Loss of Natural Habitats: High-Altitude Vegetation, Pantanal Perda de Habitats Naturais: Campos de Altitude, Pantanal Perdida de Hábitats Naturales: Campos de Altitud, Pantanal

The objective of this study is to quantify the spatiotemporal variation of high-altitude grassland (Campos de Altitude) in two different study areas between 1985 and 2020. First, the types of vegetation present in both areas were classified using a false-color (Red, Green, Blue) composition, with the channels replaced by mid-infrared, near-infrared, red, respectively. Following the definition of vegetation classes and the creation of polygons over them, they were superimposed on the NDVI product (1985 and 2020) to obtain the maximum and minimum values of the index and to identify which value ranges each class would be in. Next, a multitemporal false-color composition was carried out with the modification of the Red channel by the NDVI 1985 product, the Green channel by the NDVI 2020 product and the Blue channel without any changes. The results show that the total area of Campos de Altitude decreased between 1985 and 2020 in both study areas, with area 1 losing 31.93% and area 2 losing 35.12%, whereas the vegetation of denser strata (arboreal) increased in its spatial distribution in both study areas, with area 2 gaining the most (230%). There was a noticeable increase in arboreal vegetation in the two study areas during the years analyzed. The tendency for the fraction of arboreal vegetation cover to increase is in line with the decrease in Campos de Altitude vegetation, and this phenomenon may be related to environmental factors such as temperature variations, amount of rainfall, solar radiation, and soil composition.


Introduction
Native vegetation plays a key role in terrestrial ecosystems as it affects land cover, provides ecosystem services such as water availability and quality, stores and sequestrates carbon, contributes to climate regulation, controls erosion and sedimentation, and helps to maintain biodiversity (Newbold 2018, Jeong et al. 2011. However, the native vegetation of all ecosystems has changed dramatically, owing primarily to changes in land use and climate change. Alterations in land use, such as the conversion of natural areas to provide food, fiber, water, and shelter for the population, are the primary drivers of native vegetation loss (Foley et al. 2005). Climate change can also affect the pattern of distribution and growth of vegetation through precipitation, temperature, and radiation, which can alter available energy and water, both of which are essential for plant growth, impacting the process of carbon accumulation in the water cycle as well as the decomposition and conversion of organic carbon (Linscheid et al. 2020;Tai et al. 2020;Pan et al. 2018).
Numerous studies in various regions, such as the Tibetan Plateau, have shown that vegetation changes exhibit spatial and temporal heterogeneity, and the effects of climate change on vegetation reflect its spatial and temporal variation (Tai et al. 2020;Zhang et al. 2019). Exploring the spatial and temporal pattern of vegetation and discussing the role of climate factors has been one of the main topics of current global change research, with important theoretical and practical implications for assessing the quality of terrestrial ecosystems and regulating ecological vegetation processes (Tai et al. 2020; Wamelink et al. Research, Society and Development, v. 11, n. 3, e2911326242, 2022(CC BY 4.0) | ISSN 2525-3409 | DOI: http://dx.doi.org/10.33448/rsd-v11i3.26242 2018Mu et al. 2020).
The change in the spatial and temporal distribution of vegetation is also an important indicator of the dynamic responses of ecosystems to climate change (Zhang et al. 2019;Jeong et al. 2011). As a result, research on the spatiotemporal distribution of vegetation is relevant not only for improving the understanding of vegetation responses to climate change, but also for increasing the accuracy of predictions (Pouliot et al. 2009;Zhang et al. 2019).
In this context, mapping the distribution and composition of vegetation is a fundamental practice for assessing ecosystem dynamics and their response to climate change on a large scale. Traditional field surveys are frequently limited because they do not provide spatial and temporal coverage data due to significant water dynamics, complex types of coverage, and vast dimensions of the system (Han et al. 2015;Mu et al. 2020). Thus, several methods for obtaining the spatiotemporal distribution of vegetation using remote sensing images have been developed, as the first works used visual interpretation to discriminate and map types of coverage using compositions of true or false colors (Johnston & Barson 1993).
The development of rule-based models, such as machine learning algorithms, has vastly improved classification accuracy over time (Mu et al. 2020;Hu et al. 2015;Chen et al. 2018). Supervised classification has been widely used in the differentiation of various plant phenology (Berberoglu et al. 2004;Han et al. 2015;Dronova et al. 2012). Another technique that has been employed in the spatiotemporal analysis of vegetation is the multi-temporal classification, the same used in the work by Peres et al. (2016) who, through false-color composition techniques of the RGB spectral bands of NDVI (Normalized Difference Vegetation Index) data, were able to generate information on areas that suffered phytomass reduction, which tend to red, as well as those that had an increase, which appear in green, and areas without changes in vegetation cover, which are visualized in yellow.
The Pantanal is the world's largest tropical floodplain; however, there are some isolated massifs in the middle of the Pantaneira plain, such as Urucum, with an elevation of 1,160 meters near the city of Corumbá, and Serra do Amolar. The biome is the most preserved in Brazil, with more than 85% of native vegetation (SOS PANTANAL et al. 2017), but it is also dependent on activities carried out in the Upper Paraguay Basin Plateau, which is undergoing rapid conversion of land use (Roque et al. 2016;Guerra et al. 2020). The region also experiences severe droughts as a result of climate change, which affects the biome's water regime (Marengo et al. 2021).
Serra do Amolar and Morraria do Urucum have a typical high-altitude vegetation cover known as Campos de Altitude (high-altitude grassland), which plays an important role in the regional ecological environment. The elevation gradient in Serra do Amolar results in distinct vertical distributions of climate and vegetation. The goal of this study is to quantify the spatiotemporal variation of the vegetation present in these two areas of analysis between 1985 and 2020, thus providing evidence of vegetation variation that may be linked to environmental or anthropogenic factors. Based on such analysis, it is possible to raise questions about what caused this vegetation to undergo modification over the years.

Study area
The study area comprises two sample areas, namely Serra do Amolar and Maciço do Urucum, both inserted in the Pantanal sul-mato-grossense, in the municipality of Corumbá-MS ( Figure 1).  Serra do Amolar is located in the Pantanal sub-region named Paraguay (Silva & Abdon 1998;Mioto et al. 2012), because of the Paraguay River that borders the Serra, having peculiar characteristics of the region as it is a border area with the Bolivian Chaco and because of its presence. of large permanently flooded bays (Larcher et al. 2017;Junk et al. 2006), constituting a composition between the Pantanal plain and the Serra do Amolar, representing one of the greatest biological diversity assets in Brazil (IHP 2012). It is located 180 km from the municipality of Corumbá and has an extension of approximately 40 km along the border between Brazil and Bolivia (Larcher et al. 2017), with Pre-Cambrian geological formations, which exercise geological control over the flow of water of the Pantanal (Souza & Souza 2010). Serra do Amolar has the highest point at around 1,000 meters of altitude, and in this area the vegetation is mainly formed by lowland ecosystems up to Campos de Altitude which, together, form a potentially important biological and geographic corridor (ECOTROPICA 2003, Larcher et al. 2017. The Urucum massif is located near the urban area of Corumbá, formed by residual hills located on the southwestern edge of the Pantanal, having the highest points in the state of Mato Grosso do Sul, with altitudes above 1,000 meters (Urbanetz et al. 2012). The hills are surrounded by the Pantanal floodplain, a contact region of different phytogeographic provinces, such as the Amazon, Cerrado, Chaco and Southern Forests, in addition to being of special interest for conservation, with endemic species such as Aspilia grazielae in the fields of altitude and Gomphrena centrota in the lateritic benches (Pott et al. 2000(Pott et al. 2011Urbanetz et al. 2012).
It is worth noting that in Maciço do Urucum there is mining activity, as it has a large reserve of iron ore and manganese. These activities have been modifying the landscape of the morraria, putting the plant community and the stability Research, Society and Development, v. 11, n. 3, e2911326242, 2022 (CC BY 4.0) | ISSN 2525-3409 | DOI: http://dx.doi.org/10.33448/rsd-v11i3.26242 5 of the land at risk.

Reporting period, Landsat 8 and Landsat 5 images
Landsat 8 (LANDSAT 2020) and Landsat 5 (LANDSAT 1985) images were used in this study, extension .TIF, both projected in the UTM SIRGAS 2000 Geographical Projection System, which was obtained for free from the website https://espa.cr.usgs.gov/index/. (US Department of the Interior US Geological Survey), which provides all spectral bands as well as pre-calculated vegetation indices. Both images received radiometric correction and treatments for better use and standardization of images, since we are working with different sensors (OLI and TM) which results in small differences in the spectral information of the images. Only the NDVI vegetation index was obtained for this study, along with the other products.
Here, the years of analysis are 1985 to 2020, and the dates of the selected images were September 12, 1985 for Morraria do Urucum, August 27, 1985 for Serra do Amolar, and September 28, 2020 for both areas. The image dates correspond to Pantanal's dry season, when there is a greater contrast between phytophysiognomies and a smaller water surface (PERES et al. 2016).

Classification of vegetation types
To identify the vegetation types in the study area, a false-color RGB (Red, Green, Blue) composition was created in QGIS 3.10 (QGIS Development Team 2021), with the Red channels replaced by MIR (mid-infrared), the Green channels replaced by NIR (near-infrared), and the Blue channels replaced by R (red) (Figure 2). Table 1 shows the bands used and channels assigned to each satellite.  . In Morraria do Urucum, Scene A is from 2020 (28/09/2020), and Scene B from 1985Scene B from (27/08/1985. In Serra do Amolar, Scene C is from 2020, and Scene D from 1985. The images were all taken during Pantanal's dry season. The difference in water volume between images C and D is the result of a severe drought which occurred in Pantanal in 2020. Source: Images from Landsat 8 (U.S. Department of the Interior U.S. Geological Survey, 2020). This same composition was used in a study conducted in Pantanal by Peres et al. (2016), who differentiated the arboreal/shrubby and undergrowth classes using the same modified RGB composition mentioned in the preceding paragraph.
Thus, we decided to apply it in the present study to carry out the analyses, as we intended to differentiate the same classes.
Following the composition procedure, it was possible to identify the areas with arboreal vegetation, grassy vegetation (Campos de Altitude, high-altitude grassland), and exposed soil for both study areas ( Figure 3). Research, Society and Development, v. 11, n. 3, e2911326242, 2022 (CC BY 4.0) | ISSN 2525-3409 | DOI: http://dx.doi.org/10.33448/rsd-v11i3.26242 After defining the vegetation classes, cuts were made from sample areas (Figure 4) of grassy vegetation classes (Campos de Altitude), arboreal, and exposed soil using the polygon tool in QGIS 3.10, to be superimposed on the NDVI product (1985 and 2020). The goal was to determine the index's maximum and minimum values, as well as which value ranges each class would fall into. Although NDVI values can range from -1 to 1, in finished products downloaded from Landsat 5 and 8, the values are distributed with a multiplier factor of 20,000 for Landsat 5 and 10,000 for Landsat 8. Therefore, such values were used in the current study to facilitate the visualization and discussion of results.  Research, Society and Development, v. 11, n. 3, e2911326242, 2022 (CC BY 4  The presence of exposed soil due to mining activities stands out in Morraria do Urucum. The company Urucum Mineração, which was later acquired by CVRD, has been conducting mining activities in Morraria since the mid-1970s (Brito 2011). As a result, the exposed soil class was also measured, and the total area of the class measured in the 1985 NDVI (areas where mining activity had not yet occurred) was subtracted from the 2020 NDVI (areas with exposed soil but no post-mining vegetation), so that the class was not overestimated in those locations. Following the determination of maximum and minimum NDVI values for the analyzed classes, the NDVI product was reclassified on QGIS 3.10 using the r.recode tool for the defined intervals, resulting in a product of clustered pixels. The data was then converted from raster to polygon and the of each class areas (in hectares) were calculated using the field calculator. This allowed for the quantification of pixels in terms of area.
Because this process was carried out for the two NDVI products from both years of analysis (1985 and 2020), the difference in coverage of the classes (in area) could be calculated to verify variations between these years.
Morraria do Urucum also has pasture areas. These areas were manually classified by polygons as the NDVI values are similar to the NDVI values of Campos de Altitude, because it is of grass type and the NDVI ranges are similar or close. In this case, the pasture was classified as Campos de Altitude, and it was subtracted from the total Campos de Altitude area after being manually selected. It is worth noting that this area was classified due to the presence of Campos de Altitude in this location in 1985 ( Figure 5).

Delimitation of the Morraria do Urucum and Serra do Amolar areas
To avoid errors in classifying the classes and overestimating the pixels of NDVI maps (1985-2020) that were not within the area of interest, particularly the Campos de Altitude class, which can be confused with other NDVI pixels, the two study areas were delimited using a QGIS 3.10 procedure. The altimetry data (TOPODATA 2008) of the study areas (available on the INPE website at: http://www.webmapit.com.br/inpe/topodata/) were reclassified to values greater than 400 meters, since from this attitude the entire area of interest for this study was expressed, containing the highest parts with the presence of the Campos de Altitude, which was the area with vegetation that interested us in the analysis. Following reclassification, the raster was converted into a shapefile ( Figure 6). Research, Society and Development, v. 11, n. 3, e2911326242, 2022 (CC BY 4.0) | ISSN 2525-3409 | DOI: http://dx.doi.org/10.33448/rsd-v11i3.26242

Multi-temporal false color composition
A multi-temporal false-color composition was created by changing the Red channel with the NDVI 1985 product, the Green channel with the NDVI 2020 product. This composition was used to visually validate the findings of the temporal variation analysis in the Campos de Altitude area. This occurred because the areas that underwent some change in vegetation, whether in phytomass or where the vegetation was replaced by another type of phytophysiognomy or land cover, present a red color in the image. Green areas correspond to an increase in phytomass or the presence of another type of more robust vegetation. Unmodified sites are represented in yellow. Table 3 shows the minimum and maximum NDVI values for the two areas during both periods. lower when compared to arboreal vegetation, and values tend to increase as vegetation becomes more robust. Therefore, because of this difference in value scales for the different classes, it was possible to identify and differentiate the classes analyzed in this study. This same pattern of values for classes and for different sensors was observed in the study by Peres et al. (2016), where the same land cover analysis method was used. Figure 7 and Figure 8 depicts the NDVI product of both study areas, with darker shades (dark gray) indicating areas with higher plant biomass, lighter shades (light gray) indicating grassy vegetation (Campos de Altitude), and white tones indicating areas without vegetation (exposed soil). Table 3. In A and B, Serra do Amolar in 2020 and 1985, respectively.

Figure 7 -NDVI product images of the analysis areas with the maximum and minimum NDVI values as shown in
Source: Authors.  Table 3. In C and D, Morraria do Urucum in 2020 and 1985, respectively.
The color gradient in Figure 7, corresponding to the values by land cover class is illustrated, as shown in Table 3. For the Serra do Amolar area, in Figure 7A (2020) the values between 2500 and 4000 are illustrated in the higher regions of the Serra do Amolar (above 500 meters, see Figure 1). In these regions are the Campos de Altitude. The same is seen in Figure 7B   In order to evaluate in more detail the evolution of the loss of altitude field area and increase in dense vegetation areas (trees), a multitemporal composition was carried out, which illustrates well the areas where reduction (colors in red), increase occurred (colors in green) and areas that showed no difference (colors in yellow) in plant biomass (Figure 8 and Figure 9).  authors observed an obvious spatial heterogeneity and an increasing trend in plant biomass growth as a reflection of environmental (meteorological) variables, and they concluded that the spatial distribution of AGB grassland is sensitive to precipitation, while the AGB temporal dynamics were significantly correlated with temperature.
Changes in plant biomass are more visible in Figure 8, which contrasts the multitemporal false-color composition with the two years of analysis. Figures A1-A3  Using Landsat images, this study examined the changes in alpine vegetation cover (Campos de Altitude) between 1985 and 2020. During the years studied, there was a noticeable increase in arboreal vegetation in both study areas. The tendency for the fraction of arboreal vegetation cover to increase is in line with the decrease in Campos de Altitude vegetation, and this phenomenon can be attributed to environmental factors such as temperature, rainfall, solar radiation, and physicalchemical changes in the soil. The main factor for the increase in arboreal vegetation cover, as observed in other studies (Zhou et al. 2016;Bai et al. 2020;Zhang et al. 2007;Li & Qu 2018;Ma et al. 2019;Shen et al. 2015), is an increase in temperature and precipitation. In Morraria do Urucum, the decrease in Campos de Altitude is linked to anthropogenic pressures such as mining activities and cattle grazing.

Conclusion
This research approached the distribution of vegetation cover in two mountainous areas of Mato Grosso do Sul.
Although the anthropogenic influence on land cover changes in Morraria do Urucum was observed, the spatiotemporal variation relationship of Campos de Altitude in areas where there was no anthropogenic pressure, both in Morraria do Urucum and in Serra do Amolar, has yet to be explained. Nevertheless, the results showed that there was clearly a difference between the years of study; therefore, further research is required to understand the environmental factors that led to this difference in vegetation cover, as well as the dynamics of such factors, so as to better contemplate the Campos de Altitude vegetation, which is an important ecosystem of Pantanal, and to determine whether climate change has an impact in this region.
For future research, we intend to use the same study areas, however seeking to understand which environmental factors led to differences in the coverage and distribution of Campos de Altitude. We also aim to automate this method for a more refined/accurate classification that is automatic without intervention in land cover classification. The research carried out in this study can be expanded and replicated to other study areas that are not necessarily the Pantanal biome, since the Campos de Altitude are not restricted to this biome, and can also be used for other types of vegetation that belong to the same class than the Campos de Altitude (grasses).