Tracking Climate-Induced Shifts in Rainfall Regimes Using a Hybrid Hidden Markov--Copula Framework.

Understanding and forecasting changes in rainfall regimes is critical for climate adaptation in regions facing increasing hydroclimatic variability. We develop a hybrid modeling framework combining non-homogeneous Hidden Markov Models (nHMMs) and multivariate copulas to detect and simulate climate-driven shifts in precipitation patterns. Applied to daily rainfall data from 462 stations in Chile (1980--2021) during austral winter (May--August). The nHMM identifies hidden climatic states conditioned on large-scale atmospheric covariates. Dynamic Time Warping (DTW) clustering partitions stations into seven distinct climatic zones, enabling the construction of 35 zone-state-specific copulas that capture spatial dependencies within each hidden state. The framework reveals five robust precipitation regimes and documents a 14.45\% decline in the persistence of the wet state over 42 years, strongly associated with rising global ocean temperatures. By integrating temporal state dynamics and spatial coherence, the hybrid nHMM-Copula approach significantly improves the realism of probabilistic rainfall simulations over traditional models. This methodology offers a scalable and transferable tool for tracking precipitation regime shifts and forecasting hydroclimatic variability under climate change. Our findings highlight an accelerating drying trend across Chile, with critical implications for regional water resource management and long-term climate resilience planning.