Local Climate Zones Classification Applied to a Brazilian Amazon City

Abstract

Urban elements influence atmospheric flow, turbulence, and the local microclimate, modifying the transport and composition of pollutants. However, although this focus on cities is crucial for managing climate change, our knowledge of most cities in the world is still quite limited. Thus, the classification of Local Climate Zones (LCZs) aims to increase the accuracy of urban studies and has already been applied in various regions of the world, including, more recently, in Brazil. This article aimed to apply the LCZ to Santarém, a city in the Brazilian Amazon. The methodological procedures included the digital mapping protocol of the World Urban Database and Access Portal Tools (WUDAPT-Level 0) and the supervised classification tool of the LCZ Generator application, resulting in 78 polygons representing 10.02% of the training area and 13.94% of the study area (urban zone). The research identified 7 of the 17 main LCZ classes in Santarém. The use of the NDVI was essential for assessing the vegetation in each class, highlighting variations in green areas and emphasizing that vegetation is reduced in built environments. This approach enhances the understanding of urban morphology and enables future research into urbanization and the climate in the Amazon.

1. Introduction

The rapid urban growth occurring on a global scale has led to a concerning density of inhabitants, resources, and capital in metropolitan areas. This process poses a substantial threat to both the ecological balance of cities and the physical and mental health of urban populations [1,2,3]. As urbanization intensifies, there is an accelerated expansion of built-up areas, resulting in substantial transformations of the urban landscape [4,5].

Urbanization and climate change are interconnected trends that significantly influence local climate patterns and urban planning [6]. Urbanization, while acting as a driver of transformation by promoting economic progress and social mobility through improved access to essential services such as education, health, and employment [6], also directly contributes to rising temperatures in urban areas. This occurs due to the increase in paved surfaces and the decrease in green areas, leading to intensified greenhouse gas emissions and the formation of Urban Heat Islands (UHIs) [7]. UHIs are traditionally defined as the temperature variations (ΔT) between the urban perimeter and the rural zone, and they are intensified by the arrangement of buildings, the materials used in construction, and population density as emissions from anthropogenic pollutants that affect the radiation transfer, the cloud formation, and the local and surrounding precipitations. Furthermore, the UHIs are intensified by the reduction in natural surfaces such as parks and green spaces [8,9].

In addition, climate change can exacerbate heatwaves, droughts, flooding, and the deterioration of air quality, putting the most vulnerable groups, such as elderly people and children, at risk [6]. Understanding these dynamics between urbanization and climate change allows planners to develop effective mitigation strategies, such as the implementation of green spaces and the selection of urban materials that help to reduce the intensity of UHIs, promoting a more sustainable and comfortable environment in cities [7]. Taking this into consideration, urban planners are becoming increasingly aware of the need to strengthen the resilience of cities, as the effective mitigation of and adaptation to climate change will largely depend on the actions taken in these spaces [10].

However, despite this focus on cities being crucial for managing climate change, current knowledge about most cities in the world is still quite limited [11,12]. Furthermore, understanding the evolution of urban systems over time and space is essential for promoting sustainable urban development [13]. In contrast to urban areas, rural landscapes are often viewed as being less densely populated, with fewer built structures and more natural space available for agricultural activities [14]. The use of the terms “urban” versus “rural” took its first steps in studies on heat islands without a clear definition of these words. This has led to methodological challenges that affect the accuracy of results, such as the imprecise classification of research sites and the lack of scientific rigor in analyses and observations [15]. To address these inconsistencies, ref. [14] proposed a classification system based on the analysis of urban geometry and land cover, called Local Climate Zone (LCZ). This system has become an essential tool for characterization and planning in different climatic contexts, eliminating contradictions in the definitions of urban, suburban, and rural areas [5]. The LCZ creates a link between urbanization and its thermal impacts on the landscape while standardizing the documentation and global sharing of data on urban temperature [14]. Each area is classified into a specific local climate zone if the influencing factors, such as land cover, building geometry, and human activities, are homogeneous [14,16].

Additionally, the Normalized Difference Vegetation Index (NDVI) can be used to establish the interrelations with urban morphology for each LCZ class [17]. The NDVI also plays a crucial role in analyzing the health and density of vegetation in agricultural areas [18]. By correlating these variables with LCZs, it is possible to identify patterns of the manipulation or recovery of urban vegetation. This enables a more robust analysis of the interactions between habitat, urban climate, and sustainable development, contributing to planning and mitigating environmental impacts in urban and peri-urban areas [17]. Considering research that relates urbanization and the urban microclimate, ref. [19] emphasizes the relevance of more localized studies on UHIs and climate classifications, as an increase in urbanization, combined with a reduction in green areas and the growth of built surfaces, intensifies UHIs. However, these effects vary between regions, depending on factors such as climate, geography, and urban development [19]. Moreover, regarding the innovative use of LCZs in Brazil, it is important to highlight that the concept, while widely used globally, is still in the process of adaptation and implementation in various Brazilian cities. Due to its vast diversity of climates and urban characteristics, Brazil has a great potential to contribute to refining the LCZ concept. Recent projects, such as the mapping of LCZs in cities like São Paulo, Brasília, and Fortaleza, have demonstrated the effectiveness of this methodology for understanding microclimatic variations in the Brazilian context, especially in monitoring phenomena such as urban heat islands [20,21,22]. This reinforces the need for more national and local investigations that consider the peculiarities of the Amazon biome and its interactions with the urban environment, such as in Santarém, Brazil, where the impact of urbanization on the climate has not yet been widely explored.

Thus, studies on urban climate and LCZ classification in Santarém are essential for developing adaptation and mitigation strategies in response to climate change in the Amazon. In light of this, the present article aims to classify the urban zone of the municipality of Santarém in the Brazilian Amazon into Local Climate Zones, making this study pioneering. This approach allows for in-depth studies on urban thermal issues, providing essential theoretical support for urban planning and aiding in the development of more effective strategies.

2. Materials and Methods

2.1. Characterization of the Study Area

The study area is located in the municipality of Santarém, in the Lower Amazon region, in the state of Pará (Figure 1a).The municipality covers a territory of 22,887.08 km², of which 97 km²corresponds to the urban area (Figure 1b). Geographically, the city is situated at an altitude of 43 m, at coordinates of 2°26′22″ S and 54°41′55″ W, approximately 700 km from the capital, Belém. The region is influenced by the Amazon and Tapajós rivers, which play central roles in the local economy and way of life, as well as contributing to the formation of floodplains and alluvial plains typical of the Amazon landscape [23,24].

The climate of Santarém is characterized as hot and humid tropical, with an average annual temperature between 25 °C and 28 °C and an average relative humidity of 86%. Its annual precipitation is high, with an average of 1920 mm, and the region’s hydrological regime follows a cycle of flooding (December to April), high water (May and June), low water (July to September), and a dry season (October and November) [25]. According to the Köppen–Geiger classification, the climate is of type Am (Figure 1c).

According to the Brazilian Institute of Geography and Statistics (IBGE) [26], the population of Santarém is 331, 942 inhabitants, with a population density of 18.55 inhabitants per km², making it the most populous municipality in the Lower Amazon region. The increase in urbanization, combined with a lack of adequate public policies, has led to the emergence of irregular settlements. The city has expanded from the initial neighborhoods of Aldeia and Prainha to 48 neighborhoods, as shown in Figure 2 [27,28].

Santarém has undergone a significant process of urban expansion, registering a 55.80% increase in its urban area between 1984 and 2020 [23]. Economic, social, and physical transformations were largely driven by the construction of highways BR-163, PA-370, and Fernando Guilhon in the 1960s and 1970s. This infrastructure facilitated urbanization and the creation of new neighborhoods further away from the main rivers, contributing to population growth and the diversification of economic activities, which include agriculture, fishing, commerce, and tourism [24].

2.2. The LCZ Classification System

LCZ is a classification system based on the physical and thermal characteristics of urban surfaces, and it is widely used in climate studies [14]. It is described as “regions with uniform surface cover in terms of structure, material, and human activity, ranging from hundreds of meters to several kilometers in horizontal scale” [29]. This system excels in evaluating the effect of urban heat islands (UHIs), taking into account factors such as building height and spacing, impermeable surfaces, vegetation density, and soil moisture [14]. The classification encompasses 17 categories, divided into 10 “built types” and 7 “land cover types”, providing a detailed analysis of the spatial patterns in different urban thermal environments [14] (Figure 3, p. 1884).

Since its creation in 2012, the LCZ system has become the new standard for characterizing urban landscapes, considering the physical properties of the soil and its cover. In 2015, the WUDAPT project developed a free protocol for mapping cities into LCZs, utilizing the LCZ Generator, a web-based application that employs a classified algorithm to perform supervised classifications automatically, pixel by pixel. Using Google Earth Pro for training areas (TAs), visual and numerical information is collected, allowing for the removal of data from Landsat 8 satellite images. Using the Random Forest method, the program categorizes the area of interest based on the types of Local Climate Zones. This mapping process enables a detailed climatic analysis of cities [30,31].

The LCZ mapping follows the following three main steps: data collection from sample areas, defining zones for thermal measurements, and selecting the LCZ that best corresponds to the observations. This system simplifies the understanding of the characteristics of urban and natural surfaces, making them accessible to both specialists and the general public [21].

2.3. Method for LCZ Classification

The procedure for creating the LCZ classification map for the municipality of Santarém was conducted according to the method described by [30], following the instruction guide of the LCZ Generator web application (https://lcz-generator.rub.de/. Accessed on 4 August 2024). First, the training area was defined in the Google Earth Pro digital environment, based on an analysis of typologies and morphological parameters, as described by [21]. The sky view factor was not explored in this work. Google Street View was utilized instead. Subsequently, the training areas were digitized in Google Earth Pro following the WUDAPT protocol, as shown in Figure 4.

After the supervised classification in Google Earth Pro, the next step involved accessing the submission tab in the LCZ Generator (web application) and entering the necessary personal and training information. After successful submission, the LCZ classification and quality control processes were initiated on the back end to generate an LCZ map, complete with quality control measures, metadata statistics, and labels for suspicious polygons. In the final step, the compressed results were emailed and simultaneously updated in the online submission table.

The submission of input data was performed through a form on the LCZ Generator webpage (https://lcz-generator.rub.de/, Accessed on 23 April 2021), as previously described. The form includes personal information from the author, such as their name and email, as well as details about the area to be classified, including the training area file, city/country/continent, reference date, and any observations the author wishes to include about the work. At this stage, a series of checks created by the developers are conducted. First, the file containing the training areas must have the extension “.kml” (Keyhole Markup Language) or its compressed version with the extension “.kmz” (Figure 5). Another check verifies whether the training area file complies with the WUDAPT standard for proper class labeling. During this stage, each polygon receives an identifier, which is necessary for conducting the automated accuracy test.

Accuracy is evaluated using the following metrics: overall accuracy (OA), overall accuracy only for urban LCZ classes (OAu), overall accuracy only for built vs. natural LCZ classes (OAbu), weighted accuracy (OAw), and the F1 class metric. Overall accuracy indicates the percentage of correctly classified pixels. The OAu reflects the percentage of correctly classified pixels solely from urban LCZ classes, while OAbu is the overall accuracy for built vs. natural LCZ classes, ignoring internal differentiation. Weighted accuracy (OAw) is obtained by applying weights to the confusion matrix, taking into account the (dis)similarity between LCZ types [31].

For the final stage, the free software QGIS (version 3.36.3) was used to perform spatial operations, image processing, and statistical calculations, using the products obtained from the LCZ Generator classification.

The vegetation cover of each LCZ was analyzed using the Normalized Difference Vegetation Index (NDVI), obtained through the open-source platform ClimateEngine (https://app.climateengine.com/climateEngine. Accessed on 19 September 2024), utilizing data from the Sentinel-2 SR sensor, which offers a spatial resolution of 10 m [18]. The data were extracted for each polygon representative of the previously identified LCZs (Figures 10–14) in the urban area of Santarém, for the same classification period, representing the morphology in the year 2023. Subsequently, descriptive statistics were generated, allowing for a comparative analysis of the average NDVI for each LCZ.

3. Results and Discussion

3.1. Accuracy and Frequency in LCZ Classification in Santarém: Challenges and Limitations

The overall accuracy obtained for the classification on the LCZ Generator platform was 82% (OA), the accuracy for urban land classes (OAu) was 86%, and the overall accuracy for urban versus natural classes (OAbu) was 91%. The weighted accuracy (OAw) was 95%, as can be seen in (Figure 6).

The accuracy metrics used are based on the work of [12], where the overall accuracy for urban classes is 0.78%; the overall accuracy for built versus natural LCZ classes is 0.92%; and finally, the weighted accuracy obtained by applying weights to the confusion matrix, considering the dissimilarity among LCZ types, is 0.95% [33]. Ref. [34] notes that many authors set a general accuracy target of 85% for thematic mapping. However, the results available on the WUDAPT portal range from 60% to 90%.

Refs. [12,35] emphasize that a high overall accuracy does not guarantee that a map is correct. This can happen for several reasons. The accuracy assessment is conducted only for the pixels included in the training area (TA) set, meaning that pixels outside these areas are not subject to quality control. Additionally, the polygons in the TA may contain errors due to incorrect interpretations of the landscape by the user, such as confusion between similar classes (for example, LCZs 3 and 7, A and B, or 8, 10, and E) or the use of non-persistent land cover. Another factor to consider is that confusion among LCZ classes is related to the number of classes, as insufficient discrimination among LCZ types in the training sample can result in an artificially high overall accuracy. To facilitate classification and improve accuracy, one can use 3D structural urban data, such as Google Street View images and the sky view factor [36].

According to the aforementioned graph (Figure 6), it is observed that the urban classes LCZ 6 and LCZ 9 showed a lower accuracy than LCZ 3, which demonstrated the best performance, confirming expectations based on the training area analysis. During classification, similar characteristics were observed between the classes, making distinction difficult, so differentiation was only possible based on land cover, i.e., whether there were more, fewer, or no trees or any vegetation types. The difficulty in classifying some areas is also noted in the study by [36]; the authors exemplified that LCZ4 is defined by 20–40% building cover, 30–40% impervious surface cover, and 30–40% permeable surface cover. However, these properties cannot be directly extracted during human interpretation from high-resolution images. Furthermore, some urban types are difficult to distinguish without the aid of additional data sources, for example, the similarity between LCZ4 and LCZ5, which have similar surface fractions but differ in the height of roughness elements. LCZ4 has a height greater than 25 m, while LCZ5 varies between 3 and 10 m [36].

Another possible reason for the low accuracy and performance of some LCZ classes for the urban area of Santarém may be related to the existence of small training areas, which tend to provide limited information, increasing the likelihood that an automatic classifier will not correctly identify them.

A relatively low accuracy was also observed in the LCZ B class, which corresponds to sparse vegetation. This low performance may have been related to the lack of homogeneity and representativeness of this type of landscape in the study area. In Santarém, landscapes that fit this class are rare, mainly limited to the city park. This suggests a scarcity of green areas such as parks, indicating the need to increase tree planting and create spaces corresponding to this class. According to [15], the more (and larger) training areas for each type of LCZ, the better the classification. A greater number of training pixels provides a better representation of the variety of spectral characteristics associated with an LCZ type.

The frequency of the identified classes in Santarém can be observed in Figure 7. According to the criterion adopted regarding the number of representative areas, some LCZ classes were not identified due to their low representativeness in the studied area. These results align with [37], who mentioned that the morphological characteristics of a city have a significant influence on the quality of the generated classification, resulting in regions being better or worse classified, depending on their construction patterns. In this context, some classes represented by the LCZ did not fit well in Santarém and its metropolitan region due to the lack of homogeneous regions that would allow for a correct classification. The algorithm performed well in classifying urban regions and classes G (water) and A (dense trees), clearly delineating the boundaries between each class.

In light of this, it is important to note that Santarém is undergoing significant development and verticalization, exhibiting mixed characteristics of urban elements. According to [38], verticalization follows certain patterns, occurring in areas of the city with greater infrastructure, leading to a concentration of higher-income populations in specific parts of the city. This can be observed in Santarém, where this configuration of buildings is present in more central neighborhoods. According to [38], in their study of verticalization in Belém, the city’s expansion occurs through a process of uneven spatial production.

These observations can be confirmed by accessing the platform on the WUDAPT portal, where it can be verified that most classifications are applied in urban areas of Europe, while there are few studies in Latin America and Africa. This highlights that it is easier to perform classifications in countries where the level of development and type of urbanization are more uniform, facilitating the identification of patterns. In contrast, regions like Latin America and Asia experience disordered and fragmented urbanization, making it difficult to apply classification techniques. To adapt the WUDAPT method, the use of LCZ subclasses is suggested, considering heterogeneity and informal occupation [39].

3.2. LCZ Classification

Following the WUDAPT protocol, out of the 17 types of LCZ, the following 7 were identified in Santarém: LCZ 3, LCZ 6, LCZ 9, LCZ A, LCZ B, LCZ D, and LCZ G, comprising 3 types of urban classes and 4 types of natural/rural classes. The set of training areas produced for the mapping resulted in 78 polygons, covering a total area of 13.523 km². This value represents 13.94% of the study area (urban area, obtained from image editing using QGIS software) and 10.02% of the total area of polygon TAs. The spatial distribution of the LCZs over the Region of Interest (ROI) produced by the LCZ Generator can be seen in Figure 8.

Since the study area was delineated based on the urban macrozone of the municipality of Santarém, the mapping image was cropped, as shown in Figure 9. With this cropping, it was observed that LCZ 6 and LCZ 9 were the most frequent types of urban land use, while LCZ 3 was the least frequent. This highlights low-altitude land use as a predominant characteristic in Santarém. The natural area types LCZ A and LCZ D were the most common within the urban macrozone. However, LCZ A was located around this urban macrozone and may be associated with existing natural barriers in the municipality, such as river boundaries and preservation areas like the Tapajós National Forest (FLONA). On the other hand, LCZ D was primarily found in the floodplain area, thus functioning as a natural pasture. An overview of the distribution of the LCZs for the urban area in percentages can be seen in Figure 9.

According to the generated maps presented in Figure 8 (LCZ Generator) and Figure 9 (created after processing the image in QGIS software), there was an increase in the percentage of classes with the presence of buildings, meaning urban classes. LCZ 3 increased from 2.15% to 14.41%, LCZ 6 increased from 4.92% to 19.67%, and LCZ 9 saw an increase of 5.68%, rising from 9.27% to 14.95%. The results for the municipality of Santarém align with the findings by [33] in their study on the variability of the daytime surface temperature among Local Climate Zones (LCZs) in the urban area of Brasília. However, the author notes that a total of 16 out of the 17 primary classes were identified in their study area. The author identified the following distribution percentages of LCZs in the study area for the same building classes: LCZ 3 at 14.65%, LCZ 6 at 12.96%, and LCZ 9 at 25.13%. After reclassification, the distribution percentages of LCZs in the study area were the following: LCZ 3 at 12.75%, LCZ 6 at 11.08%, and LCZ 9 at 12.42%. It is worth noting that each city behaves differently, and Brasília, being a planned city, follows a distinct structure compared to most cities, especially when compared to an inland city in the Amazon. Therefore, it is important to consider these particularities when analyzing urban behavior.

On the other hand, the results of this study differ from those obtained by [40], where, in their research on the characterization of an urban area in the Legal Amazon, using Local Climate Zones (LCZs) in the municipality of Nova Mutum, located in the northern macro-region of the state of Mato Grosso with in the Amazon basin, they found values of 0.55% for LCZ 3, 0.91% for LCZ 5, 6.03% for LCZ 6, 15.97% for LCZ A, 0.50% for LCZ B, and 48.07% for LCZ D, with areas of herbaceous vegetation, predominantly representing the future expansion area of the urban perimeter, currently used for agriculture. Furthermore, densification points in LCZ 3 are already easily visible in the municipality, especially in central zones, resulting from the local colonization process, along with occasional vertical developments, which may adversely affect the local microclimate due to the formation of heat islands [40]. Thus, they concluded similarly to [41], who stated that the increase in classes due to urbanization and climate change is a natural process; this is compatible with Santarém, which has only recently begun the process of accelerated urbanization.

3.3. Relationship between LCZ and NDVI

The NDVI has been widely used to map vegetation distribution in urban and rural areas, aiding in the analysis of urban heat islands and their interactions with land use. In Santarém, the application of the NDVI demonstrated a clear difference between urban and natural areas, as reflected in the LCZ classification. Studies indicate that urbanized areas tend to have lower NDVI values, reducing vegetation density, which contributes to increased surface temperatures and the intensification of heat islands. This can be seen in the study by [42], who indicated that, in large urban centers, there is a temperature difference, with variations of about 0.1 °C between built urban areas in vegetated and non-vegetated environments.

Other studies investigating the NDVI in urban areas have found similar results, showing an inverse relationship between the index and the intensity of heat islands. For example, Ref. [43] examined the relationship between vegetation, surface temperature, and urban morphology in the São Paulo Metropolitan Region, describing that urbanized areas exhibited the lowest vegetation index values, while higher values were observed in vegetated areas with less human occupation. The presence of vegetation in cities plays a crucial role in mitigating these effects, as observed in the LCZ zones of Santarém, where areas with higher vegetation (LCZ A and B) had higher NDVI values, as can be seen in

Table 1. Normalized Difference Vegetation Index (NDVI) according to LCZ classes in the urban area of Santarém.

NDVI/LCZ Class	Mean	Standard Deviation	Minimum	Maximum
NDVI/LCZ 3	0.31	0.16	0.16	0.79
NDVI/LCZ 6	0.40	0.17	0.22	0.83
NDVI/LCZ 9	0.58	0.13	0.39	0.87
NDVI/LCZ A	0.78	0.11	0.54	0.91
NDVI/LCZ B	0.71	0.07	0.59	0.85
NDVI/LCZ D	0.59	0.07	0.43	0.78
NDVI/LCZ G	−0.42	0.25	−0.98	−0.10

and Figure 10, Figure 11, Figure 12, Figure 13 and Figure 14.

The data from this study also align with the findings of [44], where, in their research on the relationship between urban form and vegetation indices in a subtropical megacity, they identified 14 out of the 17 existing LCZ classes, consisting of 7 urban classes and 7 natural classes. Their results show a relationship between the average NDVI and the respective Local Climate Zones, as follows: LCZ 3: NDVI = 0.31, LCZ 6: NDVI = 0.49, LCZ 9: NDVI = 0.65, LCZ A: NDVI = 0.78, LCZ B: NDVI = 0.60, LCZ D: NDVI = 0.57, and LCZ G: NDVI = −12, which are similar to those found in the urban and natural areas of Santarém.

Similarly, ref. [45] characterized the city urbanization of Manaus using the WUDAPT portal, identifying six main local climate zones, consisting of five urban and one natural. This classification aligns with reality, considering that Manaus and Belém, as capitals, have simultaneously undergone an intense urbanization process. They also evaluated and correlated the morphology with the NDVI, and according to their findings, the NDVI values revealed that Manaus experienced a significant vegetation cover removal process (NDVI < 0.3) in the years 1987, 1998, 2007, and 2015, corroborating the description of the urbanization process occurring in the northern region of Brazil over the years.

Thus, the analysis of land cover and urban areas in Santarém reveals a diverse range of Local Climate Zones (LCZs) that characterize the spatial and environmental dynamics of the city, reflecting a complex urban landscape with varying degrees of built and natural elements. The results of this study reinforce the findings of [25]. The author describes the city layout, noting that it has slowly expanded parallel to the Tapajós River from the central area towards the Prainha and Aldeia neighborhoods, also densifying inward. This expansion is characterized by a transition from LCZ 3 to LCZ 9, moving from more densely built areas to less built areas.

The LCZ 3—Compact Low-Rise was the least common pattern in Santarém, and is represented by densely constructed residential and commercial areas, with buildings ranging from one to three stories and a predominance of impervious surfaces with minimal vegetation. It represented 14.41% of the study area, with an NDVI of 0.31. This class was mainly found in the northern part of the city, where the urban area begins and where the historic center is concentrated, as shown in Figure 10.

The LCZ 6—Low-rise Open was the most common urban type in Santarém, a conclusion also reached by [41] in Japan and [46] in Lisbon, Portugal, with the majority of the population residing in these areas. This class is characterized by mixed land use, with medium- and high-standard residential areas and simpler constructions. It represented 19.67% of the study area, with an NDVI of 0.40. It was found both near and far from the city center, in peri-urban areas such as the neighborhoods of Área Verde, Juntaí, Santo André, and Floresta. In these locations, isolated buildings and regular spacing were observed, along with more abundant vegetation (Figure 11).

The LCZ 9—Sparse Occupancy with Small and Medium Buildings was the second most common urban type in Santarém, characterized by a dispersed arrangement of low buildings of varying sizes. Ref. [47] describes this class as a peri-urban transition zone between urban and rural/natural areas, primarily residential or rural, with a permeable surface abundance. It represented 14.95% of the study area, with an NDVI of 0.58. In Santarém, this class was primarily observed in neighborhoods such as Urumanduba and Mararu, and also in the regions of Pérola do Maicá and Maicá (Figure 12).

These aspects are reinforced by the analysis of [48], which identified variations in the NDVI in Santarém between 1999 and 2010 (ranging from 0.54 to 0.87 in areas with dense vegetation, while urbanized areas showed lower values, between 0.14 and 0.53), mainly due to the expansion of the municipality with the creation of new neighborhoods, which resulted in a decrease in vegetation cover and, consequently, a reduction in NDVI values.

The LCZ 9, classified as sparsely built, represents a transition zone between urban and rural areas. This classification highlights the importance of considering the local microclimate. It is important to note that this class can undergo significant modifications and may eventually be classified as another LCZ, especially due to the likelihood of urban expansion. For example, one study indicates that Santarém has undergone a significant process of urban expansion, registering a 55.80% increase in its urban area between 1984 and 2020 [23]. This demonstrates the potential changes and shifts in classifications that may occur as urban areas continue to grow.

The LCZ A—Dense Tree Cover is defined by a dense arrangement of trees that creates a significant obstruction of the sky at eye level. This class may be characteristic of forest remnants, protected areas, urban forests, tree cultivation areas, green belts, and urban parks. It represented 18.62% of the study area, with an NDVI of 0.78. As mentioned, it was found in the surroundings of the urban and peri-urban areas of Santarém (Figure 13a).

The LCZ B—Sparse Tree Cover is defined by a dispersed arrangement of trees. This class may also be characteristic of forest remnants, protected areas, urban forests, tree cultivation areas, green belts, and urban parks. It represented 5.77% of the study area, with an NDVI of 0.71. In Santarém, this natural class was the least common, appearing rarely in the urban area and represented in this study only by the city park (Figure 13b).

Santarém is developing its first urban forestry plan. This plan aims to guide urban planning by establishing procedures for vegetation implementation and maintenance in public spaces and streets of the city [49]. Currently, only 25% of Santarém’s urban area is covered by trees, and this percentage is in constant decline. Creating a positive agenda for urban forestry is essential to reduce and mitigate related issues, as tree masses play a crucial role in thermal and environmental comfort in cities, especially in tropical and equatorial regions. In the absence of vegetation, there is an increase in direct solar radiation, air temperature, alteration sin rainfall cycles, decreases in humidity, and modifications of wind speed and direction, causing thermal discomfort to residents of urban areas, as also addressed by [50]. In Santarém, the climatic situation with rising temperatures is concerning. Ref. [51] studied the traditional knowledge of the Borari people of Alter do Chão related to weather and climate issues, demonstrating an increase in temperature of 0.43 °C or 1.66% per decade in the region.

From a geodescriptive perspective, the floodplain is considered to be a floodplain formed by a variable-width strip along the Amazon River, which can reach up to 24 km in Santarém [52]. Topographically, the floodplain can be divided into low and high areas. In the former, the land is periodically flooded during the Amazon River’s flood cycle, featuring vegetation that alternates between pasture (natural field) and forest. The latter is a higher and relatively stable area of the floodplain, flooded during major floods and covered by tree vegetation. In this study, LCZ D is associated with natural pastures used for cattle grazing during the “low-water” period.

Finally, the LCZ G—Water class was prominently represented along the city’s front, including the Tapajós and Amazon Rivers and the lakes surrounding Santarém, as well as small reservoirs in the peripheral area, representing 17.60% of the study area, with an NDVI of −0.42 (Figure 14b).

The Tapajós River, technically known as the River of Clear Waters, has waters that vary from dark to bluish or greenish, depending on the day’s conditions. This coloration is caused by the high concentration of organic acids resulting from the decomposition of vegetation along its banks. In contrast, the Amazon River, or River of Muddy Waters, is characterized by its muddy waters, resulting from the large amount of sediment transported from its source in the Andes Mountains in Peru. The differences in composition, acidity levels, flow temperature, velocity, and density between the two rivers prevent their waters from mixing, creating a clearly visible contrast, as illustrated in (Figure 14b).

4. Conclusions

The classification method of Local Climate Zones (LCZs) was used for the first time to characterize the city morphology of Santarém (Pará) in the Brazilian Amazon. This method made it possible to identify seven of the seventeen main LCZ classes and relate them to the Normalized Difference Vegetation Index (NDVI). The classes were (i) LCZ 3, characterized by a dense mix of low buildings (1–3 stories) with few or no trees, a land cover that is predominantly paved, with construction materials including stone, brick, tile, and concrete, representing 14.41% of the study area, with an NDVI of 0.31; (ii) LCZ 6, characterized by an open arrangement of low buildings (1–3 stories) with abundant permeable vegetation cover (low plants and sparse trees) and construction materials of wood, brick, tiles, and concrete, representing 19.67% of the study area, with an NDVI of 0.40; (iii) LCZ 9, characterized by a sparse arrangement of small or medium buildings in a natural environment, with abundant permeable vegetation cover (low plants and sparse trees), representing 14.95% of the study area, with an NDVI of 0.58; (iv) LCZ A, characterized by a dense cover of trees, representing 18.62% of the study area, with an NDVI of 0.78; (v) LCZ B, characterized by a sparse cover of trees, representing 5.77% of the study area, with an NDVI of 0.71; (vi) LCZ D, characterized by the presence of ground-level vegetation, occupying 8.98% of the study area, with an NDVI of 0.59; and (vii) LCZ G (water), representing 17.60% of the study area, with an NDVI of −0.42.

The adoption of this classification system, still relatively new in the Amazon region, has proven valuable for representing the urban fabric on a local scale, being crucial for future studies on the expansion of riverside cities and issues related to climate and climate change. The method applied in Santarém could be expanded to conurbated areas, considering the rapid urban growth occurring and the integration of the urban fabrics of neighboring cities, and could also be enhanced by using the sky view factor.

This study provides insights for both urban planning professionals and public policy makers, highlighting the importance of improving green spaces and managing urban expansion efficiently. Future research should focus on more detailed assessments of microclimates within the Local Climate Zones (LCZs) and how these microclimates affect local environmental quality and urban heat islands. Integrating these findings into urban planning can promote more sustainable and resilient development in Santarém. Despite the challenge of representing urban form for application in urban climate studies, especially in cities of low- and middle-income countries like Brazil—where most lack planning and the urban fabric is heterogeneous, with various types of buildings mixed in the same area—this study confirms the relevance of the LCZ system for classifying urban landscapes in the Amazon region and suggests its potential for application throughout the area. Thus, the use of the LCZ system not only contributes to a better understanding of urban morphology, but also establishes a solid foundation for studies related to urbanization and climate, facilitating sustainable urban management and planning.