Typhoons or heavy rainfall events account for a significant portion of annual extreme rainfall in southern Taiwan. This study built a dynamic time series model for long duration extreme rainfalls to investigate common trends in annual maximum precipitation. The annual maximum rainfall depths for 24-h duration (AM24. h) and annual maximum rainfall depths for 1-h duration (AM1. h) of the same event were observed at 16 rainfall stations in the Kaoping River watershed, Southern Taiwan. The relationship between short and long duration observations is difficult to directly evaluate from comparative statistical techniques. Dynamic factor analysis (DFA, a dimension reduction technique and can be adopted for analysis of non-stationary time series) was therefore used to identify underlying common trends of AM24. h time series and significant explanatory variables (AM1. h) contributing to variations of AM24. h. The results of best dynamic factor models (DFMs) show that the AM1. h of nearby rainfall stations at both high elevation (coefficient of efficiency, Ceff=0.924) and low elevation (Ceff=0.909) successfully describe the AM24. h in the Kaoping River watershed. The AM1. h observed at high elevation, close to the study area center significantly relates to the AM24. h of most rainfall stations. The common trends of extreme precipitation at most stations are upward in the Kaoping River watershed. Due to this successful application of DFA, predicting AM24. h in the Kaoping River watershed should account for the dominant rainfall stations with annual maximum rainfalls for short durations, to provide important information to determine precipitation trends for long-term planning.
All Science Journal Classification (ASJC) codes
- Water Science and Technology