Spatio-Temporal Dynamics of Urban Growth in Latin American Cities: An Analysis Using Nighttime Lights Imagery

The impact of urban form on economic performance and quality of life has been extensively recognized. The studies on urban form have focused in developed countries; only a few cities in developing countries have been studied. This paper utilizes nighttime lights imagery and information on street networks, automatically retrieved from OpenStreetMap, to calculate a series of spatial metrics that capture different aspects of the urban form of 919 Latin American and Caribbean cities. The paper classifies these cities into clusters according to these spatial metrics. It also studies the relationship between the urban form metrics and some factors that can correlate with urban form (topography, size, colony, and economic performance) and performs a spatio-temporal analysis of urban growth from 1996 to 2010. Among the results, the paper highlights the identification of five typologies of cities, the tendency of a group of cities to grow at a steeper slope, some worrying cases of urban growth over protected areas, and a trend toward increasing sprawl in some Latin American and Caribbean cities.


Policy Research Working Paper 8702
The impact of urban form on economic performance and quality of life has been extensively recognized. The studies on urban form have focused in developed countries; only a few cities in developing countries have been studied. This paper utilizes nighttime lights imagery and information on street networks, automatically retrieved from OpenStreetMap, to calculate a series of spatial metrics that capture different aspects of the urban form of 919 Latin American and Caribbean cities. The paper classifies these cities into clusters according to these spatial metrics.
It also studies the relationship between the urban form metrics and some factors that can correlate with urban form (topography, size, colony, and economic performance) and performs a spatio-temporal analysis of urban growth from 1996 to 2010. Among the results, the paper highlights the identification of five typologies of cities, the tendency of a group of cities to grow at a steeper slope, some worrying cases of urban growth over protected areas, and a trend toward increasing sprawl in some Latin American and Caribbean cities. This paper is a product of the Social, Urban, Rural and Resilience Global Practice. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/research. The authors may be contacted at nlozano@worldbank.org and prestrepocadavid@worldbank.org.

Introduction
In 2007 the global urban population exceeded the global rural population, and it is projected that urban dwellers will double the rural dwellers by 2050 (United Nations, 2014). Latin America and the Caribbean (LAC) is the most urbanized region on the planet, with 80% of its population living in cities (UN-Habitat, 2012). This growth has not stopped, according to the World Development Indicators, LAC cities added 206 million people between 1960, and 191 million people between 1990(The World Bank, 2017. This rapid level of urbanization imposes challenges for local authorities, as many cannot provide services and infrastructure for the new urban areas at the same speed as the urban growth. Urban form has an important influence on the emergence of agglomeration economies and congestion costs, and on cities' economic productivity, as well as quality of life and environmental sustainability (Squires, 2002). The economies of agglomeration is strongly associated with intraurban connectivity levels, the level of compactness, and land use patterns (Ciccone and Hall, 1996;Cervero, 2001;Rosenthal and Strange, 2004). LAC cities are still spatially growing despite the deceleration in population growth (UN-Habitat, 2012).
However, there is a lack of empirical studies about the characteristics of LAC cities and their changes over time. The challenges of topography may have prevented dense development in some cases, as many Latin American cities are located in rugged topographies with natural barriers. Some of the urbanization in the region is in fact occurring over steep terrain, with slopes greater than 15%, and we verified the trend of urbanizing even steeper terrains in recent years. According to Gencer (2013), these levels of urban concentration of population increase the potential for disasters and most of the cities exposed to at least one natural hazard are located in Asia and Latin America. In some instances, the urbanization occurs at the cost of occupying adjacent protected areas. More than 2,400 square kilometers of protected areas have been lost to urban growth in the region from 1996 to 2010 according to the findings in this work. Rapid urbanization combined with limited infrastructure investments may have also led to urban forms that pose a barrier to the rise of agglomeration economies, limit firm interaction, and make it hard for workers to reach their jobs. A better understanding about the dynamics of urban form in the region can shed light on whether urban policy has a role to play in supporting city productivity, quality of life, and urban sustainability, while mitigating the increased vulnerability that usually comes with urban expansion. 3 In this paper we use nighttime lights (NTL) radiance-calibrated annual image composites, for 1996, 2000 and 2010, to provide a standardized characterization of the urban form of 919 Latin American cities. We make a descriptive analysis of the urban form variables, and their relationships with the potential factors that affect the urban form. We then look for differences in the data across countries and regions and identify the trends of change over time.
The rest of the paper is organized as follows. Section 2 presents a literature review. Section 3 describes the data used in this analysis, and the definitions of urban form variables and factors.
Section 4 shows the descriptive analysis of the set of LAC cities, the differences across countries, and the interplay with the factors that affect urban form. Section 5 shows the space-time analysis of urban growth in the region and Section 6 presents the main conclusions of this work.

Literature review
The importance of urban form has been extensively recognized in the urban economics literature. The form of cities affects their economic performance (Parr, 1979), sustainability levels (Breheny, 1992;De Roo and Miller, 2000), quality of life (Squires, 2002), commuting costs (Wheeler, 2001), costs of transporting intermediate goods (Ciccone and Hall, 1996), the level of knowledge spillovers through human interactions (Lynch, 1981;Jaffe et al 1993 andGlaeser 1998), and the level of matching of firms and workers (Ciccone and Hall, 1996;Rosenthal and Strange, 2004;Cervero, 2001). All this together makes urban form a key factor from an urban policy standpoint.
However, the literature on urban form is not extensive, and it is mainly focused in North American cities (Hasse and Lathrop, 2003;Ewing et al., 2002;Filion, 2001;Tsai, 2005); European cities (Schwarz, 2010;Kasanko et al., 2006, Antrop, 2004Cheshire, 1995;CEC, 1992;Champion, 1992), and Asian cities (Deng and Huang, 2004;Lin, 2002;Sorensen, 2000). In these studies, the main focus has been to measure the spatio-temporal patterns of urban sprawl. 4 J.C. Duque, N. Lozano-Gracia, J.E. Patino, P. Restrepo, W. Velazquez / LCR Flagship Report on Cities and Productivity Developing countries have been much less studied. Most of them treat specific cases such as Lagos, Nigeria (Barredo et al., 2004); Morelia city, Mexico (Lopez et al., 2001); or Cairo, Arab Republic of Egypt (Sutton and Fahmi, 2001). From the best of our knowledge there is evidence of three crosscountry studies that include LA cities in their sample: Huang et al. (2007) study 17 Latin American cities that are part of a total sample of 77 metropolitan areas in Asia, the United States, Europe, and Australia. Using a semi-automatic process for delineating the urban extents from Landsat ETM imagery, the authors calculate seven spatial metrics that capture five dimensions of urban form: compactness, centrality, complexity, porosity and density. Angel et al. (2010a) include 17 Latin American cities in a total sample of 120 cities from all around the world. Using also satellite images the authors calculate four urban fragmentation metrics: edge index, openness index, core open space ratio, and city footprint ratio. Finally, Inostroza et al. (2013) use Landsat imagery to study urban sprawl and fragmentation in 10 Latin American cities. The metrics for measuring the sprawl are: built-up area, density, spatial configuration, and speed. The common finding in these three studies is that Latin American cities are more compact and denser than their counterparts in either Europe or North America.
In our paper we contribute to the literature on urban form in three ways: First, we combine the use of NTL imagery, for urban extent delineation, with recently developed software on computational geometry and network topology (Angel et al., 2010b;Boeing, 2017), to generate a highly automated procedure that makes possible to analyze 919 urban extents in Latin America and the Caribbean. This is the most comprehensive study of urban form for this region. Second, the indicators used to describe urban form are tightly connected to the definition of urban form from the urban planning perspective, which describes urban form as a combination of external shape and internal structure (Whyte, 1968).
In this paper we implement a definition of urban form that includes three dimensions: Shape of the urban extent, internal urban structure and land use pattern (Prosperi et al., 2009;Batty, 2008). This definition allows us to go beyond the debate sprawl vs. compact and cover other important aspects of urban form. Finally, we explore the relationship between urban form and factors such as topography, size of the city, economic strength (measured with NTL), and heritage of colonial urban planning.

Study region
This analysis focuses on Latin American and Caribbean (LAC) cities with more than 50,000 people in 2010. According to (UN-Habitat, 2012), Latin America and the Caribbean is the most urbanized region on the planet with more than 80% of the population living in urban areas, and built-up area continues to expand despite deceleration in population growth. The LAC region is characterized by large diversity of climatic conditions, including from tropical to temperate climates; and the terrain also shows high variability in altitudes due to the presence of the Andean Mountain Range. That diversity in climate and topographic conditions translates into a high diversity in the shape that urban areas have taken. On the one hand, there are cities located in the mountains constrained by hilly terrain; on the other hand, there are also some cities located in the lower lands in flat areas, constrained only by the presence of large rivers or lakes, and yet other cities have grown located at the Caribbean and Pacific coasts, whose shapes reflect the coastline. Speaking of urbanized area, in the region it accounts for 0.4% of the land, a much lower value than Europe (2.3%) and the United States (1.2%), somehow closer to Asia (0.7%), but higher than the former Soviet Union region (0.2%) and Oceania (0.1%) (Y. Zhou et al., 2015). When compared to other regions of the world, LAC cities show denser populations than Europe and the United States, greater regularity and lower open space (Angel et al., 2010a;Huang et al., 2007). However, Inostroza et al. (2013) reported a general trend towards low density growing for some of the largest LAC cities.

Use of Night-time Lights Data for urban extents delineation
The nighttime lights data (NTL) are based on nighttime imagery recorded by the Defense Meteorological Satellite Programs -Operational Linescan System (DMSP-OLS), which reports the recorded intensity of Earth's surface lights. Nighttime lights products have high correlation to human activities (Hsu et al, 2015), and have been previously used for regional and global analysis of urbanization (Cheng et al., 2016;Pandey et al., 2013;Sutton et al., 2006;Zhang and Seto, 2011;Zhou et al., 2015a;Zhou et al., 2015), population modeling (Anderson et al., 2010;Lo, 2001), and economic performance (Cao et al.,2016;Forbes, 2013). There are two different nighttime light products from DMSP-OLS that can be used to delineate urban areas: the stable or ordinary product (NTL), and the radiance-calibrated (RC) product. For this work we use the latter, given that such data include a correction for the saturation issue likely to be an important issue for most large cities in the region. Further, previous work has suggested that the RC information provides a better proxy for socioeconomic variables than the stable products (Hsu et al., 2015;Ma et al., 2014). Radiance-calibrated annual composites for 1996, 2000 and 2010 obtained from NOAA National Centers for Environmental Information 1 were used to delineate urban extents.
These composites have a spatial resolution of 30 arc-seconds (about 1 km at the Equator).
Another known issue of the DMSP-OLS products is the "overglow" effect: dim lighting detected from light in surrounding areas of cities because of the scattering of lights in the atmosphere . A novel deblurring process was applied to address the issue of over glow in the radiancecalibrated products. This process involves the use of two sequential filters, a standard deconvolution and the frequency of illumination maxima, to withdraw the light from the surroundings back and restacking it vertically on its source pixels at city centers (Abrahams et al., 2016).

Calibration and interannual correction of deblurred RC data
Deblurred DSMP-OLS RC annual composites for the years 1996, 2000 and 2010 were previously inter-calibrated and corrected for a multi-temporal analysis of urban form and city productivity in Latin America (Duque et al., 2017a). In that work, the three nighttime images were used to delineate urban extents in each year for most of the Latin American and Caribbean cities that had more than 50,000 people in 2010 using a threshold approach. Figure

Measuring urban form
In this paper we adopt an integral definition of urban form from the urban planning literature.
According to contributions such as Whyte (1968), Batty and Longley (1994) and Prosperi et al. (2009), the characterization of urban form should include information on the following three dimensions: shape of the border, urban texture, and land use patterns. This multidimensional definition of urban form allows going beyond the differentiation between sprawl and compactness into a definition that differentiates between natural/organic and planned/regular/artificial/geometric cities. In this paper we propose a series of seven indicators that to cover these three dimensions.
According to Angel et al. (2010b), the geometric characterization of a shape includes three aspects: degree of roundness, smoothness of the perimeter, and fullness. Using a perfect circle as the perfect benchmark shape, the degree of roundness is measured with the Exchange_index that measures the 9 degree in which the shape of a polygon deviates from its equal-area circle. It is calculated as the share of the total area of the urban extent that is inside the equal-area circle about its center of gravity (Angel et al., 2010b). An Exchange_index equal to 1 corresponds to a perfect circle. As the Exchange_index moves towards 0, the shape of the polygon becomes irregular, elongated, and noncompact. The smoothness of the perimeter is measured with the Perimeter_index, which indicates how smooth is the perimeter of the urban extent. It is calculated as the ratio of the perimeter of the equal-area circle and the perimeter of the shape (Angel et al., 2010b). A Perimeter_index equal to 1 indicates a totally smooth perimeter, found in a perfect circle. A Perimeter_index close to 0 indicates a highly irregular perimeter, which is very common in natural/organic cities located in rugged topography. Finally, the level of fullness is quantified with the Fullness_index, which measures the presence of built-up areas within the urban extent as a fraction of the urban extent area. It is calculated as the fraction of the total area of the urban extent that is built-up. This measure can be understood as the complement of measures such as porosity or sprawling (Burchfield et al., 2006). A Fullness_index equal to 1 corresponds to a totally compact/built-up city. A Fullness_index close to 0 represents a sprawling city with lots of empty spaces within its borders. In this study, the Fullness_index was Next, we study two dimensions of a city structure related to two dimensions of connectivity: first, the structure given by the layout of the road network, and second, the connectedness of its segments. We measure urban structure with the Circuity_avg, which takes values close to 1 when the streets in the network are mostly straight lines, and greater than 1 when the streets are curvier/organic. This metric is calculated as the average ratio between an edge length and the straight-line distance between the two nodes it links (Boeing, 2017). Regarding the connectivity we applied the two most popular metrics for assessing connectivity: Intersection_density and the Street_density; Intersection_density is calculated as the sum of all edge lengths divided by the area (Boeing, 2017), and Street_density is calculated as the sum of all edges in the undirected representation of the graph (in km) divided by the area in Km 2 (Boeing, 2017). Both metrics inform about the ease of movement across the city (Boeing, 2017). High values of these two measures are associated to high rates of walking and use of nonmotorized transport modes (Cervero, 1997). Finally, in order to measure the third dimension, the land use pattern, we use the metric proposed by Fallah et al (2011), the Sprawl_index, which approaches the land use pattern by measuring the level of evenness in the distribution of population within the urban extent. When the population is evenly distributed, the Sprawl_index takes values close to 0; when the population is highly concentrated in a portion of the urban extent, the Sprawl_index takes values close to 1. It is calculated as the normalized difference between the share of areas with population density below the regional average density and the share of areas with population density above the regional average density (Fallah et al., 2011). We implemented this metric using population counts at pixel level retrieved from GHS population raster layers for 1990, 2000 and 2015 at 250 meters of spatial resolution Pesaresi et al., 2016) within the urban extents extracted from DMSP-OLS NTL RC images.
As one of our goals is to generate highly automated methods, we used the "Shape Metrics Toolbox", which is a Python script that runs in ArcGIS, to calculate two of the three the shape metrics of the urban extents (Exchange index, and Perimeter index). This software is intellectual property of the Center for Land Use Education and Research (CLEAR) at the University of Connecticut (http://clear.uconn.edu/tools/Shape_Metrics/index.htm). We used the OSMnx Python library to compute three network topology variables (Circuity, Intersection_density, and Street_density) (Boeing, 2017). This library uses OpenStreetMap to retrieve, automatically, the street networks within a given polygon, and calculates a series of network-based metrics. 11 Table 1 presents examples of urban areas with high, medium and low values of these seven variables that describe urban form; and Figure

Factors that can affect urban form
We explore the relationship between the urban form variables and four factors that may affect the way in which an urban extent grows. The selected factors and the respective rationale are the following:  Available land: The natural constraints to urbanization play an important role in defining urban form and structure. Cities with rugged topography located in the mountains should show more organic street networks and a less smooth perimeter than those with flat topography.
 Size of the city: As the city grows, the "best" land for urban growth may start to become scarce. In this situation, the city may consider occupying new areas that would not consider before, and its indicators of urban form may begin to deteriorate.
 Culture/Colonization history: The colonization history in Latin America is seen as proxy for cultural differences that translate to urban planning practices and ultimately to the resulting urban form today. Most historic centers of cities that were Spanish colonies show a more regular urban layout than cities that were colonies of other countries. We expect that the colonization differences may translate to differences in the urban structure in Latin American and Caribbean cities, with cities that were Iberian colonies showing more regular and organized urban layouts.
 GDP: Proxied with the density of radiance within the urban extent, as recorded by DMSP-OLS RC nighttime imagery. We expect that cities with high GDP can have the financial 15 resources to ensure a planned and orderly urban growth and to provide, in a timely manner, the new areas with the required public services.

Descriptive analysis
This section describes the whole set of LAC cities in terms of its urban shape, texture and land use variables. Figure Figure 5a) values tend to be high, with median and mean values above 0.5 in the three dates.

Cluster Analysis
In this section we classify the cities into homogeneous groups according to their profile of values on the seven metrics. For this analysis we adopted the sequence proposed by Huang et al. (2007): First, we determine the number of clusters based on a visual inspection of the dendrogram that results from running a hierarchical clustering algorithm. The dendrogram in Figure 6 shows that four would be a proper number of clusters. Since the hierarchical cluster is a highly constrained algorithm (i.e., once two sets of areas are collapsed into one cluster, they are forced to remain together in further solutions, which can lead to suboptimal solutions). The second step consists of running a K-means clustering algorithm to aggregate the cities in four clusters. The profile of each cluster is presented in the parallel coordinate plot in 21 Figure 7. Cluster 1 is composed of cities with a highly organic street network, with poor connectivity and sprawled. Cluster 2 is composed of cities with organic perimeter, but with a street network with high level of connectivity and low sprawl. Cluster 3 is composed of compact cities with good connectivity and low sprawl. Finally, Cluster 4 is composed of compact cities with moderate level of connectivity and highly sprawled.
The mean and standard deviation of the variables for each cluster are presented in Table 3. Figure 8 shows the map of the spatial distribution of the cluster groups and Table 4 shows the proportion of cities in each cluster by country. Even though a clear spatial pattern is absent, there are some subtle patterns that help to understand the differences across countries shown in Figure 5.

23
In this section we analyze the urban form variables in regard to each of the factors that can correlate with urban form: topography, size, colony, and economic performance. Figure 9 shows the distributions of urban form variables (rows) grouped by factors (columns). Factor slope was grouped in two classes to differentiate between cities located in flat terrain and cities located in the mountains or rugged terrain: below 10% and above 10%. Factors size of the city (Area_km 2 _2010) and economic performance (using the density or NTL radiance values within the urban extent as proxy) were grouped into quartiles to inspect differences between quartile groups. Factor colony was assigned according to the colony that the area was part of in 1700: Portugal, Spain, England, the Netherlands, and France (Gascoigne, 2001).

Topography
We analyze the topography using the mean slope (%) within the urban extent (Figure 10), first column). While roundness variables nExchange and nPerimeter do not show important differences between the two classes, Fullness does show a small but legible difference: urban extents with slopes above 10% tend to lower values than those with slopes below 10%. This indicates that the cities located in rugged terrain have more empty spaces than the cities located in flat terrain. Urban texture graphs corroborate that cities with steepest slopes have lower values of street and intersection densities, as well as more organic urban layouts, with higher values of Circuity_avg. This is an expected behavior because the street network in rugged terrain should adjust to the topography with 27 curvy roads. Also, according to Batty and Longley (1994), organic cities, which area associated with high levels of Circuity_avg, fit their natural landscape. That explains the high and positive correlation between Circuity_avg and Slope (0.64). The relation with Sprawl indicates that cities with steeper slopes have a less uniform distribution of their population within the urban areas than cities located in flat terrain.

Size of the city
We classified the urban extent area in square kilometers into quartiles and built kernel density plots of the distribution of urban form variables for each group (Figure 10  while street density and intersection density distributions show small differences indicating that larger cities have higher street and intersection densities, as expected from the inspection of fullness distributions. Sprawl distributions also show an interesting trend: larger cities have lower values than the rest and a low negative correlation (-0.25), which indicates that the population in those cities tends to be more uniformly distributed than in the smaller urban areas.

Culture/Colonization history
The colonization process influences urban design and growth trends. We classified the urban extents in LAC region according to the colony they belonged in 1700 (Gascoigne, 2001), and built the kernel density plots of urban form variables for each class (colony). Portugal, Spain, England, the Netherlands and France were the countries that had colonies in Latin America and the Caribbean by 1700. The third column of Figure 10 shows

GDP (proxy: density of radiance in NTL imagery, dr_ntl)
We used the density of radiance from the nighttime imagery as proxy for GDP (dr_ntl). We classified the LAC cities in quartiles according their registered radiance density in 2010.

Fragmentation
We

Growth vs. initial spatial size
Initial spatial size of the urban extent does not seem to play an important role in the magnitude of urban growth. As shown in Figure 15, the initial area of the urban extent does not show significant differences between quartile groups in regard to urban growth: all four quartile groups have similar central values and their distributions have very similar shapes in both periods. 33 Figure 15. Initial urban extent size (area in Km2, classified in quartile groups) and urban growth (as percentage of initial size).

Growth vs. population density
In regard to the relationship between urban growth and population density, we analyzed the distributions of the initial population density (for years 1996 in the period 1996 -2000, and 2000 for the period 2000 -2010) divided in quartiles groups ( Figure 16). Initial population density seems to play a weak role in the spatial growth of the urban extents: the denser urban areas (Q4 group) tend to grow about 0.1% to 0.2% more than the rest of urban extents. This was expected as the less dense urban areas can accommodate new people up to a saturation point without spatial growth. When the population density is close to reaching its physical limit or saturation point, spatial growth to new areas must accommodate the new population. Figure 16. Initial population density (people/Km2, classified in quartile groups) and urban growth (as percentage of initial size).

Growth by saturation
As stated above, growth for saturation takes place when the urban population growth does not imply new spatial growth until a top population density is reached. This situation can take place when the 34 J.C. Duque, N. Lozano-Gracia, J.E. Patino, P. Restrepo, W. Velazquez / LCR Flagship Report on Cities and Productivity population density in a city is below its upper limit for its urban carrying capacity. 2 In this scenario, the city can increase its population by infill of available spaces within the urban area, or via an increase in population density, before the occupation of new areas. When the population density has reached its maximum or saturation point, the only option left is growth on new adjacent areas. We analyzed if the urban growth is a consequence of saturation by comparing the highest population densities reached in the new urban areas in 2000 and 2010, with the population density reached in the respective date in the urban extent area for the initial date (1996). We built an indicator of saturation for each period as shown in Table 5  New urban extent between years (t-k) and t.


Population at year t Saturation values above 1 indicate that the population density in the new areas is higher than the population density in the prior-date urban extent area. This indicates low saturation, as it can be assumed that the prior urban extent area can still increase its density to a similar level to those in the new areas. Values below 1 indicate that the population density in the new areas is lower than the population density in the prior-date urban extent, indicating that the prior urban extent area is denser and fuller than the new areas; and that growth could be the result of saturation. Figure

Conclusions
Open spatial data, remote sensing imagery and automatic data retrieval systems like OSMnx, open a new horizon of possibilities of performing standardized and massive studies of urban form across cities and countries. A better, comparable and systematic study of the evolution of the form of urban extents can allow urban planners to monitor more closely the evolution of urban extents and make timely decisions on the way urban extents are evolving. 37 Based on the integral definition of urban form, from the urban planning literature, this paper approaches urban form as the combination of three dimensions: shape of the perimeter, urban structure and land use. Several indicators on each dimension where calculated using free data and highly automatized libraries for geometry and network analysis. Those indicators show that LAC cities are more rounded than elongated, with urban perimeters more smoothed than complex, and little open space. The street networks in LAC cities tend to be regular and dense. Using these indicators, LAC cities can be divided into four distinct clusters that range from compact and wellconnected cities, to highly organic, sprawled and poorly connected cities. Cities in each cluster do not show a clear spatial pattern.
Regarding the factors that can be correlated to urban form, we found that factors such as topography and historical legacies linked to urban design (colony history) are highly correlated to urban form indicators. Organic/natural urban structures are usually located in rugged topography; and LAC cities that were Iberian colonies have more regular urban layouts and denser street networks. As found in previous literature, this study finds a positive correlation between compactness/connectivity and levels of productivity.
Finally, the spatiotemporal analysis detected some signs of urban growth over steeper terrains for the cities initially located in rugged terrain, as well as a weak trend toward less fragmented urban growth between 2000 and 2010. The cases of Bolivia and the República Bolivariana de Venezuela emerged as critical cases of urban growth over protected areas, which can jeopardize sustainable urban growth in those countries.
Future research will explore the possibility of using this input data and indicators to forecast the evolution of the urban extents, which can help urban planners and policy makers to identify in advance situations of potential exposure to natural hazards, invasion of aquifer reservoirs or other non-sustainable forms of urban growth.