全球变暖对台湾极端降雨的影响
Impact of Global Warming on Extreme Rainfall in Taiwan
AI总结：
本研究利用PGEV模型结合TCCIP降雨与NCEI温度数据，分析全球变暖对台湾极端降雨的影响，发现气温升高导致降雨强度增加且存在区域差异，未来极端降雨将更频繁。
摘要：
本研究检验了全球变暖与台湾极端降雨之间的关系。分析了由 MOST 主导的 TCCIP 项目提供的台湾降雨数据。北半球参考温度数据来自由 NOAA 主导的 NCEI。研究重点关注日降雨量的年最大值，并使用 Olafsdottir 等人提出的 PGEV 模型来拟合极值并进行推断。PGEV 模型整合了常用于分析极值数据的广义极值（GEV）和超阈值（PoT）方法。相对强度和重现值被用于展示温度与极端降雨之间的联系。结果表明，台湾极端降雨的强度随气温升高而增加。然而，全球变暖对极端降雨频率和强度的影响因地区而异。在北部和南部地区，极端降雨的频率发生了变化；而在中部和东部地区，极端降雨的强度发生了变化。此外，根据重现值分析，未来极端降雨可能会更频繁地发生。为了考虑不同地点之间的差异，使用 Gaussian Process（高斯过程）对使用 PGEV 模型获得的结果进行平滑处理。此外，使用 Gaussian copula 和 Gaussian Process 进行了模拟，以确定每个 PGEV 模型的分位数置信区间。模拟表明，所有比较有协变量和无协变量模型的测试均显著。
英文摘要：
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.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
