Exploration of severities of rainfall and runoff extremes in ungauged catchments: A case study of Lai Chi Wo in Hong Kong, China

Abstract

Exploration for estimating rainfall and runoff extremes in ungauged catchments is challenging since there are no field measurements of rainfall and streamflow for confirming study results. This study proposed a systematic approach to tackle the challenge, and the approach includes field survey, rainfall data collection, frequency analysis, installation of equipment in the study area, and numerical modeling. The approach was then applied to the Lai Chi Wo (LCW) catchment in Hong Kong, China, in order to evaluate the severity of a rainstorm and flood event occurred on 11 May 2014. With the collection of rainfall data from the rain gauges near the catchment, the proxy rainfall dataset for LCW was developed. Since the time of concentration of the catchment is about 30 to 40min, this study derived rainfall intensity duration frequency (IDF) curves for 9 different durations (5min, 10min, 15min, 20min, 30min, 45min, 1h, 1.5h, and 1day) and 7 different return periods (2, 3, 5, 10, 20, 50 and 100years). Further, a hydrological model, TOPMODEL, was used to simulate streamflow process; to calibrate the model parameters, a rain gauge was set up in the catchment and a water level sensor was installed at a control cross-section of the LCW river in January 2015, and the recorded rainfall and runoff data were used to calibrate the model parameters. Using the proxy rainfall data, this study obtained the simulated streamflow for the catchment, and then derived the streamflow peaks for 7 different return periods (2, 3, 5, 10, 20, 50 and 100years). Since the time of concentration of the catchment is less than 1h, this study derived that the return periods of the rainstorm on 11 May 2014 for the durations of 30min and 1h are 2.4 and 1.9years, respectively; the return period of the daily rainfall is 9.6years. The return period of the peak flood of the event is 7.0years, and this value is between the return periods of the rainfall for the durations of the time of concentration and 1day. This study revealed that the severities of rainfall and runoff extremes are not consistent but rationally related, and the 1- and 3-day antecedent rainfalls can considerably influence flood peak severity. Overall, to achieve rational prediction of ungauged basin hydrological processes, it is fundamental to install measurement equipment and to record rainfall and streamflow data. Even though the period of the recorded data in the ungauged catchment is short, the observations are necessary for evaluating the proxy data quality, and calibrating and validating the numerical model.