Extraction of built-up area from Landsat satellite imagery time series of the city of Santarém - PA – 2012 –2002

Project title: Extraction of built-up area from Landsat satellite imagery time series for the city of Santarém, Pará (2002–2012)

Project keywords: Spatial planning; Built-up area; Remote sensing; Medium-sized Amazonian cities; Time series; Geotechnologies

Project field of knowledge: Urban and regional planning methods and techniques

Supervisor name: Carolina Moutinho Duque de Pinho [Link]

Related publication: CBERS-4A IMAGERY FOR MAPPING URBAN LAND COVER IN THE AMAZON [Link] - April 2023 | Conference: XX Simpósio Brasileiro de Sensoriamento Remoto | In: Florianópolis

Medium-sized cities are of great importance to the urban reality of the Amazon region. In the specific context of the state of Pará, an important category to be analyzed are "medium-sized cities", defined by some authors as urban centers with populations between 100,000 and 500,000 inhabitants that are not part of metropolitan areas (NUNES; JÚNIOR; TRINDADE, 2016; PEREIRA, 2004; TRINDADE JR, 2004, 2011). During the 1970s and 1980s, these cities experienced annual population growth rates exceeding 3% (RAMOS, 2014).

Medium-sized cities serve as hubs connecting the flow of information, people, goods, and political decisions, consolidating themselves as focal points in a territory characterized by a mix of traditional communities, smaller urban nuclei, settlement projects, and colonization schemes. This mosaic is made even more complex by the constant presence of the forest, rendering the boundaries between urban and rural fluid and difficult to delineate (DAL'ASTA et al., 2012). This phenomenon led Becker (1995) to coin the term "urbanized forest" (MONTE-MÓR, 2004).

Although the Amazon is frequently recognized for its vast tropical rainforest and the challenges related to its preservation, another important discussion is the region's urban question. About 70% of the Amazonian population lives in urban areas, as pointed out by the IBGE in 2010. This figure underscores the need to integrate urban planning considerations into public policies aimed at the region.

In this scenario, Remote Sensing (RS) plays a fundamental role. Combined with other geoinformation technologies, RS offers a methodologically and economically viable means to study urban phenomena across multiple scales (ALMEIDA et al., 2007). In addition, advances in RS data analysis have been made possible by various computational tools, including Decision Tree Ensembles, which generate increasingly reliable land use and land cover classifications (RICHARDS; JIA, 2006). Based on this context, this work is dedicated to analyzing the urban growth of the city of Santarém, in the state of Pará, between 2002 and 2021. To this end, a combination of satellite images, remote sensing techniques, and statistical algorithms is employed.

2.1 General objectives

This study performs a temporal analysis of images of the city of Santarém, Pará, covering the period from 2002 to 2021. The focus is on mapping built-up areas, employing advanced digital image processing techniques and taking into account the unique characteristics of cities in the Amazon region.

To conduct this work, we used open-access data from the LANDSAT and SENTINEL sensors. The analysis was conducted using the JavaScript programming language, implemented on the Google Earth Engine platform, and Python on the Google Colab platform. The latter offers the advantage of processing large volumes of data without requiring high-powered hardware. In addition, the Random Forest machine learning algorithm was used to manually categorize the images according to criteria relevant to the region under study. This process was complemented with texture metrics, as indicated in the specialized literature.

3 Theoretical framework

Through the literature review, the scientific foundations supporting this research are detailed.

3.1 Formation of Cities in the Amazon Region

The 17th century marked the beginning of Amazonian urbanization, when the first settlements in the region were established by Jesuit expeditions aiming to catechize indigenous peoples. These missionary settlements served as a "spearhead for the occupation of the territory" (BECKER, 2013, p. 11), playing a geopolitical and logistical role in the expansion of frontiers. They were destinations for those seeking the drogas do sertão (backland drugs): spices, timber, pharmaceuticals, ornaments, construction materials, and seasonings, since Portugal had lost its spice markets in the East (CORRÊA, 1987, p. 44).

Over the years, the settlements evolved through different historical and geographical processes. Even if precariously, as they were provided with infrastructure and public services that supplied agricultural areas, these settlements gave rise to villages and various Amazonian cities (CARDOSO; LIMA, 2006). A dendritic occupation pattern predominated in the Amazon—a branched form of urban network organized along a main river and its tributaries, resembling a tree composed of a trunk and branches along riverbanks, with low penetration into the territory (CORRÊA, 1987).

During the 19th century, several villages and small towns already existed in the region, employing enslaved African labor in the Amazon jungle. Between 1850 and the end of World War I, around 1920, the regional milestone of rubber extraction took place, integrating the region into the international division of labor, accounting for up to 90% of global rubber production in 1890 (CORRÊA, 1987). This was a period of high external demand for rubber for the production of pneumatic goods such as bicycle and car tires, causing rubber to surpass the economic importance of the region's traditional products like tobacco, cocoa, and hides.

Until the end of the rubber boom, the typology of Amazonian cities was closely tied to waterways and previous economic dynamics, formed along a "river-floodplain-forest" logic (PORTO-GONÇALVES, 2001; SAKATAUSKAS, 2020). This category matches what Trindade (2015) and Becker and Stenner (2008) call "Cities of the forest"—cities that are more integrated into nature's dynamics and connected with other neighboring cities, yet less integrated into national and global political and economic systems.

It was between the 1960s and 1980s, especially during the military dictatorship, that the State began to act aggressively in the region, though based on a frontier economy paradigm—that is, an economic development model that focuses on continuous expansion into new geographic areas or sectors. In this case, it was associated with the exploitation of natural resources and the incorporation of "undeveloped" or "underutilized" lands into the formal economy. In this program, economic growth was based on the incorporation of land and the exploitation of natural resources perceived as inexhaustible. Although the federal government's discourse emphasized the organization of national integration, the occupation of demographic voids, and economic development, on the ideological plane it justified the capitalist incorporation of the Amazon (BECKER; MIRANDA; MACHADO, 1990; CORRÊA, 1987).

Notably, the "National Integration Program" (PIN), initiated in 1970, aimed to promote economic development and territorial integration, especially in the North and Central-West regions of Brazil (KOHLHEPP, 2002). The program included the construction of major infrastructure, such as highways and hydroelectric plants, and encouraged migration from other regions of Brazil to the Amazon. Although the stated goal was development and national integration, these projects often led to environmental degradation, the displacement of indigenous and local communities, and the unsustainable exploitation of natural resources.

In Santarém, one of the main infrastructure projects associated with PIN was the BR-163 highway, also known as Cuiabá-Santarém. This highway connects the city of Cuiabá, in the state of Mato Grosso, to Santarém, in the state of Pará. The construction of this highway had a significant impact on the region, facilitating the transport of grains such as soybeans and corn from the Central-West to the port of Santarém, from where they are exported. Furthermore, the program encouraged migration to Santarém and other areas of the Amazon, which led to population growth and urban sprawl, but also to problems such as deforestation and land conflicts (KOHLHEPP, 2002).

With these regional integration policies, Amazonian urbanization transformed, and cities following the "Highway-terra-firme-subsoil" typology emerged (PORTO-GONÇALVES, 2001; SAKATAUSKAS, 2020), referring to the introduction of highways and fronts of economic expansion—areas subject to the repercussions of more recently established economic projects.

According to the categorization by Trindade (2015), these formations operate more along the logic of "Cities in the forest", where external demands, such as raw material extraction, become the priority. In this scenario, local traditional values are often replaced by modern urban paradigms. This term shares a certain parallelism with "Highway-terra-firme-subsoil", as both describe cities that are more subject to external influences.

Thus, two contradictory spatial patterns emerge in the Amazon region: the "River-floodplain-forest" model and the "Highway-terra-firme-subsoil" model (PORTO-GONÇALVES, 2001). The former reflects older cities, closely tied to nature and less subject to external influences. This model is also described as "Cities of the forest". The latter corresponds to more recent cities, intimately associated with external demands and with less integration into the natural environment, also categorized as "Cities in the forest".

These typologies provide a valuable lens for understanding the evolution and complexity of urbanization in the Amazon. The exploitation of local wealth has often been driven by external forces, backed by foreign capital and, consequently, conducted in a predatory and homogenizing manner. The result is an impoverishment of the multiple forms of life that previously coexisted in the region, since this external approach lacks the local knowledge required for a more equitable and sustainable management of the territory.

3.1.1 Typologies of the City of Santarém

Shortly after the years following authoritarian centralization in the Amazon and the beginning of the 2000s, there was a tendency toward concentration in large cities, with state capitals absorbing a significant share of the population relative to smaller municipalities, causing accelerated and disorderly urbanization. Currently, the Amazonian urbanization process has shifted and is characterized by the proliferation of small towns and the growth of medium-sized cities, which serve as nodes between the different types of settlements in the region and are of great importance in the urban network (TRINDADE JR, 2011).

Historically, Santarém was a city more aligned with the "City of the forest" and "River-floodplain-forest" typology; due to its strategic location at the confluence of the Tapajós and Amazon rivers, its earliest and still densest urban growth occurred along the riverbanks, which house what is known today as the Port of Miritituba, one of the main port facilities in northern Brazil, having undergone several phases of expansion and modernization over the years.

However, with the implementation of infrastructure projects such as the BR-163 highway (Cuiabá-Santarém), the city began to change and incorporate aspects of the "Highway-terra-firme-subsoil" typology, growing along the highway and also toward the Maestro Wilson Fonseca Airport, which opened in 1977 and is accessed via the Fernando Guilhon Highway, which begins at the Gerardo Monteiro Viaduct over the Santarém-Cuiabá Highway (BR-163) and terminates at the airport.

Therefore, Santarém is a city that encapsulates the transitions and tensions present in Amazonian urbanization. It shifted from being predominantly a "River-floodplain-forest" and "Cities of the forest" city to one that also incorporates characteristics of "Highway-terra-firme-subsoil" and "Cities in the forest", reflecting shifts in public policies as well as geographical and economic conditions.

The restructuring of the urban network and the new roles played by medium-sized cities lead to a new productive and labor market profile, alongside the importance of Amazonian urban centers, breaking old patterns of spatial organization. The most predominant urban logic is that of "Cities in the Forest", yet there is an "urbanodiversity" within the Amazonian territory (TRINDADE JR, 2015)—that is, diverse and plural urban realities in the region, with different types and tiers of cities. Conventional urban mapping frequently fails to capture the complex 'urbanodiversity' of the Amazon. While remote sensing data are effective for medium- and high-density population areas (GONÇALVES et al., 2021), they prove insufficient for low-density settlement patterns common in the region (DAL'ASTA et al., 2012).

This gap in data production is concerning, since currently implemented public policies tend to enforce an inappropriate urbanization model. Such policies promote the adoption of structures and lifestyles that are not only incompatible with local conditions, but also result in wasted resources and deteriorated quality of life (SAKATAUSKAS, 2020).

Still within the discussion of Amazonian city typologies, Trindade (2015) proposes the concept of "cities for the forest", which advocates for integrating local geographical and cultural characteristics into urban planning policies. This makes room for a more diversified and conscious development, in contrast to the blind replication of conventional urban models that lack Amazonian specificities.

3.3 Remote Sensing in built-up area categorization

Data obtained through remote sensing (RS) offer numerous opportunities for urban mapping and monitoring, serving as a basis for physical, climatic, and socioeconomic indicators. They provide quantitative data that are temporally and spatially consistent, complementing traditional surveys such as demographic censuses. Once translated into information and updated regularly, these data can support urban planning (GONÇALVES et al., 2021).

In the context of remote sensing (RS), Amazonian specificities stand out right from the categorization of built-up areas. In the international literature, for instance, the term "urbanized areas" often refers to spaces with high soil imperviousness (Feng et al., 2017). Methodological approaches using RS techniques to study Brazilian cities generally focus on metropolitan areas in the Southeast or regions with high building density, following international literature. However, a distinctive feature of Amazonian cities is their relatively low impervious surface intensity.

3.3.1 Sensors, Biophysical Indices, and Image Classification

The Landsat mission is widely used in urban studies, as it represents the longest satellite imagery historical record; its spatial resolution is sufficient to track the expansion of the urban footprint, but can be limited for more detailed mapping. Gonçalves et al. (2021) state that recent intra-urban target studies have turned to finer imagery, such as that provided by the Sentinel-2A and 2B satellites, which offer images with spatial resolutions between 10 and 17 meters.

Both Sentinel-2 and Landsat series data are open data, meaning they are freely accessible to the public. These missions are funded by different government organizations. Sentinel-2 is part of the Copernicus program, an initiative of the European Union in partnership with the European Space Agency (ESA). Landsat is funded and operated by the United States National Aeronautics and Space Administration (NASA) in partnership with the United States Geological Survey (USGS).

To analyze the imagery generated by these satellites, it is necessary to perform image composition through band manipulation, since the blue portion of the color spectrum is affected by atmospheric scattering. Therefore, different bands must be combined to obtain a natural RGB (Red-Green-Blue) color assignment model or other arrangements (CAMPBELL; WYNNE, 2011).

Figure: Example of spectral behavior of materials

The technique for urban analysis is based on the proper selection of spectral bands during image composition, making it possible to distinguish materials such as asphalt and concrete or identify different types of roofs, depending on the spectral behavior of each material.

Biophysical indices are derived from spectral band analysis and play a key role in identifying materials such as healthy vs. unhealthy crops, asphalt surfaces, concrete, and built structures, among other fine distinctions between objects. Combining bands and biophysical indices offers a comprehensive view of urban features. However, while these metrics are reliable, other factors such as cloud cover, shadows, and sensor angle can affect readings. Therefore, contextualizing the analyzed data is essential.

NDVI

The NDVI Index (Normalized Difference Vegetation Index) is an indicator ranging between -1 and 1. It is used to estimate the presence and condition of vegetation in an area based on reflectance in the near-infrared and red wavelengths.

Values close to 1 generally indicate areas with healthy vegetation, as healthy plants reflect more light in the near-infrared spectrum and less in the red band. Values close to zero or negative generally indicate the absence of vegetation and may represent urban areas, water bodies, or bare soil.

NDWI

Frequently used in hydrological studies, the NDWI (Normalized Difference Water Index) is an index primarily used to identify the presence of water, with values ranging from -1 to 1. In general, higher values indicate a greater probability that the area consists of water. These values are calculated using the near-infrared (NIR) and green bands.

3.4 Random Forest Algorithm

Random Forest is a machine learning method that involves building multiple decision trees during the training phase; each tree is built from a bootstrap sample of the training data, and the individual trees vote to make a final classification decision.

The choice of the number of trees can affect model performance. Generally, more trees make the model more robust and less prone to overfitting. However, this can also increase computation time and required resources. A value of 50 was chosen for this analysis, but this hyperparameter should be tuned and validated using a systematic approach, such as cross-validation, to ensure optimal model performance on unseen data.

4. Methodology

Data collection and analysis were organized into 4 stages, as summarized in Figure X. First, the spatial extent of the city of Santarém, Pará, was defined, along with the period between 2002 and 2022 as the temporal reference.

4.1 Materials and Methods

For the research, one image was used for each year of the analysis.

• One LANDSAT 7 image for 2002;

• Nine LANDSAT 5 images for the time series between 2003 and 2011;

• Eight LANDSAT 8 images for the time series between 2014 and 2021;

• Seven SENTINEL 2 images for the time series between 2016 and 2022.

Their spectral bands can be found in Table 1. Images from 2012 and 2013 were not collected or analyzed. For collection, handling, and analysis, the following software and techniques were used:

• Google Earth Engine platform for satellite image preprocessing and unsupervised classification, using JavaScript and the Random Forest algorithm;

• Google Colab platform and Python language for analyzing the generated classifications;

• QGIS 3.28 for producing thematic maps.

Table 1 - Spectral bands of the LANDSAT 5, 7, 8 and SENTINEL 2A and 2B satellites used.

The city of Santarém is a medium-sized city in the state of Pará, a state with an area of around 1,248,000 km², roughly equivalent to 23% of the Legal Amazon (AL) area and contributing about 40% of the region's GDP (FUNDAÇÃO AMAZÔNIA DE AMPARO A ESTUDOS E PESQUISAS DO PARÁ, 2017). It is the most populous state in the North Region, with about 69% of the population living in urban areas.

MAP OF SANTAREM

According to the 2022 Demographic Census, the population of Santarém exceeds 300,000 inhabitants and is the third most populous municipality in Pará, behind only Belém and Ananindeua. However, despite its importance in the regional urban network and access to technologies, there is a lack of adequate road infrastructure, especially in recently expanded areas (COSTA, 2012).

Table 2 - Population composition in 2010 (Census) and 2020 (estimate) of Santarém

Source: IBGE 2010 and 2022

In this study, the concept of "built-up areas" is adapted to capture the specificities of Amazonian cities. We chose to use the methods developed by Gonçalves et al. (2021), which define built-up areas as those modified by human intervention, encompassing various elements of the urban fabric such as houses, streets, lots, buildings, green spaces, and road systems, among others, but which, in the Amazonian context, also include areas with grasses or bare soil, as they are part of a context of subdivisions and dirt roads, considering them manifestations of the urbanization process.

4.1.3 Supervised classification

In this study, the classification approach employed involved creating manually drawn polygons to categorize pixels. These polygons were used as training data for a Random Forest classifier. After training on the training data, the classifier gains the ability to apply this knowledge to test images, assigning categories or classes to each pixel in those images.

The processing area was defined using the shapefile of the municipal boundaries of Santarém provided by IBGE and its geographic coordinates. This shapefile was loaded into the Google Earth Engine (GEE) platform to delimit and clip the satellite images used exclusively to the area of interest, speeding up processing.

In the preprocessing stage, Sentinel-2 (S2) and LANDSAT 5, 7, and 8 satellite images were prioritized through libraries available directly in Google Earth Engine (GEE). For LANDSAT images, "Collection 1" and "Tier 1" quality tiers were chosen. The Sentinel and LANDSAT images were organized into separate collections, and all described processing steps were executed on both collections, albeit independently. In both groups, images with lower cloud cover were selected.

Pixel classification

In the context of thematic categorization, polygons were defined as markers to represent different classes of interest. The Sentinel-2 satellite images were categorized following the same classes that Gonçalves et al. (2021) used in their analysis of Amazonian cities. These classes include water, herbaceous vegetation, shrub vegetation, bare soil, and built-up areas. Due to the lower image resolution of LANDSAT imagery, we opted to use only three main categories: water, built-up areas, and vegetation.

To enhance the reliability of the analysis, we adopted the strategy of using only polygons that maintained consistent classification throughout the study period. These polygons were considered 'stable'. All images were processed against these polygons to extract the spectral band values corresponding to each one.

This approach aims to improve analysis accuracy in two ways. First, it reduces the impact of temporal irregularities, such as cloud presence, which could distort data in specific years. Second, using stable polygons over an extended period increases data reliability, since the stability of these polygons acts as a control, ensuring that observed changes in data reflect genuine landscape transformations rather than being influenced by seasonal variations, sensor calibration errors, or other sources of spectral noise.

The biophysical indices NDVI and NDWI were included in the loop for higher accuracy in categorizing vegetation and water, respectively.

3.5 Results analysis and data visualization

Analysis of the results was carried out using the Python libraries rasterio, numpy, pandas, and matplotlib, used to perform simple percentage calculations and create graphs. Additionally, thematic maps were produced using QGIS software. These approaches enabled better data visualization for comparing and complementing results obtained from Landsat and Sentinel images.

5 Results and discussion of results

The city of Santarém expanded especially in the areas adjacent to the Fernando Guilhon Highway, where state highway PA-457 terminates, and around the roads extending east toward the airport. In addition, expansions occurred along state highways Santarém Curua-Una (PA-370) and Rodovia Everaldo Martins (PA-457), as well as along federal highway BR-163, which leads to the city's waterfront, where urbanization was already well consolidated in 2002, the beginning of the analyzed time series.

These results are aligned with findings in the literature, which indicate that Amazonian cities have recently grown toward the 'Highway-Terra Firme-Subsoil' typology pattern. The literature also points to cities with hybrid typologies, as in the case of Santarém, which continues to grow near the waterfront belt.

5.1 Choice of Sensors

Images obtained from Landsat 8 and Sentinel-2 can be used in a complementary way to analyze urban dynamics.

Table – LANDSAT 8 and SENTINEL-2 images side by side at 1:15,000 scale

As Table X shows, both at 1:15,000 scale, Landsat 8 images display lower spatial resolution—they are more pixelated, yet sufficient to evaluate the distribution and direction of urban sprawl, particularly because the LANDSAT mission covers many more years and is far older than SENTINEL, making it indispensable for time series.

On the other hand, Sentinel-2 images, with their higher spatial resolution, are more suitable for intra-urban identification and analysis, such as analyzing densification within the urban area; they are also more accurate for distinguishing between different types of vegetation and identifying bare soil areas. Therefore, when classifying LANDSAT images, this study chose to use only 3 classes, and for SENTINEL images, 5 classes.

5.2 Choice of LANDSAT mission imagery

Initially, we chose to use the Landsat 7 image series; however, during the course of the study, it was discovered that the Enhanced Thematic Mapper Plus (ETM+) sensor, the mission's primary sensor, failed in 2003. With this failure, images began showing data gaps, as shown in Image X, compromising their usefulness for the present analyses.

Figure: Example image

Given this limitation, only one LANDSAT 7 image was used for the year 2002, and for the 2003–2011 period, LANDSAT 5 images were used instead. No images were used for 2012 and 2013, and from 2014 to 2021, LANDSAT 8 images were used. Both Landsat 5 and 7 share the same 30-meter spatial resolution in the visible and infrared bands, with minor differences in the spectral range of some bands, which allows for this substitution without significant losses.

5.5 Comparison between LANDSAT and SENTINEL results

Both sets of results show the sprawling of the urban area.

Comparing the percentage of each category across both images, we obtain the following pixel percentages:

LANDSAT 8 - Year 2016 image

• Water: 7.57% of pixels

• Built-up Area: 20.98% of pixels

• Vegetation: 30.15% of pixels

SENTINEL-2 – Year 2016 image

• Water: 7.97% of pixels

• Herbaceous Vegetation (HV): 9.66% of pixels

• Built-up Area (BA): 25.56% of pixels

• Bare Soil (BS): 0.50% of pixels

• Arboreal Vegetation (AV): 15.01% of pixels

6 Conclusions and perspectives for future work

This study provided a deeper understanding of the complexity of urban growth in a city located within the Amazon Rainforest, highlighting the importance of tailored approaches and methodologies for characterizing urban areas through mapping based on Remote Sensing (RS) data and tools.

The classification proposed by Gonçalves et al. (2021), which defines the "Built-up Area" class not only as impervious surfaces but also as other elements indicative of human activity such as backyards and roads with exposed soil, proves to be the most accurate approach when dealing with Amazonian cities.

Despite inconsistencies in the generated data, including significant errors such as identifying built-up areas in flooded regions, the procedures proposed for identifying urban areas—such as selecting cloud-free images, collecting pixel samples for each class, manipulating bands to generate biophysical indices, and running tests with Random Forest models—are essential and sound steps toward the intended analysis. To further improve results, texture metrics need to be incorporated into the analysis, and ultimately, conducting an accuracy assessment of the results obtained is imperative.

References

ALMEIDA, C. M.; PEREIRA, M. N.; GONÇALVES, C. D. A. B.; SOUZA, Í. M.; MOREIRA, J. C.;

MIGLIORANZA, E.; NOVAES JUNIOR, R. A.; FLORENZANO, T. G. O sensoriamento remoto como

instrumental para o mapeamento da urbanização dispersa. In: SEMINÁRIO INTERNACIONAL

SOBRE URBANIZAÇÃO DISPERSA E MUDANÇAS NO TECIDO URBANO, 2007. Anais... 2007

BECKER, Bertha K.; MIRANDA, Mariana; MACHADO, Lia O. Fronteira Amazônica: questões

sobre a gestão do território. Brasília: Ed. UnB; Rio de Janeiro: Ed. da UFRJ, 1990.

BECKER, B. K. Undoing myths: the Amazon-an urbanized forest. Man and the biosphere series,

v. 15, 1995.

BECKER, Bertha Koiffmann; STENNER, Claudio. Um Futuro para a Amazônia. 1. ed. São Paulo:

Oficina de Textos, 2008. 152 p. ISBN 978-85-8623-877-2.

BECKER, B. K. Cidades Amazônicas : surtos econômicos e perspectivas. Espaço Aberto, PPGG -

UFRJ, v. 3, n. 1, 2013.

CARDOSO, A. C. D.; LIMA, J. J. F. Tipologias e padrões de ocupação urbana na Amazônia

Oriental: para que e para quem? In: CARDOSO, A. C. D. (Org.). O rural e o urbano na Amazônia:

diferentes olhares em perspectivas. Belém: UFPA, 2006.

COSTA, T. D. C. S. A relação cidade e rio na Amazônia: mudanças e permanências frente ao

processo de urbanização recente, o exemplo de Santarém (PA). 2012. 163p. Dissertação

(Mestrado em Planejamento do Desenvolvimento) - Universidade Federal do Pará, Belém,

2012. Disponível em:

<http://repositorio.ufpa.br/jspui/bitstream/2011/11030/1/Dissertacao_RelacaoCi

dadeRio.pdf>.

CORREA, R. L. A periodização da rede urbana brasileira. Revista Brasileira de Geografia Física, v.

49, n. 3, p. 39–68, 1987.DAL’ASTA, A. P. et al. Identifying Spatial Units of Human Occupation in the Brazilian Amazon

Using Landsat and CBERS Multi-Resolution Imagery. Remote Sensing, v. 4, n. 1, p. 68–87, 4 jan.

FUNDAÇÃO AMAZÔNIA DE AMPARO A ESTUDOS E PESQUISAS DO PARÁ. Estimativas e

projeções do PIB Paraense 2015-2020. Belém, 2017. Disponível em:

http://www.fapespa.pa.gov.br/upload/Arquivo/anexo/1332.pdf?id=1546554944.

GONÇALVES, G. C.; OLIVEIRA, L. M.; DAL’ASTA, A. P.; AMARAL, S. Geoinformação para a

visibilidade das áreas urbanas de cidades amazônicas geoinformatics for the visibility of urban

areas in the amazon cities. Revista Geoaraguaia, p. 1–15, 2021.

INSTITUTO BRASILEIRO DE GEOGRAFIA E ESTATÍSTICA - IBGE. Censo demográfico 2010.

Disponível em: http://censo2010.ibge.gov.br/. Acesso em: 12 jul. 2023

KOHLHEPP, Gerd. Conflitos de interesse no ordenamento territorial da Amazônia brasileira.

Estudos Avançados, São Paulo, v. 16, n. 45, p. 61, 2002.

MONTE-MÓR, R. L. Modernities in the Jungle: extended urbanization in the brazilian Amazonia.

[S.l.]: Universidade da California, 2004.

NUNES, D. A.; JÚNIOR, S.-C. C. DA T.; TRINDADE, G. O. DA. Cidades médias na Amazônia

brasileira: da centralidade econômica à centralidade política de Marabá e Santarém (Estado do

Pará). Confins. Revue franco-brésilienne de géographie/Revista franco-brasilera de geografia,

n. 29, 2016.

PEREIRA, J. C. M. Importância e significado das cidades médias na Amazônia: uma abordagem a

partir de Santarém (PA). Belém: [s.n.].

PORTO-GONÇALVES, C. W. Amazônia, amazônias. São Paulo: Contexto, 2001.

RICHARDS, John A.; JIA, Xiuping. Remote Sensing Digital Image Analysis: An Introduction.

Berlin: Springer-Verlag, 2006. ISBN 978-3-540-25128-6.

SAKATAUSKAS, Giselle de Lourdes Bangoim. Especificidades da Precariedade Habitacional na

Amazônia Ribeirinha: um olhar sobre a região do Baixo Tocantins. 2020. 281 p. Tese

(Doutorado em Planejamento e Gestão do Território) - Programa de Pós-Graduação em

Planejamento e Gestão do Território, Universidade Federal do ABC, Santo André, SP, 2020.

Orientadora: Profa. Dra. Flávia da Fonseca Feitosa; Coorientadora: Profa. Dra. Rosana Denaldi.TRINDADE JÚNIOR, Saint-Clair Cordeiro da. A questao urbana, o estatuto da cidade e o

desafio da sustentabilidade na Amazonia. In: Politicas publicas e gestao local. 1. ed. Belém:

FASE, 2004. p. 55–66.

TRINDADE JÚNIOR, Saint-Clair Cordeiro da. Cidades médias na Amazônia Oriental: das novas

centralidades à fragmentação do território. Revista Brasileira de Estudos Urbanos e Regionais,

v. 13, n. 2, p. 135, 2011.

TRINDADE JÚNIOR, Saint-Clair Cordeiro da. Pensando a Modernização do Território e a

Urbanização Difusa na Amazônia. Mercator, Fortaleza, v. 14, n. 4, p. 93-106, dez. 2015. ISSN

1984-2201.