Identificação de áreas suscetíveis à ocorrência de secas agrícolas usando dados censitários, sensoriamento remoto e análise multicritério

Abstract

Climate change has increasingly impacted ecosystems and economic activities, especially in the Northeast region of Brazil, where irregular rainfall intensifies the occurrence of agricultural droughts. This study's main objective was to identify areas at greatest risk of agricultural drought, focusing on corn, cowpea, and cassava crops, for the period 2001 to 2022, based on the Agricultural Drought Risk Index (IRSA). The IRSA measures agricultural drought risk, ranging from 0 (very high) to 1 (very low). The methodology used climate data (precipitation, evapotranspiration, and temperature) obtained via MODIS and CHIRPS satellites, as well as data from the Agricultural Census (SIDRA/IBGE), processed with support from Google Earth Engine and QGIS. Based on the Analytic Hierarchy Process (AHP) method, the IRSA was developed, which allowed spatial mapping of risk levels in different microregions. In this study, the IRSA was applied to microregions of the state of Paraíba and integrated remote sensing, census, and Hierarchical Analysis Method (AHP) data. This study used temperature (MODIS), evapotranspiration (MODIS), and precipitation (CHIRPS) data estimated by remote sensing, as well as harvested area in hectares (SIDRA/IBGE). Statistical validation of the index was performed using Receiver Operating Characteristic (ROC) and Area Under the Curve (AUC) curves, using 3,000 regular points extracted from raster maps for each crop. Spatial analysis indicated that the Sertão Paraibano and Borborema mesoregions presented the highest risks of agricultural drought, with 2012 standing out as the most critical year. The microregions of Patos, Piancó, Sousa, Catolé do Rocha, Seridó Oriental, and Ocidental were the most affected throughout the historical series. In contrast, the Cajazeiras, Sousa, Patos, Piancó, Itaporanga, and Serra do Teixeira microregions presented the lowest risk of agricultural drought. Corn crops demonstrated greater resilience to drought, while cassava was the most affected. Maps generated based on the 21-year average indicated that approximately 32.80% of the total studied area presented moderate risk, with a predominance of pixels in the 0.58 to 0.84 range, indicating low to very low risk. The results demonstrated excellent performance in identifying spatial risk patterns that can guide future actions, with AUC values ranging from 0.97 to 0.99, indicating high discrimination capacity of the model between areas with and without agricultural drought. This work demonstrates the efficiency of integrating geospatial and multicriteria methods in assessing agricultural drought risk, providing technical support for territorial management, agricultural planning, and public policies for drought mitigation in the semiarid region of Paraíba.