A new multivariate agricultural drought composite index based on random forest algorithm and remote sensing data developed for Sahelian agrosystems

This manuscript aims to develop a new multivariate composite index for monitoring agricultural drought. To achieve this, the AVHRR, VIIRS, CHIRPS data series over period of 40 years, rainfall and crop yield as references were used. Variables include parameters vegetative stress (SVCI, PV, SMN), water (PCI, RDI, NRDI), heat (SMT, TCI, STCI), variable related environmental conditions was calculated through normalized efficiency index. Then, random forest algorithm used determine weights each component model by considering interannual fluctuations in cereal yields an impact variable. The compared VHI, NVSWI SPI-12 indices validation. results show large spatiotemporal concordance between MDCI validation with maximum correlation 0.95 VHI highly significant p value (< 2.2e-16). Validation shows significantly higher statistically relationship than that observed SPI NVSWI. P range from 3.531e-05 6.137e-06 correlations vary 0.6 0.64 depending on station. It is also correlated Palmer drought severity (PDSI) climatic deficit (CWDI), R = 0.85 < 5.8e-10 0.72 1.9e-6, respectively. Finally, study provides direction modeling should be further explored under various agroclimatic conditions.