Exploring seasonal drought patterns with climate data and machine learning models

This study examines seasonal drought persistence in Punjab, Pakistan, by analyzing precipitation patterns, spatial variability, and drought trends from 1981 to 2023. The Standardized Precipitation Index (SPI-3) is used to assess drought and wetness cycles across upper and lower regions, revealing significant seasonal and inter-annual variability influenced by monsoon dynamics and climate phenomena (El Niño/La Niña). Given its three-month timescale, SPI-3 effectively captures short-term drought variations, making it particularly sensitive to seasonal precipitation shifts, influencing drought onset and recovery. Spatial analysis shows a distinct north–south gradient, with greater precipitation and variability in northern areas. Drought persistence correlations indicate a weak relationship with precipitation (−0.03) but a strong negative correlation with temperature (−0.74) in the lower region during the summer-to-winter transition, emphasizing temperature’s dominant role. In contrast, the upper region is more influenced by precipitation (−0.19) and temperature (−0.28), with soil moisture strongly correlating with temperature (0.80) in the spring-to-summer transition, underscoring its role in drought continuation. To predict drought persistence during seasonal transitions, various machine learning models-including Logistic Regression, Random Forest, Gradient Boosting, Support Vector Machine, K-Nearest Neighbors, and Naïve Bayes are evaluated. The models exhibit varying predictive performance across seasons, with Logistic Regression achieving the highest accuracy in the spring-to-summer transition (0.962). Gradient Boosting shows the weakest performance in the winter-to-spring transition (0.538). Naïve Bayes demonstrates strong predictive capability in the autumn-to-winter transition, achieving an accuracy of 0.885 in both the lower and upper regions, reinforcing its reliability for seasonal drought forecasting. These insights highlight how seasonal climate patterns influence drought behavior, suggesting the importance of considering climate variations in water resource planning and drought management. Our findings help improve the understanding of drought persistence and may support efforts toward sustainable agricultural and water management practices in changing climatic conditions.