The water vulnerability of metro and megacities - Wiley Online Library

2 downloads 0 Views 140KB Size Report
The water vulnerability of metro and megacities: An investigation of structural determinants. Thomas Bolognesi. Abstract. Urban areas are becoming increasingly ...
Natural Resources Forum 39 (2015) 123–133

DOI: 10.1111/1477-8947.12056

The water vulnerability of metro and megacities: An investigation of structural determinants Thomas Bolognesi Abstract Urban areas are becoming increasingly subject to and vulnerable to water-related natural disasters. Urban areas are a kind of socio-ecological system wherein the human development dynamics co-evolve with the natural dynamics. Most of the literature is focused on the impact of natural disasters on human development; we evaluate the impacts of human development on natural disasters and present an analysis of this phenomenon in the context of megacities. The approach is exploratory and begins with the construction of a database on the 595 existing megacities in the world. Multifactor analysis is then used to determine the main characteristics of these megacities. Finally, three structural components (maturity, anthropization and centrality) are identified and then correlated with data on water-related hazards, distinguishing groups of cities according to their structure and factors of vulnerability to water-related risks. Keywords: Cities; vulnerability; natural hazards; water; multifactor analysis.

1. Introduction When hurricane Katrina hit New Orleans, it resulted in total losses to the economy of US$ 149 billion (Hallegatte, 2008). In Melbourne, Australia, US$ 2.55 billion have recently been invested in a desalination project to ensure that this drought-affected city receives adequate and reliable supplies of drinking water (Government of Victoria, 2011). These two examples illustrate the enormous social and economic impacts that natural water hazards can have on megacities. They are part of a general upward trend both in the number of hydrological hazards affecting urban areas and in the vulnerability of cities to water-related risks (Visser et al., 2012). Climate change, which has been shown to have an impact on extreme hydrological events (Min et al., 2011), is fuelling this phenomenon. Public sector decision-makers are calling on the scientific community to help them determine the vulnerability of towns and cities to natural hazards. As Adger (2006:269) shows “in all formulations, the key parameters of vulnerability are the stress to which a system Thomas Bolognesi is double-affiliated at the University of Geneva, Institute for Environmental Sciences, group Politics, Environment and Territories, Genève, Switzerland. E-mail: [email protected] and at Université Grenoble Alpes, Centre national de la recherche scientifique (CNRS), Politiques Publiques, Actions Politiques, Territoires-Economie du développement durable et de l’énergie (PACTE-EDDEN), Grenoble, France. © 2014 The Author. Natural Resources Forum © 2014 United Nations

is exposed, its sensitivity, and its adaptive capacity”. Vulnerability is approached from two different perspectives in the case of cities (Global Environmental Change, special issue, 2008). One approach examines the impact of natural hazards on cities (Praskievicz and Chang, 2009), whereas the other looks at the impact of urban development on the population (Pelling, 2003; Manuel-Navarrete et al., 2007). Most research into the vulnerability of towns and cities has been in the form of case studies. In general, studies have focused on one of three topics: cities in developing countries, cities in developed countries, or specific hazards that have occurred. The literature on cities in developing countries demonstrates their fragility and the likely impact of climate change (Adger, 1999; Adger et al., 2003; Varis et al., 2006a; 2006b). In the case of cities in less developed countries, the development of risk assessment methods to assist policymakers with hazard management is a major scientific challenge (Cross, 2001). Finally, when a particular type of urban hazard occurs, studies focus on the impacts and the possibilities for recovery (Campanella, 2006; Hallegatte, 2008). This approach provides insight into the cross-cutting aspects of vulnerability research as well as the diversity of outcomes. However, it produces contingent results that are difficult to apply to other situations. The absence of a systematic method producing transposable results on the basis of a large number of cases is restricting the acquisition of precise knowledge

124

Thomas Bolognesi / Natural Resources Forum 39 (2015) 123–133

on the urban phenomenon (Otto-Zimmermann, 2011). Consequently, the purpose of this paper is to analyse megacities from a global perspective rather than in a specific context. This approach enables us to identify the structural determinants of the megacities’ vulnerability to hydro-meteorological hazards, and we focus specifically on floods and droughts. As a starting point, we assume that the social and the biophysical systems are interlinked in a complex co-evolutive system (Gunderson and Holling, 2002; Ostrom and Janssen, 2004). In this respect, we are situated in the socio-ecological system approach of vulnerability, in the tradition of human ecology1 (Adger, 2006). The study is presented in five parts. In the second part we describe the methodology, our choice of variables and sources, and the statistical techniques and methods chosen to analyse the database. The results obtained are presented in the third part of the study. Three structuring factors emerge from the multifactorial analysis, and these factors enable us to distinguish the principal classes of an urban typology on the basis of vulnerability to water risks. Before concluding, we discuss the results in relation to the current literature.

2. Methodology The research project was divided into three stages. First, we constructed a database containing the information needed for the study (2.1). A multifactorial analysis of this information was then conducted to identify the characteristic features of the sample (2.2). Finally, the results of the multifactorial analysis were used to create a typology and to assess vulnerability factors (2.3). 2.1. A database on megacities and water vulnerability We used a comparative static analysis to examine the vulnerability of cities to water risks. First, a database was constructed containing the information needed to compare megacities. Small towns were excluded since they do not involve the same mechanisms in terms of vulnerability to natural hazards as megacities (Cross, 2001). Our sample was selected by consulting demographic data published by the United Nations, and it includes all cities with a population of over 750,000 inhabitants identified by the organization in 2010 (UN-Habitat, 2010). Floods and drought are among the most frequent natural hazards in the world. They represent 37% of natural hazards identified between 1990 and 2007 (EM-DAT, nd). The present study therefore focuses on these particular types of hazards. 1

This tradition is also labelled political ecology, and focuses on the explanation of environmental constraints or opportunities for individuals through the lens of social change and decision-making theories.

The database enabled us to compare megacities across the world. The analysis only concerns major urban trends, thus we only take into account structural and natural variables. With 31 variables, the database informs all the 595 cities (Table 1). These variables, providing information on the structure and vulnerability of the cities, reflect five major aspects of megacities: country, city, climate, risks and river basin (or watershed). Four variables form the “country” category including: GDP, purchasing power parity GDP (indicating the level of national development), the urbanization rate for 2010, and the projected urbanization rate for 2020. “City” variables include city size, in terms of the number of inhabitants in 2000, 2010 and projections for 2020, and geographical location (coast, delta or other). Economic data comes from the World Bank and demographic data comes from the UN-Habitat (2010). The “climate” category only includes the climate variable and indicates the type of climate for the area where the city is located according to the KöppenGeiger classification and the map produced by Peel et al. (2007) of the University of Melbourne. We chose this classification and map because their accuracy is acknowledged among climatologists (Kottek et al., 2006). Furthermore, the information required for the analysis was available. The “risks” category reflects the vulnerability of the megacity. It includes information about flood and drought occurrences in the time between 1980 and 2009, along with the human and economic costs. Analysing a 30-year period enables us to identify over 3,000 occurrences worldwide, providing us with a total of 7,000 events in our database.2 The raw data comes from EM-DAT CRED (EMergency Events DATabase of the Centre for Research on the Epidemiology of Disasters). These data provide details of the main characteristics of each hazard (date, duration, location, impact, etc.). The CRED does not use towns and cities as geographical units so adjustments were made before entering the information in the database. For our purposes, we attribute the same hazard characteristics to all cities situated in a same geographical area.3 Eight variables are presented in the “risk” category, and they indicate the number of hazards, the number of deaths, the number of people affected by the hazard, and the cost of the damage in US dollars, for the two categories of floods and drought. For the purpose of this study, we collect data on world hazards following the CRED’s methods. The data is

2

Hazards identified by CRED affect a zone, so that they may concern several cities at the same time. 3 This method has one major limitation. The area may be much larger than the city, so the impacts would not necessarily be limited to the city. However, since we wanted to determine the degree of vulnerability of the area in which the city was located rather than the damage actually suffered, we used the data. © 2014 The Author. Natural Resources Forum © 2014 United Nations

Thomas Bolognesi / Natural Resources Forum 39 (2015) 123–133

125

Table 1. Content of database Major categories of variables

Country

City

Climate

Risks

Number of variables Variables

4

4

1

8

GDP; PPP-GDP; Urbanization rate

Population; Geographic situation

Type of climate

Floods; Droughts

Number of cities included Scale

595

595

595

595

$/person; $/person; %

Thousands of inhabitants; coast, delta, other

Köppen-Geiger classification

Source

World Bank (2008); United Nations (2010)

United Nations (2010)

Peel et al. (2007)

Unit; Number of deaths; Number of people affected; Damage in $US CRED EM-DAT

summarized below.4 Floods are hydrological hazards, while droughts are climatological hazards. The database includes events according to their consequences. Each event included meets at least one of the following four criteria: 10 or more deaths recorded, at least 100 people affected, a declaration of a state of emergency, and an appeal for international aid. The number of deaths is the official figure. The number of persons affected is an estimate based on the number of households affected multiplied by the national average size of a household. Finally, there is no standard method for calculating the cost of a hazard. The CRED bases its calculations on information available from various sources such as international agencies, governments, and NGOs. The watershed scale is based on the Integrated Water Resources Management (IWRM) paradigm. Fourteen variables inform the river basin category including: area, density, number of cities with over 100,000 inhabitants,5 per capita water availability, degree of fragmentation of rivers, number of dams of over 15 metres and eight variables relating to land cover. We get the information from the Watersheds of the World (nd) application created by several international water authorities under the direction of the World Resources Institute. This application provides access to the main characteristics of the 154 largest watersheds,

4

For more details, see the following CRED page: http://www.emdat.be/ explanatory-notes, (consulted 09-03-2013) and the articles by Sapir and Below (2006) and Scheuren et al. (2008). 5 The threshold for river basing cities is 100,000 and not 750,000 as in our database, because it is the one available in the Watersheds for the World application. This is not a limitation to the study because this variable aims to inform the watershed in which the cities studied are located and not the cities themselves. © 2014 The Author. Natural Resources Forum © 2014 United Nations

Watershed 14 Area; Density; Number of cities > 100,000 inhabitants; Per capita water availability; Degree of fragmentation of rivers; Number of dams > 15m; Rates and type of land cover 192 Various

World Resources Institute

utilized by the 192 cities in our database. As far as we know, it is the data collection with the broadest range of watershed. All of this information was then used in the subsequent analysis. 2.2. Multifactorial analysis of characteristics of megacities Multifactorial analysis is particularly appropriate when dealing with large amounts of information (Hair et al., 2010). We are able to synthesize the data by creating principal components that summarize the behaviour of the sample, after which we can extract meaningful information. This type of analysis is appropriate for our study since it decrypts the information in the database. Three stages are conducted in the analysis: preparation of data, selection of analysis procedure, and choice regarding dimensionality. First, the variables were classified to make them suitable for multifactorial analysis. When they occurred in the literature, we adopted consensus-based thresholds (Biswas, 2006; World Bank, 2008; United Nations, 2010). Otherwise, quintiles of each variable were used to create the different categories. The scale obtained was then validated by means of Chronbach’s alpha, a coefficient for assessing the internal consistency of a scale. A value above 0.7 justifies an exploratory factor analysis, whereas a value of over 0.8 corresponds to basic research (Nunnaly and Bernstein, 1994). Next, we selected the analysis procedure on the basis of two critical points. The first concerned the representativeness of the variables considered. Variables were incorporated or rejected depending on their degree of representativeness. The second point concerned the method of extracting data. The minimum level of representation, or

126

Thomas Bolognesi / Natural Resources Forum 39 (2015) 123–133

extraction, was set at 0.65. With this value, it is agreed that the components include a fairly large amount of the information contained in the variables (Hair et al., 2010). We chose to extract data using Varimax rotation, which is an orthogonal rotation of the factor axes. By using the rotation method, it minimizes the number of values with high contributions on an axis and simplifies the interpretation of the factors. Furthermore, since the plane remains orthogonal, the factor structure is very clear to analyse. This method concentrates on all the influential variables, and is therefore well suited to our exploratory factor analysis. Finally we chose the level of dimensionality, which is a compromise between the clarity of the message and the accuracy of the synthesized data. In fact, the lower the level of dimensionality the greater the clarity of the message, and the smaller the likelihood that peripheral phenomena, which would result in a loss of information, will appear. The number of factors6 selected depends on this compromise, and it is based on an analysis of variance. This analysis requires simultaneously observing the total explained variance and the variance explained by each additional factor. We selected the smallest number of factors explaining at least 60% of the total variance (Hair et al., 2010). We then further sorted the variables in order to create representative uncorrelated factors. Thus, all the variables with an explanatory coefficient above 0.3 for at least two factors and without any coefficient above 0.5 were eliminated. The Kaiser-Mayer-Olkin (KMO) index then validated or rejected the factor structure obtained. This index assesses the coherence of the data set and measures the efficiency of the factors. Essentially, it is a test for partial correlation. The test accepts the solution if the index is above 0.3 but below 0.7, beyond which the model becomes self-determining. The analysis continues using the factors calculated in this way. 2.3. Typology of megacities based on the different factors The factors in the multifactorial analysis describe the different facets of the cities studied. They can identify the reasons for a phenomenon. In our case, we call them factors of vulnerability, and we use them as a basis for creating a typology. Each component in the multifactorial analysis identifies one feature of the cities. When placed on a graduated axis, the components classify the cities in relation to each other. In the VARIMAX orthogonal method used, the axes are not correlated between themselves. When the axes intersect, they provide information about a phenomenon or a new characteristic. Thus, types of cities

6

In multifactor analysis, the terms “component”, “factor” and “dimension”, or “notion”, are almost synonymous. The component is the object of the analysis. It becomes a factor when the variables are positioned. Finally, the term “dimension” is used when one moves away from numbers and the meaning becomes more important.

emerge from the analysis, enabling a typology to be proposed, which can then be used in a comparative positive analysis. Then, the multifactorial analysis can be used to characterize the cities and build stylized facts that can be used in a subsequent explanatory approach. The correlation between the factors in the analysis and the risk variables reflects, at least in part, the vulnerability of the cities. Thus, if a significant relationship is found between a component of the multifactorial analysis and a type of risk, we conclude that there is causality. The causality, which is comprised of the factor relationships, represents the structural dimensions of the cities and therefore is not the result of random uncertain events. The factors thus exacerbate or attenuate the vulnerability of the cities to water-related risks. A type and level of vulnerability is then associated with each group of cities defined by the typology. For verification (Adger et al., 2004), we give insights on specific cases, and the discussion provides a comparison of the findings with the relevant literature.

3. Results Three factors resulted from the multifactorial analysis, and they are presented below according to levels of significance, meaning, and correlation with risk (3.1; 3.2; 3.3). Finally, in section 3.4 we present a typology based on these factors. 3.1. Factor 1: Level of maturity of megacities The factor solution selected produced three factors (Table 2), which are comprised of eight variables. The table shows how each variable contributes to the factors’ determination. The solution explains 61.3% of the variance of the total sample (595 cities) and 83.2% of the variance of the final sample (192 cities), which excludes all the incomplete data rows. The KMO index was 0.627. It is therefore within the range of values indicating that the factor based solution is a coherent set of values adequately measuring the factors. For the most part, the quality of representation of the variables in the factors is excellent, and in every case the quality is at least good. The factors are not correlated among themselves since each variable used only contributed to the construction of one factor. The first factor explains 33.4% of the total variance, and it reflects the most important trend in the sample. Factor 1 reflects the maturity of the city within the country. It is made up of city and country variables: the GDP/inhabitants in 2008, the rate of urbanization in 2010 and the projection of the city’s dynamism between 2010 and 2020. The dynamism projection is based on the measurement of the ratio of the city’s demographic growth to national demographic growth. This enables an assessment to be made of the movement of population to the city. Assuming homogeneous natural demographic growth © 2014 The Author. Natural Resources Forum © 2014 United Nations

Thomas Bolognesi / Natural Resources Forum 39 (2015) 123–133

127

Table 2. Matrix of components after rotation Components

Explained variance (%) GDP/inhab 2008 % urbanization 2010 Size of city/urban pop 2010 Size of city/country 2010 Crops Irrigated crops Dynamism of city Density of watershed

1 Maturity

2 Anthropization

3 Centrality

33.4 0.939 0.942 −0.189 0.022 0.214 −0.128 0−.875 −0.202

29.8 0.064 −0.139 −0.104 −0.090 0.784 0.780 0.038 0.872

18.8 −0.082 −0.085 0.955 0.972 0.089 −0.258 0.013 −0.097

Notes: Method of extraction: Principal components analysis. Rotation method: VARIMAX with Kaiser normalization. Shading represent variables and coefficients that determinate each component.

over the national territory as a whole, differences are then perceived as migrations. The first factor represents this variable well (0.768), and the other two variables even better (0.893 and 0.913). The country variables and the dynamism variable act inversely. Thus, the different situations considered range from a city that is ultra-dynamic in relation to the country in which it is located, which has undergone little or no development, to a city that is simply part of a country with a high level of development. The maturity axis represents an urban dynamic, which can be considered a continuous process. It reflects the development of the country and the place of the city in this development, as well as the role of the city in the country’s urbanization phase. The axis measures the extent to which the development of a city acts as a driving force for development in the country as a whole. The example of the United States provides insights into the phenomenon of the polarization and diffusion of development. The US is a rich country and it is highly developed, so all of the cities within the US score high on this factor. Every city scores at least 0.7, with the average around 1.2. In comparison, Mogadishu, the capital of Somalia, scores −1.24 on this axis since it is in a country with a very low level of development. The differences in the contribution of cities to national development can then be revealed by comparing each city’s score with its country’s national average. When the city is below the national average it is very dynamic, and vice-versa. In the US, Pittsburgh has the highest score with 1.43. This city experienced an industrial boom in the 19th century thanks to its mines, the development of the steel industry, and its railways. At that time, it was a dynamic economic centre and development hub for both Pennsylvania and the entire country. However, in the second half of the 20th century economic activity declined, people began leaving the area, and the city lost its dynamism. Since then, the city of Pittsburgh has developed in a similar manner to the country as a whole, and sometimes at a slightly slower rate, hence it has a slightly higher than average index. © 2014 The Author. Natural Resources Forum © 2014 United Nations

El Paso, in Texas, finds itself in the opposite situation. Its low score of 0.9 reflects the dynamic nature of the city. Since the ratification of the North American Free Trade Agreement in 1993, this region has become very dynamic, mainly through the development of the maquilladoras. Living conditions are more difficult in El Paso, due to its proximity to the desert; therefore, the city’s established infrastructure provides a strong attraction for economic activity. Polarization of economic activity in El Paso, in comparison to the US as a whole, has therefore increased, resulting in its lower than average index. This factor is negatively correlated with the risk variables (Table 3). The diffusion of development over a territory reduces the vulnerability of cities. This trend does not have the same impact on each variable.7 Thus, in terms of hazards, the probability of a drought event decreases with the spread of development. A negative coefficient of −0.222** links the two phenomena.8 However, whether caused by floods or drought, the human impacts become less marked if there is a spread of development. The correlation between factor 1 and the number of people affected by flooding is −0.220** and the factor increases to −0.251** in the case of drought. The costs resulting from droughts are negatively correlated with factor 1 (−0.182*). The vulnerability of cities decreases as the country develops and the development is shared among the inhabitants. We reviewed two Chinese cities that had a comparable level of people affected by flooding, and one finding was that the number of risks varies by a factor between 1 and 2 depending on the level of maturity of the city. For example, Guangdong is a Chinese city with a relatively high level of maturity for China, −0.45 compared to a 7 To facilitate interpretation we present only the significant impacts. Thus, the correlation between the variables not cited and the factor not dealt with is close to zero. 8 By convention, * indicates a correlation that is significant at the 0.05 level (bilateral) and ** a correlation that is significant at the 0.01 level (bilateral).

128

Thomas Bolognesi / Natural Resources Forum 39 (2015) 123–133 Table 3. Correlation between factors and risks

Floods

Deaths

People affected

Damage in $

Droughts

Deaths

People affected

Damage in $

Pearson correlation Significance (bilateral) N Pearson correlation Significance (bilateral) N Pearson correlation Significance (bilateral) N Pearson correlation Significance (bilateral) N Pearson correlation Significance (bilateral) N Pearson correlation Significance (bilateral) N Pearson correlation Significance (bilateral) N Pearson correlation Significance (bilateral) N

Maturity

Anthropization

−0.064 0.379 192 0.039 0.595 192 −0.220** 0.002 192 −0.099 0.170 192 −0.222** 0.002 192 −0.050 0.493 192 −0.251** 0.000 192 −0.182* 0.011 192

0.265** 0.000 192 0.068 0.350 192 0.301** 0.000 192 0.248** 0.001 192 0.085 0.243 192 −0.002 0.975 192 0.342** 0.000 192 0.095 0.191 192

Centrality −0.422** 0.000 192 0.153* 0.034 192 −0.279** 0.000 192 −0.280** 0.000 192 −0.197** 0.006 192 0.099 0.171 192 −0.240** 0.001 192 −0.289** 0.000 192

Notes: ** The correlation is significant at 0.01 level (bilateral). * The correlation is significant at 0.05 level (bilateral). Source: Author’s elaboration.

national average of −0.7. Over the study period, more than 400 million people suffered from 28 floods. Over the same study period, Luoyang, with a low level of maturity (−1.02), was subjected to 12 floods and a comparable number of people were affected. We can see that the impact intensity of the events decreases as the maturity of the city increases. 3.2. Factor 2: Level of anthropization The second factor, reflecting the level of anthropization of the river basin in which the city is situated, explains 29.8% of the total variance. This second factor is made up of three watershed variables. The density of the watershed is very well represented (0.812), and has the strongest influence on the factor (0.872). The other two variables, the proportions of the area used for crops and irrigated crops, are also well represented (0.668 and 0.691 respectively) and have a similar impact (0.784 and 0.780). This factor reflects the land use of the river basin by human activities. Thus, the higher a city scores on this factor, the more its territory has been modified and occupied by human activity. The situations range from a river basin that is devoid of any human activity to one where the land has become totally artificialized (a polder, for example). Cities that are situated in river basins with high densities are found on the right of the axis. All the European cities, as

well as some Asian cities, score highest on the x-axis. The case of Europe provides a good example of the anthropization phenomenon where land use control is a major concern for the development of European societies. In France, the decree of the 14th of Frimaire (4 December 1793) was enacted to drain the marshes of the Republic to halt the reproduction of mosquitoes that spread diseases and to develop agriculture (Mathevet, 2011). Similarly, the second pillar of the common agricultural policy of the European Union recognizes the multi-functionality of agriculture and appoints farmers to maintain the natural landscape, which creates human intervention. In the US, however, the trend has been more towards the creation of large national parks, limiting any human footprint through the creation of organizations such as Leave no trace (http:// www.lnt.org/, accessed September 2013). The balance between this trend and growing tourism of river basins looks more and more difficult to maintain (Runte, 1997). The American river basins thus have a low level of anthropization. Consequently, US cities are situated on the left of the graph. The anthropization of river basins increases water-related risks. Thus a positive correlation of 0.265** was found between the number of floods and the level of anthropization. Similarly, the impact of floods increases with higher anthropization levels. The number of people affected and the damage caused are positively correlated © 2014 The Author. Natural Resources Forum © 2014 United Nations

Thomas Bolognesi / Natural Resources Forum 39 (2015) 123–133

129

with anthropization (respectively 0.301** and 0.248**). Thus, anthropization not only places cities at greater risk from flooding, but it also increases the potential impacts of such flooding. This phenomenon makes cities more sensitive to droughts. The correlation coefficient between anthropization and the number of people affected by droughts was measured at 0.342**. Thus, anthropization exacerbates the vulnerability of cities to water-related risks. By way of illustration, the 37 cities with an anthropization index of less than −1.1 were subjected to an average of 4 floods over a period of thirty years. The 31 cities with an index of greater than 1 were subjected to an average of 10 floods over the same period. Furthermore, in the first case, the group is fairly homogeneous in terms of known hazards (three cases of above 6), while in the second there is considerable dispersion (numerous cases above 20, and a maximum of 37).

coefficient of −0.153* as a function of the city’s centrality. However, centrality considerably reduces the probability of flooding. The correlation coefficient is −0.422**. To a lesser extent (−0.197**), the effect is the same with regard to droughts in that centrality reduces the impact of hazards. From a human welfare point of view, the number of people affected by floods is negatively correlated with centrality (−0.279**), as is the number of people affected by droughts (−0.240**). The costs resulting from hazards decrease as centrality increases. This is reflected in negative correlations of −0.280** in the case of floods and −0.289** in the case of droughts. The analysis therefore shows that the degree of centrality of cities is a factor that attenuates the vulnerability of urban centres with respect to waterrelated risks regarding seven of the eight risk variables. It should also be noted that the impact can vary, since the factor increases the number of deaths from flooding.

3.3. Factor 3: Degree of centrality in the urban network

3.4. Typology of cities based on the axes of the multifactorial analysis

The third factor explains 18.8% of total variance. With about ten points less than the second factor, the third factor expresses the degree of centrality of a city within the country. It provides an excellent representation of the two variables it comprises: the size of the city in relation to total urban population (0.959) and the size of the city in relation to the population of the country as a whole (0.953). The reason why this factor explains relatively less of the total variance is that it is only made up of two variables, while the other factors represent three variables.9 Here, the scale of observation is the city where the two variables act in an almost identical manner in creating the factor (0.955 and 0.972). This factor measures the weight of the city in the country as a whole, and the range of possibilities extends from an insignificant city within the country to a City-State. For example, in France a strong asymmetry appears in the distribution of the population. Paris has one of the highest degrees of centrality in the entire sample (1.6) while other French megacities are all situated at the level of the y-axis. This strong centrality of the capital is the result of the country’s centralizing tradition. In Russia, however, this centrality phenomenon is not so evident in the urban structure. Here, urban weight is sparingly distributed between the megacities. Their degree of centrality is between −0.5 and −0.8. This balance in the settlement patterns of the Russian territory is the result of deliberate Soviet policies. The central character of a city significantly reduces its water-related risks, apart from one exception. It was found that the number of deaths due to floods increases by a

9

Besides, this lower level of significance reveals that commonly a good factor is based on three factors. © 2014 The Author. Natural Resources Forum © 2014 United Nations

The three factors presented above provide the first elements of a typology. Based on our information, we can classify 192 cities. If each city is classified on a binary scale, the typology is based on eight city profiles. Since a graphical presentation on a single plane would be difficult to read, we prefer to present our results in terms of central and noncentral megacities. However, the distribution of cities according to their characteristics provides an interesting overall view. In 54.2% of cases, the cities are in a country where development is well diffused over the territory, with a mature urban network. In 43.2% of cases, the river basins are characterized by a significant level of anthropization. Finally, in 40.1% of cases, the cities occupy a central place in the country. The eight city profiles correspond to four levels of potential risk: very low, low, high, very high. The typical city that best attenuates risks is a central city where human activity has had little impact on the river basin, and in a country where the level of development is high and shared by all. Only 12 cities have these characteristics, but they are scattered over the globe without any apparent logical link. At the other end of the spectrum, 42 cities meet all the conditions for increasing water-related risks. These cities are all located in Asia, mostly in India and China. The axes enable us to create eight categories of classification (Figure 1). Taking into account the geographical location of cities, certain similarities may be noted among cities in the same zone. Thus, there are six geographical groups of cities: Asian, North American, African, Russian, Western European and Eastern European. The Asian cities combine a negative factor 1 with a positive factor 2 where development is concentrated around a few dynamic cities, and the river basins have been considerably modified by human activities. The North American cities are characterized by a positive factor 1 and a negative factor 3, where the diffusion of development is thus relatively

130

Thomas Bolognesi / Natural Resources Forum 39 (2015) 123–133 C entr al cities V ery l ow risk

Non centr al cities L ow risk

High risk

L ow risk Maturity

Maturity Western Europe

USA

E astern E urope

Russia

USA

Russia

A nthrop.

A nthrop.

Asia

Asi a

A frica

L ow risk

A frica

High risk

High risk

V ery high risk

Figure 1. Distribution of cities according to structural characteristics and corresponding level of flood and drought risk. Source: Author’s elaboration.

homogeneous over the national territory and the cities are not very central. African cities are characterized by negative factors 1 and 2, where the cities concentrate the little development there is and they are generally located in territories where the landscape has undergone little anthropization. Russian cities are also not very central (negative factor 3), but they share in a higher level of development (positive factor 1). Western Europe has a number of central cities in territories that reflect a high level of anthropization, and the development is well diffused. The cities of Eastern Europe have similar characteristics to Western Europe, apart from the fact that development remains concentrated and limited.

Discussion In this section, we discuss our results in two directions. First, we focus on the quality and the way to improve our quantitative characterization of megacities and vulnerability (4.1). Then, we present some implications of our results in terms of governance (4.2). 4.1. Measuring the structural potential of vulnerability This multifactorial characterization of megacities and their vulnerability to hydro-meteorological hazards is exploratory. The quantitative results obtained enable us to establish an international comparison of megacities and their sensitivity to different water risks. As a starting point for further research, we can use them to create a classification of megacities according to their structural characteristics. The matrix of correlations provides the equation of this classification:

Vulnerability Flooddev . … Floodcentr . ⎤ ⎡ ⎥ = ∑⎢    ⎥ ⎢ ⎢⎣ Damage drought dev . … Damage droughtcentr. ⎥⎦ ⎡ F1 ⎤ × ⎢ F2 ⎥ . ⎢ ⎥ ⎢⎣ F3 ⎥⎦ Then, by comparing this classification with reality, other variables will appear that will help refine the measurement of the structural megacities’ vulnerability. The sample of hazards should also be widened to more than 30 years of data collection. Floods and droughts do not follow a linear dynamic (Simonovic, 2009) and megacities are vulnerable to some major events that might not occur in a 30 year time period, such as a centennial flood. The integration of such events would reveal additional facets of vulnerability. For example, the variables representing the risks of floods and droughts in terms of deaths are not statistically significant in the model. We could reasonably assume that an extension of the period analysed will increase the significance of the correlation of these risks to our three factors. The broader integration of past events will strengthen the robustness of the present findings. In the future, it would be necessary to examine how climate changes will impact vulnerability and its structural determinants. The analysis shows that the structural characteristics of megacities and urban development impact the megacities’ vulnerability. Spatial analysis with satellite data appears to be complementary to this first study, and it represents a fruitful way to make the correlations more precise. We © 2014 The Author. Natural Resources Forum © 2014 United Nations

Thomas Bolognesi / Natural Resources Forum 39 (2015) 123–133

identify two relevant variables for further study, which include urban sprawl and water footprints. The level of maturity factor reflects the dynamic of expansion and contraction of megacities. Cities located to the right of the national average stagnate or tend to shrink, while those to the left grow significantly. The examples of El Paso, Pittsburgh and the cities of the east coast of China mentioned earlier are illustrative of this process. Urban sprawl and, at the opposite end of the scale, shrinking cities, are a major concern in recent urban dynamics (Wassmer, 2000; Galster et al., 2001; Johnson, 2001). As an illustration, Nechyba and Walsh (2004) reveal a marked change in the distribution of the American urban population between 1950 and 1990. During this period, it changed from 65% in the centres and 35% in the suburbs to 20% in the centres and 80% in the suburbs. At the same time, the area occupied by suburbs increased by 250%, while the population density of the centres fell by 40%, from 10,000 inhab/km2 to 6,000 inhab/km2. Since the beginning of the 1990s, urban sprawl has become generalized in Europe (EEA, 2006; Patacchnini et al., 2009). Between 1991 and 2001, the area occupied by European cities increased by an average of 7.46%, while their population only increased by 0.34% (Pacchinini et al., 2009). Analyses in terms of water footprint corroborate our results on the anthropization and artificialization of land (Jenerette and Larsen, 2006; Jenerette et al., 2006). Megacities shape the territory around them in a significant manner in order to satisfy their needs, a process found to be particularly powerful in Western Europe and on the east coast of China. According to their estimates, the urban water footprint is increasing in the world. The estimated average in 1950 was 29,937 km2 per city, and by 2000 had increased to 35,397 km2, representing an increase of 18%. It is expected to reach 38,400 km2 in 2015, representing an increase of 28% since 1950. Throughout the world, the growth of the water footprint of megacities is accelerating. In 2015, Indian and Chinese cities will have the greatest water footprints. The European cities are slightly lower than the average. However, since European urban networks are more densely meshed than in other parts of the world, the European megacities appear to be leaving their water footprint on the entire territory of Western Europe. So, water footprint highlights and measures the process of artificialization on the local territory. This complements our results and argues that, in the future, land use will be increasingly determinant on the megacities’ vulnerability. Therefore, variables such as water withdrawal and pollution (which were not the main concern of our approach) could be useful in defining different artificialization paths. 4.2. Governance concerns As a starting point, we assumed that megacities are socioecological systems and that a social sphere co-evolves with a biophysical one. Accordingly, vulnerability does not © 2014 The Author. Natural Resources Forum © 2014 United Nations

131

always result in a catastrophic water-related event and is mediated through the governance process (Adger, 2006). In this respect, we discuss our results on two points: the significance of centrality in terms of governance and the path dependency of megacities’ structural characteristics. Factor 3 reveals the centrality of the different megacities studied. Thus, it contributes to the analysis of the hierarchical structure of the urban fabric. A city is a system in its own right, but at the same time it is also part of a larger network or networks; the factors place a city within a larger system. (Bretagnolle and Pumain, 2010; Pflieger and Rozenblat, 2010). This concept is related to the structural aspects of urban systems. Thus, centrality determines in large part whether the territory is structured by several equivalent cities or by a major one that dominates all the others. We have shown that variables of risks are significantly correlated to the degree of centrality. Consequently we suppose that a high centrality gives to megacities a set of resources that enables them to reduce their vulnerability (Arrow et al, 1995). This set enlarges their governance alternatives and strengthens their resilience. The central megacities are more connected and integrated in multi-scalar networks. Managers are able reduce the dependence to some particular function thanks to this location at the heart of decentralized systems. As an illustration, Ahern (2011: 342) explains that “redundancy and modularization spread risks — across time, across geographical areas, and across multiple systems”. The first factor confirms that megacities are path dependent and that this path dependency determines their vulnerability. It shows that the level of development and its associated dynamism modifies the impacts of risks. Consequently, in developing countries the vulnerability to hydrometeorological hazards is mainly a matter of underdevelopment and of economic emergence. For developed countries, land use management seems to be the critical point. The level and the form of vulnerability is resulting from a trade-off between the preservation of natural space (and its ecosystem services) and a land use that is economically-oriented. Theories of institutional dynamics enlighten these links between path dependency and environmental vulnerability (Vatn, 2005; Ostrom and Janssen, 2004; Young, 2010). They show that each socioecological system is engaged in a specific adaptive cycle, and that the potential of resilience or vulnerability is evolving according to the phase of development. Thus, governance has to propose a continuous transition between these different phases. The modalities of this transition are contingents to each system and the process is as important as the finality. Bolognesi et al. (2013) support this idea, and they propose a conceptual framework to address water security in terms of water securing path. This path conceptualizes a model of governance aiming at achieving water security goals through participation and adaptive governance. This kind of analysis seems to be useful in order to extend our study.

132

Thomas Bolognesi / Natural Resources Forum 39 (2015) 123–133

4. Conclusions and further research

References

It should be emphasized that these groups only classify the cities considered in the multifactorial analysis. There is no certainty that they apply to all cities with more than 750,000 inhabitants. Moreover, cities may appear as statistical exceptions. For these two reasons, this typology is only a starting point. Nevertheless, given the large number of cities included (about one third of existing megacities), it seems to us that generalization is possible. Before the typology can be completed, the cities and their territories will need to be examined in more detail. For example, case studies relative to the different categories and groups would improve understanding and validity, and increasing the number of variables would produce the same effect. Nevertheless, the typology as it stands already informs us of the existence of urban profiles according to continent. The use of multifactorial analysis proved to be effective in comparing the world’s megacities. The analysis of a sample of 595 cities, described by 31 variables, resulted in the identification of three major factors that characterize 192 different cities: maturity, level of anthropization of the river basin concerned, and degree of centrality of the city. They enable the cities to be divided into six groups. After conducting correlation studies between city groups and water-related risks, we obtained an overall function of urban vulnerability, as well as four different levels of intensity. These results provided the basis for constructing a typology. We observed six geographical groups of cities with homogeneous characteristics: Asia, North America, Africa, Russia, Western Europe, and Eastern Europe. This result confirms the findings in the literature on city structure. Thus, the continents often provide a basis for classification. Our contribution lies in the large number of cases considered in the study, namely 192 different cities, or one third of the world’s megacities. Consequently, a generalization of the results to other cities does not seem unreasonable. Our study also confirms the important structuring role of economic development and centrality in the characterization of cities. A third dimension, anthropization, also appears to be equally determining. These dimensions tend to reflect the findings in the literature on urban sprawl, the artificialization of land, and networks of cities. Nevertheless, we believe our results are not simply a reiteration of past findings, but are complementary to those of earlier studies. Thus, we identify and measure the different phenomena on the basis of a structural analysis of megacities, which generates a certain level of generality. We intend to continue our research in three different directions: conduct more in-depth studies on certain parts of the sample in order to refine the typology; compare results with observations from case studies, to avoid concealing reality in statistics; and incorporate governance considerations into this research.

Adger, N., 1999. Social vulnerability to climate change and extremes in coastal Vietnam. World Development, 27(2): 249-269. Adger, N., 2006. Vulnerability. Global Environmental Change, 16(3): 268-281. Adger, N., Huq, S., Brown, K., Conway, D., Hulme, M., 2003. Adaptation to climate change in the developing world. Progress in Development Studies, 3(3): 179-195. Adger, N., Brooks, N., Bentham, G., Agnew, M., Eriksen, S., 2004. New indicators of vulnerability and adaptive capacity. Technical Report Tyndal Centre for Climate Change Research 7. Ahern, J., 2011. From fail-safe to safe-to-fail: Sustainability and resilience in the new urban world. Landscape and Urban Planning, 100(4): 341-343. Arrow, K., Bolin, B., Costanza, R., Dasgupta, P., Folke, C., Holling, C.S., Jansson, B., Levin, S., Mäler, K., Perrings, C., Pimentel, D., 1995. Economic growth, carrying capacity, and the environment. Ecological Economics, 15(2): 91-95. Biswas, A.W., 2006. Water management for major urban centres. Water Resources Development, 22(2): 183-197. Bolognesi, T., Buchs, A., Renou, Y., 2013. Water no get enemy. How could water security shape sustainable water governance? Ecological Economics and Institutional Dynamics, 10th international conference of the European Society for Ecological Economics. ESEE, Lille, 18-21 June. Bretagnolle, A., Pumain, D., 2010. Simulating urban networks through multiscalar space-time dynamics (Europe and United-States, 17th-20th centuries). Urban Studies, 47(13): 2819-2839. Campanella, T.J., 2006. Urban resilience and the recovery of New Orleans. Journal of the American Planning Association, 72(2): 141-146. Cross, J.A., 2001. Megacities and small towns: Different perspectives on hazard vulnerability. Environmental Hazards, 3(2): 63-80. European Environment Agency (EEA) 2006. Urban sprawl in Europe. EEA Report, n° 10. EEA, Copenhagen. EM-DAT: The OFDA/CRED International Disaster Database nd. Available at http://www.emdat.be (accessed 31 October 2014). University of Louvain. Galster, G., Hanson, R., Ratcliffe, M.R., Wolman, H., Coleman, S., Freihage, J., 2001. Wrestling sprawl to the ground: Defining and measuring an elusive concept. Housing Policy Debate, 12(4): 681-717. Global Environmental Change, 2008. Local evidence on vulnerabilitie and adaptations to global environmental change, 18(special issue): 539-597. Government of Victoria, 2011. Desalination Project. Government Programs. Available at http://www.water.vic.gov.au/programs/ desalination (accessed 9 February 2012). Gunderson, L.H., Holling, C.S., (Eds.) 2002. Panarchy: understanding transformations in human and natural systems. Island Press, Washington, DC. Hair, J., Tatham, R.L., Anderson, R.E., Black, W., 2010. Multivariate data analysis. Prentice Hall, Englewood Cliffs. Hallegatte, S., 2008. An adaptive regional input-output model and its application to the assessment of the economic cost of Katrina. Risk Analysis, 28(3): 779-799. Jenerette, D., Larsen, L., 2006. A global perspective on changing sustainable urban water supplies. Global and Planetary Change, 50(3): 202-211. Jenerette, D., Wu, W., Goldsmith, S., Marussich, W.A., John Roach, W., 2006. Contrasting water footprints of cities in China and the United States. Ecological Economics, 57(3): 346-358. Johnson, M.P., 2001. Environmental impacts of urban sprawl: A survey of the literature and proposed research agenda. Environment and Planning, 33(4): 717-735. Kottek, M., Grieser, J., Beck, C., Rudolf, B., Rubel, F., 2006. World Map of the Köppen-Geiger climate classification updated. Meteorologische Zeitschrift, 15(3): 259-263. © 2014 The Author. Natural Resources Forum © 2014 United Nations

Thomas Bolognesi / Natural Resources Forum 39 (2015) 123–133 Mathevet, R., 2011. Resilience of wetlands with multiple uses — example in Petite Camargue. Workshop Resilience, Water and Foresight, 18-19 January, Montpellier. Manuel-Navarrete, D., Gomez, J.J., Gallopin, G., 2007. Syndromes of sustainability of development for assessing the vulnerability of coupled human–environmental systems. The case of hydrometeorological disasters in Central America and the Caribbean. Global Environmental Change, 17(2): 207-217. Min, S.K., Zhang, X., Zwiers, F.W., Hegerl, G.C., 2011. Human contribution to more-intense precipitation extremes. Nature, 470(7334): 378-381. Nechyba, T.J., Walsh, R.P., 2004. Urban sprawl. The Journal of Economic Perspectives, 18(4): 177-200. Nunnaly, J.C., Bernstein, I.H., 1994. Psychometric Theory. McGraw-Hill, New York. Ostrom, E., Janssen, M., 2004. Multi-level governance and resilience of social-ecological systems. In Spoor, M. (Ed.), Globalisation, Poverty and Conflict. Kluwer, Dordrecht. 239-259. Otto-Zimmermann, N., 2011. Opening speech. Resilient Cities 2011 Conference, Bonn. Patacchnini, E., Zenou, Y., Henderson, J.V., Epple, D., 2009. Urban sprawl in Europe. Brooking-Wharton Papers on Urban Affairs, 125-149. Peel, M.C., Finlayson, B.L., McMahon, T.A., 2007. Updated world map of the Köppen-Geiger climate classification. Hydrological Earth Systems Science, 11(5): 1633-1644. Pelling, M., 2003. The Vulnerability of cities: Natural Disasters and Social Resilience. Earthscan, London. Pflieger, G., Rozenblat, C., 2010. Introduction. Urban networks and network theory: the city as the connector of multiple networks. Urban Studies, 47(13): 2723-2735. Praskievicz, S., Chang, H., 2009. A review of hydrological modelling of bassin-scale climate change and urban development impacts. Progress in Physical Geography, 33(5): 650-671.

© 2014 The Author. Natural Resources Forum © 2014 United Nations

133

Runte A., 1997. National Parks: The American Experience. The University of Nebrasca Press, Lincoln, NE. Sapir, D.G., Below, R., 2006. Data on disasters: easier said than done. Asian Disaster Management News 12(2): 9-10. Scheuren, J.M., le Polain de Waroux, O., Below, R., Guha-Sapir, D., Ponserre, S., 2008. The Numbers and Trends 2007. Annual Disaster Statistical Review. CRED: Brussels. Simonovic, S.P., 2009. Managing Water Resources. Methods and Tools for a Systems Approach. UNESCO publishing, Paris. UN-habitat, 2010. World Urban Prospect. Available at: http://esa.un.org/ unpd/wup/index.html (accessed 21 March 2012). Varis, O., Biswas, A.K., Tortajada, C., Lundqvist, J., 2006a. Megacities and water management. International Journal of Water Resources Development, 22(2): 377-394. Varis, O., Kummu, M., Keskinen, M., 2006b. Editorial. International Journal of Water Resources Development, 22(3): 395-398. Vatn, A., 2005. Institutions and the Environment. Edward Elgar, Cheltenham. Visser, H., Bouwman, A., Petersen, A., Ligtvoet, W., 2012. A Statistical Study of Weather-related Disasters: Past, Present and Future. Netherlands Environmental Assessment Agency, The Hague. Watersheds of the World nd. http://www.wri.org/publication/watersheds -of-the-world (accessed 21 March 2013). Wassmer, R.W., 2000. Urban sprawl in a US metropolitan area: Ways to measure and a comparison of the Sacramento area to similar metropolitan areas in California and the U.S. CSUS Public Policy and Administration Working Paper No. 2000-03. World Bank 2008. http://econ.worldbank.org/ (accessed 31 December 2013). Young, O., 2010. Institutional dynamics: resilience, vulnerability and adaptation in environmental and resource regimes. Global Environmental Change, 20(3): 378-385.