha Düring who supervised me during my master thesis, and together with Nariddh Khean shared their academic knowledge and experiences, which enriched this journey even more. I appreciate all my friends and colleagues for their good humour, excellent advice, and support during this time. To conclude, I am endlessly grateful to my family: my mother and my siblings for their love, support, and encouragement. Finally, I am most thankful to my father for laying the foundation of all my knowledge and being the main reason for who I am now. He was my biggest supporter and would have been very happy to have seen this moment.

Abstract With rapid urbanization, the necessity for sustainable development has skyrocketed, and urban agendas foster the call for more sustainable development. A plethora of researchers has already made substantial contributions to this discussion of measuring sustainability in urban design. In parallel, recent advances in computing performance of urban layouts in real-time allow for new paradigms of performance-driven design. In this context, data mining, generation, and analysis gain more significance, especially in early design phases. As beneficial as utilizing multiple layers of urban data may be, it can also create a challenge in finding relevant datasets. Data can prove deficient because just quantifying a design's performance does not necessarily provide insight nor guidance as to why it performs better or worse and on how to improve the urban layout. To operationalize data, a sound understanding between the design parameters and the different performance criteria is crucial. In response to that, this thesis presents an integrated computational framework to measure sustainability and operationalize and interpret the urban form's performance data using generative design methods. The framework is applied to a new site development in Hernals, as a strategic location of the northwest Vienna. The performance data is clustered into three pillars of sustainability: social, environmental, and economical. Linear correlation between the design parameters and performance indicators is applied, and several actionable rules of thumb are derived from the research. A big advantage of the framework is that it can be used as a discussion table in participatory planning processes, since it could be easily adapted with interactive environments. Due to the complexity of both urban systems simulation models and the quantification of sustainability, more research is needed in the field. This thesis follows the call of urban agendas, by fostering the research on quantifying sustainability in the early design stages.

Contents 1. Introduction .................................................................................................................. 11 2. Literature Review .......................................................................................................... 17 2.1 The Concept of Sustainability: Historical Background .................................................... 17 2.1.1 Sustainability in Urban Design ................................................................................. 17 2.1.2 Quantifying sustainability in urban design .............................................................. 18 2.2 Urban Configurations ..................................................................................................... 24 2.3 Generative Design .......................................................................................................... 26 2.4 Public Participation in Urban Design .............................................................................. 27 3. Methodology: A framework for quantifying urban sustainability through social, environmental and economic dimensions ......................................................................... 29 3.1 Related Tools .................................................................................................................. 29 3.2 Architecture of the Framework ...................................................................................... 30 3.3 Data Structure ................................................................................................................ 32 3.4 Generating Urban Designs .............................................................................................. 32 3.5 Performance Indicators: Social, Environmental and Economic Dimensions .................. 38 3.5.1 Social Sustainability Evaluation Method ................................................................. 40 3.5.2 Environmental Sustainability Evaluation Method ................................................... 44 3.5.3 Economic Sustainability Evaluation Method ........................................................... 47 3.6 Additional Potential Algorithms and Performance Data Aggregation ........................... 52 4.Application of the framework: Hernals Vienna, Austria .................................................. 53 4.1 Why Hernals ................................................................................................................... 54 4.1.2 Vienna Urban Development .................................................................................... 55 4.1.3 Vienna and Sustainable Developmet Goals Strategies............................................ 55 4.1.4 Hernals Data ............................................................................................................ 56 4.2 Data Preparation ............................................................................................................ 57 4.3 Design Space Exploration: The Generative model ......................................................... 57 4.4 The performance space: Social, environmental and economic ..................................... 58 4.4.1 Social........................................................................................................................ 60 4.4.2 Environmental ......................................................................................................... 63 4.4.3 Economic ................................................................................................................. 68 5. Conclusion and reflection .............................................................................................. 76 6. List of Figures and Tables ............................................................................................... 82 7. References .................................................................................................................... 85 Affidavit ............................................................................................................................ 89 9

1. Introduction Given the changes the world has undergone over the last decades, the necessity for sustainable development has skyrocketed. Sustainability is the ability to be maintained, as well as being the integration of measures and acts focused on three pillars: environmental, social, and economical. Yet, to balance social equity and economic potential with environmental-focused design is quite a lofty challenge. The urban agendas foster the call for a more sustainable development followed by tools for action. As a response, this research is an attempt to contribute to existing methods towards sustainable development, focusing on a discussion of the three aforementioned pillars of sustainability. Furthermore, many attempts towards quantifying quality in the domain of sustainable development are ongoing. A plethora of researchers have already made substantial contributions to this discussion (Cotgrave & Riley, 2013), presenting answers for how sustainable a project can be, and what impact such a sustainable project has on the world. The traditional evaluation method of urban spaces until now was either by analyzing and interpreting secondary data such as open-source data, GIS, which follows the objectives approach1, or the subjective approach where data is generated from surveys and focuses on humans behaviors and more on an individual level (Marans & Stimson, 2011). Similarly, there have been several checklists on measuring sustainability introduced and assisted in urban design. It is quite challenging to specify, quantify, and translate the urban performance in terms of sustainability into an operationalizable format. On the other hand, frameworks that can estimate the impact of changing elements within the urban systems are in high demand (Achary et al., 2017). It makes the planning process more manageable, where designers combine their design instinct with performance data. In this context, data mining, generation, and analysis gain more significance, especially in early design phases (Nembrini, 2012). By providing an enormous number of analysis parameters that influence specific performance metrics, we can increase a designer’s capacity to achieve design goals based on urban performance (Lira, 2012). The rrecent advances in computing the performance of urban layouts' have opened new paradigms of performance-driven and evidence-based urban design (Keller, 2006). Applying performance-driven design, we could create a workflow that allows us to have overall feedback on the design and its performance and have an iterative process of advancements based on the formation of ideas and their evaluations (Lawson, 2006). A significant advantage of computing design options, and provide their performance indicators, instead of using conventional design approaches, is its speed and practicality. It allows the ability to test vast design options in a short manner of time. This topic falls under the generative design umbrella, which collates digital computation and humans' creativity. The most significant advantage of cognitive computing that generative design offers is the possibility to manage the urban system's complexity. Interpreting urban data, either mined from open source data platforms, generated from simulations, or different computational methods, would assist the early stages of design as guidance towards urban design. To operationalize data, to understand why a design performs in the way it does and avoid mistakes in the planning process, a sound understanding between the design parameters and the different performance criteria is crucial (Bielik et al., 2019). This could lead towards extracting knowledge on actionable guidance for designers.

Data is based on known valid evidence, e.g.: existing demographic data

However, analyzing many aspects of an urban space leads to several data layers and an almost unmanageable amount of information. As beneficial as having many urban data layers might be, it also creates a challenge in finding relevant datasets (Ribeiro, 2015). This way, data can also be confusing if the approaches toward dimensionality reduction and data interpretation are inadequate. Even more, impracticable data can be deficient because quantifying a design's performance does not necessarily provide insight or guidance on why it performs better or worse and how to improve the urban layout. Consequently, it is evident that there is a research gap on the approaches toward measuring sustainability, operating with multiple data layers, and deducting actionable guidance for designers based on data. Therefore, more research is needed in:   

quantifying urban sustainability in new urban development projects; filtering the information and clustering it in an operationalizable format, which is interpretable and easy to understand by stakeholders; extract knowledge on actionable guidance for designers based on the performance data.

Existing models and methods for the prediction and measurement of sustainability are contributing to the performance driven design area. On the other hand, urban parameters to build and read an urban layout are essential when composing new projects. Additionally, computational approaches on generating the design options and simulating their performance are also valid in the digital era. Therefore, a combination or reshaped form of these methods could be a valid model to deal with the aforementioned urban challenges. Especially in urban redevelopment sites, where interdisciplinary knowledge is required and achieving a rich design that provides social equality, offers better mobility, economic potential and urban vitality, also pedestrian comfort in urban space at the same time, is quite challenging. In addition, knowing which design parameters are causing an increase or decrease in the design's performance could give general instructions on improving the design option. The density metrics have been a relevant parameter in urban development, starting with small settlements' growth until their evolutions into villages or cities. Le Corbusier (1967) states that the cities need a specific density for the machine-age man, which would provide short travel time between housing, jobs and other attractive locations in the city. Furthermore, Jan Gehl (2010) states that urban density is a critical factor in understanding how cities function. With COVID-19, the global pandemic faced in 2020, the debate around whether highly dense areas and compact cities are the goal of urban planners, comes into question. However, defining the specific density is seen as an essential input parameter for an urban form. Spacematrix, a tool to analyze the built environment's spatial configurations, sets the density as a correlation of the density and the built mass (Nes et al., 2012). It is focusing on measuring various types of urban density, floor area ratio (FAR), the ground space index (GSI), and network density. Another way to read and understand the architecture and morphology of a city pattern is to explore the orientation of the cities' streets. Regardless of whether the street network was designed through a top-down approach or if it evolved organically through time, their configurations and orientations assist in defining the spatial logic and spatial order of the city (Boeing, 2019). Relating these design parameters with the performance indicators could lead towards actionable guidance for designers. Understandably, simple correlation doesn't necessarily 12

indicate causation. However, available suggestions that can inform the early stages of design, an action that could improve its performance, would doubtlessly contribute to the urban design world. Derived from the provided argumentations this research hypothetically correlates the density metrics and street network orientation with the performance indicators. Two main hypotheses are presented: 



the density metrics, FAR and GSI, are not only significant parameters to understand the urban form, but are is key parameters that affect the urban performance in social, environmental and economic dimensions of sustainability; the orientation of the streets in an urban layout doesn't only play an essential role in exploring a city while walking; but it also affects the urban performance in social, environmental and economic dimensions of sustainability;

In response to the research gaps, and to test the hypotheses this thesis presents an integrated computational framework to operationalize and interpret the performance of the urban form using generative design methods. The research goals of the thesis are:

1. To understand how we can ensure sustainability in urban design by quantifying the performance in the early stages of urban design. By generating design options and simulating their performance based on the performance indicators corresponding social, environmental and economic sustainability. Six performance indicators are defined, each of them informs us on the sustainability level of the design. Having them among each other tends to be a more comprehensive evaluation of urban layout. An overview on how the research narrows down each pillar of sustainability and derives the measurable indicator is given below: The social dimension of sustainability states that a system of social organization tackles the equalities and relations between humans. Societies do more than only exist in space (Hillier& Hanson, 1984), they act and interact with each other. And it is exactly those mutual relationships between individuals and human interactions that make their behaviors social (Weber, 1978). The study adapts space syntax metrics with demographic data to measure social integration and accessibility to significant quarters of the designed urban layout. On the other hand, with the growing threat of climate change, and the significant impact of the built environment on climate, microclimate analysis is a must in the early stages of design. Geo-located wind speeds and solar radiation analyses are the only microclimatic parameters that depend widely on urban planning (Reiter, 2010). Therefore, this study focuses on sunlight hours and wind comfort performance for each design option. To have instant feedback on the performance, the research uses pretrained machine learning models to predict the microclimate analysis. Regarding the economic potential of a designed option, the suggested framework elaborates upon the tight connection between the economic potential with the time dimension and focuses more on how the cities work rather than how they look like. Agent-based models and analytical economic models cannot be integrated easily in the design process (Karimi, 2012); however, this research puts together a hypothetical economic dynamic model, combining 13

three existing workflows to understand the resilience of the city as a complex system and provide an indication on the value return for investors. The workflow consists of, first, revealing initial hotspots (determined by the level of mix use index); leveraging this information to the evolved configurational properties of Bielik et al. (2019); creating an endless cycle of simulating the pedestrian movement flows and accordingly the land use distribution; finally, interpolating the results with construction costs. The outcomes of the three dimensions are combined, input and output parameters are correlated, and trade-offs between them are elaborated further. 2. To filter and operationalize the input parameters and performance indicator data By selecting subset of urban layout input parameters and operationalize the performance indicator data. As argued above, FAR, GSI and street orientation are valid parameters, and this thesis will use them to generate urban forms. The performance indicator data of each metric, and the input parameters are aggregated into single values, which simplifies the complexity of multilayer data. A clear interpretation for each indicator is provided, which operationalizes the data in a format which is easy to understand by stakeholders and assists in the guidance towards improving the urban layout. 3. To correlate and analyze input data that was used to generate the urban layout, with output data that indicates the performance of the urban layout; This study correlates the FAR, GSI, and street network orientation data with the performance data of each indicator and tries to find relation that could inform on actionable guidance towards urban design.

In the form of research questions that the thesis attempts to answer, we can state the following:   

How to ensure sustainability in urban design by quantifying the performance for early stages of urban design Can we relate the performance of urban layouts to FAR, GSI, and street network orientation? Can we deduce actionable rules of thumb for urban design problems?

As a method, to answer those question, this research proposes a computational framework, which:   

Builds a generative urban design model using FAR, FSI, and street orientation as input parameters; from where hundreds of urban layouts are generated; Adjusts, combines, and remodels existing methods to quantify urban performance in each pillar of sustainability, which is followed by its data aggregation; Correlates the input design parameters with the outputs that indicate the performance and further data analysis, which assists in deducing rules of thumb to guide the design configurations.

The framework is applied to a new site development in Vienna. By understanding the context conditions, challenges, and design goals, the suggested framework simulates the spatial performance by augmenting the design and corresponding performance explorations. As 14

stated before, the performance data is clustered into three pillars of sustainability: social, environmental and economic. General guidance on how to improve the urban layout in each dimension is given after every performance indicator based on the data analysis. A significant advantage of the framework is that it can be used as a discussion table in participatory planning processes, since it can be easily adapted with interactive environments. As promising as the framework appears, and it presumably can be a supplement to the existing tools, it does not cover all issues related measuring sustainability. But it doubtlessly follows the call of urban agendas, by fostering the research on measuring sustainability.

Thesis Structure The thesis consists of five main chapters. The introduction chapter gives insights about the background problem and argues the significance of measuring sustainability. It clearly states the research questions and hypotheses within the research scope and intrudes briefly on the applied methods. Chapter two is the literature review, and it elaborates on the existing methods of measuring sustainability. It starts with the historical background and explores until state-of-the-art researches on the topic. It also justifies the methods used in the framework. Chapter three, the used methodology to quantify urban sustainability, explains the architecture of the designed framework in detail. It includes the related software and existing tools that are used. It covers all the generative design study steps and the design of the workflow of the performance indicators in social, environmental, and economic dimensions. The central part of the technical work of the research is elaborated in the methodology chapter. In chapter four, the framework is applied to an urban redevelopment project in Vienna. It starts with the data gathering and preparation methods, and it continues with the actual application of the framework. The results of the generated cases with their performance scores are used to correlate input and output parameters. Additionally, deduced rules that can guide the early stages of design are given. In the conclusion chapter (chapter five) is given a reflection on the framework. The chapter explores the limitations and potentials of the designed framework. It is followed by the overall conclusions of the work and methodologies in cognitive urban design computing.

2. Literature Review 2.1 The Concept of Sustainability: Historical Background The topic of sustainability was brought into the discussion over 200 years ago by the political economist Thomas Malthus. He raised whether natural resources will max out among the growing world population, which brought into debate the term sustainable development (Dixon, 1989). The rapid urbanization, the digital era, and the shortening of natural resources of the last decades brought a global concern on renewability and the questions that Malthus raised. In previous decades, economists debated whether the contemporary course of economic development is ‘sustainable’ (Baisago, 1999). These global concerns regarding the renewability of natural resources and sustainability were still being developed in decades. In A Blueprint for Survival, the editors of the Ecologist 1972, a significant Brattish panel, stated that our ‘industrial way of life with its ethos of expansion’ is not ‘sustainable’ (Baisago, 1999). As a conclusion from the panel was indicated that a stable society should have under control or maintain the population growth, maximize the conversation, and minimize the ecological disruption. It was stated that ‘Our task is to create a sustainable society, which will give the fullest possible satisfaction of its members.’ This statement can interpret that the word sustainability started to be shaped, and its significance while planning was becoming higher and higher. During the 1980s, the founder of WorldWatch Institute, Brown, made impactful progress in defining the term ‘sustainability’ by writing and stating the topics of overpopulation, the harm that industrial production is causing to the natural system the lack of natural resources (Baisago, 1999). With the progress of involving sustainability and framing it in urban design, the field got more interdisciplinary, giving possibilities to consider many urban scale dimensions in the design phase. The new dimensions that were considered in urban design were energy efficiency; material flows, including recycling and upcycling; social aspects such as social equity and integration; sustainable economic growth; and, in general, sustainable and resilient urban growth.

2.1.1 Sustainability in Urban Design In urban planning, sustainability is more often seen as a strategy, while planning and sustainable urban development is the way to sustainability. Chronologically, in the form of ‘sustainable development,’ the term was used for the first time in 1980 at the World Conservation Strategy written by UNEP – the United Nations Environment Programme and IUCN – The International Union for the Conservation of Nature. The appearance of the term ‘sustainability’ in the science of development has triggered urban planners and architects to apply evolving notions of ‘sustainability’ to the contemporary debate on how cities and regions should grow, be revitalized, redeveloped, and reformed (Baisago, 1999). Chasing up, in the upcoming years with the urban agendas and international development policies, the topic of economics, society, and the environments under the umbrella of sustainability was being tackled. Later, these three topics became the three conceptual pillars of sustainable development, stated by Kahn (1995) in the Agenda 21. Khan elaborated on the social, environmental, and economic component of sustainability and fostered them to be integrated and nested. To achieve more comprehensive sustainability, the three pillars 17

should be balanced and in harmony since the social, environmental, and economic substrates of sustainability relate to each other (Khan, 1995). With the urban agendas, sustainable development - as the only way towards sustainability – has been a high aspect of urban planning strategies. However, validating projects in terms of sustainability is an ongoing topic still nowadays. Researchers and specialists have done already contributions to the integrations of measuring sustainable development into their practices (Cotgrave & Riley, 2013) (Newton, 2012). Even though the meaningful measurement of sustainability requires assessing a wide range of phenomena, it can hardly be done comprehensively due to the set of involved factors (Neuman & Churchill, 2015). The upcoming chapter will elaborate on several existing approaches to measuring aspects of the sustainability of urban form.

2.1.2 Quantifying sustainability in urban design The traditional evaluation method of urban spaces until now was either by analysing and interpreting secondary data such as open-source data, GIS, which follows the objectives approach2, or the subjective approach where data is generated from surveys and focuses on human’s behaviours and more on an individual level (Marans & Stimson, 2011). Lately, many research centres and offices are trying to identify the most relevant indicators for evaluating urban form and smartness. One of them is the Finnish Technical Research Centre with The CITYkeys project, which is a collaborative work of VIT3, Austrian Institute of Technology (AIT), and TNO4, where they define needs, analyze outcomes and develop a recommendation for the use of performance indicators (Garau & Pavan, 2018). In the project, they mostly focus on smart cities and the corresponding indicators at the city level. In the study, both approaches, objective and subjective, are combined. They identify the critical stakeholders and their benefits from the project as a starting point, them being: cities itself, policymakers, solution providers, industrial stakeholders, and citizens. They categorize the indicators into groups such as input indicators (e.g., policies, human resources, materials, etc.), process indicators (includes: holding of planned events and meetings), output indicators (has: distribution of smart meters, number of electric busses, etc.), outcome indicators '(includes: the overall number of dwelling, they are results of quality and quantity of the implemented activities) and impact indicators(includes: long term results such as energy use, air pollution, etc.) (Huovila et al., 2017). However, as complete as the framework is, there are downsides to use it in new developments. In the new quarter development, the CITYkeys framework's disadvantages are the very general metrics oriented on the city level and the long-lasting data collection phase. Besides, all the data layers are not relevant for small scale projects and might be difficult in concluding in overall general evaluation. Another research that aims to translate the subjective perceptive in an objective vision and define the performance with a number on a scale from 1 to 100 is done by Delsante et al. (2014). He indicates that the indicators should consider housing, social and collective services on a smaller scale, landscape, and environmental conditions. The research finds significant relations between health, wellbeing perception, and urban quality. The hopeful prospect of 2

Data is based on known valid evidence, e.g.: existing demographic data VTT Technical Research Centre of Finland Ltd 4 The Netherlands Organisation for applied scientific research 3

the study is that they categorize the indicators into four groups: architecture group (architectural values, identity, etc.), fruition group (services and mobility), environment group (environmental system and landscape), and the social group (public and collective functions and services) (Delsante et al., 2014). Similarly, there have been several checklists on measuring sustainability introduced and assisted in urban design. The Territorial Agenda 2020 – TA2020 fosters sustainable territorial development, including equal opportunities for citizens and enterprises; it supports economic growth and environmental quality (Territorial Agenda, 2020). With the release of the 17 Sustainable Development Goals from the UN in 2015, the world moved into a 'global age of sustainable development.' The call for evidence-based methods with impact evaluations, especially in European cohesion policies, is becoming higher every day. Methods such as life cycle analysis, balancing techniques, or urban metabolism already inform the design process about material flows and interpret urban systems' complexity. However, it is quite challenging to specify, quantify, and translate the urban performance in terms of sustainability into an operationalizable format. On the other dimension of the checklist and collective subjective evaluation, this study aims to contribute to the existing measurable indicator collection and interpret an overall conclusion of design options. Based on existing methods, the study seeks to put together metrics that are relevant to urban sustainability. Aligned with existing research, the study suggests an inductive way to come to the overall score of design performance. Starting from the design's performance in each pillar of sustainability, a combined overall score will give the general performance rate of the design. Remarkably, quantifying an urban layout's performance in each pillar of sustainability, and having an overall score is a challenge that still requires a lot of contribution. This research aims to achieve a more holistic evaluation based on indicators – as key elements of evidence-based urban design, planning, and management. Comprehensible with sustainability is meant the ability to be maintained, and it is the integration of measures and acts focused on three pillars: environmental, social, and economical. This research will focus on finding significant measurable indicators for each of those pillars that depended on urban form. Undoubtedly in each of these pillars, there are spatial aspects that cannot be quantified. Therefore we try to quantify the relevant and quantifiable aspects. While exploring existing tools to measure sustainability, the study will be focused on the most relevant indicators of each pillar of sustainability that can inform the early stages of design. It is crucial to keep in mind that these three pillars are social nested, interrelated, and bearable interdependent. While: Environment + Social + Economic = Sustainability; Environment + Social = Bearable; equitable sustainable Economic + Environment = Viable; and Social + Economic = Equitable. viable

environment

In the following chapters we will elaborate separately this three pillars Figure 1 Venn diagram representing the standard dimensions and corresponding existing methods for of sustainable development. Adapted from Tanguay, 2009, quantifying their key aspects, potential and referencing concepts proposed in WCED, 198 and performance in urban form. economic

Social Sustainability The social dimension of sustainability states a social organization system that tackles the equalities and relations between humans. Societies do more that only exist in space (Hillier & Hanson, 1984); they act and interact with each other. Precisely those mutual relationships between individuals and human interactions make their behaviours social (Weber, 1978). Under the umbrella of sustainability, the social dimension is a result of efficiency and equity (Alexander, 1994). Many urban issues that reflect inequalities such as ghetto formations or segregation of a particular group of people result in significant consequences that affect the whole site as a complex urban system. Equality in social aspects of space indicates many occasions prioritizing equal access to services or basic needs and public spaces. The opportunities between people who are part of a city or a neighbourhood should be similar. This equality is access to a fundamental necessity to achieve social sustainability. To fix the present urban social issues and better plan the future, there has been significant research on the consequences of social behaviours such as crime, segregation, and unemployment with the spatial dimensions and urban form (Bielik et al., 2019). Therefore, it is highly recommended to measure the equality of access to the city's significant parts, or to the lively streets, from all populations of the new design quarters. Although there have been much empirical research and studies on how the allocation, urban typology, and other spatial configurations can be traced back to several malicious social behaviours, it is still hard to conclude the topic (Bielik et al., 2019). Some of the social dimensions, such as culture and feelings in urban space, are not easily quantifiable, not in the current decade. But the hypothesis, that unequal access to public areas and zones of interaction can lead to segregated and disadvantaged zones in the city there can be dimensions based on street network configurations that can be quantified. Hillier states that the connection between space and society cannot be limited to cultural and lifestyle (Hillier & Hanson, 1984). The classical approach to measuring segregation and aggregate this information in geolocation would be the E-I index by Krackhard and Stern 1988, dedicated to measuring the homophily. Other methods would be the Assortativity Coefficient, Gupta, Anderson, May's Q, Odds-ratio for within-group ties, Freeman's segregation index, etc. However, in an unbuilt city or quarter with a lack of demographic data, a simulation method is required. If we start with access to the city's liveable spaces, one of the urban morphological elements that lead to quantifiable indicators is the street network. Hillier, 2009, using street network configurations, gives two calculations on how to quantify spatial sustainability at the social level: the community formation and the security (Hillier, 2009). While the community formation indicates the accessibility and integration within the surrounding neighbourhoods’ and safety, security is based on the neighbourhood’s internal pedestrian flow distribution (Hillier, 2009). Moreover, Bielik et al. (2018), bringing these metrics on a different level, developed a method that helps capture the street network's social potential, named the Aggregated Social Accessibility ASA. The ASA method -a graph-configurational method based on street network configurations- starts with estimating the pedestrian movement flows based on the street network configurations. It continues with the accessibility to these movement flows calculated for any given location (Bielik et al., 2018). ASA can serve as a tool to detect the excluded zones and the zones that don't have equal access to liveable streets and spaces with high potential for human interactions.

Environmental Sustainability Environmental sustainability advocates society to 'live within the limitations of the biophysical environment' (Goodland, 1995). Many indicators can be quantified with the environment related to temperature, oxygen, CO2, pH, water, traffic noise, wind or sunrelated, etc. In this study, we tight near the environmental factors twitch the microclimate indicators since they are more constant and are the main factor in habitat selection or other ecological activities. Lately, with the threat of climate change, and its significant impact on the built environment on climate, microclimate analysis and performance-driven design is becoming a movement in the Architecture, engineering, and construction world. The microclimate is the set of environmental variables that includes indicators like wind speed, sunlight, temperature, and moisture. Considering that this study will be applied in urban design options, where the parameters that matter and can be optimized are the building geometry, open space, street network, etc., the study will focus on parameters that hardly depend on the urban form. According to Reiter, the geo-located wind speeds and solar radiation are the only microclimatic parameters that depend widely on urban planning (Reiter, 2010). One of the most relevant indicators is solar radiation, the leading natural energy input in terms of energy in urban environments. Along sunlight hours, it has an impact on several human social behaviors as well as other living beings and ecosystems in general. Also, solar radiation has both direct and indirect effects on human health example: some of the indirect effects are the quality of food, its impact on terrestrial plants and ecosystem, deterioration in air quality, as well as energy-related issues that are drivers for the world economy (Seckmeyer et al., 2012). Sun exposure is a significant measure, and it needs to be elaborated in the early stages of design. However, the physiology and human bodies are tightly bonded with daylight, which is an essential general outcome that emerged from the last decades (Mardaljevic & Tregenza, 2017). In the same paper, regarding guidance for daylighting buildings, Mardaljevic cites Socrates (469-399 BC): 'Now in the houses with a south aspect, the Sun's rays penetrate into the porticoes in the winter, but in summer, the path of the Sun is right over our heads and above the roof so that there is shade. If, then, this is the best arrangement, we should build the south side loftier to get the winter sun and the north side lower to keep out the cold winds.' Socrates (Xenophon, 1979)

Moreover, some studies tried to find the association between temperature, sunlight hours, and alcohol consumption, even though the researchers, Hagström, Widman, and Seth found no association between colder climate with increased alcohol consumption socio-economic factors justify the suggested relation (Hagström et al., 2019). But similar hypotheses and researches are still being elaborated, so the debate whether there is a correlation or not between sunlight hours and alcohol consumption is still open to discussion. Currently, there are studies on building mass optimization in terms of sunlight hours. The simulation technology has made it possible to understand the performance of potential designs. Case in point, including solar analysis in the early stages of design, will facilitate natural daylight design and benefit from reducing costs and energy for artificial lighting (Si & 21

Wang, 2015). So, the paramount significance of quantifying solar radiation is to understand a specific location's energy potential. On the other hand, wind comfort is also an integral part of the environmental strategy (Othmani Marabout & Anton, 2019). Pedestrian-level wind in microclimate conditions is one of the first microclimatic issues to be considered in modern urban and architectural planning and design (Wu & Kriksic, 2012). Many urban authorities grant construction permission for new high-rise buildings only after a wind-comfort study has been done (Blocken et al., 2016). Because a wind nuisance issue after the design is finalized despite the time consuming, it is costly (Fadl & Karadelis, 2013). Accordingly, it is also complex to make significant modifications and improve pedestrian comfort quality in the wind aspects after constructing building structures. Nevertheless, there are several existing methods to measure postoccupancy wind performance. However, in the design stage, a physical or mathematical measurement is not possible because simulation-based methods and theoretical calculations are frequently used while allowing space for specific design modifications before the final project execution (Othmani Marabout & Anton, 2019). A method to evaluate pedestrian comfort and microclimate safety in urban spaces is the CFD (Computational Fluid Dynamics) simulations, since it is seen as a powerful tool and assess in quantifying the wind comfort levels. The advantage of CFD simulations is that it results in qualitative and quantitative wind flow representations by simulate in the whole space and not based only on measurement points (Reiter, 2010). Like SimScale, applications, or online platform-based, various software provides the CFD simulations, among other computeraided engineering-related urban matter. To have instant feedback on the design in terms of wind comfort and solar radiation, and sunlight hours, the Austrian Institute of Technology brought two different domains together in practice: Deep Learning (DL) and Computational Environmental Design (CED) workflows. This approach makes it possible to predict pedestrian comfort instantly and keep it in the loop while designing.

Economic Sustainability Measuring the economic sustainability of a district is not novel. Starting with the Millennium Development Goals MDGs OF 2000-2015 and now with the Sustainable Development Goals SDGs, there is a massive improvement and progress in the economic resilience of places (UnitedNations, 2020). Economic sustainability was a highlight in the Urban Agenda by fostering the reduction of economic inequalities and promoting economic growth. Despite simple checklists, the EU has made several approaches on how to quantify economic activity. Leading indicators are the unemployment rate and economic growth. Measure forms of the indicator Unemployment could be Underemployment, employment, unemployment rates, Percentage of green jobs in the local economy, Average professional education years of the labor force. In contrast, example measures of Economic growth is the Annual GDP growth rate, Annual GNP growth rate, Net Export Growth rates (% increase of country's total exports minus the value of its total imports per annum; Foreign Direct Investments (Capital/Earnings accrued from listed FDI's per annum) (EuropeanCommission, 2018). 22

Jointly among demographic data, to understand the economic potential of a space, it requires to understand the interdependence between economically involved elements. Example: the rent price of a place in a specific location, the costs of the investment, its potential value and income, the location of the target group costumers. Those and other relevant data can help build a dynamic model and see the potential trade-offs and investment strategies. In this aspect, the time aspect becomes an essential component of space performance; a simulation-based method could help understand the performance of the space at a specific time in the future. However, involved parameters in measuring economic potential, in this case in a newly generated city demographic data such as unemployment, or location-based GDP, are not there yet, so we select measurable indicators directly related to the building geometries their land use as well. In terms of spatial sustainability, Hillier (2009) indicates that correlating accessibility values on tax bands and the center's location can assist in measuring economic sustainability. Later on, Long (2017) conducted an extensive study using open data regarding urban economic vitality. He correlated several physical properties with spatial data on economic activities; the definition's involved parameters were the no. of amenities, mix-use index, population density, and distance-related metrics (Long 2017 in (Düring, 2019)). Similar studies based on the morphological elements of the urban layout have been developed as tools to read or understand the spatial properties and urbanity making in urban design. The urban and mixed areas are considered more heterogeneous, aiming to provide vitality in urban space. Land use diversity is a topic that has been tackled in the field of urban planning among urban vitality, urbanity, and many layers of urban potentials. “Urban vitality is the synergism of a sizeable number of varied and somewhat unique, commercial and experimental opportunities, and a relatively dense and socially heterogeneous pedestrian population, which animates certain city areas, almost continuously, throughout each day and evening.” (Maas, 1984).

According to Paul MAAS and his theories about urban vitality, the key factor that makes a space vital is the heterogeneity, either land use or user-based. However, each layer of diversity in space is interrelated and interdependent from each other. Simultaneously, specific land use might be necessary for one particular group of people; a totally different land use might be a resource for another target group. In Le Corbusier's 'The City of Tomorrow and its Planning,' where Le Corbusier criticizes the European cities because of their chaos, he focuses on having diverse town users—defining the groups of 'city-users' led him to form a practical program in town planning. By now, metrics that appear to be correlated with economic vitality can be considered: mixed-use index (MXI); potential number of new business; potential number of job places; accessibility to jobs and services (% of amenities and work or services use on highly accessible streets); local retail locations; land value; the construction costs of the design proposal; etc. To calculate the construction costs are already several tools, based on meter square, cubic meters. One of the primary references used in Austria is the BKI (Das Baukosteninformationszentrum Deutscher Architektenkammern). However, these metrics 23

are not used in dynamic models. In other words, they just track the current stage of the design. Karimi, 2012, states that agent-based models, analytical models, and economic models, cannot be integrated easily into the design process. So, any suggested framework can still be experimental and further adapted in the future. One of the urban elements that can be integrated into a dynamic framework is the street network; the accessible streets might have a higher potential to be frequented by customers. Another design parameter directly related to the economic performance is the land use, including the mix-use index. Academic experimental research based on Hillier's movement economy theories, done by Bielik et al. (2019), evolves the configurational properties creating an endless cycle by simulating the pedestrian movement flows and accordingly the land use distribution. The study is based on the space syntax theories, which states that the betweenness centrality influences the pedestrian movement flows. These movements can attract a special type of land uses (Bielik et al.,2019). So, the economic dimension can be focused more on how the cities work rather than how they look. As a result, the street network, land use, density, and costs can be considered involved parameters in a hypothetical dynamic urban economic model.

2.2 Urban Configurations There are several spatial analysis methods to understand the spatial configurations and properties of the built environment. In urban design literature, urban element refers to the features that combined up create a city fabric. Respectively they are the physical buildings itself, the groups of buildings, the space between them, and it becomes meaningful when combined in a specific rule with another urban element (Richthofen, et al., 2018). They are also tools to read an urban form; in this sense, urban density and street networks are elements that contribute a lot in understanding a city. Density has been relevant metrics in urban development, starting with the small settlements' growth until their evolutions into villages or cities. With the machine-age, many planners were elaborated on the topic of compact cities and a specific density, one of them being Le Corbusier. In the book, 'The Radiant City', Le Corbusier et al. (1967) state that cities need a specific density for the machine-age man, which would provide short travel time between housing, jobs, and other attractive locations in the city. Even though density has played an essential role in the urban design sphere, its definition has varied considerably in the last 60 years (Nes et al., 2012). Spacematrix enlightens the design world by setting the density as a

A point type, low rise B street type, low rise C block type, low rise D street type, mid rise E block type, mid rise F hybrid point/street type, high rise

24Figure 2: Types of urban areas in the FSI‐GSI plane of the Spacematrix (Berghauser Pont and Haupt 2010)

correlation of the density and the built mass. Spacematrix is a tool to analyze the built environment's spatial configurations, focusing on measuring various types of urban density, floor area ratio (FAR), the ground space index (GSI), and network density. Fig.2 based on empirical counting in Berlin, Barcelona, and the Netherlands, visualizes where different types of urban form are located in the space matrix plane based on their FSI-GSI and OSR properties. Different building typologies corresponding to their density configurations are distributed in the graph, and clusters of points can be seen. The buildings with low GSI and high FAR correspond to post-war buildings, inspired by Le Corbusier's La Ville Radieuse design principles (Nes et al., 2012). Ye et al. (2013), applies the spacematrix method as well to read the city and categorize the building structures into different categories of "urbanity," but using only two indicators, the floor area ratio (FAR) and the ground space index (GSI). Furthermore, Jan Gehl (2010) states that urban density is a crucial factor in understanding how cities function. With Covid 19, the situation faced in 2020, the debate about whether high dens areas and compact cities are the goals of urban planners is quite discussable. To conclude, defining the specific density is seen as an essential input parameter for an urban form.

Figure 3: Street Network Orientation in major US cities by Boeing (2019)

Spacematrix indicates that the density of the street network assists in understanding the urban form too. Boeing (2019) stated that regardless of the street network design on a topdown approach or it evolved organically thought time, their configurations and orientations help define the spatial logic and spatial order of the city. Furthermore, he indicates that a way to read and understand the architecture and morphology of a city pattern is to explore the orientation of the cities' streets. Boeing studied the orientation of the street nodes in several cities in the US (see fig.3) and represented them by a polar histogram. The graphs make it very visible to spot the gridded street networks like Atlanta, Buffalo, Chicago, etc. and the not gridded ones like Boston or Charlotte. Boeing indicates that the street network orientations assist in understanding the histories of cities' urban development and evaluating the current network system so researchers and planners could be critical and explore further alternatives and come up with new infrastructure proposals. Knowing the number and the corresponding length of streets that are oriented in a particular direction could assist not only in reading the urban morphology but also in correlation with performance indicators such as connectivity or wind comfort. But this is part of the hypothesis and needs to be tested.

2.3 Generative Design A large number of design options are a stronger foundation, where we have to choose the best perfuming one after quantifying their qualities. However, drawing many design options might be either time consuming, or it required many people working on the generation of urban layouts. In this research, we want to have many designs that vary from each other, explore the geometrical solution space, and select the best performing designs. For this reason, a computational generative design method is used. The framework of generative design puts together the digital computation and the creativity of humans. This cognitive computation of urban design makes it possible to manage the complexity of urban systems, navigating the trade-offs, and structuring and operationalizing data mathematically, making the process transparent and untestable for all stakeholders. Kean Walmsley, a platform architect and working for Autodesk Research, states that generative design: 'involves the integration of a rule-based geometric system, a series of measurable goals, and a system for automatically generating, evaluating, and evolving a very large number of design options (Walmsley, 2017).' The generative design framework is consisted out of three components: generate, evaluate, evolve. The generated part has to do with the generation of design solution space, possible geometrical solution. The number of potential solutions shall be broad, allowing the user to select among the options, even more for further optimization processes. The second part is the evaluation, wherefrom specific challenges, measurable indicators can be developed, and the design is evaluation in those specific criteria. And the third component evolves generations of designs through evolutionary computation.

One of the framework's main advantages is that it is flexible, and it is practical to be used as a communication tool in participatory planning. The participation of citizens and other stakeholders in the design process became more significant every day; therefore, in the upcoming chapter, we will explore the topic of co-designing and public participation.

2.4 Public Participation in Urban Design Often, the term urbanization came across as an authoritarian word, and perhaps the process of urbanization happened authoritatively sometimes too. A top-down approach proved not to satisfy the needs of citizens. Therefore, citizens' involvement in the planning processes is a crucial factor in long-term social sustainability when it comes to urban development. In this study, it is given significant importance to the public participatory possibility in the generative design method and the evaluations and explorations of the design options. In other words, the co-design and evaluation with stakeholders, visualizing performance data in real-time, allows better communication between stakeholders: planners, designers, citizens, and other relevant stakeholders in the project. To see the community's participation in planning and understand its impact in the world of urban design, we need to go back to history. In the 19th century, we can see the anarchist origins of community development and involvements in decision-making processes. Two anarchist intellectuals that contributed to this movement were: Elisee Reclus, a geographer, who believes that 'authoritarian, power-based institutions of society conspired against human freedom and nature; and the Russian Prince, Peter Kropotkin who ends in jail for advocating anarcho-communism, he planned self-sufficient cities, produced from many participators (Walters, 2007). Local communities' involvement in designing processes movement is followed by other intellectuals, such as Ebenezer Howard, with his concept of garden cities, where the idea is to develop self-contained communities surrounded by "greenbelts," containing proportionate areas of residences, industry, and agriculture. These cities would be self-governed and have community identity and will be managed locally. He puts the human as the center of the design, between three magnets, Town, Country, and Town Country (Howard, 1974). When talking about public participation, it is inevitable to escape Jane Jacobs and the neighborhood topic. Jacobs's ideology against monolithic plans influenced many citizens to understand their power and get involved in urban renewal projects. More specifically, she states that "neighborhood is any region in which can be organized enough to act as an independent polity and campaign against the larger government in which it is nested." (Jacobs, 1992). However, community participation has many ladders, and building community trust was another challenge that was faced, and it is a big challenge nowadays. Relph Erskine made an immense contribution in this field who; during the redevelopment in Byker, R. Erskine, and his Partner Gracie lived on-site for many years in a flat above the drawing office set up in an old corner store (prev. funeral parlor). They kept a daybook of visitors, complaints, and solutions and tried to solve all the design requirements. The best part of this was that people felt heart. Planning and designing is not the only way to involve citizens in designing processes; Christopher Alexander, for example, wrote a practical book, "A Pattern Language," where all the patterns describe a problem, and there is a solution offered for each of them, a solution that ordinary people can use to design or improve their neighborhood, town, etc. Given the aforementioned, it is still a vivid challenge to co-design without facing any issue or misleading euphemisms in Participatory Design. In the digital era, the need to engage the 27

citizens in urban planning is becoming higher. Therefore the design tools, among other innovations, are being adapted to engage non-experts in the design and evaluation phases of projects. This research suggests applying methods and tools to make the data visual, practical, and understandable for each stakeholder. In this context by data is meant the urban design itself and the performance indicators. The generative model discussed in this research and the suggested evaluation methods foster all stakeholders' involvement and participation, either in the design or evaluation process. The used methodology is practical to be used as a discussion table in meetings with communities, using augmented reality, interactive tables where all the workflow can be elaborated together. After each design, the user can see its performance and better understand how it might work and what the tradeoffs are in each case.

3. Methodology: A framework for quantifying urban sustainability through social, environmental and economic dimensions Among many existing tools and approaches towards quantifying urban form performance, the suggested toolbox will combine the ones that were determined as relevant for this research domain, respectively. Such tools are directly conducted from the urban form. The main benefit of the framework is that spotting the potential segregated areas provides higher social equity and social cohesion, which in parallel offers better mobility, higher economic potential and urban vitality, and pedestrian comfort in urban space. The first step of the study is to define a design space and a generative model that can explore it. The framework aims to give an overall score and overview of the design options through a generative design, graph network analysis, dynamic urban simulations, and microclimate climate predictions. It will show the impact and the precise number, the success factor for each category. Measuring an existing site's performance can be done efficiently, mainly if open source GIS data is provided. However, to understand the impact that a new urban space design might have, a computational approach is required, where the framework assists in understanding the trade-offs. This live model can be an essential tool in participatory design meetings, where all stakeholders can have an adequate debate on equal ground. At the same time, it is easy to explore and elaborate on the design among performance metrics. To apply this framework is used a generative method. The generative method serves not only to generate cases and several urban forms but also to find the performance domain bounds in each measured category. Therefore, the initial step of the framework is the generative urban model, followed by developing indicators to measure the performance in terms of social, environmental, and economic sustainability. A further step is the exploration of performance space and data analysis.

3.1 Related Tools The framework is implemented using several existing tools and programs. While GIS data is prepared using QGis, the framework is implemented within the Rhino/Grasshopper5 environment. The latter is a visual programming language specialized on geometric and spatial computations. Grasshopper provides the possibility to build the generative algorithms of the urban design, and performance simulations, as well as data analytics. To build the generative model as well as for data correlation part is used the grasshopper plug in decoding spaces toolbox6. To visualise the performance data is used flourish.studio7 which is an online data visualization tool.

While Rhinoceros is 3D computer graphics and computer-aided design application software, Grasshopper is the visual programming language and environment that runs within it. 6 Decoding Spaces Toolbox is a collection of analytical and generative components for algorithmic architectural and urban planning for more info https://toolbox.decodingspaces.net/ 7 A data visualization and storytelling tool founded by Clark and Houston. https://flourish.studio/

3.2 Architecture of the Framework The framework is consisted out of three main parts: 1. Generative urban model 2. Performance Simulation 3. Data Evaluation - Solution Space Exploration The detailed content of each part of the flow chart is explained chronologically in the following chapters.

Figure 4: 30The architecture of the framework presented in a flowchart. Source: Author

ASA Aggregated social accessibility % of favorable and unfavorable places within the analysis grid

3.3 Data Structure Data type in the whole framework is mostly within the grasshopper environment. It works with basic spatial data objects such as streets, plots, buildings, and regular analysis grid to relate these different spatial units to one another. Streets: To lay a foundation for network analysis this study represents the streets and roads in a street network. Graph theory8, as a tool to model pairwise relations between objects, is a suitable method to analyse the street networks. In the scope of this research simulations that will be applied are closeness centrality and betweenness centrality, and they can only be computed with graphs. All the streets in the designed workflow are converted into graph. Regardless the initial input of the street network (rectangle, curve or line) using decoding spaces components it is turned into a line segment. The graph is built by interpreting every segment as a node and as edges all the connections to other segments – endpoints of segments respectively (Bielik, et al., 2012). Buildings: The solid geometries such buildings, are represented as volumes. In the frame of graph theory, to find the shortest paths, buildings are assigned to the closest street segment so the computation of distances between buildings is simplified as distances between street segments Analysis Grid: For each analysis, different results and data representations are generated. While the social dimension of sustainability lies on the street network, it is a node-based; the environmental dimension is assessed for any point in space covering all the unbuild space; while the economic one is a combined between street network nodes and analysis grid. As a result, they need to be translated in the same format to operate with them. Therefore as a method to work with the generated data, it is used an imaginary rectangular grid where every grid cell will contain a value from the corresponding metrics and categories. This grid serves as a two-dimensional spatial configuration to aggregate data between different spatial units. To aggregate the nodes or points in the analysis grid, the nearest neighbor is detected. The type of analysis mostly defines how far we search for neighbors, so the radius is elaborated individually for each performance indicator. By assigning land use or construction costs in the buildings, the mapping to the analysis grid was done using subsample form, subsampling. The grid is relatively fine-grained, but it is flexible and can change the size depending on the location that it is applied. An adequate grid size where we can operate with the suggested measurable indicator's performance data uses a 5 x 5 m cell size grid. It works fast and still saves the data accuracy.

3.4 Generating Urban Designs To be able to explore the geometrical solution space and select the best performing design, the research employs a generative design method. This method enabled the possibility to explore different complex urban forms. In parametric modelling, the geometrical output comes from the algorithmic process and specific input parameters (Woodbury, 2010). Since the urban forms, generated from the model, will serve as a basis to run analysis and try to find relation with the input parameters and output performance indicator, the model's goal was to output various clearer urban forms. 8

Graph Theory is a study of graph, where structures used to model pairwise relations between objects. It is consisted out of vertices (or nodes) and edges. In the urban context, while the vertices can be the locations like buildings or points of interest, the edges represent the connection between the nodes (Hillier & Hanson, 1984).

Types of inputs: The parametric model is structured along three steps. It starts with generating the street network; as a next step based on ground space index, the building footprints are generated, and the last step mostly depended on the floor area ratio is the step that generates the building volumes. The combined given FAR and GSI leads also to a block typology, which we will elaborate further in this research. The given inputs in the model are grouped into three categories: geometric, numeric, and regulations. As geometric input can be given the site's boundary, the existing street network, and other geometrical context information if available. As numerical input is the floor area ratio and the ground space index and the street network rotation angle. Regulations input data, is location specific and it contains the setbacks and offsets of buildings in response to the plot. Logically, regulation data could be numeric at the same time. Other inputs that can be modified but are not used further for correlation data are: average building length, depth and depth variation, and the spacing between them; in the level of street network generation: grid type and size, and thresholds in minimum and maximum block size.

The street network generation: There are two methods tested to generate the street network. One of them follows the organic existing patterns of the existing street structure. It tries to align with the existing streets and generated organic super blocks. Afterwards the algorithm checks the size of the organic blocks and subdivides the large ones into regular rectangles or forms. To achieve this form decoding spaces components are used. The main streets generated in the first iteration of the algorithm- the organic ones, can be rotated, and this act allows us to have a vast amount of street network variations. The large amount of options can be used to analyse data based on the performance of the options, as well chose the one that performs the best and proceed further with design. The second option to generate the street network, is an alternative solution. I can be applied if the location where we apply the parametric model is more rectangular and has a grid structure. It is also influenced by the existing street network, the given boundary and the minimum and maximum thresholds. In the scope of the research the first approach is used to generate the street network. The street network is one of the key parameters for the spatial analysis, and it has an impact on the overall performance of the design. In this research, street network orientation is used to understand the form of urban layout and to test its correlation with the performance indicators. From each design option a representative indicator is derived as case representative. The method to extract this number is described in the following paragraph.

Figure 5: Street network generation steps. Source: Author.

Street Network Orientation Data Aggregation The method to get the representative indicator of each designed street network is as it follows: The direction of each street node in the generated street network is defined. They are classified into four categories depending on their closest direction:

   

North-South Northwest –Southeast West-East Southwest-Northeast

Figure 6a: Compass representing 4 orientation categories. Source: Author.

As a next step, the overall length of the all streets in each orientation is summed up. The number ratio of the list is expressed in percentage. In conclusion we have the amount of street length in each direction for every design option from the generative model.

Figure 6b: Street networks obtained from the generative design, visualised based on the orientation category. Source: Author.

Building- Geometries Generation The density metrics have been always relevant in the urban development and defining the certain density is seen as an important input for a generative urban model. There are two main steps of the building generation algorithm, footprint generation and height generation. The footprint generation part gets as input the plot boundaries; GSI, setback distance; depth and depth variation; optionally also areas where no split in the building can happen, with the possibility to adjust this position geometrically can be given as input. The part of the algorithm outputs the buildable area. There are two categories on how buildings can be generated according to the size of the plot. The first approach is: the buildings are created from a centreline with an offset function, which reminds us on a block typology. This would be the initial state of the generation. This line later on gets an width and becomes a surface which will be the footprint of the building. 34

Depending on the amount of built area that we want to have on plot – depending on the ground space index, the line gets split into parts. So only the amount of the surface that reaches the aimed GSI, turns into a building footprint. The height of the building is extracted from the given FAR. Alternatively, if the plot cannot reach a certain given GSI, the generation process becomes different: only with an offset function, it will create a base on the plot depending on the value of GSI. An on top of that regular towers will be extruded based on the aimed FAR. So, the Ground Space Index (GSI) or coverage, which demonstrates the relationship between built and non-built space; and FAR - Floor Area Ratio, the ratio of built area to the lot area which reflects the building intensity independently of the programmatic composition, play a massive role in the generation of the building footprint (Pont & Haupt, 2009). By defining it, we can get different building typologies. In the fig.7 are given some examples generated from the parametric model. The main input parameters to generate these urban layouts are the GSI and FAR.

GSI 0.9 FAR 1.9

GSI 0.5 FAR 1.0

GSI 0.2 FAR 1.0

GSI 0.25 FAR 3.0

Figure 5: Urban Layout generation based on the given FAR and GSI. Source: Author

FAR and GSI Data Aggregation: After generation the design option, the FAR and GSI were recalculated to validate the input. Additionally, to map the FAR and GSI information in the analysis grid, we calculate the FAR and GSI radius based for each point in the analysis grid. To aggregate this numbers into single values the average FAR and GSI within the plot is calculated. These two numbers are used further for the correlations with the performance indicator data. 35

Land use Distribution and MXI Calculation

housing

workplaces

amenities

Figure 6: Land use visualisation on the site. Source: Author.

There is a clear impact of the land use patterns on the liveability of the neighbourhoods, even on health of the residents (Sallis JF, 2009).The order of the neighbourhood and the land use patterns emerge naturally from the interaction of the citizens and planners with the urban space (Lenormand, et al., 2015). So, in a feedback loop the function or usage of a land is a key actor when it comes to the experience of residents and visitors as users of the city (Humphries, 2012). Users can have a big impact in the redistribution of the land use and reallocation of functions in a city, and planners as well. For example, the planners plan a school in a certain neighbourhood; as result many students are attracted to that area and are around; in parallel cafes and bar targeting the student age as main costumer can be opened by private investors. In living structures or cities as complex network of actors, the need of each actor affects the whole network system. In this workflow, the land use information is used in dynamic model where the pedestrian movement patterns, and the potential land use distribution happens in an iterative process within a loop. Therefore, we suggest a random initial land use distributed randomly to see how fast this will change on the dynamic model. Three main land uses were distributed Housing, Work places and Amenities9. While the dwelling includes the various residential shelters such as apartments, single houses or townhouses; work places include every type of offices, labs or factories; and the amenities include public facilities such as school and universities as well as all kind of commercial facilities such as shopping and retail. The mix used index (MXI) served to find the value of how mixed every area in the space is, which in the thesis in considered as initial attractiveness for destinations.

Housing, Work places and Amenities, as the three main land uses were chosen based on the Mix Use Index (MXI) metrics of Van den Hoek.

Mix Use Index (MXI) Data Aggregation The MXI information is aggregated in the analysis grid and the street network. The process for aggregating it in the analysis grid is as stated: from each point in the analysis grid in a 60 meters radius (120m diameter), it was searched for the closed five buildings. The 120m diameter was used because the framework is designing to be applied in an urban settlement., where facilities are closer to each other size of the buildings' footprints is relatively small. If a residential building average footprint is between 600-1000, in a 60m radius can be found up to 10 buildings. However, there is also spacing between them, so we assume that a lower number of buildings can be detected. Up to five buildings will be enough to show the dominant land use; still, each analysis point in the grid gives local information for the surrounding buildings. It worth mentioning that based on the location where the framework has applied, the radius and no. of buildings can be modified. In case of parks, greenery, or empty areas, no building was found in the 60m radius. As operationalizable number, we calculated each function's amount and checked the absolute difference from the 30%. If, from an analysis point, three buildings are located in a 60m radius, and each of them has a different function, the difference would be 0. But if this is not the case, the number would be higher. This score indicates the MXI in the scope of the research. However, the performance was mapped in a normalised list from 0 to 1 (a negation was calculated for each value), where 1 is the spot where the area is a multifunctional area, and between 0.30-0.70, the area is a bifunctional area and values below 0.30 indicate that the space is monofunctional. The MXI information is aggregated similarly in the street network, too, using the same radius.

Input Parameters Outline Summary:

Description

Floor Area Ratio (FAR)

Ground Space Index (GSI)

Street Network Orientation

The ratio of built area to the lot area. It reflects the building intensity independently of the programmatic composition (Pont & Haupt, 2009).

The coverage, which demonstrates the relationship between built and non-built space.

the orientation of the streets of the cities

Involved Parameter s

Data Aggregation Method

Output

Building Geometries

Plot based FAR is predefined in the input parameter of the generative model. 1. Re-calculate the FAR from each point in the analysis grid in a certain radius (radius based) 2. Calculate the average FAR from all the points within the analysis grid

The average FAR

Building Geometries

The plot based GSI is initially predefined in the input parameter of the generative design. 1. Re-calculate the GSI from each point in the analysis grid in a certain radius (radius based) 2. Calculate the average GSI from all the points within the analysis grid

The average GSI

Street Network

1. Categorize the orientation of each street node into the closest orientation to: North-South, Northwest – Southeast, West-East, Southwest-Northeast. 2.Get the amount of street length in each direction.

4 numbers that represent the % of the amount of street lengths in each category

Table 1: Outline of the Input Parameters, and the methods how the data was aggregated. Source: Author.

3.5

Performance Indicators: Social, Environmental and Economic Dimensions

aggregated social indicator ASA

This chapter describes how each performance indicator is calculated. The indicators are categorised in the three pillars of sustainability: social, environmental and economic. The developed indicators and the workflow derive from the literature review related with social equity and segregation; comfort in terms of weather as well as economic growth and potential of space. The outline of the indicators of the framework e shown and explained in detail below (see table 2). Following this table, a subchapter is dedicated to each performance indicator.

Table 2: Outline of the performance indicators and the methods how the data was aggregated. Source: Author.

3.5.1 Social Sustainability Evaluation Method This section aims to quantify the performance of an urban layout on the social pillar of sustainability, on community formation with a focus on social equities and social integration.

The social integration measurement starts with examining problems in a location. It aims to detect unfavourited areas, that don’t provide equal access to liveable streets and spaces with high potential of interaction. This indicates that the approach should be based on a larger scale than the site context, and it should consider the long-lasting aspect of space. Street are urban elements that can carry this function, since they are the backbone where communities are built on (Marshall, Garrick, & Marshall, 2014). To lay a foundation for network analysis this study represents the streets and roads in a street network. Given the aforementioned, including the reasoning in the literature review, the method that this research will apply to measure the social potential of the space, is based on the street network configurations. The method is adapted from the Aggregated Social Accessibility (ASA), which is a graphconfigurational method based on street network, that estimates the pedestrian movement flows and measures the accessibility to these locations, as the liveable parts of the city. The method has two main steps: 1. The estimation of the pedestrian movement flows using betweenness centrality which measures the numbers of times a node is passed along the shortest path between two other nodes (Freeman, 1977). The nodes that have a higher number of times they were used as a bridge are considered as the lively parts, and the space where people have more potential to interact with each other. It is defined as (1): which, on the shortest paths (spij), calculates how many times a node K is passed ( spij(k)) by connecting each node with all other nodes in the street network (Newman, 2010). Since this thesis uses weight to calculate the betweenness, the equation (2) is applied (where w indicates the initial weight). 2. Measuring the accessibility to the lively parts of the city using closeness centrality, which taking into consideration the attractivity of a destination-in this case the lively partcalculates the proportion of accumulated destination weights to weighted path distance (Bielik, et al., 2018). It is defined as: (3) where C indicates the Closeness centrality, di,j the distance between node i and j. Since the thesis uses the weighted closeness centrality the equation (4) corresponds to the approach that is implemented, where w indicates the weight.

(1)

(2)

(3)

(4)

Based on the study of Bielik et al. (2018), in this study to measure long term social sustainability betweenness and closeness centrality are used. In difference from the ASA10 approach this research takes into calculation the population number as origin weight. The population density is estimated from the building volumes generated from the parametric model. We assumed that 50m2 is an adequate space per person. Regarding the radius within which shortest paths are searched is based on walkable distance, similarly with the approach of morphocode. Morphcode refers to the pedestrian shed while exploring the five minutes’ walk, which indicates the catchment area within a walkable distance which is about 600 meters. The 5-minute distance is stated as a considerable distance to walk (morphocode, 2018) also in the neighbourhood unit by Clararence Perry, who was involved in communities based social activists. Therefore, the 600m meters as a five min walk distance is used as a radius. The same radius is used in the calculation of the betweenness centrality, and in the next step in the closeness centrality. The approach takes as input the street network and the estimated population density, and as output we get the value on how favourited each street network in terms of access to lively spots within 600m is. This information was aggregated from the street network to the analysis grid and categorised further.

Social Sustainability Data Aggregation to the Analysis Grid Regarding the social sustainability aggregation several approaches were tested, and the methods are elaborated and reasons why we proceed with the chosen one were given (see table 2). As a common step to get the overall performance from the spatial network centrality analysis the framework uses a large extended boundary of the location, to avoid the edge effect11. The performance data of the street within the analysis grid, and on the boundary, edges is used to aggregate the information in the analysis grid. In all three tested approaches, each street contains a performance value that needs to be aggregated to the grid. To have a more accurate aggregation in the analysis grid, multiple points were populated in the street node, and the performance value for each was inherited from the street node. (in the implementation, each 10 meter a point was generated). The third approach was used to proceed further with the data, since it gives more accurate results, and due to the interpolation, it gives a more cohesive result. The outcome global data from all generated cases will serve to categorise the local performance data of each design option.

Aggregated Social Accessibility (ASA) a graph-configurational method based on street network configurations, which measures the accessibility to the estimation of the pedestrian movement flows 11 Edge Effect: Gil, 2015 in the Proceedings of the 10th International Space Syntax Symposium examines the ‘edge effects’, the sensitivity of spatial network analysis to boundary conditions. The study is related with the general sensitivity or robustness of centrality measures of the street network. Based on experiments and empirical data set of real-world street network, it suggests a correct way on defining the boundary in a form that is significant to the research question and give accurate results in the focus area. More precisely it suggests abandoning global analysis and focus on egocentric analysis. To eliminate the edge effect, the research suggests that a local radius can be used, or by selecting relevant streets nodes only up to a certain depth and create a certain catchments area is a larger applied context in spatial network (Gil, 2015).

The first approach that was tested, is from every grid cell, in a radius of 100 meters, get the average of the twenty highest values. It results with a more coherent gradient, and the transition is smoother. However, it based on the results, it can be indicated that a lot of information gets lost, and the overall score is high since the list is sorted and checks only the highest values in the surrounding. If there is no point in a 100m radius, by default the grid gets a 0 value.

The second approach is again aggregating the information from the nodes to the grid, but this time we use a larger radius to search for values and we use only the values of the five closest points to the grid cell. After getting the sum of the value of the five closest points, the overall values are normalised between 0 and 1. Again to ensure that every cell gets a value, if no point is found in a 150m radius, a 0 value is given to the cell.

On the third approach from every centre of the cell, we search for points in a 80m radius only. Six closest points are used, and the average of those values was aggregated in the cell. In this approach the replacements of null values are more significant, because there are spaces where no streets are in a 60m distance. The advantage of this approach is that the results are more accurate, and the data in the grid is interpolated, so the results on the heatmap looks smoother and cohesive.

Table 3: Social Data Aggregation Approaches. Source: Author.

Categorisation of the social dimension performance data -Jenks Natural Breaks12 As the study uses a comparative method to choose the best performing option, it is considered that the scope of performance is affected by the surroundings and the existing structure. Especially when it comes to measuring it by the street network, which doesn’t change a lot and not very often in real life, and the domain or scope of the solution space can be tight. Therefore a general value to compare might not be helpful, so we need to select the best performing solution out of the large number of designed generated from the parametric model. To find the general domain of the data, and use the proper categorization for all cases, the social sustainability performance value of hundreds of cases of the generative model was simulated. The global data from the simulation of all cases is categories used Jenks natural breaks. This method was chosen to minimise each class’s average deviation from the class’s mean value, and at the same time maximising each class’s deviation from the mean values of other groups (Yu Ye, 2013).

Jenks Natural Breaks is a range finder algorithm. While ‘natural breaks’ are methods to split up ranges, by minimizing the variation within the range; Jenk assigns data to the groups, so that distances within the group are minimized. The method is done in an iterative process, starting with the calculation of the sum of squared deviations from the class means - SDCM. Afterwards the SDAM the sum of squared deviations from the array means is calculated. And the third step are the actions of the units they move from classes with larger SDCM to one with lower SDCM. This iterative process is repeated until the sum of the within class deviations reaches a minimal value (Jenks, 1967).

Figure 7: Workflow of operationalizing the social performance of the design options. Source: Author.

To acquire the general categorization, we used all the dataset generated the performance of hundreds of cases from the performance space of the generative model. As result, from the global results the overall domain is found, and the bounds of the five categories will be defines. The categories will be used to visualise the performance data in a histogram, and to see the distribution of data for each single case. Based on the four natural break classification method, five subcategories were defined. The first category includes all the spaces that are already located near a street with potential of interaction. The second category also is a safe area where the attractive streets are reached easily. The third category includes all the parts where the spaces with interaction potential can be reached in 5 minutes. The fourth and fifth category are the ones which need to be reconsidered since the basic and fundamental needs for a social sustainable development are not met. This line is also used a threshold to define the good and bad performing spots. In this research they are named favourable and unfavourable locations. Despite the numerical values, from this information we can also spot the location on map where the spatial conditions are not balanced and equality in not reached in social level, therefore we can conclude that physical interventions in the area are needed.

Data Aggregation: To have an overall comparable number, that we can use as representative for each case we used the information from the histogram and the data distribution according the categorisation. The sum of the percentages above the threshold which logically is a number from 0 to 100, is divided by 10, so the performance domain can vary from 0 to 10. A low value means the site doesn’t offer equal potential for social interaction and integration, while a high value indicated the opposite – the design option offers a wellintegrated site. This social dimension of sustainability, ASA, indicates the social interaction potential; in exact terms: the residents’ accessibility to the liveable parts of the district. It reflects the score of the design in social dimension, respectively it is based on the average access to liveable spots from each analysis point within the grid. 43

3.5.2 Environmental Sustainability Evaluation Method This section aims to quantify the performance of an urban layout on the environmental pillar of sustainability, based on environmental metrics that highly depend on the urban form.

Based on Reiter’s (2010), statement that the geo-located wind speeds and solar radiation are the only microclimatic parameters that depend widely on urban planning, therefore in the category of environmental sustainability this research suggest application methods to measure wind comfort and sunlight hours as the main environmental indicators. To measure and analyse the performance of environmental indicators, instead of using conventional simulation methods to compute microclimate analyses, this research uses the deep learning models for microclimate urban analysis of the “Intelligent Framework for Resilient Design -inFraReD”, developed by the City Intelligence Lab at the Austrian Institute of Technology. The CIL employs a novel deep learning model (DL) 13, which reduces computation time by the order of several magnitudes, which allows the user of this models to go back and forth between design and performance evaluation (Düring, Chronis, & König, 2020). The microclimate prediction models of CIL were trained with simulations of different locations, which made the comparison of the models in different geographical locations possible. This includes different climate conditions related to solar radiation, different building typologies and urban morphologies (Galanos et al., 2019). Since the results are encoded in a bitmap, there is a process to get them as values in the structure of the analysis grid. The bitmaps are sampled, and the output prediction values are converted to HSV (hue, saturation, value) where each colour represents a wind speed. This value, the wind factor, is normalised within 0 and 1; as a next step the csv file is written. The csv file can be imported easily in GH and allows the user to proceed further with design or evaluate in the same environment-GH- as urban form generation. So, in other words the CIL has development GH scripts with costume Python components to deploy the models and the results of real time predictions for any generated urban form within grasshopper. A detailed overview of their methodology can be found in their paper ‘Best of both worlds – using computational design and deep learning for real-time urban performance evaluation’ (Galanos et al., 2019). The main advantage of their approach is to reduction of computation time, and instant feedback on the performance within fraction of seconds instead of minutes of even hours. Another reason why it is suggested to be used in this research is that it can all run within grasshopper, and stay alongside with the generative model, and other relevant analysis or simulations.

Based on the work of Isola et al. (2017), titled Image-to-Image Translation with Conditional Adversarial Networks’, presented at the IEEE Conference on Computer Vision and Pattern Recognition The approach is based on Generative Adversarial Networks (GAN) which trains a conditional generative network and it used for imageto-image translation tasks. Using conditional adversarial networks as a general-purpose deep learning model, which generates a corresponding output image depending on an input image. By now there are already many applications of this method, and it has wide applicability domain, e.g.: colorizing images, reconstructing objects from edges of maps, manipulating photos from label maps, or microclimate predictions.

Wind Comfort To be able to predict wind comfort and evaluate it afterwards, since it is location based local meteorological data, local wind conditions and specific comfort criteria are required. The workflow based on the work of the City Intelligence Lab (CIL), allows us to set as input the wind direction and wind speed. As wind direction input was used the south, and the windspeed is set to an average stormy day in Vienna. As result it shows the performance categorised in six comfort level groups based on the Lawson Criteria’s14. The Lawson Criteria has been accepted and adopted for the elaboration and evaluation of the wind environment associated with buildings in many locations (Fadl & Karadelis, 2013). The comfort level categories are as it follows: sitting long where the windspeed is higher than 2.5 m/s, sitting short, walking slow, walking fast, uncomfortable zones and dangerous zones. The percentage of the uncomfortable zones where the windspeed is higher than 8m/s and the dangerous ones, where the windspeed is higher than 15m/s, are summed up, and this shows the overall % unsafe areas in the analysis canvas. Therefore, the methodology despite the numerical performance values, assists also in finding the exact location of the dangerous areas. Data Aggregation: The prediction outputs the % of unsafe area within the analysis grid. Several factors can affect the performance and mathematically the outcome can vary from 0 to 100, but assuming that there are always also comfortable sides in the area, we run all the cases and find the minimum and maximum unsafe area %data in both categories. The bounds of each case are used as source domain to remap the numbers in a new numeric domain varying from 0 to 10. The inverse value was calculated so the higher the value indicates more pedestrian comfort. This numbers will is used when comparing the designs with each other in the environmental dimension.

Figure 8: Wind Comfort Prediction, using the Intelligent Framework for Resilient Design (Infrared). Prediction by Author.

Lawson Criteria is a criterion to assess the pedestrian-level wind environment. It provides wind speed threshold values and frequency ranges for pedestrian comfort and safety based on pedestrian activities (Lawson, 1975).

Sunlight Hours Sunlight hours indicate the number of hours of direct sunlight. A performance data corresponds to each point within the analysis grid, which informs about the number of vulnerable areas lacking daily sunlight hours. Similarly with wind comfort, to increase the speed of the simulation, sunlight hours are predicted using InFraRed, which leverages state of the art neural networks. As initial step we run the sunlight hours prediction and got the results in the bitmap visualised as in fig.11. The values were extracted from the image and operationalised further. The minimum hours of sunlight that an area needs to be can vary from the purpose of the usage of that zone. In this study, based on the work of Serrano et al., (2017), published in the Science of The Total Environment journal, it is indicated that for human beings, to obtain the required vitamin D dose in winter it required 2 hours of sunlight; while for vegetation up to six hours15. Several other approaches also indicate a minimum threshold, in this study 5,5 as an urban area is used as a minimum. The % of areas under this number are considered as vulnerable areas with lack of daily sunlight hours. What could be added to this section is the sunlight hours impact on the population. As exercise, we searched around every building the performance values and based on the average value among the points we could spot which buildings are location in agglomeration with shortage of sunlight hours. The estimated population numbers are used to find the percentage of the new population that will be allocation in these buildings. An example case from the generative study in fig.12 shows that 19% of overall population (850 out of 4483 people), lives in one of the disadvantaged buildings visualised in red in the fig.12.

Figure 11: Sunlight hours prediction using Infrared. Red indicates high sunlight hours, blue indicates low daily sunlight hours.

Figure 12: Impact of lack of sunlight hours on the population. Source: Author.

Data Aggregation: After predicting the sunlight hours for each analysis point in the grid, the % of areas with lower than 5,5 sunlight hours per day was calculated (this number can change, based on the purpose of the site). The percentage is express in a decimal form. The process is repeated for each case, and one single value was output from each case. The inverse of the values is calculated, so the higher value indicates more sunlight hours per day. All this data is remapped in a domain from 0 to 10. The lower the value indicates larger the area with lack of sunlight. The higher the value indicates the better performing design in terms of sunlight hours. 15

It depends on the type of the flora. e.g.: while Kale, or Chards demand 3-4 hours of sun exposure per day, patatas need 6 hours of sunlight per day.

3.5.3

Economic Sustainability Evaluation Method

This section aims to quantify the performance of an urban layout on the economic pillar of sustainability, with a focus on urban dynamics including system sturdiness, investment worth potential and footfall potential. The goal of the economic metrics is to understand if the investments done in a certain location are sustainable and resilient. This get higher significance when the time factor is included in the equation. As economic sustainability is hardly related to investments and value of plots and their correlations in this chapter to measure the economical dimension the parameters that are were considered as significant are value of plot over time and construction prices. As elaborated in the literature review, the economic dimension can be focused more on how the cities as complex system work, rather than how they look like. The street network, the land use, density, and costs can be considered as involved parameters in a hypothetical dynamic urban economic model. This research will use two different methods to measure economic sustainability: the first one is the economic dynamic model, and the second one is the footfall potential of each location. From the economic model we get two output indicators: the sturdiness, system change indicator; and the investment worth prediction score. It worth to mention that the modelling is based on soft computing16, where the approximate results give a general overview and performance information in the early stages of design. The coming chapter will elaborate the methodology of the economic dynamic model and the footfall potential.

Economic Dynamic Model The economic dynamic model is based on a significant study regarding the city as a complex system is done by Bielik et al. (2019), which is evolving the configurational properties. The study is based on simulating a dynamic urban model, where the impact of the street nodes configurations on the evolution of the relationship between movement and land use allocation over time is elaborated. It is clearly understandable that these changes happen per a long period of time, and not all functions can be relocated, such as train stations, bridges, etc., but the theoretical hypothesis can show the influencing force of the street network on the land use attractivity locations. The approach of Bielik demonstrates how studies based on the simulation of the interaction between movement flow and land use over time, can be operationalized. The paper ‘City planning using integrated urban modelling Jeddah structure plan’, tests a centres strategy to distribute population, employment and supporting facilities along with a public transport strategy for the city (Achary A., 2017); similarly Bielik’s study in the economic level uses the pedestrian flow, and accordingly distributes the land uses but this time in an iterative process. For the movement flow and land use allocation, betweenness centrality (equation 1, page 40) and closeness centrality weighted by betweenness (equation 4, page 40) is used in a loop. Two main trajectories were examined in the Bielik’s approach, and the main difference is that the initial land use of the location is or is not taken into consideration. If yes, that is used as initial weighing for the dynamic model, 16

Soft Computing: a computing method that deals with approximate models, to solve the realistic complex problems. It is therefore tolerant of imprecision and approximations. It uses a combination of several methods including fuzzy logic, neural networks, and genetic algorithms. While soft computing uses an approximate model, and approximate reasoning techniques, conventional hard computing on the other hand uses exact models. The main advantages of using soft computing modellings are the low-cost, less data needed and faster process (Ibrahim, 2016).

if not all street segments have equal weighing as starting state. If there are assigned initial weights to the streets segments the system is considered as disrupted network. Based on the Bielik study, 28 iterations are simulated, as the experiment finds a state of equilibrium and no further change in the system was observed. The approach in this research corresponds to the disrupted network method since for the initial weightings we use existing information. Bielik concludes that there is no linear multiplier effect observed in the non-equally loadeddisturbed street network for the centrality analysis. So, in this case to obtain the analysis of movement potentials and rearrangements of the land uses a proper dynamic simulation is needed. The main inputs for the model are the street network, the MXI and population density, the output indicators elaborated in this paper are the system sturdiness and the investment worth potential. While the system sturdiness gives an overview for the whole area, the investment worth potential gives information for the local area within the analysis grid. Corresponding result for each analysis point indicating the investment worth potential are generated. The simulation of the model starts with defining the initial weighting: the origin weights and the destination weights. As next steps the dynamic models run though all 28 iterations (calculating first the betweenness centrality and afterwards the closeness centrality weighted by betweenness; the result for each street segment serves as weigh for the next iteration; this process is repeated until there is no change in the outcome numbers anymore), and all the data from each iteration of the dynamic model are saved to calculate the system change value -the sturdiness. The last state of the simulation is used to interpret the value of the plots in relation to the initial construction costs, respectively of each cell in the analysis grid.

The system change value – Sturdiness The sturdiness- system change value gives information on how robust the or resilient the design is. If in a loop the land use distribution follows the pedestrian movements, how many iterations does it take for the system to reach a static state. A higher amount of iterations indicates a more robust the system. The economic dynamic model calculates the betweenness centrality and afterwards the closeness centrality weighted by betweenness in a loop of 28 iterations. The output of every iteration is a list with values for each street segment in the map. The values of the streets indicate corresponding accessibility for each street segment. After each iteration, if the land use follows the pedestrian movement, the city gets more centralised up to a point where it stays static. As Bielik stated, there is an impact of street network configuration on movement potential and land use allocation. In this research regarding the economic robustness value, we assume that if the system reaches in a different stage very quickly the system was not robust, and the suggested design is less resilient. In this research this economic robustness is named ‘sturdiness’. To calculate the sturdiness level, we keep trap of the system change. In a similar approach with the work of Bielik, this study suggests a method to track the system change. In an altered way, this research uses relative differences, and traces f the performance after each iteration.

Figure 13: Economic Dynamic Model Outcomes after the 1st ,2nd ,3rd ,6th and 20th iteration.

The results from every sequence of the iterations are summed up, and the overall change is quantified in one single number. In other words, it is checked how many iterations does the system need to go through until it stays on a static state. This is measured on a dynamic model, with Figure 14: The system change indicator through 28 iterations 28 repeated iterations. The changes are shown in the fig.13, in the graphs. The first graph shows the sum of the centrality measures after each iteration, whereas the second chart in a sorted setup, shows the performance curve after each iteration. We assume that if the designs performance after each iteration changes fast based on the movement patterns of pedestrians, indicates that the initial design was not planned properly to match the needs of citizens and it is affected easily by pedestrian movements, so it is not resilient or sturdy. We assume that the slower the system changes the sturdier the system is. Data Aggregation: The performing numbers from all cases were remapped in a scale from 0 to 10. It worth to mention, that in difference from the other performance indicators, the sturdiness level is not specific for any point in the analysis grid, and it takes into consideration all the streets in the site. It is rather an indicator for the overall site, or district than for the area within the analysis grid. While a low sturdiness value means the system is not resilient to changes and changes fast, a higher value indicates a more robust design for the whole site.

Investment worth potential Investment worth potential indicates the predicted potential for investing in a place by combining the initial construction costs with the plot's changed value over time. A performance data corresponds to each point within the analysis grid. Financial system, and cost related matters play a critical role in economy, therefore to construct or invents in a place a certain cost should be considered. The costs are directly related to the physical feature of urban form, such as size of the building geometries, category of land use, quality and location. All this information can reveal also expected increases decrease in land values. To predict this fluctuation, the increasing or decreasing of the land value the output of the economic dynamic model was used. Since the result of the model were stored in street nodes, the data was aggregated from the street network in the analysis grid, to have a more cohesive result. A difference map was generated from the values of the last iteration of the model and the values of the first iteration of the model. All the values within the analysis grid were subtracted correspondingly. The output value of each analysis point will serve as a coefficient to multiply the initial construction costs price. Based on the increasing or decreasing value of the plot, and the size of the initial investment, we could hypothetically infer whether the investment was worthy or not and assign an economic importance value to each cell. Hence, we want to analyse the investment decision we need to know the construction price of each building volume created from the generative model depending on their use-function.

From the land use distribution model, we got four cost categories: housing a, housing b, work places and amenities. Based Construction Costs Index17 (BKI, 2020), the corresponding construction costs were included in the model. This information was used after assigning the land use of the buildings, to generate the cost of each as it follows: we use two different housing construction cost variables, the construction price of a one- and two-family houses, with a cellar is 970 per m2, and the other type of housing construction cost was considered 1105 Euro per m2. The price of an office or administrative building is 2340 Euro per M2. And institution buildings, that in this study we consider the amenities construction cost per m2 is 2700. To have the data in the same category, the building use category was aggregated in the analysis grid (see Land use Distribution chapter). In the open spaces - locations without building volume a zero-construction cost value was assigned. Within the scope of this research the outcome of this calculation is named ‘Investment worth potential’.

Figure 15a: Aggregated data in analysis grid based on the last iteration of the economic dynamic model. Source: Author.

Figure 15b: Difference map, calculated from the subtraction of last iteration data with the first iteration output data. Source: Author.

Figure 16: Normalised difference map, including initial construction costs per each point in analysis grid. Source: Author.

Data Aggregation: To aggregate the data into only single value, first all the data within the the analysis grid was summed up. This process gives an overall performance indicator for each case. The same method is applied to all cases of the generative model. The dataset is remapped in a domain from 0 to 10. However, the result is an inverse of linear correlation of the investment worth value, therefore we calculate the investment worth IW score by: IW = 10 – x. In this case, the lower the value, the less likely offers the place opportunities for investments. The higher the value the worthier can become the place over time.

Figure 17: Calculation and aggregation of the investment worth potential indicator. Source: Author. 17

BKI – BauKostIndex is a building construction cost database developed by the building cost information center of the Chambers of Architects in Germany. The BKI cost database includes thousands billed projects for new buildings, old buildings or outdoor facilities. The construction costs for each category of building are meaningfully included in the database.

Footfall potential Footfall potential indicates the number of people crossing a street node, tracing shortest paths using population as origin weight. This calculation is done by using weighted betweenness centrality (equation 2, page 40). Betweenness measures how many times a node was passed on the shorted paths, while connecting every node with all nodes in the street network (Newman, 2010). Therefore, betweenness gives information on the footfall potential of each street node, which is one of the main driving factors for the location choice of businesses. As result, the method could be applied as a proxy for economic potential for commercial uses or to roughly estimate traffic flows (Düring S. , 2019). In this research we use this method to understand the footfall potential of the designed quarter, as well as to compare the local maximum of each design option from the generative model, with the overall global maximum footfall potential. This metrics will be a supplement, among other economic metrics. To make meaning out of this indicator, firstly 5000 people were distributed in all the existing nodes of the site based on the density of the plots. This was taken as weights of the origins (which defines the attractiveness of the origins). We assume that every individual, takes a path to go to one certain – one random destination. Therefore, there is considered equal attractivity in all possible destinations - all street nodes. Similarly, with the other metrics, as a 5 minutes’ walk neighbourhood, the radius within which shortest paths are searched is defined as 600m. The simulation of the betweenness centrality, where the origin weights is based on the population number, gives us information on how many pedestrians crossed for all the street nodes. This information serves us to see how many pedestrians we can capture in the new site development (out of the initial distributed 5000 people). We calculate the proportion of the highest value in the site- local maxima, with the highest value of the whole site- the global maxima, to see how good the design option performs in comparison with the surrounding context. This calculation is expressed in percentage e.g. In the 61st case of the generative model the max. local measured footfall potential is 3654 people, which is 87% of the overall maxima. The higher the percentage, the more attractive is the place for shops and retail to locate in the quarter. On account of the generative model, this proportion is computed for all design options, and it makes it possible to compare the design options with each other.

Data Aggregation: To aggregate the performance into one single indicator, the overall maximum inside the boundary site is divided by the maximum most frequented street in the context. This process is repeated for all cases, where a single output from all cases create a dataset. The data is remapped in a domain from 0 to 10. The lower the value the less liveable and vital the design is. The higher the value indicates that the design can attract and get more people in its the area, whom can consider as potential customers for retail.

Example: Calculation of the footfall potential using betweenness centrality. The population number was used as the origin weight, and the destination weight was considered the same for every destination. case Study ID

max. on-site footfall

3654 people

2572 people

2777 people

% in comparison to overall max.

87.68%

62.06%

67.87%

Table 4: Footfall potential exploration

3.6

Additional Potential Algorithms and Performance Data Aggregation

Based on the predefined challenges and design objectives, one can select the best performing design. This approach is achieved by setting a simple weighing system for the framework. This way, the framework is adapted and is more project-specific so that the user can weigh the importance of the performance metrics. Eventually, one can set the goals for each metric, and this can help find which design is the most adequate based on the pre-set design challenges. Despite the numerical number, the performance data corresponds to a physical location in the analysis grid. This information was used to spot the problematic zones in the location. Based on the information that we want to reveal; several indicators, also potential further algorithms can derive from the results: -

The percentage of area with dangerous wind in socially segregated locations; Amount of people living in segregated areas with lack of daily sunlight; Amount of people living in a zone with high footfall potential; Pedestrian comfort in locations with high economic potential; etc.

It all depends on the initial goals where the framework is applied. However, the discussion for further developing the framework is given in the conclusion chapter.

4. Application of the framework: Hernals Vienna, Austria This chapter explores the application of the framework in a case study. The framework is applied to Vienna's city in Austria, in the northwest part of the city in the 17th district in Hernals. The chapter consists of four parts. The first part argues why the chosen location is a great spot to apply the framework, and it gives a summary of the urban development and data related to the location. The second part deals with location-based data gathering methods. The third part explores the design space derived from the generative model. The most significant share of the application chapter is the performance space exploration, where all performance indicators are elaborated in detail, performance data is visualized, and correlations between the generative model's input parameters with each performance indicator were applied. Each performance space exploration (social, environmental, and economic) is followed up by a conclusion based on the generated data.

Figure 18: Vienna city map. Conducted from OSM (june, 2020); adapted and visualised by author.

4.1 Why Hernals With the Austro-Hungarian block typologies and its city urban layout morphology, Vienna is a typical European Town. With all the political and urban developments of the last century in the Europe, Vienna’s position within this structure has shifted from an edge tonw, towards a hub, or a centre within European Cities. Its rapid growth affects the life quality of the citizens and the surroundings as well. That’s why, to achieve proper sustainable development, the Vienna City Council has already undertaken many actions, among them being adapting the Smart City Wien Framework Strategy 2050. One of the Viennas districts, where the population has increased 9% over a decade, and it is known for its greenery and parks in the city is Hernals. Hernals is the 17th district of Vienna, facing migration flows, but the population number gets increasing every day. This increase indicates that the site's planning and designing should meet the residents' needs and foresee the location's potential. The site's development should be based on its residents' social aspects, offer them job opportunities and a healthy and sustainable environment. Logical following up is that Hernals development should be based on specific sustainability criteria. Therefore testing the research and measuring sustainability by developing the site using generative design methods in Hernals is considered significant and a good base for the study. Among the already foreseen goals for the location, and Vienna in general, are its integration with the city, diversity, mobility provision, and urban vitality achievement. Based on these challenges, strategies on how to reach them are derived. The framework developed in this research can cover them by defining performance indicators and the involved urban parameters.

Figure 19: Hernals close up map. Conducted from OSM (june, 2020); adapted and visualised by author.

4.1.2 Vienna Urban Development Vienna's position within the European Union is becoming more significant with the last century's political development in Europe. The political situation mainly indicates the fall of the iron curtain in 1989/90 and Austria's membership in the EU in 1995, leading Vienna to be a geographical bridge and a cultural bridge in the region (Lichtenberger, 1993). With all the advantages of the opening up of eastern Europe that Austria gained, the challenges were highly visible. One of the biggest challenges was the rapid urbanization and urban growth, especially in Vienna's urban areas. The fast and development have led to discussing the city limits and free space in Vienna (Pfefferkorn, 1998). The rapidly grown population number makes it also tricky for Vienna to provide equal infrastructure and services to its residences, which indicates the quality of life in the city. On the other hand, climate thread as one of the global challenges calls for significantly stronger sustainable future development. To achieve proper sustainable development, the Vienna City Council has already undertaken many actions to adapt to the Smart City Wien Framework Strategy 2050. This adaption happened in 2014, and it is a long-term umbrella strategy that provides significant orientations to long-term development

4.1.3 Vienna and Sustainable Developmet Goals Strategies Vienna is considered one of the most livable cities in the world. In this sense, livability includes many dimensions, such as infrastructure, services, education facilities, green spaces, safety, and gender equality (Cohen, 2013). One of the most impactful initiatives to plan Vienna's development in the upcoming decades is the significant Smart City Wien Initiative launched in 2011. The framework strategy's main points are keeping the social quality of life, safety in the city, reducing resource consumptions, growing and expanding the research and business opportunities, etc. Regarding social inclusion, the Smart City Vienna aims to offer equality regarding the needs of all residents. By recognizing its own population's heterogeneity, the strategy seeks to balance the services, access to services, and provide safety for everyone. The heterogeneity includes gender, age, nation, preferences, and diverse groups of residents in the city. Considering the quality of life of all citizens in the city, the development strategy – Smart City Vienna aims to include all of the residents to participate and engage in the development process, as well as provide the right environment, comfort, and possibilities such as business or jobs. The agenda indicates that economic attractiveness, affordable housing and high-quality living should be fostered. The key traits of quality of life mentioned in the Smart City Wien Framework Strategy involve political, social, economic and environmental aspects. Regarding economic potentials, one of the goals is to become a city of short distances. With the global climate crisis, resilience was another topic brought under the Smart City Wien Framework Strategy umbrella. It suggests adapting the infrastructure and building structures to the Climate Change (InKA18) . Therefore this research puts together the goals by trying to find the best performing solution in social, environmental and economic dimensions, as the main pillars of sustainability.

InKA (Infrastrukturelle Anpassung an den Klimawandel) a city-internal implementation program, that coordinates the implementation of projects in the field of climate change adaptation. It is commissioned by the Competence Center for Green and Environmentally Related Infrastructure in the Vienna Building Department and led by the same and the Vienna Environmental Protection Department.

4.1.4 Hernals Data The location's significance lies in its geographical position and connectivity with the city, especially on a large scale, and its potential impact on the urban development of the growing city of Vienna. Therefore, its development must be based on international quality standards. Fifty-seven thousand twenty-seven people live in Hernals, which is 3,0% of the whole population in Vienna. The gender distribution is relatively equal, where 50,9% are women and 49,1% in men. It worth mentioning that the population of the district is increasing among the population growth in general. Since 2010 over 4,5 thousand people have moved to Hernals, which indicated a 9% population growth over a decade. To this population growth contributed also the natality, which is higher than the mortality rate in Hernals. Based on the Vienna City official data during 2019, there were 616 new-borns and 403 deaths in total. The migrations are also playing an essential role in the population density of the district. In 2019 the city recorded 7767 emigrations in Hernals, where the number of immigrations towards another place was higher; it was 8239. These migrations, in general, decreased the overall district population per 472 residents. The average age of the people in Hernals is 41 years old, the same as the average of Vienna's community in general. In Hernals, there are around 40 thousand inhabitants between 15 and 64 years old. There are approximately 5 thousand inhabitants per km2 in Hernals, while Vienna's average is 4,6 thousand.

Ground Space Index

The buildings reflect different building styles that date back from their construction time regarding its architecture and building structure. Over 1500 buildings were built before 1919, and about 1000 buildings were built between 1919 and 1944. Most of the buildings in Hernals were constructed between 1945 and 1980; there are over 1700 existing buildings in Hernals built in that timeframe. However, the construction and urban development of the quarter are still going on. Over 1600 buildings were built after 1982, and the need for proper sustainable development is becoming higher every day. The green spaces and water surfaces are still dominating, with 53%, where the infrastructure takes around 11% of the district. This amount of built environment indicates that buildings structures cover the left 36% of the district. The district is known for its green spaces and parks. Currently, 27 parks make the place children, elderly, and animal friendlier. In Hernals, there are presently 1393 dogs registered. Overall, it may be said, diverse inhabitants are living in the district, and the residents are changing with the migration flows.

Floor Area Ratio Figure 20: Main input parameters that generates the solution space

4.2 Data Preparation As an initial step, the open-source data information regarding the location was retrieved. It is worth mentioning that there is a broad spectrum of open data on the built environment, infrastructure, and demographics available regarding Vienna. Mainly the data was conducted from Open Government Vienna19 and OpenStreetMap20. The data was in shape files, and to import the files in the Rhino and grasshopper work environment, the library pyshp21 was used. The data is stored in layers in Rhino. The principal used geometries in the framework were the street network represented in lines, the building geometries, the boundary to generate the design options, and the analysis grid location. This way, in the experimental research, we could elaborate and play around with existing urban data combined with synthetic data generated from the model. The generated input parameters and performance indicators were exported as csv files for further data analysis, interpretation, and visualization.

4.3 Design Space Exploration: The Generative model Exploring all the potential solutions from the urban model would create an infinite design space; that’s why we try to semi-automatically choose the design input parameters from which we can have several output geometries. The framework is focused on generating many options that differ, rather than enormous options that don’t vary a lot, and parameters are very close to each other. As defined in the research goals, we use limited input parameters: the street network orientation, the ground space index, and the floor area ratio. The applied steps are as follows: As input parameters to generate several designs, we used the degree of the rotation of the streets; eleven subdivisions were defined. As a parameter representative for each case, the orientation of the amount of the streets in each option was defined, so for every street variant, we have a list with four items. Each indicates the % of length in meter of streets in each orientation. Consequently, each eleventh design has the same street network out of the generated hundred and one cases. Four possible data for FAR (0.10; 0.80; 1.80; 2.75) and three possible variables for GSI (0.10; 0.30; 0.60) were used parts of the blocks in the site. 25% of the plots were assigned from the variable, and the rest 75% was based on the existing context. This input is used as a constant. Data from the rotation angle list, various input data of the GSI and FAR were used to create all possible permutations (fig.20). Technically, to achieve all the combinations, for the initial input for 101 generated cases, the crossreference component within grasshopper was used. Some of the Figure 21: Geometrical outputs of the geometrical outputs from the generative urban model are shown in. generative model. Source: Author. 19

www.data.gv.at open data Austria was used to conduct shp and csv files regarding the geometries and planning regulations. www.wien.gv.at was used to conduct demographic data in csv format. 20 www.openstreetmap.org was used to conduct geometries and nodes with attributes. 21 https://github.com/GeospatialPython/pyshp

4.4 The performance space: Social, environmental and economic The performance space chapter elaborates on the generated urban forms' performance in the social, environmental, and economic dimensions.

For all the dimensions, the performance indicators are visualized, interpreted, and managed, respectively. The categorization and performance space exploration makes it easier to understand the bounds of the performance. Besides, for each performance indicator, the defined input parameters (GSI, FAR, and street orientation) are correlated, and conclusions are drawn whether the data are strongly linked together or not. It worth mentioning that unexpected results are also concluded, and further analysis can be done on the topic. Before starting with the individual exploration of the indicators, an overall overview of all the metrics is shown. In the fig.22 we can see the performance of the designs in each indicator separately, ASA, sunlight hours, wind comfort, sturdiness, investment worth potential, and footfall potential. The color of each indicator remains consistent in all data visualizations based on this figure. As argued in the method chapter, the data is remapped on a scale from 0 to 10, where 10 indicates the better performing option (for a detailed explanation, see table 2, page38). In fig.23, the graph visualizes the performance of the design in considering all the performance indicators. The x-axis shows the ID of the design variant, and the y-axis indicates its overall performance. However, this is only a mathematical sum in an equal weighting system, where we assume that all indicators are equally important. The same approach is also shown in fig.24, where we see the same results; the only difference here is that the design options are ranked from the best performing to the worst performing one. While in the fig.23, the x-axis starts with case one, two three, in fig.24 the x-axis still refers to the ID of the design, just the ID are not sorted anymore, since the sorting happens based on their performances.

Performance Score

asa

Design Option Case ID - index Figure 22: Design Performance in all metrices. Source: Author. Performance of the generated 101 urban layouts in each indicator/ metrics. The x axis indicates the case id, and the y axis indicates the performance (remapped in a range 0-10).

However, in the real-world, projects have different visions and objectives. Also, depending on the location's climate conditions, even environmental indicators might have different goals. Project-based, we can add weighing to the indicators, and the best performing case will update based on the importance ranking of the performance indicators.

Performance Score

For a proper explanation and understanding of the design, we look closer in each metric and understand the correlations of the performance with the initial input. The FAR, GSI, and street orientation for all 101 design options are saved as input data. After running all the analysis and storing the performing score, it was possible to correlate the input with the output data. The statistical measure R-square22 is used to understand how strongly input data and output performance data are linked together.

Design Option Case ID - index Figure 23: All the performance metrics visualised in one chart. Source: Author.

Performance Score

In an equal weighting system, the score of the performance of each design is shown in the chart. Each color represents one metrics, and the value from every metrics in a scale from 0-10 was added to the performance axis y. The x axis represents the Case ID from 1 to 101.

Design Option Case ID - index Figure 24: Ranked design options based on their overall performing score in an equal weighting system of all metrics In an equal weighting system, the score of the performance of each design is shown in the chart. Each color represents one metrics, and the value from every metrics in a scale from 0-10 was added to the performance axis y. The x axis represents the Case ID, from the best performing one until the worst performing one.

The R-squared value is the square of the correlation. It measures the proportion of variation in the dependent variable that can be attributed to the independent variable.

4.4.1 Social The social potential is defined through the A score, which indicates the social interaction potential - the residents’ accessibility to the liveable parts of the district. As explained in the method, using betweenness centrality and closeness centrality we measured how accessibly the liveable spots from each street segment are. The results are mapped in the analysis grid, and the performance data is aggregated into one single value, which varies from 0 to 10. In fig.25 and 26 are visualised graphically two design options, the one that performs the best and the one that performs the worst. The heatmap in the analysis area is in a gradient colour from blue to red, where blue indicates that it is hard to reach liveable spots, on the other hand red indicates that from there it is easy to access liveable spots. The area chart in fig. 27 and the bar charts in fig. 28 visualise the performance of each case, where the y axis indicates their performance and the x axis indicates their ID. In fig. 27 the cases are sorted based on their ID number and since the analysis are mainly based in the street network it is seen in the performance graphs that the results are similarly repeating it when the street network is repeated after several iterations.

Figure26: Worst performing solution in social level Case ID 43

Performance Score

Figure 25: Best performing solution in social level Case ID 50

Design Option Case ID - index Figure 27: Performance of the generated 101 urban layouts in social dimension indicator ASA metric..

Performance Score

The x axis indicates the case id, and the y axis indicates the performance (remapped in a range 0-10).

Design Option Case ID - index Figure 28: Ranked design options based on the social dimension performing score - ASA The x axis indicates the case id, and the y axis indicates the performance (remapped in a range 0-10).

Ground Space Index

Figure 29: Correlation between FAR and ASA. Source: Author.

Figure 30: Correlation between GSI and ASA. Source: Author.

ASA and street network orientaion: From the correlations of ASA value with the street orientation properties some conclusions can be drawn. In general, we can see that the higher the amount of northwest-southeast and the southwest northeast oriented street amount, the value of ASA is lower. And oppositely, if the streets are oriented in north-south or west-east the value of ASA is high. But, from the context, we can also associate that the existing network is dominating in the north-south and west-east orientation. This could be a basis to draw a rule of thumb to inform the design decisions.

Figure 31: Correlations of ASA value with the street orientation. Source: Author.

Figure 33: Aggregated social accessibility performance indicator –ASA- grouped based on the street network option. The numbers above the graphs indicate the street network type; each bar represents a case – an urban layout output from the generative mode; and the y axis indicates the performance in a score from 0 to 10

Figure 32: Street network orientation in the context. Source: Author.

There are, in total, eleven street network variants, and each of them contains several urban layouts where the floor area ratio and ground space index vary. The bar graph, visualized in fig.21, shows the aggregated social accessibility performance indicator based on the street network option. The bar graph shows that the design options with the same street configuration perform similarly in the dimension of ASA indicator. However, based on the density metrics' correlation with the ASA indicator, as low as it is, the built structures are also related to ASA performance.

Social Dimension Conclusion: There was no strong correlation in the input parameters (FAR and GSI) in the design performance at the social level. However, there is seen a firm relationship between the street network configurations with aggregated social sustainability -ASA- performance. Besides, in worth to mention that combining the outcomes from the correlations, orientation, and FAR could give more transparent results, and a higher correlation between the input parameters and the performance indicator can happen. Therefore, we suggest that a higher ASA value is to match the orientation of the existing street network with the new design ones.

4.4.2 Environmental Environmental analysis includes wind comfort analysis and sunlight hours. The research predicts the performance of these indicators using machine learning models. For wind comfort, the model was trained on computational fluid dynamics simulations, whereas for sunlight hours, it was trained based on simulations in Vienna. In the graphs both performance data are visualised, and the score varies from 0 to 10, where 10 indicates that the design option that performs better. Detail elaboration of both indicators are explained further, and a generate conclusion that can inform the early stages of design is given in the end of the environmental analysis application chapter. Sunlight

Performance Score

Wind – Pedestrian Comfort

Design Option Case ID - index Figure 34: Performance of the generated 101 urban layouts in environmental dimension, respectively wind comfort and sunlight hours. Source: Author.

Performance Score

The x axis indicates the case id, and the y axis indicates the performance (remapped in a range 0-10).

Design Option Case ID - index Figure 35: Design options based on the environmental dimension performing indicator. Source: Author.

Performance Score

In an equal weighting system of wind comfort and sunlight hours, the score of the performance of each design is shown in the chart. Yellow represents sunlight hours performance, and wind is visualized in orange. The summed-up performance is shown in axis y. The x axis represents the Case ID from 1 to 101.

Design Option Case ID - index Figure 36: Ranked design options based on the environmental dimension performing indicator. Source: Author. In an equal weighting system of wind comfort and sunlight hours, the score of the performance of each design is shown in the chart. Yellow represents sunlight hours performance, and wind is visualized in orange. The summed-up performance is shown in axis y. The x axis represents the Case ID ranked from the best performing one to the worst performing one.

Wind Comfort For wind, the threshold of dangerous areas that do not offer pedestrian comfort is set based on the Lawson Criteria. Regarding the correlation outcomes, there is seen a Figure 37: Line chart visualising the performance space of wind medium positive linear comfort based on the generated cases. Source: Author. association between the pedestrian comfort in terms of wind and the floor area ratio of the design (see fig.40). There is a small negative linear association between the pedestrian comfort in terms of wind and the design's Figure 38: Line chart visualising the sorted performance space of wind floor space index. This comfort based on the generated cases. Source: Author. correlation does not necessarily indicate that the higher the floor area ratio, the higher the wind comfort, and it does not imply causation. But, an explanation of this is that the higher the building's volume, it creates more wind shadow, which could contribute to lower the wind speed and offer higher pedestrian comfort. Besides, it is seen from the performance, open spaces tend to be less comfortable for pedestrians, and most of the time dangerous (see fig.39). So, in a sustainable urban layout, the balance between built and non-built spaces should be carefully designed.

Figure 39: Wind Comfort performance in several design options, where the dangerous category is visualised in red.

Ground Space Index

Figure 40: The Correlation between the Urban Density Input Parameters and the Wind Comfort Performance.

Figure 41: Correlations of the wind comfort value with the street orientation. Source: Author.

The fig.41 shows the correlation of wind comfort with the street orientations. The street network orientation's relation with the wind flow seems to be low based on the calculated correlation data. The highest R-squared value with wind comfort is shown in its correlation with the west-east-oriented street network, where there is a linear negative association. One can assume that having more buildings built in the west-east area could reduce the wind speed overall in the site because it blocks the wind. It worth mentioning that the wind comfort prediction includes the weather information of the location (the wind is coming from the south). There are 11 street network variants, and each of them contains several urban layouts where the floor area ratio and ground space index vary. The bar graph visualized in fig.42 aggregated wind performance indicator grouped based on the street network option. We can see that the performance data varies in each group. However, there is still an outline in each cluster, which indicates that the street networks might cause different open spaces in the design and impact the overall wind comfort performance. In group one, the performances are closer to each other, but this information doesn’t allow us to conclude correlations. Based on the correlations data and the bar chart, one could conclude that it would make sense to see the impact in a denser urban area, where the street would be the central open space within the analysis grid. Still, there is seen no direct impact of the street network on wind comfort in the research scope.

Figure 42: Aggregated wind comfort performance indicator, grouped based on the street network option. The numbers above the graphs indicate the street network type; each bar represents a case – an urban layout output from the generative mode; and the y axis indicates the performance in a score from 0 to 10

Sunlight Hours

Figure 43: Line chart visualising the performance space of sunlight hours based on the generated cases. Source: Author.

Sunlight hours analysis checks if the design option provides significant sunlight hours per day. In this research, 5,5 was used as a target daily sunlight, since despite human beings, it needs to satisfy vegetation too. The lower the value in the graph indicates a larger area with a lack of sunlight. The higher the value indicates a better performing design in terms of sunlight hours.

Figure 44: Line chart visualising the sorted performance space of sunlight hours based on the generated cases. Source: Author.

The highest R-squared value is seen in the aggregated sunlight hour performance indicator with the floor area ratio from the correlation outputs. There is a very low association between the design option's sunlight performance and the floor space index of the design. Similarly, between sunlight hours and the street network orientations, there is no clear correlation. On the other hand, the scatterplot shows a low negative linear association between the design option's sunlight performance and the design's floor area ratio. A high floor area ratio, causing a lack of sunlight, can be argued that more shadow is created from high-rise buildings, which can block the direct sunlight to particular buildings in the site. Therefore it is suggested in general to pay attention to the spacing between buildings and sunlight hours analysis of the site in the early stages of design.

Ground Space IndexPerformance. Figure 45: The Correlation between the Urban Density Input Parameters and the Sunlight Hours Source: Author.

Figure 46: Correlations of the sunlight hour’s value with the street orientation. Source: Author.

Figure 47: Aggregated sunlight hours performance indicator, grouped based on the street network option. Source: Author. The numbers above the graphs indicate the street network type; each bar represents a case – an urban layout output from the generative mode; and the y axis indicates the performance in a score from 0 to 10.

Environmental Dimension Conclusion:

Figure 48: Best (A) and worst (B) performing design cases in environmental dimension. Source: Author.

In the fig.48 are visualized the best performing and worst performing case out of all generated urban layouts. We can see that the diversity in the building heights and mixed compositing of the buildings gives better performance. While the open spaces cause higher wind speed and don’t provide pedestrian comfort, the denser areas don’t offer enough daily sunlight hours. Therefore, as a rule of thumb, a design to perform better in terms of environment is suggested to have diverse building typologies and have a reasonable proportion between the built and non-built areas. 67

Performance Score

4.4.3 Economic

Design Option Case ID - index Figure 51: Ranked design options based on their overall performing score in the economic dimension, based on an equal weighting system. Source: Author. In an equal weighting system, the score of the performance of each design is shown in the chart. Each color represents one indicator, and the value from every metrics in a scale from 0-10 was added to the performance axis y. The x axis represents the Case ID.

Investment worth

Footfall potential

Performance Score

System Sturdiness

Design Option Case ID - index

Performance Score

Figure 49: Performance of the generated 101 urban layouts in economic dimension, respectively system sturdiness, investment worth scores and footfall potential. Source: Author.

Design Option Case ID - index Figure 50: Economic performance indicators visualized in one chart. Source: Author. In an equal weighting system, the score of the performance of each design is shown in the chart. Each color represents one indicator, and the value from every metrics in a scale from 0-10 was added to the performance axis y. The x axis represents the Case ID from 1 to 101.

The three leading indicators applied are the system sturdiness, the investment worth potential, and the footfall potential. The system sturdiness and investment worth potential indicators have derived from the economic dynamic model. The output of each iteration of the dynamic model was recorded, and the system sturdiness, by keeps track on how many iterations the system is still changing, while the investment worth potential uses the difference between the first iteration and the last one data only within the analysis grid. Further, this number is multiplying with the construction costs, as indicated in the method chapter. The footfall potential shows the relation of the maximum pedestrian no within the design boundary-local, compared to the maximum in the whole area- global.

System Sturdiness The sturdiness – system change indicator informs us how fast the system reaches a static state. If the system changes slowly, we indicate that the design is more resilient (see chapter 3.5.3 Economic Sustainability Evaluation Method/ Economic Dynamic Model & The system change value – Sturdiness). The dynamic model runs for 28 iterations on each design option, and the performance difference between each iteration was tracked. The times that there were still changes were saved as a score of the design option. After getting the performance numbers' bounds, for aggregation purposes, all data was remapped on a scale from 0 to 10. It was seen from the application that the most significant change happens in the first iterations, and after five iterations, the changes are minimal. However, in some cases, there is no change even after two iterations

Figure 52: Economic Dynamic Model Outcomes after the 1,2,3,6 and 20th iteration. Source: Author.

a) Figure 53: Line chart visualizing the performance space of sturdiness level, based on the generated cases. Source: Author.

It worth to repeat, that the score informs on the sturdiness of the whole context area where the site is located. The overall value does not correspond to any analysis point, but the entire design and its impact on the district. In fig.53 and fig.54 is visualized the sturdiness level of each design variant (101 cases).

In the first graph, fig.53, the case IDs are in order Figure 54: Line chart visualizing the sorted performance space of sturdiness level (0,1,2,3,4, …,100) based on the generated cases. Source: Author. among the x-axis, whereas in the second graph, they are sorted from the best performing one to the worst-performing case. The second graph fig.54 helps to understand how the performance space (bounds and medium), while the first graph allows us to know how the cases performed in general. We can see the street network's impact on the first graph in fig.53, where whenever the street network was repeated, we get similar results. Regarding the system's sensitivity, it worth mentioning that the economic dynamic model resulted to be very sensitive. Minimal changes can lead to very different outcomes. In the fig.52 are visualized two design options with different street networks; in the (a) case the streets within the new quarter are better and more directly connected to the existing roads,

Ground Space Index

Figure 55: The correlation of the sturdiness level of the design options with (1) FAR and (2) GSI. Source: Author.

Figure 56: The correlation between the sturdiness level of the design option with the amount of street lengths oriented in North-South, Northwest –Southeast, West-East, Southwest-Northeast in %. Source: Author.

Figure 57: Sturdiness level performance indicator grouped by the street type. Source: Author. Graph description: Out of 101 design options there are 11 street types, each of them contains 9-10 different urban layouts; the number above the bar graphs indicates the id of the street option; and the y axis indicates its sturdiness level.

when in case (b) the street nodes are less well connected to existing ones because of less straight network and shorter street. The center of the district after the 28 iterations ends up in a totally different location. Regarding the density metrics, based on the outcome of the correlation, one can indicate that the built structure has a relation with the system sturdiness. With a 0.35 R-squared, there is seen a moderate negative correlation between the floor area ratio and the design's sturdiness level. On the other hand, there is a positive correlation between the ground space index with the system sturdiness. It might have happened since the building volumes are used to estimate the population number, and this data was aggregated in the street network and used as an initial weight. Nevertheless, this does not indicate necessarily impact, and further studies can be done on the topic. On the other hand, the street network has a close relationship with the system sturdiness level; this can also be seen in fig.54, where the design options with the same street network have very similar system sturdiness level. An argument for this sensitivity is that the angular shortest paths can impact how people move, which changes the urban system and its tipping points. Thus, the street orientation perhaps doesn’t capture all factors affecting sturdiness, and further parameters should be tracked from the urban form, for example, the connectivity to other areas using a mix of angular distance and shortest paths functions. It worth to mention, that further research is needed to understand the performance of the system sturdiness level.

Investment Worth Potential

Figure 58: Line chart visualizing the performance space of investment worth potential, based on the generated cases. Source: Author.

The investment worth potential compares each analysis point's value in the initial state of the design with the value of the last state of the dynamic model. The difference number is weighing by the construction price of each point in the analysis grid. The sum of all values in the analysis grid gives general information on how worthy it is to invest in the designed option. The dataset of the

Figure 59: Line chart visualizing the sorted performance space of investment worth potential based on the generated cases. Source: Author.

performance of each design is remapped on a scale from 0 to 10. There is no correlation between the average floor area ratio and the worth investment metric; however, there is seen a moderate negative relationship between the floor space index and the investment worth potential.

Figure 60. Best (A) and worst (B) performing cases based on their investment worth potential. Source: Author.

Ground Space Index

Figure 61: The correlation of the investment worth potential of the design options with (1) FAR and (2) GSI. Source Author.

Figure 62: The correlation between the investment worth potential of the design option with the amount of street lengths oriented in North-South, Northwest –Southeast, West-East, Southwest-Northeast in %. Source: Author.

Figure 63: Investment worth potential performance indicator grouped by the street type. Source: Author. Graph description: Out of 101 design options there are 11 street types, each of them contains 9-10 different urban layouts; the number above the bar graphs indicates the id of the street option; and the y axis indicates its sturdiness level.

Based on the simulated change of the plot values over time (based on the dynamic economic model, see above section 3.4), built structures with higher construction cost indices (BKI) should be allocated to areas where the value remains stable or appreciates. In order to reduce investment risk, built structures with lower construction cost indices (BKI) should preferably be allocated in worse-performing areas, in favour of the high construction costly buildings. Nevertheless, we should keep in mind that this indicator doesn’t consider the market values, but only the construction cost. The complexity of simulating such models is higher, so in the research scope, it was not considered, which is also the limitation of the investment worth potential of this research. However, it would doubtlessly augment the economic dynamic model if data on market values would be integrated.

Footfall Potential

Figure 64: Line chart visualizing the sorted performance space of footfall potential based on the generated cases. Source: Author.

Figure 65: Line chart visualizing the performance space of footfall potential, based on the generated cases. Source: Author.

Footfall potential indicates the visitor’s potential in a location, which can indicate potential customers. This indicator contributes to the economic potential of a space. In the exploration of 101 designs, we see that the results after 11 iterations are very similar. This explains why in the fig.64 the graph repeats itself at a particular time, and it indicates the impact of the street network on the footfall potential.

Ground Space Index

Figure 66: The correlation of the footfall potential of the design options with (1) FAR and (2) GSI. Source: Author.

Figure 67: The correlation between the footfall potential of the design option with the amount of street lengths oriented in North-South, Northwest –Southeast, West-East, Southwest-Northeast in %. Source: Author.

Figure 68: Aggregated footfall potential performance indicator, grouped based on the street network option. Source: Author. The numbers above the graphs indicate the street network type; each bar represents a case – an urban layout output from the generative mode; and the y axis indicates the performance in a score from 0 to 10.

There is a low linear correlation between the footfall potential of a location with the average floor area ratio or the design's ground space index. Similarly, the streets' orientation also doesn't correlate strongly with the footfall potential aggregated performance indicator. However, the bar graph shows that design options with the same street network configuration sometimes perform very similarly. There are seen escalating values in some of the groups with the same street network configuration. That is explained because to calculate the footfall potential, we use the population number as weigh, which depends on the building volumes. This implies that the urban layout, including street networks and building geometries, should impact the footfall potential. But to understand what is precisely is causing the changes, further studies can be done. Multiple linear regression models can be applied to correlate the performance data further by isolating the effects of either FAR, GSI or the street network orientation. In conclusion, we can assume that it would contribute to the footfall potential if the designed streets are aligned with the existing street network.

Economic Dimension Conclusion

Figure 69: Best and worst performing cases based on their overall economic potential. Source: Author.

A general outcome from the indicators is that the designed streets should be aligned with the existing street network. This street matching effect is also slightly reflected in the best performing case in economic dimensions compared to the worst-performing case visualized in fig.69. However, this research's urban parameters to correlate the performance data (FAR, GSI, Street network orientation) do not cover all factors of system sturdiness, investment worth potential, and footfall potential. Additionally, simple linear correlations might be deficient in interpreting complex systems such as economic sustainability. 75

5. Conclusion and reflection Throughout this research, a framework to measure the sustainability of urban form was developed. This process was done through the use of generative design methods, making it possible to generate hundreds of cases parametrically and analyze their performance data. The interdisciplinary field, where the research takes place, made the work more interesting. It starts with social sciences and urban sociology; dives into the urban design world, its regulations, and parameters; it continues with algorithmic design and ends up in data science. The advancement of design tools made it possible to build this framework using mostly visual programming languages. To build this framework, information on the concept of sustainability in urban design is conducted. Existing methods on how to quantify and measure the quality of urban design are elaborated. All the gathered information lead to a method which adjusts, and remodels existing methods to quantify urban performance. The thesis presented six performance indicators corresponding social, environmental and economic sustainability:      

The social performance indicator that focuses on social equities and integration – ASA- aggregated social sustainability; Wind comfort; Sunlight hours; The system change value indicating the economic robustness, which in this thesis is named 'sturdiness'; The investment worth potential, related with the economic sustainability; and Footfall potential indicating the visitor’s potential in a location.

Figure 70: Comparative performance indicators aggregated in an analysis grid. Source: Author.

Alongside the society, environment and economy topics, urban configurations such as input parameters and urban elements are another key aspect of the thesis. To quantify the performance an urban layout was required, and to understand which input parameter affects the performance many urban layouts were requires; and that was the point where generative design methods was brought into discussion. Using generative design methods many urban design projects are generated and corresponding performance indicators are simulated. The simulated data falls into the category of synthetic data, which served for further data analysis and correlation. The framework is applied to a new site development in Vienna, in Hernals. The significance of the location lies in its geographical position and connectivity with the city, especially in large scale and its potential impact in the urban development of the growing city of Vienna. 11 Street network layouts were generated, each of them contains several urban layouts where the floor area ratio and ground space index vary. For each performance indicator: 



 

The performance space was simulated and visualised, which assists in understanding the performance minimum and maximum, and the potential of solutions in that corresponding indicator. The correlation of the performance with the density parameters: FAR and GSI, which informed about the relevance of the density and building mass in the corresponding indicator The correlation of the performance with the street network orientation amount in each direction, The performance indicators were grouped in clusters within the same street network configuration, which shows how much the street network can affect the performance in the corresponding indicator.

In the social dimension, where is measured the accessibility to liveable spots on the neighbourhood it is seen a strong correlation with the street network orientation. With the R-squared over 0.7, the streets that are oriented similarly with the orientation of the existing streets of the context, (in our case mostly west-east) correlate positively with ASA indicator. In addition, from ASA performance indicator grouped based on the street network option fig.30 can be seen that the design options with the same street configuration perform similarly in the dimension of ASA. Regarding environmental aspects, some conclusions are drawn. Wind comfort is positively correlated with the floor area ratio with a R-squared 0.57 and negatively with the ground space index with a 0.25 R-squared. An expiration of this is that the higher the buildings volume, the more wind shadow and more pedestrian comfort is provided, and this could contribute to lower the wind speed; contrary the more open spaces the higher wind is created. On the other hand, a negative correlation is seen between the sunlight hours and floor area ratio, with a R-squared 0.26. High floor area ratio indicates high buildings, which causes lack of sunlight, more shadow is created, and this can block the direct sunlight to particular buildings in the site. To sum it up, regarding environmental dimension, while the open spaces cause higher wind speed, and don’t provide pedestrian comfort; the denser areas don’t offer enough daily sunlight hours. In the economic dimension, high correlation between input parameters and performance indicators are seen too. The sturdiness level appears to be a very sensitive indicator, where very small changes can lead to very different outcomes, and it is mostly tightly related with 77

the street network. The correlation outcomes show other associations too, but the scope of this research could not cover them all, and further research is needed to understand the performance of the systems sturdiness level. The foundation of the investment worth potential indicator lays in the same economic dynamic model and it is combined with construction costs. With a 0.33 R-squared a negative correlation is seen between investment worth potential and ground space index. However, the outcome results are deficient to argument the method. Applying the method in real location, with existing land uses and not only on the land use was distributed computationally, could augment the scatterplot. Regarding footfall potential, there was no strong relation with the input parameters (FAR, GSI, and street network orientation) and performance indicator. However, there is seen that design options with the same street network configuration sometimes perform very similarly. Multiple linear regression models can be applied to corelate further the performance data, by isolating the effects of either FAR, GSI or the street network orientation.

When coming to the research questions and hypothesis of the research, some conclusions and answers can be made: Question 1: Can we relate the performance of urban layouts to FAR, GSI, and street network orientation? There is strong relation, and many aspects of urban performance can be interpreted using FAR, GSI and street network orientation, but additional parameters are needed to explain certain phenomena in the urban performance. Street network orientation seems to be a solid and reliable parameter to describe the social dimension and ASA indicator, respectively. Floor area ratio and ground space index, on the other hand, are highly related to the environmental performance of a design. In reverse, both built geometries and street networks affect the performance at the economic level, but only the FAR, GSI, and street network orientation parameters seem to be deficient in explaining the causations. Getting a larger dataset from the generative model, including other parameters, or applying multiple linear regression models can augment the framework.

Question 2: Can we deduce some actionable rules of thumb for urban design problems? Yes, there are several actionable rules of thumb derived from the research. However, for some of the indictors it was hard to conclude on general thumb rules. Regarding the social dimension, people would have more access to liveable spots if we match and align the existing street network's orientation with the new design ones. A design to perform better in terms of environment is suggested to have a diverse building typology and have a reasonable proportion between the built and non-built areas. While high and large buildings create wind shadow, at the same time they block the sunlight. So, compromises should be done, and the optimum solutions based on the initial goals could assist which design is more sustainable for the location. The framework has potential to include weighing for each indicator, and that would assist in evaluating the designs based on the goals of the project. As a general rule, the economic aspect could be that the designed streets should be aligned with the existing street network. However, it could not be defined what exactly should be 78

changed in order have a design that would have higher investment potential. Similarly, the system sturdiness with the input parameters show a high correlation, but it is hard to draw conclusions since this does not indicate necessarily impact, and further studies can be done in the topic.

Question 3: How to ensure urban design sustainability by quantifying the performance in the early stages of urban design? This research concludes that the selection of adequate performance indicators is a first step. While measuring sustainability it is crucial to include indicators that correspond to social, environmental and economic sustainability at the same time. Synthetic data, simulated from the design option might not be very accurate, due to the simplification of the model while measuring it, however it can inform the early stages of urban design. Another crucial aspect that contributes towards quantifying sustainability is the large amount of design options. This can be practically achieved using generative design methods. that’s why this research concludes that to ensure urban design sustainability generative design methods could be a valid approach. They allow for the rapid performance testing of many variants in a short period of time, providing an almost-instantaneous indication as to which designs are performing as desired. Keeping in loop the three components of generating urban form: generate, evaluate, evolve; doubtlessly augments the urban design world, and performance data informs us about the quality of the design. It makes the planning process more manageable, where designers combine their design instinct with performance data. Reflection on synthetic data Generative design methods provide many urban designs with an understanding of the performance for all design options. Input parameters and performance indicators were aggregated, and the data analysis was done using this information, meaning that this research deals with synthetic data. Despite it being fabricated, the data is created algorithmically, and its roots lay in real-world phenomena and values. The generated urban designs, for example, aim to look as realistic as possible, and their density and street network parameters indicate real values. The advantage of working with synthetic data was that it made it more practical to have a large dataset for the same location with the same climate, social and economic conditions. It can be challenging and time-consuming to gather and mine real data manually. Besides, synthetic data enables possibilities to test the sensitivity of the performance indicators. Another strong point of using sensitive data lies exactly under the umbrella of public participation in planning processes (several options can be tested together with all stakeholders). Their impact on the environment can be seen immediately from their performance indicators, making it easier to find common ground. However, if the data is generated for a complex system, such as an urban system, it is challenging to create high-quality data. Many parameters and factors can affect urban performance. A way to understand if the model is generating realistic data is to validate it with the real-world. The validation of the framework could be an extension of this research. Performance indicators or urban form parameters could be related to real-world phenomena. The performance indicators in the economic dimension could be tested in real locations to validate their relevance.

Discussion As promising as the framework appears, it does not cover all issues related measuring sustainability. However, it presumably is be a supplement to the existing tools. A big advantage of the framework is that it can be used as a discussion table in participatory planning processes. It could be easily adapted with interactive environments and devices such as interactive tables or augmented reality headset.

There are several directions how the framework could be augmented or improved: 

 



Input parameters: only floor area ratio, ground space index and street network orientation seem to be deficient to explain the urban performance; additional input parameters could be added. Performance indicators: further research could be done in the economic dimension; its performance indicators could be tested in several real locations. Data Analysis: simply linear correlations tend to be insufficient to explain some phenomena’s in performance of the design, therefor multiple linear regression models can be applied to corelate further the performance data. Additional Potential Algorithms: Despite the numerical score, the performance data corresponds to a physical location in the analysis grid too. Additional data aggregation could be done spotting the problematic zones in the location. Project specific several other information could be extracted from the framework if needed, like: - The percentage of area with dangerous wind in socially segregated locations; - Amount of people living in segregated areas with lack of daily sunlight; - Amount of people living in a zone with high footfall potential; - Pedestrian comfort in locations with high economic potential; etc.

Due to the complexity of both urban systems and quantification of sustainability, on top of that due to their complex simulations models further research is needed to verify the outcomes of this thesis. Further research would increase the quality of measuring sustainability and would provide a better understanding of the relationship between essential design input parameters with relevant performance indicators.

6. List of Figures and Tables Tables: Table 1. Outline of the Input Parameters Table 2: Outline of the performance indicators Table 3: Social Data Aggregation Approaches Table 4: Footfall potential exploration Figures: Figure 1: Venn diagram representing the standard dimensions of sustainable development Figure 2: Types of urban areas in the FSI‐GSI plane of the Spacematrix Figure 3: Street Network Orientation in major US cities Figure 4: The architecture of the framework presented in a flowchart Figure 5: Street network generation steps. Source: Author. Figure 6a: Compass representing 4 orientation categories Figure 6b: Street networks obtained from the generative design Figure 7: Urban Layout generation based on the given FAR and GSI Figure 8: Land use visualization on the site Figure 9: Workflow of operationalizing the social performance of the design options Figure 10: Wind Comfort Prediction using Infrared Figure 11: Sunlight hours prediction using Infrared Figure 12: Impact of lack of sunlight hours on the population Figure 13: Economic Dynamic Model Outcomes after the 1st ,2nd ,3rd ,6th and 20th iteration. Figure 14: The system change indicator through iterations Figure 15a: Aggregated data in analysis grid Figure 15b: Difference map Figure 16: Normalized difference map, including initial construction costs Figure 17: Calculation and aggregation of the investment worth potential indicator. Figure 18: Vienna city map Figure 19: Hernals close up map Figure 20: Main input parameters that generates the solution space Figure 21: Examples of the generated urban layouts Figure 22: Design Performance in all metrices. Figure 23: All the performance metrics visualized in one chart. Figure 24 Ranked design options based on their overall performing score in an equal weighting system Figure 25: Best performing solution in social level Figure 26: Worst performing solution in social level Figure 27: Performance of the generated 101 urban layouts in social dimension indicator ASA metric Figure 28: Ranked design options based on the social dimension performing score – ASA Figure 29: Correlation between FAR and ASA Figure 30: Correlation between GSI and ASA Figure 31: Correlations of the ASA value with the street orientation. Figure 32: Street network orientation in the context Figure 33: ASA performance indicator, grouped based on the street network option Figure 34: Performance of the generated urban layouts in environmental dimension, respectively wind comfort and sunlight hours Figure 35: Design options based on the environmental dimension performing indicator 82

Figure 36: Ranked design options based on the environmental dimension performing indicator Figure 37: Line chart visualizing the performance space of wind comfort based on the generated cases. Figure 38: Line chart visualizing the sorted performance space of wind comfort based on the generated cases. Figure 39. Wind Comfort performance in several design options Figure 40: The Correlation between the Urban Density Input Parameters and the Wind Comfort Performance Figure 41: Correlations of the wind comfort value with the street orientation. Figure 42: Aggregated wind comfort performance indicator, grouped based on the street network option. Figure 43: Line chart visualizing the performance space of sunlight hours based on the generated cases Figure 44: Line chart visualizing the sorted performance space of sunlight hours based on the generated cases Figure 45: The Correlation between the Urban Density Input Parameters and the Sunlight Hours Performance Figure 46: Correlations of the sunlight hours value with the street orientation Figure 47: Aggregated sunlight hours performance indicator, grouped based on the street network option Figure 48: Best (A) and worst (B) performing design cases in environmental dimension Figure 49: Performance of the generated 101 urban layouts in economic dimension, respectively system sturdiness, investment worth scores and footfall potential Figure 50: Economic performance indicators visualized in one chart Figure 51: Ranked design options based on their overall performing score in the economic dimension, based on an equal weighting system Figure 52: Economic Dynamic Model Outcomes after the 1,2,3,6 and 20th iteration Figure 53: Line chart visualizing the performance space of sturdiness level, based on the generated cases Figure 54: Line chart visualizing the sorted performance space of sturdiness level based on the generated cases Figure 55: The correlation of the sturdiness level of the design options with (1) FAR and (2) GSI Figure 56: The correlation between the sturdiness level of the design option with the amount of street lengths oriented in North-South, Northwest –Southeast, West-East, SouthwestNortheast in % Figure 57: Sturdiness level performance indicator grouped by the street type Figure 58: Line chart visualizing the performance space of investment worth potential, based on the generated cases Figure 59: Line chart visualizing the sorted performance space of investment worth potential based on the generated cases. Figure 60: Best (A) and worst (B) performing cases based on their investment worth potential Figure 61: The correlation of the investment worth potential of the design options with (1) FAR and (2) GSI Figure 62: The correlation between the investment worth potential of the design option with the amount of street lengths oriented in North-South, Northwest –Southeast, West-East, Southwest-Northeast in % Figure 63: Investment worth potential performance indicator grouped by the street type Figure 64: Line chart visualizing the sorted performance space of footfall potential based on the generated cases 83

Figure 65: Line chart visualizing the performance space of footfall potential, based on the generated cases Figure 66: The correlation of the footfall potential of the design options with (1) FAR and (2) GSI Figure 67: The correlation between the footfall potential of the design option with the amount of street lengths oriented in North-South, Northwest –Southeast, West-East, Southwest-Northeast in % Figure 68: Aggregated footfall potential performance indicator, grouped based on the street network option Figure 69: Best and worst performing cases based on their overall economic potential Figure 70: Comparative performance indicators aggregated in an analysis grid.

References

Achary A., Karimi K., Parham E., Guven A., Uyar G. (2017): City planning using integrated urban modeling. Jeddah Structure Plan. Proceedings of the 11th Space Syntax symposium Alexander, W. M. (1994). Humans sharing the bounty of the Earth: hopeful lessons from Kerala. Proceedings of the International Congress on Kerala Studies, Kerala, India, Aug. 27-29, 1994 Baisago, A.D. (1999). Economic, social, and environmental sustainability in development theory and urban planning practice. In Environmentalist 19, pp. 145–161. Berghauser Pont, M., & Haupt, P. (2009). SPACEMATRIX; Space, Density and Urban Form. Lund. Bielik, M., Koenig, R., Fuchkina, E., Schneider, S., & Abdulmalik, A. (2019). EVOLVING CONFIGURATIONAL PROPERTIES – Simulating multiplier effects between land use and movement patterns. Presented at the 12th Space Syntax Symposium, Beijing, China. Bielik, M.; König, R.; Schneider, S.; Varoudis, T. (2018): Measuring the Impact of Street Network Configuration on the Accessibility to People and Walking Attractors. In Networks and Spatial Economics 18 (3), pp. 657–676. DOI: 10.1007/s11067-018-9426-x. Bielik, M.; Schneider, S.; Kuliga, S.; Griego, D.; Ojha, V.; König, R.; Schmitt, G.; Donath, D. (2019). Examining Trade-Offs between Social, Psychological, and Energy Potential of Urban Form. SPRS Int. journal of geo-information. Bielik, Martin; Schneider, Sven; Koenig, Reinhard (2012): Parametric Urban Patterns Exploring and integrating graph-based spatial properties in parametric urban modelling. Blocken, B.; Stathopoulos, T.; van Beeck, J.P.A.J. (2016): Pedestrian-level wind conditions around buildings: Review of wind-tunnel and CFD techniques and their accuracy for wind comfort assessment. In Building and Environment 100, pp. 50–81. DOI: 10.1016/j.buildenv.2016.02.004. Boeing, G. (2019): Urban spatial order: street network orientation, configuration, and entropy. In Appl Netw Sci 4 (1). DOI: 10.1007/s41109-019-0189-1. Cohen, B. (2013). Smart Cities Ranking “Worldwide Quality of living Survey. State of the World’s Cities : United Nations Human Settlements Programmed. Retrived on August 2020. Cotgrave, A; Riley, M. (2013): Total sustainability in the built environment. Basingstoke: Palgrave Macmillan. Delsante, I., Bertolino, N., Bugatti, A., & Cristina, M. (2014). Indicators for urban quality evaluation at district scale. In Proceedings of the World Sustainable Building, (pp. 28–30 October). Barcelona, Spain. Dixon, J.A and Fallon, L.A. (1989): The Concept of Sustainability: Origins, Extensions, and Usefulness for Policy, in Society and Natural Resources Vol.2, pp.73-84. Düring, S. (2019). Between plan and reality: tracing the development dynamics of the Lanzhou New Area - a computational approach. Weimar, Germany: Bauhaus University Weimar. 85

Düring, S., Chronis, A., & König, R. (2020). Symposium on Simulation for Architecture and Urban Design. Vienna: Optimizing Urban Systems: Integrated optimization if spatial configuration. EuropeanCommission. (2018). Indicators for Sustainable Cities. Science for Environment Policy. Retrieved on August 2020. Fadl, M. S., & Karadelis, J. (2013). CFD Simulation for Wind Comfort and Safety in Urban Area: A Case Study of Coventry niversity Central Campus. International Journal of Architecture, Engineering and Construction, 131-143. Galanos, T., Chronis, A., Vesely, O., Aichinger A., & Koenig, R. (2019). Best of both worlds – using computational design and deep learning for real-time urban performance evaluation. 1st International Conference on Optimization Driven Architectural Design. Garau, C.; Pavan, V.M. Evaluating Urban Quality: Indicators and Assessment Tools for Smart Sustainable Cities. Sustainability 2018, 10, 575. Gehl, J. (2010). Cities for people. Washington DC: Island Press. Gil, J. (2015). Examining ‘edge effects’:Sensitivity of spatial network centrality analysis to boundary conditions. 10th International Space Syntax Symposium. Goodland, R. (1995). The Concept of Environmental Sustainability. Annual Review of Ecology and Systematics, 26, 1-24. Retrieved October 28, 2020, from http://www.jstor.org/stable/2097196 Hagström, H.; Widman, L.; Seth, E. von (2019): Association between temperature, sunlight hours and alcohol consumption. In PloS one 14 (9), e0223312. DOI: 10.1371/journal.pone.0223312. Hillier, B. (2009) Spatial sustainability in cities: organic patterns and sustainable forms. In: Koch, D. and Marcus, L. and Steen, J., (eds.) Proceedings of the 7th International Space Syntax Symposium. (pp. p. 1). Royal Institute of Technology (KTH): Stockholm, Swede Hillier, B., & Hanson, J. (1984). The social logic of space. Cambridge: Cambridgeshire: Cambridge University Press. Humphries, C. (2012). The science of cities: Life in the concrete jungle. In Nature 491 (7425), pp. 514–515. DOI: 10.1038/491514a. Huovila, A. & Airaksinen, M. & Pinto-Seppä, I. & Piira, K. & Bosch, P. & Penttinen, T. & Neumann, H.& Kontinakis, N. (2017). CITYkeys smart city performance measurement system. International Journal for Housing Science and Its Applications. 41. 113-125. Ibrahim, D. (2016). An overview of soft computing. Application of Fuzzy Systems and Soft Computing, ICAFS (pp. 34-38). Vienna: Procedia Computer Science 102. Isola, P.; Zhu, J.; Zhou, T.; Efros,A. (2017). "Image-to-Image Translation with Conditional Adversarial Networks," 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Honolulu, HI, 2017, pp. 5967-5976, doi: 10.1109/CVPR.2017.632. Jenks, G. (1967). The Data Model Concept in Statistical Mapping. In International Yearbook of Cartography 7 (pp. 186–190.).

Karimi, K. (2012). A configurational approach to analytical urban design: ‘Space syntax’ methodology. In URBAN DESIGN International 17 (4), pp. 297–318. DOI: 10.1057/udi.2012.19. Khan, M. A. (1995). Sustainable development: The key concepts, issues and implications. Keynote paper given at the international sustainable development research conference, 2729 march 1995, Manchester, UK. In Sustainable Development 3 (2), pp. 63–69. Available online at https://EconPapers.repec.org/RePEc:wly:sustdv:v:3:y:1995:i:2:p:63-69. Lawson, B. (2006). How designers think. The design process demystified / Bryan Lawson. 4th ed. Amsterdam, London: Architectural. Lawson, T.V. (1975). The determination of the wind environment of a building complex before construction". Bristol University, Department of Aeronautical Engineering, Bristol. Le Corbusier., Knight, P., Levieux, E., & Coltman, D. (1967). The radiant city: Elements of a doctrine of urbanism to be used as the basis of our machine-age civilization. New York: Orion Press. Lenormand, M., Picornell, M., Cantú-Ros, O. G., Louail, T., Herranz, R., & Herranz, R. (2015). Comparing and modelling land use organization in cities. Royal Society Open Science, https://doi.org/10.1098/rsos.150449. Lichtenberger, E. (1993). Vienna: Bridge Between Cultures. Wiley: Social Science. Marabout, L. O., Vladut, A. C., Cosoiu, C. I., & Anton, A. (2019). Pedestrian Wind Comfort Evaluation for Sun Valley Campus. Materials Science and Engineering471 (2019) 09200. oi:10.1088/1757-899X/471/9/0920. Marans, R., & Stimson, R. (2011). Investigating quality of urban life: Theory, methods, and empirical research. Soc.Indic.Res., (pp. 45,1-29). Mardaljevic, J., & Tregenza, P. (2017). Daylighting buildings: Standards andthe needs of the designer. In Lighting Res. Technol 2018 (pp. Vol.50: 63–79). The Society of Light and Lightining . Marshall, W. E., Garrick, N., & Marshall, S. (2014). Street Networks. In D. B. Editors R. Hickman, International Handbook on Transport and Development. Edward Elgar Publishing. Morphocode. (2018). Location Explorer https://morphocode.com/morphocode-exploreroverview/ retrieved on September 2020. Nes, A., Berghauser Pont, M., & Mashhoodi, B. (2012). Combination of Space Syntax With Spacematrix and the Mixed Use Index. The Rotterdam South test case. Eighth International Space Syntax Symposium. Santiago. Neuman, M., & Churchill, S. W. (2015). Measuring sustainability. Town Planning Review, 86(4), 457–482. Newman, M. E. J. (2010). Networks: An introduction / M.E.J. Newman. Oxford: Oxford University Press. Newton, P. (2012). ‘Liveable and sustainable? Socio-technical challenges for twenty-first. Journal of Urban Technology 18, 81-102.

Othmani Marabout, L.; Anton, A. (2019). Wind Tunnel Experiment for Pedestrian Wind Comfort in Sun Valley Campus. In IOP Conf. Ser.: Mater. Sci. Eng. 603, p. 52072. DOI: 10.1088/1757-899X/603/5/052072 Pfefferkorn, W. (1998). Vienna is Different (!?) — a Critical View of the Current Development of City Limits and Landscape in the Vienna Urban Area. In Urban Ecology. DOI: 10.1007/9783-642-88583-9_67. Reiter, S. (2010). Assessing wind comfort in urban planning. In Environment and Planning B Planning and Design 37(5) (pp. 857-873). RePEc https://doi.org/10.1068/b35154. Richthofen, A. von; Knecht, K.; Miao, Y.; König, R. (2018): The ‘Urban Elements’ method for teaching parametric urban design to professionals. In Frontiers of Architectural Research 7 (4), pp. 573–587. DOI: 10.1016/j.foar.2018.08.002. Sallis JF, S. B. (2009). Neighborhood built environment and income: examining multiple health outcomes. Soc. Sci. Med. 68, , 1285–1293. Seckmeyer, G., Zittermann, A., McKenzie, R. L., & Greinert, R. (212). Solar Radiation solar radiation and Human Health solar radiation human health. In Encyclopedia of Sustainability Science and Technology. 10.1007/978-1-4419-0851-3_455. Serrano, M.-A., Cañada, J., Moreno, J. C., & Gurrea, G. (2017). Solar ultraviolet doses and vitamin D in a northern mid-latitude. Science of The Total Environment, 574: 744. doi:DOI: 10.1016/j.scitotenv.2016.09.102 Si, F., & Wang, T.-H. (May 2015). Building Massing Optimisation in Early Design Stage- A Performance-Driven Design Based on Total Sunlight Hours. Emerging Experience in Past, Present and Future of Digital Architecture (pp. 583-592). Hong Kong: CAADRIA 2015. Territorial Agenda EU 2020. (2011) UnitedNations. (2020). The Sustainable Development Agenda. (accessed on 15 May 2020). Walmsley, K. (2017). https://www.autodesk.com/research/projects/project-discover. (Retriever on September 2020) Weber, M. (1978). The Nature of Social Action. In W. Runciman (Ed.) & E. Matthews (Trans.), Max Weber: Selections in Translation (pp. 7-32). Cambridge: Cambridge University Press. doi:10.1017/CBO9780511810831.005 Woodbury, R. (2010). Elements of Parametric Design. London: Routledge. Wu, H., & Kriksic, F. (2012). Designing for pedestrian comfort in response to local climate. Journal, 104-106, 397–407. Xenophon. (1979). Memorabilia – Xenophon in SevenVolumes. (Trans: Brownson CL). Cambridge: MA: Harvard University Press. Yu Ye, A. v. (2013). Measuring urban maturation processes in Dutch and Chinese new towns: Combining street network configuration with building density and degree of land use diversification through GIS. Journal of Space Syntax, 18-37. Zemljič, B., & Hlebec, V. (2005). ‘Reliability of measures of centrality and prominence’. In Social Networks, Vol. 27(1), p.73–88.

Affidavit

I hereby aﬃrm that this Master’s Thesis represents my own written work and that I have used no sources and aids other than those indicated. All passages quoted from publications or paraphrased from these sources are properly cited and attributed. The thesis was not submitted in the same or in a substantially similar version, not even partially, to another examination board and was not published elsewhere.

_______________ Date

________________________________ Signature

Master Thesis | 2020 | Diellza Elshani Integrated Urban Development & Design | Bauhaus Universität Weimar