Avaliação de índices espectrais para a identificação da salinidade do solo na bacia do Pajeú - Semiárido de Pernambuco

Abstract

Soils affected by salts is one of the main threats of soil degradation occurring worldwide, especially in arid and semi-arid regions. This study evaluates the feasibility of identifying salinized areas in different soil salinity classes using spectral indices in the Pajeú river basin, in the semi-arid region of northeastern Brazil in the state of Pernambuco. Fifteen spectral indices that encompassed vegetation and soil salinity indices were extracted from Landsat 8 satellite images. Statistical analyzes performed were based on descriptive statistics and second-order polynomial regression (p < 0.05) between Electrical Conductivity (EC) of the soil and the spectral reflectances of the indices. The study area was dominated by non-saline areas. Statistical correlation between EC field measurements and spectral indices showed variable results for each assessed salinity class. However, the Vegetation Soil Salinity Index (VSSI) which is used to discriminate between soil and vegetation stress provided correlation with EC across all salinity classes being considered to predict areas affected by salts across the watershed. The regression model considered obtained predictions resulting in a coefficient of determination (R2) of 0,70 and a mean squared error (RMSE) of 3,33. The use of this spectral index to assess soil salinity proved to be an interesting option to be considered in soil monitoring programs. However, further research is needed to improve the watershed level salinity assessment model. Finally, the efficiency of using remote sensing, followed by site observations, is outlined as a powerful tool in detecting areas affected by salts.