Hybrid point and interval prediction approaches for drought modeling using ground-based and remote sensing data

Drought as a severe natural disaster has devastating effects on the environment; therefore, reliable drought prediction is an important issue. In the current study, based on lower upper bound estimation, hybrid models including data preprocessing, permutation entropy, and artificial intelligence (AI) methods were used for point and interval predictions of short- to long-term series of Standardized Precipitation Index in the Northwest of Iran. Ground-based and remote sensing precipitation data were used covering the period of 1983-2017. In the modeling process, first, the data processing capability via variational mode decomposition (VMD), ensemble empirical mode decomposition, and permutation entropy (PE) was investigated in drought point prediction. Then, interval prediction was applied for tolerating increased uncertainty and providing more details for practical operation decisions. The simulation results demonstrated that the proposed integrated models could achieve significantly better performance compared to single models. Hybrid PE models increased the modeling accuracy up to 40 and 55%. Finally, the efficiency of developed models was verified for Normalized Difference Vegetation Index prediction. Results demonstrated that the proposed methodology based on remote sensing data and VMD-PE-AI approaches could be successfully used for drought modeling, especially in limited or non-gauged areas.