Accurate estimation of runoff from extreme precipitation is crucial for hydraulic design and flood management. Although the SAC-SMA conceptual model is widely used, its assumption of stationary parameters and reliance on traditional volume-based calibration limit its ability to simulate extreme flows under non-stationary climate conditions. This study develops an integrated framework combining dynamic parameterization and multi-objective calibration to improve the realism and predictive performance of SAC-SMA for extreme hydrological events. Key parameters are expressed as functions of the Normalized Difference Vegetation Index (NDVI) and climatic anomalies. Calibration is performed using Particle Swarm Optimization (PSO), hydrological signatures, and weighted emphasis on extreme flows. The approach is evaluated in the Nazluchay and Siminehroud watersheds in the Urmia Lake basin over 2002–2022. Results show that the enhanced model outperforms the classical SAC-SMA in reproducing hydrographs, baseflow, and peak discharges. Average performance improvements are 32.04% for Nazluchay and 31.19% for Siminehroud. NSE for extreme flows increases by more than 101% in Nazluchay and about 53% in Siminehroud. Concurrent reductions in RMSE and increases in KGE and R² indicate reduced systematic errors and improved representation of nonlinear hydrological behavior under variable climatic conditions. Overall, dynamic parameterization of SAC-SMA combined with signature-guided multi-objective calibration significantly improves simulation of extreme events and provides a robust framework for hydrological analysis in semi-arid, climate-impacted watersheds.