Machine Learning for Drought Prediction; Review, Bibliometric Analysis, and Models Evaluation

Drought is among the most destructive natural disasters with a drastic impact on the economy, ecosystem, biodiversity, and agriculture. Climate change and global warming are greatly contributing in increasing the severity, irregularity, and frequency of droughts throughout the world. Accurate prediction of the droughts empowers the policies made for mitigation, resilience and adaptation. Machine learning models have shown outstanding performance with promising results in predicting the drought. This paper aims at presenting a systematics state of the art review on the major machine learning advancements using a novel taxonomy. Machine learning models have been classified in five major groups based on the method used, i.e., neural networks-based, decision tree-based, support vector-based, hybrids and ensembles. The survey shows that emerging ensemble and hybrid models deliver higher performance where single methods are often outperformed.