Characterizing Inter-Seasonal Meteorological Drought Using Random Effect Logistic Regression

Abstract

Sustainable watershed development focuses on building resilience to drought through better water resource management, ecosystem protection, and adaptation strategies. In this study, the spatiotemporal dynamics and inter-seasonal characteristics of meteorological drought across Ankara Province, Turkey, were investigated and compared using a conditional fixed effect logistic regression model (CFELogRM) and a random effect logistic regression model (RELogRM). To assess the statistical validity and effectiveness of these models, we conducted significance tests, including the log-likelihood ratio chi-square, and Wald chi-square tests. The obtained p-values associated with both the RELogRM and CFELogRM models for the selected seasons demonstrate their statistical significance. Additionally, we conducted the Hausman test (HT) to compare the efficiency of the RELogRM and CFELogRM models. Remarkably, the results of the HT suggest that RELogRM is the optimal model for modeling fall-to-winter season drought dynamics across the study area. Notably, the significant coefficient derived from RELogRM indicates a statistically significant negative correlation between spring moisture conditions and the probability of summer droughts. Specifically, the odds ratio of 0.2416 reflects a 24.16% reduction in the likelihood of transitioning to a higher drought category, emphasizing the crucial role of antecedent moisture conditions in influencing drought propensity.