Modelagem de secas usando um modelo de expansão polinomial evolutiva interpretável com seleção de características

Abstract

Droughts are natural phenomena capable of generating negative consequences for both the environment and human society, such as water scarcity, crop loss, forest fires, and even soil desertification. The literature provides various indices for monitoring droughts, such as the Standardized Precipitation Index (SPI), which aims to determine dry and wet periods on time scales ranging from 3 to 48 months. This paper presents a novel approach to drought modeling by introducing an Evolutionary Polynomial Expansion (EPE) model with integrated feature selection techniques. The EPE model, a powerful mathematical tool, is employed to enhance the accuracy of drought prediction. The EPE was evaluated on a dataset of drought data from Turkey. It can potentially be a valuable tool for drought forecasting and mitigation. The EPE can capture the complex dynamics of drought and identify the most important features for drought prediction. This information can be used to develop early warning systems for drought and strategies for mitigating drought impacts.