Influência dos métodos de classificação supervisionada do solo na parametrização de modelos hidrológicos - hidráulicos

Abstract

The changes witnessed in the way man develops in society show the need for adequate planning of cities. Environmental issues such as the issue of flooding in cities illustrate the importance of preventive management. Therefore, remote sensing methodologies and computational modeling allow for the analysis and anticipation of natural phenomena in a non-face-to-face manner. The objective of this work was to classify the land use for the watershed of the Maratoan creek, located in an urban-rural transition zone, in the municipality of Crateús, Ceará, and attest in hydrological - hydraulic modeling the behavior of peak flows in natural drainage. of the microbasin, using tools incorporated in geographic information systems (GIS). Supervised classification of Landsat-8 satellite images was carried out according to three statistical methods: minimum distance, maximum likelihood and Gaussian mixture model. With a view to checking the most appropriate method, the accuracy of the methods was measured using the global accuracy indices, producer accuracy, user accuracy and Kappa index. The minimum distance method showed better results in terms of accuracy and visual inspection of classified images compared to real scenes. Finally, the classification results provided input data for hydrological – hydraulic modeling in the SWMM software for simulating peak flows. It was concluded that parameters such as sealing rates generated by the classification process had little influence on flow determinations in the applied model.