The Experimental Detecting and Adjusting of the Precipitation Data Homogeneity in the Yangtze Delta

Annual and monthly precipitation data over 36 stations in Yangtze Delta are chosen in the experimental homogeneity studies. It is expected to find a method to homogenize the annual and monthly historical precipitation series over China. Two road maps are adopted to test the homogeneity of the precipitation series, one is a traditional standard normal homogeneity test (SNHT) and the other is a method based on EOF of filtered normalized monthly precipitation series which do not rely on the reference series. In the first method, the first step is to create a homogeneous reference series for each station with series from those stations that are near the checking station and have good correlations with the candidate series. Then the second step is to test the homogeneity of the candidate series with SNHT. Testing results show that some inhomogeneities exist in the precipitation data series in the selected region. Although the stations distribution is well-proportioned and the landform is relatively even in the study region, the density of the stations in the region is not enough, the references series might be questioned. Then the later references-independent method is used to check the potential discontinuities detected by SNHT. The precipitation data are normalized by calculating the cubic root of the monthly value, and the variations under inter-annual time scales are filtered. Then EOF is conducted to the fields where month is used as variables, the abrupt change points in time coefficients of the significant first eigenvectors are detected as the discontinuities of the series. Only those discontinuities which are detected by both of the two methods are regarded as real inhomogeneities. Finally, the confirmed discontinuities are adjusted in both annual and monthly series. Comparison analysis shows that the adjustment is reasonable and acceptable. Based on the earlier studies and operational practices of the network in China, a combination of the two different methods is proved to be necessary and possible, and it shows reasonability and application values in the homogeneity study of precipitation series in China, especially for the regions where the station densities are not enough. Further more, stations metadata are relied little on by these methods. In some context, the puzzles of lack of station metadata of precipitation data are avoided by these methods.