Spatial Precipitation Frequency Analysis Based On L-moment Approach

Frequency analysis approach has been developed for many years.Lots of researchers have done work about choosing frequency distribution functions,estimating the parameters and studying the methods of regional precipitation frequency analysis.Spatial precipitation frequency analysis is nowadays a hot and trendy issue in the hydrological field.This study improved and used two methods to implement the spatial precipitation frequency analysis,which were interpolation method and regional spatial precipitation frequency analysis,based on the L-moment approach.One method used co-kriging which considers the site elevations to interpolate the design storms of different return periods in the whole region.The other method consisted of the following steps:dividing the hydrologysimilar regions,screening the data using the discordance measure,homogeneity testing using the heterogeneity measure,distribution function selection using the goodness-of-fit measure, and finally assessing the design storms using the index-flood method.These two methods were applied in Hanjiang River Basin and 21 stations were chosen as the basic stations(Sanmenxia station is outside the Hanjiang River Basin).The annual maximum daily precipitations in each station were used as the basic data.The mann-kendall method was then used to test the time series whether they have any trends.Based on the L-moment approach,the study calculated each station's design storm of different return periods using the P-â…¢distribution function.The kriging and co-kriging methods interpolated the design storm in the whole Hanjiang River Basin.The results showed that these two interpolation methods can both give reliable results while the co-ckriging method showed better performance.Regional frequency analysis was also based on the L-moment approach,which divided the Hanjiang River Basin into some hydrology-similar regions.The results showed that the LN3 distribution function was suitable for the upstream data and the GEV distribution function was suitable for the mid-stream and downstream data.At last,for testing validity of the results,5 stations were chosen in the ungauged sub-basins,in which the design storms were calculated respectively using the two methods. Also,the results based on"Diagrams of Storm Statistical Parameters in Hubei Province" were used to compare and analyze the results based on the two methods. The final results showed that these two methods were both effective.The prediction results had higher accuracy particularly for small return periods.Each method had its merits and defects.The co-kriging method could get more steady results and the regional frequency analysis could get more accurate results,especially for small return periods.If more at site data could be obtained,the results based on each method would become more accurate.The study of co-kriging and regional frequency analysis methods can provide essential supports for project-design and decision-making.Especially,these two methods are useful for ungauged basins which are short of precipitation data.Also,the study in this thesis will show its significance in further spatial frequency analysis research.