Dinâmica da linha de costa e mapeamento de uso da terra no litoral central do estado de Sergipe entre os anos de 1988 e 2019

Abstract

Coastal zones are characterized by the high population density, high urban development, tourist potential, and industrial development. On the central coast of Sergipe, the land-cover and land-use have been frequently altered because of natural and anthropic origins by marine and terrestrial dynamics. Nowadays, numerous study techniques allow us to quickly and cheaply identify and estimate how this area changes. The present work evaluates satellite images in a coastal area: (1) the performance of three machine-learning algorithms in land use and land cover classification; (2) shoreline variation by automatic digitization. The original data were multi-temporal scenes from the Landsat 5 and 8 satellite. For the classification, three more data sets were generated in addition to the color composition: (1) composition with vegetation indices, (2) composition with texture metrics, and (3) composition with vegetation index and texture metrics. I selected 450 samples of eight classes of landuse and land-cover subdivided into training (70%) and validation (30%). I assessed the accuracy using the confusion matrix method. I analyzed the shoreline variation through multi-temporal mapping performed with the Digital Shoreline Analysis System tool. The shoreline variation rate and envelope values were estimated over the years in three sectors of the metropolitan region, two in the municipality of Aracaju and one in the municipality of Barra dos Coqueiros. The rates were obtained through two different methods: linear regression and endpoint. In addition, the highest accuracies were produced through Random Forest (0.88 to 0.99), followed by Support Vector Machine (0.87 to 0.99) and Networks Artificial Neural (0.73 to 0.94). The variation of the coastline presented that rates ranged between -10.99 and 18.44 m.year-1, and the accretion behavior predominated over time. The shoreline variation envelope resulted in a lower value in Maré de Apicum located in the Coroa do Meio neighborhood (4.19 m) and a higher value in Atalaia Nova in Barra dos Coqueiros (562.85 m). The classification showed an increasing trend in the urban area and a vegetation reduction and beach strips. After considering the obtained values, more studies are needed to assess the potential of the 3 classifiers. The classification showed a trend towards an increase in the urban area and a reduction in vegetation and beach strips. Given the proximity of the accuracy values obtained, more studies are needed to assess the potential of the 3 classifiers. On the other hand, the detection of shoreline variation showed a predominant trend of accretion in the three analyzed sectors. Therefore, we may conclude that the coastline had a higher proportion of accretion that allowed the growth of the urban area, to the detriment of the natural vegetation and beach strip. Finally, I conclude that the Landsat series images were useful to capture the changes over the past years. These results enabled me to identify the dynamics of the coastal area in a medium to long term and to assess possible trends in behaviors towards erosion, stability, or accretion.