LANDSCAPE CHANGE WITH THE TRANSPOSITION OF THE SÃO FRANCISCO RIVER, IN THE DOMAIN CAATINGA, PERNAMBUCO

Landscape change with the transposition of the São Francisco River, in the Domain Caatinga, Pernambuco.The objective of this study was to evaluate the dynamics of land use and cover and the landscape structure, as well as the landscape structure, of the Directly Affected Area by Integration Project of the São Francisco River with the Northeastern Hydrographic Basins, from the East Axis portion located in the State of Pernambuco, Brazil. For this purpose, TM / Landsat 5 and OLI / Landsat 8 images were used, referring to the years 1998, 2008 and 2018, periods before, during and after the transposition, respectively. The land use and land cover classes used in this analysis were savanna-steppe, savanna-steppe, anthropized and / or uncovered area, water resources and agricultural area. Image processing and classification were performed using the QGIS software. Also, studies related to the landscape structure were carried out, using different types of metrics, processed in the Patch Analyst tool, an extension of ArcGIS 10.5. As a result of the analysis of the 20 years, it was obtained that the savanna-steppe vegetation cover showed a 13.86% reduction. However, there were additions in the areas of ciliary savanna, water resources, agricultural area and anthropized area in 1.93%, 0.11%, 0.31% and 11.51%, respectively. Furthermore, there was an increase in forest fragmentation, which corroborated with the results regarding for the core area metrics, which show that there has been a reduction in the size of the fragments and a trend towards the loss of the core areas, due to the edge effect.


INTRODUCTION
As a result of anthropic and/or natural interventions in landscapes, as they acquire heterogeneous spatial conformations, characterized by a mosaic of spots that differ in size, shape, content and history (WU, 2013). As a way of monitoring these changes in the landscape, geotechnologies are used as a means of understanding the fragilities of forest fragments and their consequences for the maintenance of biodiversity, also serving as a basis for studies related to the influence of the landscape pattern in the fire regime region (LARIS et al., 2018). Through combinations of image classification techniques and metric indexes of the landscape, it is possible to evaluate the temporal and spatial evolution of forest fragmentation. (SANTOS et al., 2016).
The Brazilian semi-arid region is a world example of an area with intense changes in its landscapes, requiring constant orbital monitoring of its natural resources. This region occupies 11% of the Brazilian territory (844,453 km2), characterized by having a floristic diversity that includes about 3,150 species of plants and covering nine biogeographic sub-regions, which result from the heterogeneity of environmental conditions, such as temperature, precipitation and altitude (QUEIROZ et al., 2017;SOUZA, 2018). It is one of the regions of the world that presents high ecological sensitivity to climatic variability (SEDDON et al., 2016), with only 9.36% of its protected areas (77537,64 km2), corresponding to 208 Conservation Units (UCs) (UNCT, 2020).
Among the current socioeconomic activities that have generated controversy in relation to the conservation of the Caatinga domain, it is the Project for the Integration of the São Francisco River with the Hydrographic Basins of the Northern Northeast (PISF), implemented in the states of Pernambuco (PE), Ceará (CE), Paraíba (PB) and Rio Grande do Norte (RN). The objective of the project was to distribute a percentage of 3.5% of the waters of the São Francisco River to basins in the semi-arid Northeast region, through two waterconducting channels with a total length of 720 km, known as Northern Axis and Eastern Axis (BRASIL, 2004). The works of the referred project had great proportions and lasted for a long period (they started in 2007 and they were completed in 2017, in the Eastern Axis, and in 2020, in the Northern Axis).
Therefore, there is a need for a temporal analysis of the area, in order to diagnose the possible interactions that PISF may have caused in the landscapes. Given the context presented, this paper aims to answer the following question: Did the transposition of the São Francisco River influence the dynamics of land use and cover in the Eastern Axis? As a main hypothesis, with the construction of the transposition, there was a decrease in forest cover, as well as an increase in exposed soil sites.
Therefore, the objective of this study was to evaluate the dynamics of changes in land use and forest cover, as well as the landscape structure of the Directly Affected Area (DAA) by the São Francisco River Integration Project with the Hydrographic Basins of the Northern Northeast, from the portion of the Eastern Axis located in the State of Pernambuco, Brazil, for a period of 20 years.

Location of the study area and general environmental characteristics
The present study was carried out in the Directly Affected Area (DAA) by the Project for the Integration of the São Francisco River with the Hydrographic Basins of the Northern Northeast (PISF), of the portion of the Eastern Axis, in Pernambuco. The DAA is defined as a strip along the structures of the Project, 5.0 km wide on each side (BRASIL, 2004) ( Figure 1 PISF, a 5.0 km buffer was created, starting from the referred axis and, with that, the research area went beyond the limits of Pernambuco, reaching the municipality of Monteiro, in Paraíba. With an extension of 164.125,44 ha, the study area is located in the semi-arid Northeast and its main types of vegetation are the Wooded Steppe-Savanna and the Forest Steppe-Savanna. It presents average annual rainfall below 800 mm, average sunshine of 2,800 hours per year, average annual temperatures of 23ºC to 27ºC, evaporation of 2,000 mm per year and average relative humidity around 50%, which favors the negative water balance (BRITO; MOURA; GAMA, 2007).

Land use and cover in the Eastern Axis of the DAA
To understand the dynamics of land use and cover in the DAA of the base channel (Eastern Axis) of the PISF, the Landsat 5 Thematic Mapper (TM) and Landsat 8 Operational Land Imager (OLI) satellites were used, requested from the USGS -United States Geological Survey (https://earthexplorer.usgs.gov) platform, referring to the years of 1998, 2008 and 2018, respectively, before, during and after transposition. The orbits/points used to acquire the images were: 215/065; 215/066; and 216/066.
The acquisition periods of the images varied between September and October, as they were the months with the lowest cloud cover, which favored the analysis of the areas under study. However, as the drought months are characterized, the spectral response of the vegetation was not so expressive, which required more attention in the process of collecting spectral samples from the targets, because in the dry months the tree and shrub species stood out from their blades (deciduous).
Before carrying out the supervised classification of the study area, the classes of land use and occupation were determined, as noted below. The choice of classes was made with the aim of verifying how much the transposition of the São Francisco River would interfere in the level of the river, in the amount of vegetation, in the expansion of agricultural, urban and/or exposed soils.
Based on the classes of land use and cover of the Brazilian Institute of Geography and Statistics (2017), the classes used in the categorization of the area under analysis were determined, namely: Steppe-Savanna, Riparian Steppe-Savanna, anthropized and/or uncovered area, water resources and agricultural area.
The supervised classifications of the generated compositions were carried out in the SCP Dock tool of the Semi-Automatic Classification Plugin (SCP), where associations were made among the spectral data obtained by the sensors with the land cover classes. Subsequently, the regions of interest (training areas) were selected for each land cover class identified in the image and polygons were generated over homogeneous areas, overlapping the pixels belonging to the same land cover class (CONGEDO, 2016 After that, the image classification was performed using the Maximum Likelihood Classification, which is the most common supervised classification method used with remote sensing image data. The algorithm classifies the entire image by comparing the spectral characteristics of each pixel to the spectral characteristics of the land cover reference classes (RICHARDS; JIA, 2006;CONGEDO, 2016).
To confirm the data collected by the satellites, field visits were made in areas close to the transposition, where there was the collection of GPS waypoints in different artificial structure (highways and masonry Works) and natural structure (water resources, vegetation and exposed soil) and confirmation on-the-spot of the types of targets that have been analyzed.
Regarding the post-processing of the data obtained and resulting from the supervised classification, its precision was performed using the Kappa index, which exposes the performance of the classification and can be qualified based on the following parameters (LANDIS; KOCH; 1977): <0.00 lousy; 0.00-0.20 bad; 0.21-0.40 reasonable; 0.41-0.60 good; 0.61-0.80 very good; and 0.81 -1.00 great.
With the application of the mathematical operations of the GIS on the maps of the two decades, it was obtained: quantification of areas in hectares, referring to each class of use and coverage in each year; percentage of occupation of the DAA for each class of land use and land cover in all selected years; percentages of variation for the periods 1998-2008, 2008-2018 and 1998-2018 for the DAA (MELLO et al., 2011;MORAES et al., 2017).
Based on the aforementioned authors, the difference in area in hectares was carried out from one date to another (Equation 1), therefore, calculated to verify the increase or decrease in occupation. . The metrics were obtained using the Patch Analyst tool, an extension of ArcGIS 10.5, which provides spatial statistical estimates using vector and matrix files, based on the principles of Landscape Ecology. Regarding the analysis of the central area, a 50 m negative buffer was applied to calculate the Core Area.

Land use and cover in the eastern axis of the daa
The results of the Kappa index for the classifications carried out in 1998, 2008 and 2018 were 0.97, 0.94 and 0.95, respectively, which corresponds to an excellent classification quality (LANDIS; KOCH, 1977). The temporal analysis carried out over the 20-year period (1998 to 2018), which occurred in the Eastern Axis of the DAA of the São Francisco River transposition (Figures 3 and 4), found significant changes in the evaluated environment, as shown in Table 1    In addition to the results by year, data were also obtained regarding the variations of each class of land use in the periods from 1998 to 2008, 2008 to 2018 and 1998-2018 (Table 2). In which, the Steppe-Savanna class presented the highest percentage of change in the analyzed intervals. Landscape structure in the Eastern Axis of the ADA Continuing the spatial analysis of the landscape, the results contained in Table 3 were obtained, referring to the metrics in their different categories.