Discussion On Method Of Delineation Of Homogeneous Region In Regional L-Moments Analysis

The flood frequency calculation is important for the estimation of the design standard of flood control in China, the theoretical basis for calculation is a hypothesis that these hydrological extreme events are all independent samples in a same population of random events, and making hydrological events hydrological frequency calculation of a long enough hydrological series(flood or precipitation). Studies show that the flood are caused by rainstorm in most regions of China, the design flood is indirect calculated by the rainstorm frequency analysis and calculation, that is the most common and effective method. Regional L-moments Analysis(RLMA)is the most advanced method for hydrologic frequency analysis, which has been widely used for precipitation frequency analysis in the United States and some other countries since the1990s. Recently in China, the RLMA has been successfully applied to precipitation frequency analysis in the Taihu Lake Basin(TLB)as a pilot study.In United States and China, many studies show that the results of regional frequency analysis for precipitation quantile are more accurate and reliable than single station. However, it’s difficult to divide the study area into several or dozens of hydrometeorological region. Therefore, how to define the "Hydrological Region" is crucial for a regional frequency analysis and calculation.The Annual Maximum Series at96stations with20-year and longer of rainfall data in the TLB in East-China were used for the study. And this study discusses different partition methods, using hydrometeorological way partitioning method and the traditional clustering analysis method, respectively, to divide the same study area.The results show that the partition results of hydrometeorological way are always better than that of the traditional clustering analysis method in discordancy, heterogeneity and accuracy of100-y frequency estimates using Regional L-moments Analysis. The difference is that hydrometeorological way focuses on the "Identical" distribution of hydrologic frequency curve, and, clustering analysis, considering the similarity in mathematical statistics, is inappropriate for the actual hydrological features. In this paper, we take the partition results of hydrometeorological way as example, calculate the quantile estimation with high accuracy and reliability for different return periods in hydrological Region by Regional L-moments Analysis.However, due to the lack of the rainfall stations with sparse spatial distributions as well as the short data length, the precipitation quantiles show a spatial discontinuation between quantiles somewhere around some areas in between the homogeneous regions. It is also an unsettled problems in updates to rainfall frequency atlas of the United States made by National Oceanic and Atmospheric Administration(NOAA)Hydrometeorological Design Studies Center(HDSC). This study puts forward a meaningful and simple correction method based on spatial interpolation to solve the discontinuity by adjusting the draft quantiles. The results of adjusted precipitation quantiles that had been obtained via the RLMA for the TLB exhibit better reliablity expressed in relative error of the empirical frequency to the theoretical exceedance probability of the data, and a better spatial smoothing pattern shown as well.