Environmental analysis of land use and occupation in nine watersheds contributing to the Iguaçu River, Brazil

The land use and occupation in watersheds directly influences the water resources, and the vegetation cover is a crucial factor for water maintenance. The objective of this study was to characterize the land use and occupation in nine watersheds of the Lower Iguaçu River (Andrada, Monteiro, Gonçalves Dias, Floriano, Silva Jardim, Cotejipe, Sarandi, Capanema and Santo Antônio sub-basin rivers), using the classification from satellite images. The classification area obtained high accuracy (Kappa coefficient of 99.4 to 99.9%). The Floriano River sub-basin presented the highest degree of environmental preservation, followed by Silva Jardim and Gonçalves Dias River sub-basins, fully or partially inserted in Iguaçu National Park. Monteiro and Sarandi river sub-basins showed the smallest areas of vegetation, and Capanema, Cotejipe, Andrada and Santo Antônio River sub-basins exhibited an intermediate condition. This tendency was verified by the cluster analysis. These results can serve as a baseline for planning, monitoring and management of this area, environmental modeling, conservation of water resources, and environmental services.


Introduction
Remote Sensing, defined as obtaining information about a given object at a distance (Jensen, 2009), makes it possible to map and monitor regions, sometimes inaccessible, for various purposes. The availability of images and advances in digital processing and analysis techniques have made it possible to obtain relevant information about the type, condition, area, dimensions, and others about the areas of interest (Murmu & Biswas 2015).
Advances in remote sensing techniques, plus obtaining data in real time, produce accurate information about land cover quickly and economically (Khalil & Haque, 2018). Thus, the use of remote sensing is an important ally of environmental conservation. The anthropic influence on nature, mainly in hydrographic basins, has become an object of concern for researchers, increasing studies to qualify and quantify the impacts of this occupation (Santos, et al., 2022).
As rates of conversion of natural habitats into anthropogenic landscapes increase, the risks and concerns for environmental conservation also increase (Shimizu, 2007), with water from rivers and river basins occupying a prominent position (Tundisi, 2003, pp. 59-66;Giri, & Qiu, 2016). Agricultural, pasture and urban land use are described as major contributors to the increase of nutrients and sediments to freshwater ecosystems (Uriarte, et al., 2011).
Thus, vegetation cover, especially in riparian zones, plays a significant role in maintaining water quality and ecosystems (Mello, et al., 2010). In this way, prior knowledge of the limits and resilience capacities of natural elements has been required, in relation to land use, in search of sustainable development (Santos, et al., 2022). Given this scenario, geoprocessing techniques are efficient to classify an area in terms of land use and occupation, since they are tools that facilitate the observation of changes in the natural characteristics of the environment, in addition to enabling the understanding of possible causes and effects (Silva, et al., 2021;Valadares, 2017).
In this context, satellite images were used to characterize nine watersheds of the Lower Iguaçu River regarding land use and occupation. The susceptibility of the watersheds to environmental degradation was evaluated using the preserved vegetation as a major conservation criterion. Given the need to maintain and/or increase the resilience of water systems and, consequently, aquatic ecosystems, the maintenance of vegetation is a fundamental requirement, including for the construction of indicators of environmental fragility (Cruz, et al., 2017).

Study Area
The study area is in the South America region, on the border between the state of Paraná, in Brazil, and the state of Missiones, in Argentina ( Figure 1). It has a humid subtropical climate with hot, humid summers and moderate winters (Köppen,

Data Collection
The analyses were performed using the Geographic Information System (GIS) QGIS version 2.8 (QGIS, 2015). The delimitation of the Iguaçu River basin and sub-basins were obtained from the National Water Agency (ANA), generated through digital cartography prepared by the States of Paraná and Santa Catarina (ANA, 2015). The data were obtained during the period of the construction of the Baixo Iguaçu hydroelectric plant, implemented from 2013 to 2019, in the Lower Iguaçu River ( Figure   1).
Three bands (4 -red, 3 -green and 2 -blue) of an image from the Sentinel-2A Multispectral Instrument (MSI) sensor were used for the supervised multispectral classification, obtained by the European Space Agency (ESA), with spatial resolution of 10 m, in 2016, were used. The programs QGIS 2.8 (QGIS Mapserver 2015) and MultiSpec 3.4 (Biehl, & Landgrebe, 2002) Research, Society and Development, v. 11, n. 13, e66111334868, 2022 (CC BY 4.0) | ISSN 2525-3409 | DOI: http://dx.doi.org/10.33448/rsd-v11i13.34868 4 were used in the analysis of the images. The data obtained from the image processing were used for the classification of land use and occupation and for the grouping of the basins as to environmental preservation area.

Land Use Classification
The sub-basins were delimited for each river evidenced in the study area, to obtain its total area. After that, the areas of each basin were used to cut the raster in the satellite image, one for each basin. Regarding the Coordinate Reference System (CRS) for the World Geographic System 1984 plan (WGS 84 UTM ZONE 22S), the data were imported into the MultiSpec program for the classification of the selected sub-basins. The areas of interest for each sub-basin were sampled and categorized into (1) urban area, (2) exposed soil, (3) preserved vegetation, (4) agricultural area, and (5) pasture, following the criteria explained in Table 1. The samples were tested using the cross-validation method to internally test the performance of the classification results for each sub-basin (Morin, & Davis, 2016). By this method it is possible to locate and correct discordant samples of the spectral group to improve the effectiveness of the procedure and perform new sampling.
The Maximum Likelihood Method (MAXVER) was used to classify the areas in the image. This is a parametric classifier, which assumes a spatial probability distribution, usually a Gaussian distribution of the data, determining parameters such as mean, and covariance matrix based on spectral sampling data (Brasileiro, et al., 2016).
The classification reliability depends on the adequate choice of spectral samples, which must be representative and composed of pixels with homogeneous characteristics and distributed in classes with good spectral separability (Pinheiro, et al., 2011), among other factors. A low sample size leads the algorithm to allocate pixels in inappropriate classes, increasing the inaccuracy of the classification (Brasileiro, et al., 2016).
The classification was tested by the method of Landis and Koch (1977), which describes the strength according to the Kappa statistic and follows a classification from 0.00 to 1.00. To be considered accurate enough, the classes should obtain a value above 0.80, which represents a nearly perfect classification (Landis & Koch, 1977).
After obtaining the data, the areas of each class (urban, exposed soil, vegetation, pasture and agriculture) were defined in hectares and in percentages. Class patterns regarding land use and occupation were evaluated using Hierarchical Cluster Analysis (Clarke & Warwick 2001;Hair, et al., 2005) using PRIMER v.6 software (Clarke & Gorley, 2006) to compare subbasins. A similarity matrix was constructed using Euclidean distance, from the data of percentage of urban areas, exposed soil, vegetation, pasture, and agriculture, transformed into square root and normalized. A profile test (SIMPROF) was conducted to

Urban
Sites containing buildings and soil sealing.
Exposed soil Area with soil disturbance due to management practices.

Agricultural
Area composed of different cultures, mainly represented by soybean, corn, cassava and oats, and recently harvested sites crops.

Pasture
Soil cover composed of herbs and grasses, used for animal nutrition.

Vegetation
Vegetation areas along aquatic bodies in riparian zones, legal reserves, and other areas of shrub and tree vegetation in the watersheds.

Results and Discussion
The vegetation class showed occupation of 39.2% of the study area, which indicates a good vegetation cover that, according to the Forest Code of Law No. 12,651 (2012), must be maintained above 35% in the Cerrado biome. However, this coverage is distinct among the sub-basins. The vegetation class was the most representative in the Floriano, Gonçalves Dias and Silva Jardim Rivers. The agricultural class showed predominance in the river sub-basins of the Monteiro, et al. The pasture class was only present in the Capanema River sub-basin. The classes exposed soil and urban area were not very frequent (Table 2; Figures 2, 3A and 3B).
About 45% of the study area is contained within the Iguaçu National Park (PNI), which includes the Floriano, Gonçalves Dias and Silva Jardim River sub-basins, with vegetation cover of 99.31%, 58.31% and 45.89%, respectively - Table 2 and Figures 2, 3A and 3B. The Floriano River sub-basin is entirely contained within the intangible zone of the Iguaçu National Park, which makes it possible to use it as a reference for establishing conservation and monitoring parameters.
The appreciation of the Conservation Unit (Iguaçu National Park) and the conservation of hydrographic basins, together with the recognition of vegetation cover as a fundamental element for the maintenance of water resources and the integrity of the ecosystem (Tambosi, et al., 2015;Wurtzebach, & Schultz, 2016), although it is a very important premise for environmental conservation, they present many difficulties. the precept of conservation is explicit in the Forest Code, as one of the principles of sustainable development: "Brazil's commitment to the preservation of its forests and other forms of native vegetation, as well as biodiversity, soil, water resources and the integrity of the climate system, for the well-being of present and future generations" (art. 1, item I of Law No. 12651, 2012). However, the challenge for its management is still great.
In the Iguaçu River basin fragmentation is high (Hentz, et al., 2015). The use and occupation of the Monteiro and Sarandi River sub-basins with the agricultural (31.3%) and pasture (28.5%) classes, add up to about 60% of the area (Table 2). These data contrast with the conservation of the Floriano River. They also make explicit the importance of agropastoral activities. In these basins the vegetation occupied less than 19% of the area and the urban class about 5.6% (Table 1; Figures 2, 3A and 3B).
The agropastoral activities are favored by flat or undulated slopes, which represent about 30% of the area (Silva et al., 2022), and that according to Pinheiro et al. (2011), must have fertile soil, be arable and drained. Celestino et al. (2019) categorized the two basins as medium fragility, indicating that agropastoral activities, urban centers, especially Capitão Leônidas Marques and Realeza, with the highest densities in the region (Brazilian Institute of Geography and Statistics [IBGE], 2014), should be monitored. Vegetation areas must be conserved, mainly with a view to conserving the region's water resources.
The vegetation class, in the sub-basins of the Andrada, Santo Antonio, Cotejipe and Capanema Rivers, has an intermediate situation, occupying between 25 and 38%. And the agropastoral activities (agriculture and pasture) are between 60 and 74% (Table 2; Figures 2, 3A and 3B). In Capanema River sub-basin the pasture class occupied more than half of the basin (51.17%), it is associated with the presence of high slopes (Table 2, Figures 2 and 3B), with a predominance of strong undulating and mountainous relief (Silva et al., 2022). The urban class, in Andrada River sub-basin, was relatively higher, in it there are several municipalities in its territory and surroundings (Table 2; Figures 2 and 3A). Research, Society and Development, v. 11, n. 13, e66111334868, 2022 (CC BY 4.0) | ISSN 2525-3409 | DOI: http://dx.doi.org/10.33448/rsd-v11i13.34868 6  Source: Authors.
The Kappa coefficient values varied among the sub-basins (Table 2), indicating a good accuracy for the classes formed (from 99.4 to 99.9%), according to Landis e Koch (1977). These values were higher than those obtained by Yuan, et al., (2005), but the classes and the type of image used by them were different.
The initial challenge for the classification of the images, obtained by the Sentinel-2 satellite, were some areas that have similar wavelengths, mainly in the green scale, making it difficult to separate the agricultural and pasture classes, due to the similarity of the reflected wavelengths. vegetation and captured by the sensor. This challenge was overcome by selecting these areas in advance, at the time of analysis, and changing the wavelength values further apart between the classes. In this way, precision was increased in separating land use classes. However, the information extraction period for image creation is an important factor. Source: Authors.
In the cluster analysis, five groups were observed (Figure 4) vegetation, replaced by other activities, can weaken water systems, increasing other impacts, such as water erosion on the banks and silting of rivers.

Figure 4:
Similarity dendrogram (Cluster analysis and SIMPROF) from data matrix transformed by river sub-basins. Significant groups (p<0.05) showed in red lines, and ungrouped river sub-basin in black.

Source: Authors.
We can infer that the Floriano river sub-basin is the most preserved, followed by Silva Jardim and Gonçalves Dias. The Capanema, Cotejipe, Andrada and Santo Antônio River sub-basins possess comparatively an intermediate degree of environmental preservation and the Monteiro and Sarandi sub-basins exhibit the least degree of environmental preservation, with the greatest distance between them.

Conclusions
The supervised multispectral classification used in our study obtained satisfactory results. The grouping showed that the most conserved sub-basins are the Floriano, Silva Jardim and Gonçalves Dias Rivers. The most susceptible to degradation are the basins of the Monteiro and Sarandi Rivers.
Besides the characterization of the sub-basins, this description, which is unprecedented, provides as a baseline for planning, monitoring and management of the area. It also serves for environmental modeling, conservation of water resources, biodiversity, and environmental services.