The Effect Of Intersite Dependence On Hosking-Wallis Homogeneity Test And Relevant Correction

Regional L-moments frequency analysis has been popular among both hydrology and meteorology researchers because of its accuracy and robustness.The identification of homogeneous regions is the foundation of regional L-moment precipitation analysis.A homogeneous region should satisfy the postulate that all data of sites in the region have one identical regional frequency distribution.Hosking-Wallis homogeneity test is widely used to inspect if a region of sites are correctly divided.But inter-site dependence of data tends to affect the accuracy of Hosking-Wallis homogeneity test by decreasing the outcome of the test,which will eventually lead to loss of accuracy of quantile estimates.This paper studied the relationship between inter-site dependence,record length and distance of gauging sites using annual maximum precipitation series of 936 sites in Jiang Xi province.And then compare it with the inter-site dependence of simulated data used in Hosking-Wallis homogeneity test.Associated with the definition of Hosking-Wallis homogeneity test and Monte Carlo experiments,the effectiveness of the test was analyzed in the presence of different degree and patterns of inter-site dependence.Through generating cross-correlated data by normalized Copula,a correction of Hosking-Wallis homogeneity test was proposed.The improvement of the corrected test was then demonstrated by comparison with the original test and former correction through a series of Monte Carlo experiments.Conclusions are as follow:1)The new test performs more accurately and robustly in presence of high regional inter-site dependence.2)This correction does not introduce any experience parameters which extends its range of application.3)This correction successfully keeps the ability of detection of heterogeneous data serious.