A1528
Title: Spatial ZIEGP models with varying coefficient parameters
Authors: Leng-Yi Ku - National Central University (Taiwan) [presenting]
Chun-Shu Chen - National Central University (Taiwan)
Abstract: Understanding and modeling spatial heterogeneity is critical for climate and environmental studies, particularly in rainfall analysis where both the frequency of zero precipitation and the tail behavior of extreme events exhibit substantial geographical variation. The zero-inflated extended generalized Pareto (ZIEGP) model offers a unified probabilistic framework that simultaneously describes dry days and the full spectrum of rainfall intensities, including low, moderate, and extreme values, without relying on subjective threshold selection. However, existing applications of the ZIEGP model often assume spatially constant parameters, which restricts their ability to reflect local climatological characteristics. To overcome this limitation, a spatial ZIEGP model is developed in which all parameters vary smoothly across space through basis functions. This formulation enables flexible location-specific inference while remaining computationally feasible for large spatial datasets. Simulation studies further confirm the effectiveness of the approach in accurately recovering spatially varying parameters under different data-generating scenarios. The methodology is applied to high-resolution rainfall data from Taiwan, confirming its practical utility and effectiveness for climate risk assessment.
