Regionalization Of Precipitation Characteristics In China Using Wavelet-based Hierarchical Cluster Method

It is essential to understand the frequency and spatial distribution of precipitation as well as the estimation of monthly and seasonal precipitation in the large geographic area.In China,the variable climate patterns make the delineation of homogeneous precipitation regions very difficult.This study presents a wavelet-based clustering method by coupling the maximum overlap discrete wavelet transform(MODWT)and hierarchical clustering analysis(HCA)for establishing homogeneous precipitation regions.The method considers the wavelet variance of precipitation and wavelet correlation between precipitation and temperature as the feature variables for regionalization of precipitation characteristics by the HCA.The method can capture the multiscale variability of precipitation and the influences of temperature on the precipitation under different time scales by the use of the wavelet variance and correlation.Based on monthly precipitation,monthly temperature and location attributes in the period 1966-2015 at 580 selected stations,the precipitation characteristics in China are identified using the proposed clustering method,precipitation features for each region are analyzed and the teleconnection between regional monthly precipitation and AO/NAO are investigated by cross wavelet analysis.Results show that:(1)10 relatively uniform and different sub-regions are identified:1 Central region,2 Central-eastern plain region,3 Northeastern mountain region,4 Northeastern plain region,5 Southwestern Tibetan Plateau region,6 Northeastern Tibetan Plateau region,7 Southeastern region,8 Yunnan-Kweichow Plateau region,9 Northern plateau region and 10 Northwestern region.The stations in each of the identified clusters not only share similar precipitation features but also form a geographically contiguous region.(2)The wavelet variance of regional average monthly precipitation anomaly series across different wavelet scales for the ten regions shows that there are some clear differences.The relationship between precipitation and temperature under different time scales shows that the wavelet correlations between regional average monthly precipitation and temperature(Tmean,Tmax and Tmin)anomaly series across different wavelet scales for each of ten formed regions are different.The result of a two-sample Kolmogorov-Smimov(KS)test shows that the distributions of regional average seasonal precipitation are not the same at any two regions across four different seasons at the 5%significance level.The mean and standard deviation of regional annual and seasonal precipitation during the period 1966-2015 also indicate that the variation in regional average precipitation shows marked differences across ten delineated regions.Time series of average annual precipitation in each of the regions from 1966 to 2015 indicate that the temporal changes of the average annual precipitation are different from region to region.The intra-cluster dispersion of precipitation features regarding the seasonal and annual precipitation magnitude and the percentage of monthly precipitation in the identified regions show that the precipitation characteristics of the ten regions in China are significantly different from each other.(3)The teleconnection between monthly precipitation for each region and AO/NAO show that each region shows significant correlation with both AO/NAO in some certain time scales.The significant resonance periods for AO and other regions except for Region 2 are mainly concentrated in 64-128 month time scale,while the significant resonance periods for AO and Region 2 are mainly concentrated in>128 month time scale.The significant resonance periods for NAO and Region 1 are mainly concentrated in>128 month time scale,while the significant resonance periods for NAO and other regions are mainly concentrated in<60 month time scale.The optimum response time of monthly precipitation in different regions towards AO and NAO is different.Monthly precipitation in Region 6 towards AO and monthly precipitation in Region 1 towards NAO both shows timely responses,while the optimum response time of monthly precipitation in other regions towards AO and NAO are both longer.