Synergies and trade-offs in urban densification and environmental
sustainability on a regional scale: An Amsterdam case study
Thomas Hofman – 11066938 Credits: 30 EC
Period: 01/03/2020 – 28/07/2020
M Earth Sciences – Environmental Management Institute for Biodiversity and Ecosystem Dynamics University of Amsterdam
Supervisor: dr. ir. J. van Vliet Co-assessor: dr. L. H. Cammeraat Date: 28/07/2020
Abstract
While much qualitative research exists on the environmental impact of urban densification, quantitative research on this topic is sparse. This research tests the impact of urban densification on environmental sustainability on a regional scale in Amsterdam by plotting various aspects of environmental sustainability along a densification gradient. This study analysed the effect of urban densification on air quality, greenhouse gas emissions, ecology and environmental hazards. Based on a selection of indicators, a negative trend in environmental sustainability was observed along an urban densification gradient. With respect to the different categories of sustainability, a negative relationship was found between urban densification and air quality, environmental hazard proneness, and urban ecology. No impact of urban densification on sustainable energy use was observed. Correlations between different indicators of regional environmental sustainability were tested to identify synergies and trade-offs between these indicators. Among others, highly significant correlations were found between the urban green space(UGS) and air quality, natural capital and reduction of the urban heat island(UHI) effect. Quantification of the environmental impact of urban densification on a regional scale could improve urban planning, improving our understanding of the implications of urban densification on a regional scale. Evidence on trade-offs and synergies between different aspects of environmental sustainability could mitigate negative effects and enhance positive effects of urban planning.
Synergies and trade-offs in urban densification and environmental sustainability on a regional
scale: An Amsterdam case study ... 0
1. Introduction ... 3
1.1 Problem definition...3
1.2 Case study ...3
1.3 Research aim and objectives ...4
2. Theoretical background ... 4
2.1 History of urban densification ...4
2.2 Development of Amsterdam ...5
2.3 Urban environmental sustainability ...6
2.6 Climate change ...9
2.7 Urban ecology ... 10
2.8 Air Pollution ... 12
2.9 Natural hazards ... 13
2.5 Synergies & trade-offs ... 13
3. Method ... 14 3.1 Indicator selection ... 14 3.3 Area selection ... 17 3.4 Data Analysis ... 18 3.5 Spatial analysis ... 20 4. Results ... 20 Spatial projection ... 20 4.1 Air quality ... 22
4.2 Greenhouse gas emission ... 22
4.3 Urban ecology ... 23
4.4 Environmental hazards ... 24
4.5 synergies and trade-offs between indicators ... 25
4.6 Regression analysis of the total environmental sustainability score ... 26
5. Discussion... 27
References ... 33
1. Introduction
1.1 Problem definition
The United Nations predicts 68% of the global population to live in urban areas by 2050, an increase of 13% since 2018 (United Nations, 2019). Like several other cities, Amsterdam aims to cope with this projected population growth by expanding within the boundaries of the city (Gemeente Amsterdam 2011), a practice called urban densification. This management direction has been dominant in Europe for the past 50 years (Broitman & Koomen, 2015; Westerink et al., 2013) and is part of the 2040 vision of Amsterdam (Gemeente Amsterdam, 2011), as it would allegedly stimulate nature conservation in rural areas, reduces travel time, and reduces greenhouse gas emissions in urban areas. Moreover, the United Nations (UN) use the rate of land consumption to population density as an indicator for sustainable urban development in the UN Sustainable Development Goals (SDG) (United Nations, 2015). However, claiming such a compact city is more sustainable than the alternative, a sprawling city, can be considered an oversimplification as it is not sure to what extent urban densification affects other aspects of sustainable urban development. Some issues may be regarded as more sustainable in the compact city whereas others are more sustainable in the sprawling city. Westerink et al. (2013) called this ‘a trade-off in sustainability’. For example, whilst the net per capita emissions and pollutants might be lower in the densified city as a result of reduced travel time, accumulation of these pollutants may be higher as a result of reduced windspeeds and increased local emissions leaving more people exposed and affecting quality of life in these areas (Westerink et al., 2013).
The effects of densification on different aspects of environmental sustainability have been researched extensively. However, existing literature mainly applies a qualitative as opposed to a quantitative assessment of sustainability in urban areas (Shen et al.,2011, Lynch et al., 2011, Marzuki et al., 2011 & Westerink et al., 2013). Quantitative data for sustainability assessment are available in sufficient amounts (Gemeente Amsterdam, 2018) and multiple studies have assessed the dynamics of urbanization gradients (Frenkel and Ashkenazi, 2008, Irwin and Bockstael, 2007, Mubareka et al., 2011, Torrens and Alberti, 2000, Yu and Ng, 2007). The relation of sustainability to such a gradient however had to this point not been studied yet. Quantitative analysis of the impact of urban densification could reveal synergies and trade-offs between the different aspect of environmental sustainability such as the relation between species richness and air pollution for different levels of densification. It also allows for a trend analysis of sustainability indicators along a densification gradient.
1.2 Case study
The Netherlands has a relatively high population density compared to other European countries. Adequate urban planning is therefore required to safeguard the natural environment. Preservation of natural open spaces has been the dominant urban development strategy for the past 50 years (Broitman & Koomen, 2015). Due to its extensive history of implementing urban planning, Amsterdam is an interesting study site with an abundance of variation in densification across neighbourhoods due to environmental and cultural regulations (Broitman & Koomen, 2015).
Sustainable urban management in the Netherlands is considered successful. However, there are some incoherencies in management. Whilst Amsterdam set out to expand within the boundaries of the city, thus increasing the population density, the neighbourhood ‘IJburg’, the latest urban expansion in Amsterdam to date, has a population density very close to the city average population density (Gemeente Amsterdam, 2018). This average includes the remote industrial sites at the border of the city. If the population density in IJburg would have been higher, the current development of ‘Strandeiland’, which is the expansion of the IJburg neighbourhood, would not have been required to sustain urban growth. The development of such urban areas comes at the cost of the surrounding natural
In 2015, Amsterdam published the ‘Agenda Groen’ in which they issued the aim to increase urban green spaces to promote quality of life and ecology within the city. Between 2003 and 2016 a decrease in urban green spaces of 3.07 square kilometres occurred, 11% of the total urban green spaces (Giezen et al., 2018). This is in conflict with the Green Agenda programme which, among other targets, aimed to increase the green spaces as preventative measure against flooding. The municipality of Amsterdam reports an increase of green roofs as a measure to deal with the loss in green spaces (Giezen et al., 2018). However, it argues that the compensation is inadequate as the green spaces lost are 76 greater than the increase in green roofs. The quantification sustainability for city densification could help identify areas of conflict in relation to urban densification such as reduced protection against flooding and provide adequate alternatives.
1.3 Research aim and objectives
The aim of this research is to provide insights in sustainability trade-offs between sparsely and densely populated urban areas. The municipality of Amsterdam was chosen as a case study for its high availability of spatial environmental data, This study ultimately aims to answer the question: What is the effect of urban densification on the environmental sustainability of the city of Amsterdam? To answer this question the following objectives are set:
1. To identify quantitative indicators for urban environmental sustainability in the city of Amsterdam
2. To provide a quantitative analysis on the sustainability indicators to:
a. Identify potential relations between urban densification and individual sustainability indicators
b. Identify potential synergies and trade-offs between individual sustainability indicators c. Assess if urban densification is indeed more sustainable based on sustainability
indicators
3. To provide a spatial sustainability map of Amsterdam
Based on the widely accepted management strategy of urban densification and its inclusion as an indicator for urban sustainable development, this research hypothesized urban densification to be significantly positively correlated with environmental sustainability on a regional scale. This research furthermore hypothesized that the majority of indicators for environmental sustainability are significantly positively correlated with urban densification. Finally, this research hypothesised that synergies and trade-offs exist between the environmental sustainability indicators. Detailed hypotheses on each indicator require a more detailed description of the indicator and are thus provided in the theoretical background. Next to the expected results, this research sets out to provide a sustainability map of the municipality of Amsterdam which could potentially provide insights for urban planning in Amsterdam.
2
.
Theoretical background
2.1 History of urban densification
Urban densification itself has been interpreted in multiple ways in literature: The amount of residents per area, the amount of housing units per area and the amount of built-up land per unit area. Broitman & Koomen (2015) for example concluded urban densification throughout the Netherlands increased
based on an increased number of housing units per area. If such units include less residents per residential unit, this would not necessarily result in an increase in population per area, and if these housings are created by splitting existing units neither would the build area. Urban densification is part of the spatial structure of the city.
This spatial structure of cities is historically described from a monocentric point of view where land rents and population density are high at the center of the city and decrease monotonically with increased distance from the center (Anas et al., 1998). These densification gradients should theoretically flatten with increased higher incomes and decreased transportation costs. Most literature regarding residential expansion in cities is linked to planning and land-use change, and is mainly focused on the rate at which low density urban sprawl occurs (Frenkel and Ashkenazi, 2008, Irwin and Bockstael, 2007, Mubareka et al., 2011, Torrens and Alberti, 2000, Yu and Ng, 2007). Some literature focusses on the land-change dynamics between urban and rural areas= related to urban expansion (Bell Ka and Irwin, 2002 & Irwin and Bockstael, 2004).
In urban planning literature, densification is often considered sustainable, protecting the rural areas from urban sprawl. Daneshpour and Shakibamanesh (2011) compared sustainability in densified cities and urban sprawling and concluded that although sustainability could be part of the solution a compact city in not automatically sustainable. Moreover, since the 1990s the proposed compact city model has been challenged on three levels: whether it can deliver its benefits towards sustainability, whether it could actually be implemented in the urban environment, and if it is acceptable to the local population to implement such drastic changes (Daneshpour & Shakibamanesh, 2011). Quantitative research based on adequately assessed parameters could provide insights on the state of these models.
2.2 Development of Amsterdam
Urban land use in Europe has increased dramatically over the last decades. Resident increase is the main driver although cities that do not experience any population growth are expanding too (Broitman & Koomen, 2015). Cities appear to regain attractiveness due to the amenities they offer.
The Netherlands is one of the most densely populated countries in Europe. The major cities are relatively small with the capital of Amsterdam still just below one million residents. The polycentric nature of the Dutch urban development puts high pressure on open spaces. This resulted in the authorities on urban planning to implement the governing concept of accommodating anticipated growth whilst preserving rural and open areas for the past 50 years (Broitman & Koomen, 2015). Recent policies mainly focus on steering residential development towards large-scale urban development zones. Nature conservation laws pose another restriction on urban development in the Netherlands, the European policy for nature restoration and conservation prohibits member states to build in Natura 2000 areas. At the national level, National Ecological Network (NEN) acts as a network and corridor system to preserve nature.
Policies and urban planning in Amsterdam regarding densification and sustainability are presented in the Green Agenda (Gemeente Amsterdam, 2015). Amsterdam is growing faster than anticipated with rapid densification as a result. This densification affects urban green spaces (UGS). To overcome this decline in UGS the municipality aims to improve the attractiveness and functionality of the current green spaces.
In the structuurvisie 2040 report (Gemeente Amsterdam, 2011), a structural vision for Amsterdam in 2040, the municipality of Amsterdam states that urban densification and increased land use are important tools to create an economically strong and sustainable city. Development of high buildings in the city centre is limited as this is part of the UNESCO world heritage. As such, the cultural value of the inner city pushes residential development out of the city. Figure 1 gives an overview of the
projected building sites for 2040. Subsequently ‘special’ neighbourhoods with cultural value have a restriction to build above 30m to limit the impact of such new buildings on the architectural heritage. With regards to transportation, a key aspect of sustainable development, the municipality envisions an improved public transport network within the city, altering current roads. Bicycle and public transport are the main choices for short distance travel to date (Gemeente Amsterdam, 2011), however public transport experiences shortcomings in the middle travel distances (10km – 30km). Trains, in combination with bicycles, are more frequently used for longer distances (>30km) as an alternative to cars.
The variety in densification due to world heritage neighbourhoods, the cultural ‘special neighbourhoods’ and the new neighbourhoods at the border of the city make Amsterdam an interesting case study area. Indicators are required to assess the differences in sustainability within those neighbourhoods.
Figure 1: Planned development of high buildings in 2040: purple areas are destined for high buildings whereas red areas are prohibited and green areas are open green spaces (Gemeente Amsterdam, 2011)
2.3 Urban environmental sustainability
Addressing effects on urban environmental sustainability requires a measurable definition of what environmental sustainability is. This definition is frequently debated, and according to the US Federal Trade Commission in 2010 no clear understanding of the term existed among experts (Morelli, 2011), it was therefore argued the term cannot be defined or tested. However, the need for environmental sustainability is clear as current and future generations rely on ecosystem services for their existence. Ecosystem services are the benefits people derive from the natural environment (MEA, 2005). Morelli (2011) described sustainability as “meeting the resource and services needs of current and future generations without compromising the health of the ecosystems that provide them” (p. 6), by which he intends to operationalize the concept, increasing clarity in valuing ecological services. In 2015 the UN listed the 17 SDGs to be achieved by 2030, in which they proposed a list of indicators for urban sustainable development. The next section will discuss the indicators proposed by the UN and other literature regarding urban sustainable development.
2.3.1 The indicators for a sustainable city
To measure the effect of urban densification on the environmental sustainability of a city, a framework or model is required to assess the sustainability of a city in general. To assess the impact a city has on the environment, an extensive scope is required as cities are part of a larger infrastructure in which they are embedded and cannot be evaluated without this broader infrastructure (Ramaswami et al., 2012). To determine qualifiers for a sustainable city therefore requires knowledge on the surrounding ecosystem, the energy supply and ecology within the city. Especially in the case of city densification, the sustainable benefits of efficient infrastructure become apparent. Alberti (1996) stated that clear linkage patterns are required between urban patterns and the natural resource base to establish environmental impact. Such frameworks and linkages are based on indicators which act as performance measures of a system (Hiremath et al., 2013). The purpose of these indicators is to show how well a system is performing, what measures should be taken to address certain problems, and to reduce the amount of information required to understand the system.
Indicators for urban environmental sustainability have been assessed extensively (Hiremath et
al., 2013). An overview of indicators addressed in current literature is provided in table 1. Most of the
literature on indicators for urban sustainability is qualitative. However, some efforts have been made to quantify the indicators. Shen et al. (2011) for example produced a list of qualitative indicators for urban sustainability. While only a small amount of studies identified quantitative urban environmental sustainability indicators, the quantification of environmental sustainability in general has been done repeatedly in the past 20 years. The ecosystem services concept proposed in the millennium ecosystem assessment (MEA, 2005), the planetary boundaries concept by Rockström et al., (2009) and the sustainability indicators within the UN SDG’s (UN, 2015) all used quantitative data in the assessment of environmental sustainability. The indicators used in these frameworks overlap with indicators from literature specified on urban systems. Moreover, the elaborate UN SDG’s shared all indicators used in research specified on urban systems as projected in table 1, providing a suitable set of indicators for urban environmental sustainability explained in further detail.
.
Table 1: Overview of indicators of urban environmental sustainability in current literature
Factors Shen et al. (2011) Lynch et al. (2011) Marzuki et al. (2011) Westerink et al. (2013) UN SDG Geographically balanced x - x x x Freshwater availability x - x - x Wastewater treatment x - x - x Air/Atmosphere quality x - x x x Noise pollution x - x x x
Sustainable land use x x x x x
Waste generation and management
x x x - x
Effective transportation systems
x x - x x
Mechanisms to prepare and implement environmental plans
x x x x x
Biodiversity x x - x x
In 2015 the UN agreed on a set of 17 goals to be achieved by 2030, the UN sustainable development goals. These goals were accompanied with more specific targets and indicators to measure these targets. Within the UN SDG’s, urban environmental aspects are addressed repeatedly among different goals; within the SDG 6 of clean water and salination, SDG 7 of clean and affordable energy, SDG 9 of industry, innovation and infrastructure, SDG 11 of sustainable cities and communities, SDG 12 of responsible production and consumption, SDG 13 of climate action and finally SDG 15 on life on land. Although plenty of targets and indicators apply to the concept of environmental sustainability, only a select group of targets is subjective to regional changes. These indicators were categorized in overlapping themes in table 2. Subsequent sections will elaborate on the relation between urban densification and the urban sustainability themes derived from the UN SDG’s, identifying specific urban environmental sustainability indicators within these themes. After which, the theory on synergies and trade-offs of these indicators will be addressed.
Table 2: Recategorization of the United Nations sustainable development goals related to environmental sustainability on a local scale.
Category UN SDG indicators
Climate change ▪ 7.2.1 Renewable energy share in the total final energy consumption ▪ 13.2 Integrate climate change measures into national policies, strategies
and planning
Urban Ecology ▪ 11.3.1 Ratio of land consumption rate to population growth rate ▪ 15.1.1 Forest area as a proportion of total land area
Pollution ▪ 3.9.1. Mortality rate attributed to household and ambient air pollution ▪ 11.6.2 Annual mean levels of fine particulate matter (e.g. PM2.5 and
PM10) in cities (population weighted)
▪ 6.4.2 Level of water stress: freshwater withdrawal as a proportion of available freshwater resources
Environmental hazards
▪ 6.6.1 Change in the extent of water-related ecosystems over time ▪ 11.5.1 Number of deaths, missing persons and persons affected by
disaster per 100,000 people
▪ 11.5.2 Direct disaster economic loss in relation to global GDP, including disaster damage to critical infrastructure and disruption of basic services ▪ 1.b.1 Proportion of local governments that adopt and implement local
disaster risk reduction strategies in line with the Sendai Framework for Disaster Risk Reduction 2015-2030
▪ 13.1 Strengthen resilience and adaptive capacity to climate-related hazards and natural disasters in all countries
2.4 Climate change
Mitigating climate change is one of the key aspects of urban sustainability. One of the reasons why cities are considered an important element in the battle against climate change are their high GHG emission. In 2011 the UN estimated urban areas were responsible for 60-70% of all GHG emitted (UN, 2011). The UN SDG 11 of sustainable cities includes the target to reduce the adverse environmental impact of cities. However, most of the targets regarding climate change prevention are described in SDG7 of reliable, affordable and sustainable energy (UN,2020). A local contributor to high GHG emissions in urban areas is energy consumption by vehicles, households and industry. Distant energy consumption e.g. energy systems such as power plants act on a global scale (Kammen & Sunter, 2016). Transforming the polluting urban areas into sustainable cities can be achieved by the following two measures: 1) by reducing the energy usage in urban areas by shortening travel distances, improving walkability and isolation, and 2) by implementing renewable energy systems in urban areas (Kammen & Sunter, 2016). Indeed the UN SDG11 target 11.2 aims to achieve access to sustainable transport for all by 2030 (UN, 2020), and SDG target 7.2 sets out to substantially increase the share of renewable energy.
Urban densification has multiple advantages when it comes to energy usage and greenhouse gas (GHG) emissions. First, due to the increased proximity of work, travel distance shortens which reduces GHG emission (Broitman & Koomen, 2015; Clark, 2013). Next, densification requires highly developed infrastructure which improves walkability and public transport use (Clark, 2013). Clark (2013) furthermore, found a negative relationship of energy consumption and urban densification throughout the united states. Whilst the effect of reduced travel time and car use if often mentioned in the favour of the dense city opposed to the sprawling city, Ramaswami et al. (2012) argue energy usage and greenhouse gas (GHG) emissions are currently not well assessed in urban sustainability frameworks. They argue the effects of energy usage and its resulting GHG emission by remote powerplants is often not taken into consideration. Indeed densified urban areas could in theory reduce the amount of solar energy produced as a result of reduced rooftop surface per household.
Household energy use
In densely populated areas the average per capita housing space tends to decrease and building heights tend to increase. Both of these processes decrease the per capita housing surface area. Household energy use is, to a great extent, affected by the building properties. The greater the surface area of the building, the more heat is transferred (Danielski, Fröling & Joelsson, 2012). This affects both heating and cooling energy required by properties. As a result, the average household energy use is expected to decrease with increased urban densification. Indeed Resch et al. (2016) found that energy benefits from urban densification are most likely the result of floor area per capita rather than reductions in transportation energy. This research thus hypothesizes that urban densification has anegative effect on the household energy use.
Renewable energy
To combat the negative effects of climate change, green alternatives to fossil fuel energy sources are required. The UN aims to substantially increase the share of renewable energy in the energy mix by 2030 (United Nations, 2015). Although one of the highly developed countries in the EU, with 7.4% renewable energy in 2018, the Netherlands has the lowest share of renewable energy in their energy mix of all countries in the European Union (European Union, 2018). To overcome this low share of renewable energy, the Noord-Holland zuid region which Amsterdam is part of, documented a regional energy strategy (RES) concept. The main contributor to the renewable energy transition in the Amsterdam RES is solar energy. Amsterdam aims to increase the produced solar power to 400MW by 2030, using 50% of all the total rooftop area and providing 200.000 inhabitants in their energy demands
solar power and overshadowing effect of high buildings, the potentially available solar power decreases with increased urban densification (Lobaccaro & Frontini, 2014). This research thus hypothesizes that urban densification has a significant negative effect on the amount of solar power generated per capita.
2.5 Urban ecology
Pickett et al., (2001) described two definitions of urban ecology. In the scientific field of ecology ‘urban ecology’ is defined as the organisms in and around the city, whereas in urban planning ‘urban ecology’ is focussed on designing in such a way environmental impact is reduced. Urban environments are significantly different from rural areas and farmlands. First, temperatures are up to 5-10 degrees Celsius warmer in urban areas with increased temperatures with increase in artificial human made surface, this difference in temperature interferes with the growing seasons resulting in earlier flowering time and delayed leaf drop (Pickett et al., 2001). Next, due to increased temperature ozone levels are higher around the city and suburban areas potentially resulting in reduced crop yield by 5-10% in these areas. Subsequently, precipitation is enhanced up to 5-10% in and around the city due to a higher concentration of particulate condensation nuclei in urban atmospheres. Finally, hydrology is drastically altered in urban areas with evapotranspiration decreasing from 40% to 25% and surface runoff increasing from 10% to 30% (Pickett et al., 2001). These factors and a more hazardous environment due to highways could result in lower biodiversity.
Urban green spaces
Urban green spaces have gained a lot of attention in relation to urban sustainability and the study of urban ecosystem services in the Netherlands (Westerink et al., 2013; Derkzen et al., 2015; Giezen et
al., 2018; Paulin et al., 2020). Although placed in the category of urban ecology as UGS provides the
basic necessities for biodiversity, with regards to sustainability, UGS could have just as easily be placed under the category of air quality due to the air purification as a result of UGS, under the category of climate change due to the natural carbon sink properties, or under the hazard reduction category due to the water retention, permeable layer and heat reduction properties. The relation to urban densification is evident as increasing the population density is most likely increasing the portion of grey infrastructure (e.g. buildings and roads) at the cost of green infrastructure. This research therefore hypothesizes that urban densification has a significant negative effect on the percentage of UGS. Unsurprisingly, many studies with regards to urban densification and environmental sustainability are written from a UGS perspective, emphasizing the need for UGS in the Netherlands (Westerink et al., 2013; Derkzen et al., 2015; Giezen et al., 2018; Paulin et al., 2020). Although these approaches thoroughly analyse the benefits which these UGS offer, they do not address the potential environmental benefits from the absence of UGS, such as reduced CO2 emissions due to reduced travel time as people in closer proximity to work due to increased urban densification. These trade-offs could provide insights in the field of urban planning and urban sustainability. One could argue, acknowledging the uncertainty of the environmental benefits of UGS would undermine the use of UGS as an indicator for environmental sustainability. Nevertheless, this paper argues the ecological value of increased species richness and biomass are inherent to UGS on a regional scale, making it an important indicator on a regional scale.
Biodiversity
Urbanization has major consequences for environmental change and biodiversity. Urban densification and urban growth challenge the existence of ecosystems of conservation concern in and around urban regions and the survival of species within urban environments (Kowarik et al., 2020). Some urban environments however host high richness of animal and plant species, including endangered species (Aronsen et al., 2014; Ives et al., 2016). Although the UN SDG’s do not reflect on biodiversity in urban areas, multiple recent studies do emphasize the importance of biodiversity in cities; biodiversity
generates and supports a range of ecosystem services, promotes physical and mental health and maintains peoples connection with nature (Haase et al., 2014; Hartig et al., 2016; Ives et al., 2017). The relation between urban landscapes and biodiversity is highly dependent on the type of biodiversity. While overall biodiversity tends to increase, due to the increased number of alien species, native biodiversity tends to decrease. Furthermore, whilst native biodiversity decreases, some studies found urban areas can habitat some endangered native species by providing specific niche habitats (Ives et al., 2016). Kowarik et al.(2020) found the effect of urban densification on biodiversity to be highly region, species and ecosystem dependent. Despite these dependencies, Kowarik et al. (2020) found UGS to positively affect the biodiversity in urban systems. Due to the high dependency or region specific features, this research does not hypothesize on the effect of urban densification on species richness as this hypothesis would be unfounded.
Natural capital
In recent years, many studies have tried to put a value on nature by assessing the services nature has to offer like in the ecosystem services concept (MEA, 2005). Putting a value on nature has the advantage that different types of nature can be compared. Policy decisions can be based on this value. In some cases such as in the economics of ecosystems and biodiversity concept (Kumar & Martinez-Alier, 2011), economic impact of nature can be assessed based on a list of indicators. Fang et al. (2018) stressed the importance for natural capital in urban systems due to the high ecological footprint of urban systems. Tratalos et al. (2007) found a decreasing trend in natural services along a densification gradient by comparing multiple cities in the United Kingdom.
The municipality of Amsterdam has released several natural capital reports of which the 2016 edition is the most recent (Gemeente Amsterdam, 2016). In their analysis the put non-economic value on nature by giving a 1 to 5 score for natural value. Since the report is published multiple times, spatial and temporal analysis is possible. The following four maps form the basis for the natural capital assessment: 1) The amount of protected species in a given area, 2) The level of nature in the area, 3) The replaceability and 4) The addition to the ecological structure. Explained in further detail below.
• Biodiversity: The level of biodiversity is determined by the total number of species groups protected by the Flora and Fauna Act (mammals, birds, reptiles, amphibians and plants), corrected for the Red List value1 of each species ("index"). The map gives an indication of the main hot spots of biodiversity in and around Amsterdam. Five levels have been defined for biodiversity.
• Natural level: The level of nature is determined by the degree to which an area has a natural function. Three levels are defined by the municipality of Amsterdam, areas which do not have any natural function such as industrial sites, score the lowest and places where nature can run its course have a have high scores.
• Replaceability: The degree of replaceability is determined by the presence of old landscape elements and the presence of old trees based on these indicators, four gradations for replaceability defined. The substitutability of an area is defined as the time required to replace it. The replaceability is mainly related to the presence of old trees and the appearance of old, matured soils. Areas of low replaceability lie both in and around the city. key factors are the landscape elements, such as peat meadow areas and riparian lands, as well as areas with older planting that are difficult to replaced.
• Contribution to the ecological structure refers to the protection of national areas. In addition to the Amsterdam Ecological Structure, the national protected areas are also included in the map. For the Natura 2000 areas international agreements have been made.
Given that most urban areas do not have a natural function, are more frequently subjective to change and do not contribute to necessarily contribute to the ecological structure, this research hypothesizes urban densification significantly lowers the natural capital in a given area.
2.6 Air quality
The World Health organization (WHO) estimated ambient air pollution kills 4.2 million people annually, making it the biggest environmental threat to public health (WHO, 2016). This led to inclusion of air quality indicators in the UN sustainable development goals (SDG’s) in three ways: In the SDG3 of health by monitoring the mortality rate of ambient air pollution, in the SDG11 of urban sustainable development by monitoring the fine particulate matter (PM2.5 and PM10) in cities, and in SDG7 of sustainable energy by monitoring the proportion of the population with reliance on clean fuels and technology. The inclusion of air quality in urban SDGs is based on the notion that air quality in urban areas is lower than in non-urban areas. Recent Images by the European space agency (Figure 2) illustrate this difference in air pollution between urban and rural areas, with high concentrations of NO2 around large metropolitan areas. Rijnders et al. (2001) and Schweitzer & Zhou (2010) indeed found air quality was significantly lower in densely populated urban areas.
Figure 2 NO2 concentrations in France, Spain and Italy on march 2019, ESA 27/03/2020
Air pollution is the collective term for emitted particles which have a negative effect on health and the environment, the major pollutants being: NOx, CO, O3 & SO2. These pollutants cause strokes, heart disease, lung cancer and chronic respiratory diseases (WHO, 2016). PM2.5, a complex combination of solid, liquid, organic and inorganic particulate matter, is most often used as indicator of air pollution (WHO, 2016). However, concentrations of the main polluters, such as NO2, can be considered as marker for air quality too (Rijnders et al., 2001), NO2 is an important constituent of PM2.5 (WHO, 2016). In collaboration with the WHO, the European environmental agency (EEA) and the United States Environmental Protection Agency (EPA) set air quality standards for NO2, these standards are 40g/m3 NO2 on average annually and 99.64 g/m3 NO2 respectively (EPA, 2018; EEA, 2019). The
EEA, in their 2019 air quality report, concluded 98% of all NO2 concentrations exceeding the standard were measured in urban areas. Most of the concentrations were measured at traffic stations, NO2 is mostly emitted by traffic and is considered a urban problem (EEA, 2019). This research hypothesizes urban densification has a significant positive effect on the ambient NO2 concentration.
2.7 Natural hazards
Urbanization significantly influences the way these areas react to extreme events such as extreme precipitation or extreme temperatures. These extreme events form a potential risk for the inhabitants of these urban environments. Due to the large amount of people living in urban areas, it is vital to understand the way urban densification influences potential natural hazards. The UN SDG 11.2 pursues a strengthened adaptive capacity to environmental hazards by 2030 (UN, 2020). The UN SDG’s furthermore incorporate a range of sustainability indicators for environmental hazards listed in table 2. One of the ways in which urban densification enhances natural hazard exposure is by increased extreme temperatures in highly populated areas, referred to as urban heat islands (UHI). Urban areas can be up to 4 C warmer compared to rural areas (Oleson et al., 2011). This UHI effect has significant health implications; Baccini et al. (2008) found extreme heat significantly increased respiratory related deaths and increased the mortality rate among elderly, while Basu (2009) concluded that increased temperatures are associated with increased risk for those dying from cardiovascular, respiratory, cerebrovascular, and some specific cardiovascular diseases, such as ischemic heart disease, congestive heart failure, and myocardial infarction, based on his review on 36 articles.
Next to the health implications, the UHI effect causes extreme drought events to occur more frequently, amplified by the ongoing temperature rise as a result of climate change (Prudhomme et al., 2014). Oleson et al. (2011) note that population density gradients are likely to affect the effect of urban heat islands as the effect of high density high building areas would likely be more extreme, but state that this requires further research. The rural-urban differences occur predominantly due to differences in urban and rural partitioning of available energy into sensible, latent and storage heat fluxes controls these aspects of the heat island. Based on the extensive research on the matter this research hypothesizes urban densification to significantly increase the UHI effect of a regional scale.
2.8 Synergies & trade-offs
Most of the indicators of urban environmental sustainability addressed in the previous sections are likely to be affected by urban densification on a regional scale. Besides the influence of urban densification, environmental sustainability indicators are likely to interact. Indeed, multiple synergies and trade-offs in sustainability are recognised by the IPCC and the UN SDGs (IPCC, 2018). There are however different ways to interpret the definition of synergies. While some literature relates to synergies as the correlation between two parameters, others refer to the added effect by two correlating parameters, which is larger than the sum of its parts (Luukkanen et al., 2012).
The first interpretation is important in determining correlation between urban densification and the indicators as this indicates a potential relation between densification and the sustainability indicator. Given that such correlation exists, an advanced analysis can assess the combined effects of two or more indicators and assess if certain synergies or trade-offs are present in relation to each other. For example the added effect reduced vegetation and reduced air quality have on one-another. Luukkanen et al. (2012) provided a new way to measure such synergies, building on the Advanced Sustainability Analysis (ASA) approach developed the European framework programmes FP6 and FP7.
3. Method
This research uses the frameworks outlined for urban environmental sustainability in section 2.3.1, combined with the abundant availability of quantitative data available at regional scale by the municipality of Amsterdam (Gemeente Amsterdam, 2020) to constructively analyse the impact of urban densification on environmental sustainability on a regional scale. The indicators were assessed for each of the 98 neighbourhoods in Amsterdam, allowing for comparison of indicators per neighbourhood. Sustainability was compared to a densification gradient to identify synergies and trade-offs between densification and sustainability indicators. A flowchart of the methods applied in this study is given in figure 3. The following sections will firstly describe the indicator data used and the modification of these data, secondly the area selection will be discussed, subsequently methods used for data analysis are addressed and finally elaborate on the methods for spatial projection of urban sustainability.
Figure 3: Workflow on how to go from indicator data to sustainability maps and trade-off/synergies with urban densification
3.1 Indicator selection
Building on the framework of indicators addressed in section 2.3 to 2.7 the indicators for urban environmental sustainability used in this research are based on urban sustainability literature, the UN SDG’s and the available data in Amsterdam. This led to the selection of environmental sustainability categories used in this research. The four categories taken into account are pollution, urban ecology, environmental hazards, and climate change. For each category, indicators were used to determine the effect of urban densification on the category. Air quality was indicated by the concentration of ambient NO2 (µg/m3). Urban ecology was indicated by the amount of endangered species per neighbourhood, the percentage of UGS and the natural capital. The effect on climate change was indicated by annual household gas- (m3/y) and electricity (kWh/y) use and the amount of solar power produced (kWp/y). finally effect of urban densification on environmental hazards was indicated by the UHI island effect (℃). How these indicators were measured per neighbourhood is described in further detail below.
3.1.1 Air quality
Nitrogen dioxideData for the NO2 concentrations was downloaded from the database from the municipality of Amsterdam (Gemeente Amsterdam, 2020) in .MIF and .MID format and opened inQGIS (version
number). The NO2 database consists of 133 locations with data on the average NO2 concentration in 2019. Measurement locations are scattered throughout different neighbourhoods and required interpolation to derive the average NO2 concentration per neighbourhood. The dataset contained 5 measurements taken on a highway, as these are not representative for the neighbourhoods these values were removed from the dataset.
The point cloud was converted into a raster format using inversed distance weighting (IDW). This method bases the value of each raster value is based on the surrounding values of the point cloud with increased effect with increased proximity. The distance coefficient () of 2.085 was decided after multiple runs. The first results contained a lot of ‘bullseyes’ which indicate an overestimation of the effect distance from the measuring points. This is related to the selected distance coefficient. After testing with multiple distance coefficients a (QGIS) P value of 10 appeared most suitable to the dispersal of NO2 in Amsterdam. This P value corresponds to a distance coefficient () of 2.085 used in the Shepard’s method, which is normally used for IDW (Shepard, 1968). Mesnard (2013) argues the value of should be based on the dispersal dynamics of the specific polluter. In the case of clouds or plumes he argues should have a value between 2 and 4. Helle et al. (2011) found IDW with an of 2 was competitive with models and kriging in mapping pollution from plumes, validating the use of a value of 2.085.
A disadvantage of IDW is its inability to calculate values exceeding the upper and lower limit of the data points. When calculating the mean NO2 concentration per neighbourhood this problem can be discarded as there is no reason to assume higher values are more likely to occur than lower values (or the other way around), as such these possibilities cancel each other out.
From the interpolated raster file, mean NO2 concentrations were calculated using zonal statistics. To test for correlations between the NO2 correlation per neighbourhood and the population density linear regression analysis was conducted using R-studio.
3.1.2 Sustainable energy use
Natural gas and energy useThe data used to map the natural gas use and energy use per household has been downloaded from the website of the Central Bureau of Statistics (CBS). The CBS projected the household energy- and gas use for 2014. Not each single household is included as some households are lacking data. Furthermore, the size of each household can vary. As a result, the per capita energy- and gas use could not be calculate as some households were missing. It was therefore decided to use the mean household energy- and gas use instead of the per capita energy- and gas use. The energy and gas use data frame were clipped with the neighbourhoods data frame to deduce the average energy- and gas use per neighbourhood.
Renewable energy
This study uses the per person potential solar energy per neighbourhood as an indicator for sustainable energy. Data was used on the kWp performance from all the solar panels in Amsterdam based on detection with 2017 aerial photos with a 94% coverage, provided by the municipality of Amsterdam (Gemeente Amsterdam, 2017). The sum of the kWp solar energy of each neighbourhood was divided by the neighbourhood residents which resulted in the per capita potential kWh solar energy per neighbourhood. A correlation and regression analysis on the per capita potential kWh solar energy and population density was performed to potential trends. The 4617 solar panels detected by aerial photos by the municipality of Amsterdam were summed per neighbourhood and divided by the number of residents to derive the potential solar power in kWp per person per neighbourhood.
3.1.3 Urban ecology Urban green spaces
Data on the UGS in this study was acquired from the land use shapefile provided by the municipality of Amsterdam. The dataset contains an elaborate qualification of land use types. To derive the percentage of UGS per neighbourhood, first the different subclasses available in the land use file had to be clustered into classes of ‘green’, ‘semi’, ‘build’ & ‘water’ land use. The subclasses were reclassified accordingly in Table 3. After reclassification, the percentage of green area was used as sustainability indicator. The UGS percentage was plotted against a densification gradient perform correlation tests and regression analysis.
Table 3: Reclassification of the land use data from the 2017 land use map (Gemeente Amsterdam, 2017)
Endangered species count and natural capital
The number of species per quarter square kilometer and the natural capital data is projected by Brouwer & Timmerman commissioned by Ruimte en Duurzaamheid Gemeente Amsterdam (Gemeente Amsterdam, 2016). The map of natural capital shows the natural capital on a 1 to 5 score for natural areas in Amsterdam (The basis of this score is described in further detail in the theoretical background). Other areas which are not considered natural are left out of the map. As the percentage of natural area per neighbourhood is already taken into consideration in the UGS indicator, this indicator only indicates the average quality of the green space per neighbourhood. The average score per was calculated by splitting all the green plots at bordering neighbourhoods, resulting in a number of green plots with different scores per plot. As the plots differ in size, the score of each plot was multiplied by the size, after which the sum of all plots in a given neighbourhood was divided by the total natural square meters of that specific neighbourhood.
The data on the observed number of endangered species, protected by the Flora and Fauna Act, was projected each 500 meters as point data. This point cloud was interpolated using IDW. The IDW interpolated point cloud was then clipped by the neighbourhood shapefile to derive the average neighbourhood species count. As the neighbourhoods differ in size, the average score was used rather than the highest score per neighbourhood. Nevertheless, the highest score might be more interesting as
Category Keyword
Built-up area ▪ Industrial sites
▪ Building sites ▪ Recreational ▪ Semi paved terrain ▪ Railways
Green area ▪ Forest
▪ Open dry natural terrain ▪ Open wet natural terrain ▪ Agricultural terrain ▪ Parks
Semi green area ▪ Graveyards
▪ Sports fields
Water ▪ Lakes
▪ Water >6m wide
this indicates the specific neighbourhood might have some niche environment suitable for endangered species, which arguably, has a higher value for overall biodiversity.
3.1.4 Environmental hazards Urban heat island effect
The UHI data was derived from the department of climate adaptation at the municipality of Amsterdam. The data projects a simulated heatmap of the average temperature in Amsterdam on 01 july 2015 at 12:00-18:00. This date was selected by the national institute for public health and the environment as representative for the 1 in 1000 day max apparent temperature indication. The simulation design is based on solar radiation, wind simulation, air moisture content, shadow formation and land use among other factors. The simulation is described in further detail by the RIVM by de Nijs et al. (2019). The heatmap was available as raster data and was clipped per neighbourhood to calculate the average temperature per neighbourhood.
3.2 Area selection
To compare different types of densification and sprawling, distinct types of urban planning are required. Giezen et al. (2018) researched the effects of conflicting policies regarding densification and urban green spacing in Amsterdam using spatial analysis to assess the implications of these policies by comparing density in 2003 with density in 2016. Unlike Giezen et al. (2018), this research aims to compare different neighbourhoods in Amsterdam instead of temporal differences. This allows for a more robust way of comparing emissions and species richness as temporal effects can be neglected. This research used neighbourhoods to project the difference in urban densification is by population density per neighbourhood. The municipality of Amsterdam provides data on the resident count per neighbourhood for al 98 neighbourhoods in Amsterdam at their website (Gemeente Amsterdam, 2018).
By dividing Amsterdam in different neighbourhoods, features adjacent to multiple neighbourhoods such as urban green spaces (UGS) are sometimes attributed to one neighbourhood, neglecting the effect these UGS have on other surrounding neighbourhoods. This results in discrepancies in defining the population density in the area and in accrediting UGS percentages to neighbourhoods. The population density will be relatively lower in the neighbourhood to which the UGS is attributed and higher in areas surrounding the UGS. Another issue with attributing UGS to a single neighbourhood occurs in assessing correlations between UGS and other environmental indicators, such as air quality. The effect of air purification by UGS will most likely be present in all surrounding neighbourhoods but the UGS areal data will only be available in the attributed neighbourhood.
Figure 4 illustrates the effect of UGS on population density. Neighbourhood F86 (13,923 inhabitants) has a population density of 90/ha, whereas area E42 (10,367 inhabitants) has a population density of 247/ha. Disregarding the UGS and water from neighbourhood F86 (39.41 ha and 8.59 ha respectively) would increase the population density to 131/ha. Adding the UGS to E42 would reduce its population density to 112/ha.
Figure 4: Illustration of the uneven distribution of UGS per neighbourhood
To cope with this problem three options are available: 1) do nothing and regard the issue of unshared area as a fallacy of the neighbourhood shapes, 2) disregard all UGS areas neglecting their beneficial effect on the urban environment or 3) divide all UGS over the surrounding neighbourhoods.
Option 3 could be considered as the most ‘fair’ option however, the division would be rather arbitrary as the range to which UGS affects surrounding neighbourhoods is dependent on the type of UGS and the influence on the surrounding differs per factor, Derkzen et al. (2015) described these different effects, table 4 provides an overview of the different effects. Next to the subjectivity of the division, there are some practical issues. Imagine for example a scenario at which the 50 meter range of the bordering part of the UGS would imply the UGS score per neighbourhood. As the circumference is not a direct function of the surface this would imply the effect of long stretched surfaces would have a higher UGS score compared to more round shaped surfaces.
Table 4: Overview of ecosystem service indicators and supply rates, specified per urban green space type by Derkzen et al. (2015)
The table by Derkzen et al. (2015) also implies why option 2 is no concrete option as it illustrates all the benefits UGS offers to the environmental sustainability of urban areas. Disregarding al UGS would imply all the beneficial attributes of UGS would not be assessed in this report. In conclusion, as no measure adequately addresses the issue and the issue presents itself only in a limited number of cases, no additional steps are taken in redefining the neighbourhoods or excluding UGS from the neighbourhoods. As a result, sustainability indications for neighbourhoods whitch are bordering an UGS should be interpreted cautiously.
3.3.1 Datafile and correlation testing
Collection of the indicator data resulted in spatial data projected in a neighbourhood data frame of Amsterdam in QGIS with indicators as data points per neighbourhood. This spatial data was compiled per neighbourhood and individual regression analyses of indicators was performed to test if the sustainability indicator was influenced by densification by regression analysis. Depending on the nature of the indicator, this positive or negative regression indicates a positive or negative effect of urban densification on environmental sustainability.
3.3.2 Indicator combination
The overall sustainability score is an indication of the sustainability of one neighbourhood compared to other neighbourhoods in Amsterdam. Ideally the score would be tested against a benchmark value which indicates sustainability. However, this benchmark value would be rather arbitrary as no uniform definition of sustainability exists. This research assumes a linear relationship between environmental impact and the selected sustainability indicators. To compare the different indicators with one another, the indicators had to be normalized. Due to the occurrence of outliers, a z-score normalization was considered most suitable for normalization (2).
𝑍𝑠𝑐𝑜𝑟𝑒 = 𝑥 − 𝑚𝑒𝑎𝑛
𝑠𝑡𝑎𝑛𝑑𝑎𝑟𝑑 𝑑𝑒𝑣𝑖𝑎𝑡𝑖𝑜𝑛 (2)
Indicators for which an initial high value resembled a negative environmental impact, e.g. NO2 concentration, gas- and electricity usage, and higher average temperature, were inversed, making positive values negative and vice versa. By using this z-score standardization, values above 0 indicate an above average sustainability score for the given indicator and values below 0 indicate a below average score. The z-score normalization allows for addition of different indicators for sustainability to create an overall sustainability score to test the hypothesis of more sustainability with increased densification. A clear overview of each indicator for each neighbourhood is provided by on indicator bar plots for each neighbourhood.
3.3.3 trade-offs and synergies between indicators
The effect of urban densification on environmental indicators could partially be explained by amplifying effects of certain indicators on one another. Trade-offs and synergies between the different indicators were tested using a Pearson correlation test, resulting in the strength and the significance of the correlations. The correlation tests were conducted using the z-score data to improve the interpretability of the correlations. Positive correlations indicate a synergy with regards to environmental sustainability whist negative correlations indicate a trade-off. Correlations between indicators could explain part of the regression in overall sustainability score and urban densification.
3.3.4 Data storage
All data assembled in this research is can be found in a google drive workspace following this link:
https://drive.google.com/drive/folders/1kHMpXRBAYYKqAH0hWQzmZoHCkfsPFKBl?usp=sharin g. The Neighbourhooddata.xlsx file contains all neighbourhood properties calculated in QGIS and all z-score normalized values. The metadata information is provided in this file. Next, the R-studio script is stored as R file and should be useable (using R-studio 4.0.2) by importing the csv files form the csvfiles map into the working directory. Additional packages used in the analysis are described in the script. Finally the QGIS basemaps are shared in the allqgisdata.qgs file and are usable using QGIS 3.10.
3.4 Spatial analysis
Spatial analysis and mapping of general sustainability in Amsterdam is based on the normalized z-score data. This research used the relative effects of each indicator by normalizing them based on z-scores after which the sustainability average (sum of all indicators divided by the amount of indicators) per neighbourhood can be calculated. Neighbourhood scores >0 reflect more sustainable neighbourhoods and scores <0 are considered less sustainable neighbourhoods. These z-scores were mapped for all neighbourhoods in Amsterdam with a population exceeding 10 people per hectare using QGIS..
4. Results
Spatial projection
To provide an overview of the indicators per neighbourhood, a composite image was created showing the neighbourhood values (figure 6). The composite allows for a quick comparison between different neighbourhoods. To provide more detailed information for each indicator, a high resolution image for each indicator is added to the supplementary files.
Figure 5: Projection of all indicators per neighbourhood. High resolution images of the indicators per neighbourhood are presented in de supplementary files to improve readability of the legend.
4.1 Air quality
4.1.1 NO2 concentrations
Regression analysis revealed a significant correlation between population density and the neighbourhood average NO2 concentration (p = 0.001). The regression analysis of the linear model estimated an intercept of 25.53 g/m3 with an increase of 0.02 g/m3 per person/ha. The R2 of 0.106 indicates 10.6% of the observed increase in NO2 concentration can be explained by the increase in population density. The average NO2 concentration was plotted against the population density in figure 7.
Figure 6: Neighbourhood average NO2 concentration plotted against the population density per neighbourhood
4.2 Sustainable energy use
4.2.1 Solar power
Regression analysis revealed a significant correlation between population density and the neighbourhood per capita solar power (p-value = 9.647e-07). The regression analysis of the linear model estimated an intercept of 58.55 kWp and an decrease of 0.26 kWp per person/ha. An R2 of 0.2398 indicated about 24% of the per capita solar power could be explained by the population density of the neighbourhood. The per capita solar power per neighbourhood was plotted against the population density in figure 8.
Figure 7: Neighbourhood average solar power potential (kWp) plotted against the population density per neighbourhood
4.2.2 Household gas use
Regression analysis revealed a significant correlation between population density and the neighbourhood household gas use (p-value = 0.01). The regression analysis of the linear model estimated an intercept of 1111.8 m3/y/household gas use and an decrease of 1.12 m3/y/household per person/ha. An R2 of 0.065 indicated about 6.5% of the per household gas use could be explained by the
population density of the neighbourhood. The household gas use per neighbourhood was plotted against the population density in figure 9.
Figure 8: Neighbourhood average gas used per year per household plotted against the population density per neighbourhood
4.2.3 Household electricity use
Regression analysis revealed no significant correlation between population density and the neighbourhood household energy use (p = 0.12). The household gas use per neighbourhood was plotted against the population density in figure 10.
Figure 9: Neighbourhood average electricity used per year per household plotted against the population density per neighbourhood
4.3 Urban ecology
4.3.1 UGS
Regression analysis revealed a significant correlation between population density and the UGS percentage per neighbourhood (p = 4.98e-11). The regression analysis of the linear model estimated an intercept of 25.33% and an decrease of 0.11 % UGS per person/ha. An R2 of 0. 0.3895 indicated about 39% of UGS% in neighbourhoods could be explained by the population density of the neighbourhood. The percentage of UGS per neighbourhood was plotted against the population density in figure 11.
4.3.2 Biodiversity
Regression analysis revealed no significant correlation between population density and biodiversity percentage per neighbourhood (p-value = 0.525). The mean endangered species count per 500 meter plot per neighbourhood was plotted against the population density in figure 12.
Figure 11: Neighbourhood average endangered species count per 500m sample point plotted against the population density per neighbourhood
4.3.3 Natural capital
Regression analysis revealed a significant correlation between population density and the natural capital value percentage per neighbourhood (p = 0.0037). The regression analysis of the linear model estimated an intercept of 2.404 natural capital score and a decrease of 0.1544 per person/ha. An R2 of 0.099 indicated about 9.9% of the natural capital score could be explained by the population density of the neighbourhood. The natural capital score per neighbourhood was plotted against the population density in figure 13.
Figure 12: Neighbourhood average natural capital score plotted against the population density per neighbourhood
4.4 Environmental hazards
4.4.1 Urban heat island effect
Regression analysis revealed a significant correlation between the max. temperature and the population density per neighbourhood (p-value = 1.09*10-07). The regression analysis of the linear model estimated an intercept of 39.96 ℃ and an increase 0.009 ℃ per person/ha. An R2 of 0.267 indicated about 26.7% of the variation in temperature could be explained by the population density of the neighbourhood. The max temperature per neighbourhood was plotted against the population density in figure 14.
Figure 13: Neighbourhood average maximum temperature plotted against the population density per neighbourhood
4.5 synergies and trade-offs between indicators
A correlation test was conducted to establish synergies and trade-offs between certain parameters. Figure 15 shows all significant correlations between different indicators. Table 5 reflects all the significance scores. The strength of the correlation was reflected by the Pearson correlation test, the results of the Pearson r are presented in the supplementary files in Table 7.
Figure 14: correlation between different normalized indicators. Blue colours indicate a positive correlation and red colours indicate a negative correlation. In the case of no circle, no significant correlation was found.
Table 5: Pearson correlation test. green values have a P < 0.05
Indicators Pop/ha UGS NO2 Solar Species Electricity Gas UHI Nat. capital
Pop/ha 0,00 0,00 0,00 0,52 0,12 0,01 0,00 0,00
UGS 0,00 0,01 0,05 0,25 0,02 0,04 0,00 0,00
NO2 0,00 0,01 0,64 0,83 0,02 0,45 0,00 0,12
Electricity 0,12 0,02 0,02 0,01 0,70 0,00 0,00 0,01
Gas 0,01 0,04 0,45 0,83 0,60 0,00 0,00 0,00
UHI 0,00 0,00 0,00 0,98 0,42 0,00 0,00 0,00
Nat. capital 0,00 0,00 0,12 0,65 0,05 0,01 0,00 0,00
4.6 Regression analysis of the total environmental sustainability score
The sum of all environmental sustainability categories, e.g. climate change, air quality, ecology and hazard reduction, was used to indicate the total sustainability score per neighbourhood. Regression analysis on the total sustainability score per population density indicates the overall effect of population density on environmental sustainability. A significant negative correlation was observed between the total sustainability score and the population density (p-value = 3.3*10-9). The adjusted R2 of 0.322 indicates approximately 32.2% of the overall sustainability z-score can be explained by the population density. The regression is plotted in figure 16. The sum of z-scores per neighbourhood is projected in figure 17.
Figure 15: The neighbourhood z-score sum of all sustainability categories plotted against the population density
Figure 16: The sum of all z-score normalized sustainability categories for each neighbourhood in Amsterdam with a population density exceeding 10 residents per hectare. A Z-score >0 indicates an above average environmental sustainability for the particular neighbourhood.
The sum of all individual indicators, not summarizing the indicator categories first, resulted in a weaker correlation. Regression analysis showed a significant negative correlation between the sum of all indicators and population density (p = 1.3*10-6). The adjusted R2 of 0.282 indicates approximately 28.2% of the overall sustainability z-score can be explained by the population density. The regression is plotted in figure 18.
Figure 17: The neighbourhood z-score sum of all sustainability indicators plotted against the population density
Finally, the combined z-score for the greenhouse gas and the ecology class were calculated to describe the effect of urban densification on these two classes. No significant regression was found for the GHG emission and population density. For ecology significant regression was found (p = 1.25*10-5). The adjusted R2 of 0.18 indicates approximately 18% of the ecology score can be explained by population density.
Figure 18: : The neighbourhood z-score sum of all indicators for ecology and GHG plotted against the population density
5. Discussion
5.1 main results
This research set out to test the effect of urban densification on environmental sustainability on a regional scale, identifying synergies and trade-offs in various aspects of environmental sustainability along a densification gradient. A set of indicators was selected to infer the effect of urban densification on air quality, greenhouse gas emissions, ecology and environmental hazards. To ultimately answer the research question: What is the effect of urban densification on environmental sustainability in Amsterdam? This research hypothesized that urban densification would significantly enhance overall sustainability. This hypothesis is rejected, since , a significant negative trend in environmental sustainability was observed along an urban densification gradient. This significant negative trend implies that environmental sustainability is lower in more densified areas. This research thus concludes
