Uso de índices de vegetação (NDVI, SAVI E IAF) e de seca (PDSI) no monitoramento pluviométrico do Norte de Minas Gerais

Abstract

The Northern Region of Minas Gerais has a semi-arid climate and is affected by drought, which directly impacts crop productivity and the region's water resources. Rising temperatures and potential changes in precipitation patterns, resulting from climate change, may further aggravate this scenario; therefore, understanding these effects on natural resources and local communities is essential. Based on this premise, this research aims to evaluate the use of NDVI, SAVI, and LAI vegetation indices, together with the PDSI drought severity index, in monitoring rainfall patterns in the northern region of Minas Gerais. The study analyzed a 10-year time series using remote sensing tools and resources, including multispectral images of the study area obtained from different satellites and sensors. To achieve this, the NDVI, SAVI, and LAI vegetation indices were calculated, along with the PDSI, based on TerraClimate data. Subsequently, an analysis of vegetation indices and climate phenomena was performed using Pearson’s Correlation Coefficient, Generalized Linear Model (GLM), and Ridge Regression to monitor drought based on the correlation of the calculated indices. The proposed methodology proved valid for data acquisition and processing; however, it was inconclusive in the statistical analysis. Pearson’s Correlation Coefficient showed values close to zero in most scenarios. The highest recorded positive value was 0.24918 in the PDSI/LAI relationship in February 2022, while the lowest was -0.19304 in the PDSI/NDVI and PDSI/SAVI relationships in February 2018. The GLM model showed low performance, with McFadden's Pseudo R2 ranging from 0.00157 in the worst period (February 2020) to 0.99012 in the best scenario (August 2013). Similarly, the R2 of Ridge Regression showed low performance, with values ranging from 0.0804 in August 2013 (best period) to -0.0358 in February 2019 (worst period). To improve the methodology in future studies, it is recommended to integrate additional variables or adopt models that include hydrological and climatic variables, as well as more detailed temporal analyses to better capture the interactions between vegetation and drought or the stratification of the study area.