Max-stable process framework for intensity‒duration‒frequency analysis and climate projections in monsoon Bangladesh

Traditional statistical approaches for extreme precipitation projections suffer from systematic biases in direct climate model precipitation outputs. This study develops an innovative covariate-based max-stable process (MSP) framework that leverages physics-informed climate variables rather than biased precipitation projections to analyze extreme precipitation across Bangladesh. Systematic screening of 2500+ model combinations across 35 meteorological stations, five seasons, and ten durations (1–48 h) identified optimal MSP characterizations with seasonal specificity: Brown-Resnick processes dominate the SW-Monsoon while Extremal-t Bessel characterizations optimize Winter extremes. Geographic variables (latitude, coastal proximity, and elevation) systematically outperform large-scale climate teleconnections across all conditions, indicating that Bangladesh's extreme precipitation is controlled by regional Bay of Bengal dynamics rather than remote ocean oscillations. Seasonal intensity–duration–frequency (IDF) analysis establishes SW-Monsoon extremes reaching 37–64 mm h⁻¹ for 25-year, 1-h events, with distinct seasonal hierarchies reflecting a monsoon-dominated climate. Climate change projections under the SSP2-4.5 scenario (2025–2050) reveal substantial spatial heterogeneity (−57% to +79%), with differential responses across the probability spectrum: rare 100-year events intensify by +37% while frequent 2-year events decrease by −51%, indicating tail-stretching of extreme distributions. The results highlight that infrastructure design standards require distinct treatment of rare versus frequent extremes, with implications for climate adaptation in monsoon-dominated regions worldwide.