本研究探讨了全球变暖与台湾极端降雨之间的关系。分析了科技部主持的“台湾气候变迁推估资讯与调适知识平台计划”（TCCIP）提供的台湾降雨数据，以及由美国国家海洋和大气管理局（NOAA）主导的北半球参考温度数据。重点研究了年度最大日降雨量，并采用Olafsdottir等人提出的PGEV模型（\citep{olafsdottir2021extreme}）拟合极端值并进行推断。该PGEV模型结合了通用极值分布（GEV）和阈值峰值（PoT）方法，这两种方法常用于分析极端数据。通过相对强度和重现期值展示了温度与极端降雨之间的联系。结果表明，随着气温升高，台湾极端降雨的强度增加。然而，全球变暖对极端降雨频率和强度的影响在不同地区存在差异。北部和南部地区极端降雨的频率发生变化，而中部和东部地区的极端降雨强度发生变化。此外，根据重现期分析，未来极端降雨可能更频繁地发生。为了考虑地点间的差异，采用了高斯过程平滑PGEV模型获得的结果。同时，利用高斯copula和高斯过程进行了模拟，以确定每个PGEV模型的分位数置信区间。模拟结果显示，所有包含协变量与不含协变量的模型比较测试均具有显著性。

The relationship between global warming and extreme rainfalls in Taiwan was examined in this study. Taiwan rainfall data from TCCIP, a project led by MOST, were analyzed. North Hemisphere reference temperature data from NCEI led by NOAA. The yearly maximum of daily rainfall was focused on and the PGEV model, as proposed by Olafsdottir et al. \citep{olafsdottir2021extreme}, was used to fit the extreme values and make inferences. The PGEV model integrates the General Extreme Value (GEV) and Peak over Threshold (PoT) approaches, which are commonly used to analyze extreme data. Relative intensity and return value were used to show the connection between temperature and extreme rainfall. Results indicated that the intensity of extreme rainfall in Taiwan increases as the temperature rises. However, the effects of global warming on the frequency and intensity of extreme rainfalls varied by region. In the north and south regions, the frequency of extreme rainfalls changed, while in the center and east regions, the intensity of extreme rainfalls changed. Furthermore, according to the return value analysis, extreme rainfalls are likely to occur more frequently in the future. To account for differences between locations, Gaussian Process was used to smooth the results obtained using the PGEV model. In addition, simulations using the Gaussian copula and Gaussian Process were conducted to determine the quantile confidence intervals for each PGEV model. The simulations showed that all tests comparing with models with and without covariates are significant.