Bogotá, the capital of Colombia, is located on a high plateau at ~2,600 m in the Sabana de Bogotá, an old sedimentary basin in the center of the country [21]. The city is limited by the Andes to the east and by the Bogotá River to the west (Figure 1). Cali (Santiago de Cali) is the capital of the Cauca Department, one of the most important agricultural regions [22]. The city is located in a vast tract of alluvial soil at ~1,000 m and is limited to the west by the Andes, and bordered on the east by the Río Cauca (Figure 1). Medellín is the capital of the Antioquia Department and is the principal manufacturing and industrial area (e.g., textile production) in Colombia [23]. The city is located within the Aburra Valley at an altitude of ~1,500 m.

Quito, the capital of Ecuador, is located in the Guayllabamba river basin in the country’s northern highlands. The city is located at an altitude of ~2,850 m on the eastern slopes of the Guagua Pichincha volcano [24] (Figure 1). Guayaquil is the capital of the Guayas Province, along the southwestern coast of Ecuador. The city is located in a delta, bordered to the east by the Guayas River and to the south by the Estero River, mangrove forests, and shrimp farms [25] (Figure 1). It is the country’s principal port and main center for business and manufacturing industries.

Lima is the capital and largest city of Perú, about ten times larger than Arequipa, the country’s second city [26]. Lima is located in a coastal desert along the Pacific coast, on the banks of the Chillón, Rímac, and Lurín rivers at an altitude of ~160 m (Figure 1). It is the fifth largest city in Latin America, and one of the main financial centers [26].

The conurbation of La Paz/El Alto is the world’s highest city at an altitude of ~3,600 m. La Paz, the administrative capital of Bolivia (i.e., seat of government), is located in a steep, enclosed basin of the eastern Bolivian Andes, along La Paz River [27]. El Alto is a satellite city located on flat terrain above La Paz at an altitude of ~4,150 m on the Altiplano, a high plateau to the west of La Paz [27] (Figure 1). Santa Cruz de la Sierra is the capital of the department of Santa Cruz, and it is located in the central lowlands of Bolivia at an altitude of ~416 m. It is bordered by the Piraí River to the west, and by agricultural plains to the north, east, and south (Figure 1). Santa Cruz is a major service hub for the oil and natural gas industry and the agricultural sector [28].

The first step of the analysis of night-time lights was to establish a correlation between the DN values of the NTL images and different intensities of urban development (Figure 2). The NTL image classification and thresholding analysis used is described in detail in Alvarez-Berríos et al. [20], and is summarized below. We first defined four types of urban land-use classes using indices of urban compactness following Galster et al. [54]: no-development, scattered development, suburban development, and compact development. We then used Google Earth’s Quick Bird and IKONOS images (circa 2000–2009), to find and delineate 76 sample polygons representing the four urban land-use types across the four countries and in different years. We imported the sample polygons (kml format) into a shapefile format using GIS, and superimposed each polygon with the corresponding NTL image to record the minimum and maximum DN values of NTL within each polygon.

We used a multivariate analysis of variance to examine the range of DN values associated with each land-use class, and found that there were significant differences in the minimum and maximum DN values of NTL for each of the four land-use classes (ANOVA: minimum f = 138.30, p < 0.001; maximum f = 135.58, p < 0.001). The range of DN values for each class was: (1) DN 0–5, no-development; (2) DN > 5–20, scattered; (3) DN > 20–52, suburban; (4) DN > 52–63, compact (Figure 2). We further subdivided the last class into compact development (DN > 52–60) and central core (DN > 60–63), to distinguish the city center (i.e., most dense urban area) from other urban land classes (Figure 2). These two categories were statistically different (Mann-Whitney U-test: minimum u = 8, p = 0.02; maximum u = 1.5, p = 0.003). The central core class was established after analyzing the distribution of DN values within the city polygons, which showed that all the city centers had DN values over 60. These ranges were used to classify the NTL images for 1992 and 2009 into five light “intensity” classes, as a proxy of each land-use type (Figure 2). Finally, we mapped the distribution of the NTL classes in 1992 and in 2009 within the city polygons, and quantified the changes in area of each NTL class from 1992 to 2009 to determine the dominant patterns of urban development per city.

The time series analysis of the average light intensity or DN value per city between 1992 and 2009 revealed differences in the level of urbanization and in the rate of urban growth between cities (Figure 3). We identified three different trajectories of urban growth: (1) cities with very high levels of urbanization and little or no change in mean light intensity (Bogotá, Medellín, and Cali); (2) cities with relatively high levels of urbanization in 1992 that experienced slow urban growth (Lima and Guayaquil); and (3) cities with low levels of urbanization in 1992 that experienced rapid urban growth (La Paz/El Alto, Quito, and Santa Cruz). The analyses also showed that Quito, Santa Cruz, and particularly La Paz/El Alto grew rapidly between 1992 and 2000 (Figure 3). Overall, we identified a deceleration in the rate of urban growth between 2000 and 2009 in all cities (Figure 3).

The distribution of light classes within the city polygons, showed that in 1992, most of the built-up areas in Bogotá (83%), Cali (68%), Medellín (68%), and Guayaquil (62%) were classified as central core (DN > 60–63), indicating high levels of urbanization (Figure 4, Figure 5(a)). In contrast, the central core area of Quito (10%) and Santa Cruz (16%) were much smaller, and the majority of the built-up area was classified as suburban (DN > 20–52) (Figure 4, Figure 5(a)). In both Lima and La Paz/El Alto, the central core covered ~35% of the built-up area in 1992; however, in Lima, the suburban class also covered a large area (33%), while in La Paz/El Alto ~23% of the city polygon was classified as undeveloped (DN 0–5) or scattered development (DN > 5–20) (Figure 4, Figure 5(a)).

The major change in the Colombian cities of Bogotá, Medellín, and Cali was the conversion of the suburban and compact classes to the central core class (i.e., densification) (Figure 4, Figure 5(a,b)). By 2009, over 90% of the built-up area in these cities was classified as central core, a clear example of the densification process (Figure 4, Figure 5(b)). This densification occurred mainly along the southern and eastern peripheries of these cities (Figure 4(b)). The urban dynamics of Guayaquil, Ecuador were similar to those of the Colombian cities, characterized by a transition from the suburban and compact classes to the central core class (Figure 4, Figure 5(a,b)). The central core expanded from 117 km² to 148 km², covering 78% of the built-up area in 2009 (Figure 4, Figure 5(b)). The central core extended to the northwest of the city center, but mostly to the east across the Guayas River into areas classified as suburban in 1992 (Figure 5(b)). Urban growth in Quito, Ecuador was much more dynamic (Figure 4, Figure 5(a,b)). The central core area of Quito increased from ~32 km² to ~150 km² (Figure 5(b)), extending north, east and south of the city center into areas classified as low (DN > 5–20) and medium intensity (DN > 20–52) in 1992 (Figure 4, Figure 5(b)). Although the city experienced a dramatic increase in the central core class, more than 40% of the city remains classified as compact (~134 km²), and 14% as suburban (~47 km²) (Figure 4, Figure 5(b)).

In Bolivia, both cities experienced important urban changes, particularly La Paz/El Alto where the central core more than doubled, from 68 km² (35%) to 150 km² (76%), extending into areas classified as scattered and undeveloped to the west of the city center in El Alto municipality (Figure 4, Figure 5(b)). Santa Cruz is unique because it was the only city where the area of the compact class (~122 km²) was larger than the central core area (~74 km²) in 2009 (Figure 4, Figure 5(b)). Although the central core more than doubled, extending south and east (Figure 5(b)), it only covered 33% of the total built-up area, with the majority classified as compact (55%) and the rest as suburban (12%) (Figure 4, Figure 5(b)).

In Lima, Perú, the central core class increased from 269 km² (36%) to 414 km² (55%), expanding mostly to the south and along the northern and western edges of the city (Figure 4, Figure 5(b)). In 1992, 71 km² were classified as scattered development, by 2009 virtually all of these areas were converted into urban class with high light intensities (Figure 5(b)).

The urbanization dynamics of the eight cities varied greatly (Figure 3, Figure 4, Figure 5). The dominant pattern of urban growth in Bogotá, Cali, Medellín, and Guayaquil was high-density compact development. The populations of these cities grew considerably between 1992 and 2009 (Bogotá ~2.5 million, Medellín ~1 million, Guayaquil ~1 million and Cali ~680,000), but the urban area did not expand much, resulting in cities with very high densities (Table 1). In contrast, the dominant pattern of urban growth in Quito and Santa Cruz was low-density expansive development. These cities were the fastest growing cities, but they had the lowest population densities and the smallest central cores areas in 2009 (Table 1, Figure 4). For example, the central core of Quito increased 5-fold, but only covered ~45% of the total built-up area in 2009. Similarly, Santa Cruz, which had the highest annual population growth rate, doubled its population, but the central core area only covered ~33% of the total built-up area in 2009 (Figure 4, Figure 5(a,b)). The population of Lima increased by ~2.3 million people, but the patterns of urban development varied greatly throughout the city. The central core of Lima expanded to the south and north, while development to the east was dominated by suburban and compact development (Figure 5(b)). The most dynamic and rapidly changing city was La Paz/El Alto, which experienced the most expansive urban growth, but at relatively high densities since virtually all new development was classified as central core (~76% of the total built-up area in 2009) (Table 1, Figure 5(b)). Much of the new central core was located in the municipality of El Alto, in areas that were classified as rural or scattered in 1992 (Figure 4, Figure 5(a)).

Geography and local landscape characteristics played a major role in shaping urbanization dynamics. In Colombia, the high-density compact urban development pattern detected by NTL could be associated with the steep geography of the Andes. The cities of Bogotá, Cali, and Medellín were built in the foothills of the Andes, restricting land availability for further urban expansion and causing the densification of existing urban areas. Often, the edges of cities in this region are marked by steep slopes and terrains inadequate for building, but the pressures of population growth, due to natural increase and immigration, have increased the demand for housing and many informal settlements have been built in these marginal areas [56,57,58]. A classic example of this dynamic was the colonization of the steep slopes to the northeast of Medellín (e.g., Santo Domingo) that began in the 1950s [59]. Today, these areas are some of the most densely urbanized in Medellín, with over 400 dwellings per hectare [30]. In contrast, the majority of expansion during the last 20 years has occurred in high income municipalities (e.g., Sabaneta, Envigado) in the southern part of the city (Figure 5(b)). This growth has included new industrial areas and gated communities [60]. In Bogotá, urban growth occurred largely on the southern edge of the city, specifically in the poor neighborhoods of Usme, Ciudad Bolívar, and Soacha (Figure 5(b)). These areas were classified as suburban (DN > 20–52) in 1992 and shifted to central core (DN > 60–63) by 2009 (Figure 4, Figure 5(a,b)). Currently these neighborhoods are the fastest growing and most densely populated areas in Bogotá [57,61]. Ciudad Bolívar concentrates the poorest population (estimated between 800,000 and 2 million inhabitants), most of them immigrants and displaced people [56,61]. In Bogotá, the lack of land has also led to urban development in wetland areas; approximately 80% of Bogotá’s wetlands have been urbanized [57]. This tendency to occupy vulnerable and environmentally hazardous areas was also observed in Cali. Urban growth in Cali occurred largely to the east (Figure 5(b)), particularly in the southeast (e.g., Aguablanca). This area is the poorest and most densely populated neighborhood in Cali, dominated by informal settlements located in flood-prone areas on the marshy plains of the Río Cauca [62].

Similarly, virtually all urban development in Guayaquil has occurred in wetland ecosystems because the city is located in a low and marshy area (~4.5 m above sea level) between the Guayas River and the Salado Estuary [55]. Consequently, many informal settlements have been established in flood-prone areas without proper planning [56]. Recent urban growth in Guayaquil occurred largely to the east of the city center, spreading into the hilly hinterlands of Eloy Alfaro and Durán on the left bank of the Guayas River (Figure 5(b)). In Quito, since the city center is limited by the Andes to the east and by the Pichincha Volcano to the west, most of the urban expansion occurred to the south and east (Figure 5(b)) into agricultural lands in the valleys of Chillos, Tumbaco, and Cumbaya. Most of the development in these areas was dominated by low-density and low-rise structures [57]. Consequently, Quito has the smallest central core and the lowest urban population density of the eight cities studied (Figure 4, Figure 5(b)), yet having one of the highest population growth rates (Table 1).

Landscape characteristics are also an important factor defining urbanization dynamics in Lima, which is located in a narrow desert coastal plain. The central core of Lima expanded mostly to the south, particularly in the municipality of Villa El Salvador (Figure 5(b)). In the early 1990’s, the main land covers of this municipality were deserts and agricultural lands, but illegal land invasions have converted many areas into informal settlements [63]. For example, in January 2000, land zoned for agriculture and livestock grazing in this municipality was invaded by more than 10,000 squatters [64]. Presently, Villa El Salvador is one of Lima’s largest low-income municipalities, with over 400,000 inhabitants and where ~50% of the population lives below the poverty line [65]. Although there are no official statistics, estimates show that ~20%–30% of the city’s population live in informal settlements [63].

Urbanization dynamics in La Paz/El Alto in the Altiplano of Bolivia were also shaped by the geographic conditions of the area, specifically by the restrictions of land availability in the city center which is located in a steep and narrow valley. According to the analysis of NTL, the central core expanded west into the municipality of El Alto (Figure 5(b)), where the terrain is characterized by flat highlands with no physical barriers for urban expansion. Originally a neighborhood of La Paz, El Alto became an autonomous municipality in 1988 due to rapid urban growth caused by high immigration and the lack of space in La Paz city center [66]. The population of El Alto increased from ~400,000 in 1992 to ~1 million in 2009, accounting for over 50% of the total city population [29]. Currently, the majority of the manufacturing industries in the Bolivian highlands are located in El Alto [27]. However, it is also the poorest area, commonly known as the “indigenous city” since 95% of the population is indigenous, mostly Aymara (75%) and Quechua Indians that migrated from rural areas [27]. In 2006, about 80% of the population in El Alto lived in low-income settlements, and 70% lived on less than US$1 per day [66].

Rather the contrary occurs in the Bolivian eastern lowlands of Santa Cruz, where the vast flatlands around the city center have no natural barriers that limit urban expansion, besides the Piraí River to the west (Figure 1). The physical characteristics of the terrain and the large influx of immigrants attracted by the 1990s economic boom following the discovery of large natural gas reserves in the region fueled rapid and uncontrolled urbanization [28]. The results showed that between 1992 and 2009, the urban population in Santa Cruz doubled, having the highest annual population growth rate of all the cities studied (Table 1). The analysis of NTL revealed that the central core of Santa Cruz expanded, but the most important dynamic was the wide expansion of compact and suburban urban areas to the east and south of the city center (Figure 4, Figure 5(b)). Most of these growing areas are located outside the city’s concentric ring roads built to restrain urban growth as part of the Plan Techint urbanization scheme of 1960 [67]. Consequently, the original radial structure of the city has been altered by unplanned urban development outside the city’s rings, mostly because the municipal authorities left urban development to private developers, subject mostly to market forces [28]. The absence of geographical barriers plus the expansive and uncontrolled urban growth are probably responsible for Santa Cruz’s low population density levels (Table 1).

Landscape characteristics are also related to land value, a major driver of urban development [14]. Virtually all urban expansion in La Paz/El Alto occurred within the municipality of El Alto due to the availability of relatively inexpensive land. For example, the cost of renting in La Paz is approximately twice as expensive as in El Alto [68]. This has attracted middle-class households looking for affordable housing [27]. In the case of Santa Cruz, the high prices of rental properties, the lack of affordable housing in the city center, and the absence of policies to control land use and land speculation has encouraged illegal land dealers to sell state-owned land to rural migrants around the city periphery [28]. In the case of Cali in Colombia, the agricultural industry is one of the main economic activities for the city (e.g., sugar cane, coffee), and the high values of agricultural lands act as barriers for urban expansion [69]. Agricultural lands visibly mark the urban-rural frontier, particularly to the east. A similar dynamic is observed in Bogotá, where agricultural lands to the west of the city are mainly used for flower production. Flower exports are an important economic activity that generate over US $900 M annually, thus limiting urban expansion [70]. Furthermore, the construction sector in Bogotá has experienced a deep financial crisis since the mid-1990s that has affected the provision of formal housing throughout the city, particularly for low-income families. Approximately 11,000 affordable housing units (US $9,000–US $11,000) are constructed annually in Bogotá; however the annual demand is ~40,000, forcing many families to settle in informal developments along the periphery of the city [28]. As a result, peripheral zones in Bogotá tend to have very high population densities, mostly due to overcrowding [71]. Accordingly, while urban sprawl has been the dominant growth pattern in most developing cities worldwide, the city of Bogotá has experienced a densification process during the last decades [72]. A recent study showed that the overall density in Bogotá increased ~8% between 2001 and 2008, which represents an increase of approximately 1,400 inhabitants per square kilometer [71]. This dynamic is related to the densification of suburbs and peripheral zones, but also to the densification of areas served by the Transmilenio mass-transit bus system [71,72].