Ensemble Modeling and Threshold Analysis for Assessing Non-Linearity of Extreme Precipitation Under Future Climate Scenarios

The erratic and non-linear nature of extreme precipitation due to climate change presents significant challenges for water resource management. This study investigates disproportionate patterns of extreme precipitation under future climate scenarios (SSP 245 and SSP 585) by integrating precipitation extreme indices and advanced ensemble modeling techniques. The ensemble approach combines projections from multiple General Circulation Models (GCMs) to improve prediction reliability. Taylor Skill Score (TSS) rankings and seasonal evaluations were used to identify the most skillful models, such as BCC-CSM, CNRM-CERFACS, MPI, and MRI-ESM2.0. Weighted ensemble combinations progressively incorporated top-ranked models. Advanced regression techniques, including Artificial Neural Networks (ANN), K-Nearest Neighbors (KNN), and Support Vector Machines (SVM), optimized model merging. Ensemble performance was validated using metrics like RMSE and correlation, demonstrating improved accuracy compared to individual models.

Future precipitation patterns were analyzed under SSP 245 and SSP 585, revealing amplified extremes. Extreme precipitation indices were divided into two categories: Quantitative Upper-Tail Threshold Analysis (R90p, R95p, R99p) and Duration-Integrated Metrics (CWD, SDII, PRCPTOT). Results indicated an increase in days exceeding the 90th and 95th percentiles but a decline in days exceeding the 99th percentile, suggesting a threshold effect. Future projections show decreased CWD and increased PRCPTOT, reflecting fewer wet days but higher annual precipitation.

Correlation analysis revealed non-linear relationships. Quantitative Upper-Tail Threshold metrics showed increasing correlation with CWD under SSP 245 (336.84–500%) before declining under SSP 585 (17.78–96.88%). Their correlation with SDII increased from observed to SSP 245 (27.27–144.83%) but stabilized under SSP 585 (-1.43% to 1.41%). These findings highlight an evolving interplay between moderate and extreme precipitation events under intensifying climate conditions.

The results offer critical insights for water resource management, including optimized agricultural practices, adaptive urban infrastructure for flood management, and region-specific policies to enhance resilience against changing precipitation dynamics.