Nonstationary Distributional Changes of Annual Rainfall Indices in Taiwan

Abstract

The purpose of this study is to detect alterations of annual rainfall regime in terms of distributional changes of rainfall indices in Taiwan. A total of nine rainfall indices at eight stations with at least 72-year daily records are used to characterize annual rainfall regime. The generalized additive models for location, scale, and shape (GAMLSS) is adopted for dynamically detecting evolution of probability distributions of these indices. The results indicate that alterations of annual rainfall regime are spatial diverse and no prevalent distributional changes are observed. Approximate half indices (35/72) exhibit stationarity and WRR (wet-season rainfall ratio) is the least-affected index since only two stations (Sunmoon Lake and Chengkung) revealing nonstationarity. On the contrary, PRCPTOT (annual total rainfall), SDII (simple daily intensity index), and ADD (annual dry days) are high-affected indices due to 75% of stations exhibiting nonstationarity. Highly affected stations include Taipei, Hengchun, and Dawu because two (WRR and CRX (maximum consecutive rainfall)), one (WRR), and one (WRR) indices classified as stationarity, respectively. Taichung and Sunmoon Lake are least-affected stations since 7 out of 9 indices revealing stationarity. These facts indicate that annual rainfall regime of central-west Taiwan is less affected, and north and southeastern regions are change-prone regions.