Advancements in drought modeling: a comprehensive review of artificial intelligence and statistical models

Drought, characterized by a significant shortfall in rainfall over an extended timeframe, wreaks havoc on the environment, agriculture, wildlife, water supply systems, economy, and human lives. The importance of accurate drought forecasting cannot be overstated, as it forms the backbone of a reliable and timely warning system to mitigate the associated adverse effects. Over the past twenty years, drought prediction has seen significant advancements, particularly through the use of artificial intelligence (AI) and statistical models. The present article provides a comprehensive review of predictive models from 2006 to 2025, encompassing 241 scholarly articles. It distills essential information from these reviewed papers, including data duration, drought indices, drought categorization, model types, input parameters, performance benchmarks, comparable models, top-performing models, and year of publication. While numerous review articles have shed light on innovative developments and cutting-edge techniques to enhance drought prediction, they have not yet fully met the need for researchers for detailed, reliable guidelines on the validity and dependability of the predictive models. Furthermore, these papers predominantly concentrated on enhancing prediction accuracy, sidelining a vital consideration of the reliability of the methodologies applied. To illustrate, our paper discovered that an alarming 27.42% of the AI-based models were incorrectly constructed, thereby rendering their results misleading. The current study concludes by proposing several recommendations for future research, aiming at significantly improving the precision of drought prediction through the application of sound, robust, and relevant scientific principles.