NMDI application for monitoring different vegetation covers in the Atlantic Forest biome, Brazil

Abstract The Atlantic Forest biome, a biodiversity hotspot, has been suffering from the occurrence of fires, which are associated with extreme drought events that present an uncertain future for its vegetation. In this context, the present study evaluated and correlated the normalized multi-band drought index (NMDI) with biophysical variables in the monthly period from 2001 to 2019 in 12 classes of land use and cover. The Auto Regressive Integrated Moving Average (ARIMA) model was applied to the NMDI series and its ability to simulate data from the observed time series (2001-2019) and the future (2020-2030). The results showed a decrease in the NMDI values for the period considered dry in the State of Rio de Janeiro (SRJ), mainly for the classes of pasture and savannah, which presented greater sources of heat. The non-parametric analysis was performed using the Mann-Kendall test for all biophysical variables. The variables soil moisture and NMDI showed negative trends (Z = -1.68 and Z = -0.76), whereas gross primary productivity (GPP) showed a positive trend (Z = 1.89). The generated and validated ARIMA modeling simulated NMDI well and the Willmott coefficient (d) was approximately 1.0 for the study period. The 10-year projection (2020-2030) from NMDI for SRJ pointed to a change in class from wet to dry in the mixed forest area (D) and cultivated land (L). The ARIMA model is capable of representing the seasonality of the series. These results showed that the applicability of NMDI in predicting fire hazard conditions would be adequate in other areas of tropical forests.