Trends, change points and spatial variability in extreme precipitation events from 1961 to 2017 in China

Extreme precipitation events vary with different sub-regions, sites and years and show complex characteristics. In this study, the temporal variations, trends with significance and change points in the annual time series of 10 extreme precipitation indices (EPIs) at 552 sites and in seven sub-regions were analyzed using the modified Mann–Kendall test and sequential Mann–Kendall analysis. Three representative (extremely wet, normal and extremely dry) years from 1961 to 2017 were selected by the largest, 50%, and smallest empirical frequency values in China. The spatiotemporal changes in the EPIs during the three representative years were analyzed in detail. The results showed that during 1961–2017, both theconsecutivewetordrydaysdecreasedsignificantly,while thenumberofheavyprecipitationdays hadnosignificant trend, and theother sevenwetEPIs increased insignificantly. Theabrupt changeyearsof the 10EPIs occurred32 and40 times from1963 to 1978and from1990 to 2016, respectively, regardlessof sub-region. The extremely dry (or wet) events mainly occurred in western (or southwestern) China, implying a higher extreme event risk. The extremely wet, normal and extremely dry events from 1961 to 2017 occurred in2016, 1997 and2011with empirical frequencies of 1.7%, 50%and 98.3%, respectively. In addition, 1998 was the second-most extremely wet year (empirical frequency was 3.7%). The monthly precipitation valueswere larger from February to August in 1998, forming amuch earlier flood peak than that of 2016. The10 EPIs hadcloseconnectionswithNormalizedDifferenceVegetation Indexes during the 12 months of 1998 and 2016. This study provides useful references for disaster prevention in China. This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/). doi: 10.2166/nh.2020.095 om http://iwaponline.com/hr/article-pdf/51/3/484/698571/nh0510484.pdf er 2021 Linchao Li Yi Li (corresponding author) Haixia Lin Ning Yao Songbai Song College of Water Resources and Architectural Engineering, Northwest A&F University, Yangling 712100, China E-mail: liyi@nwsuaf.edu.cn Linchao Li Yufeng Zou Yi Li Haixia Lin Ning Yao Songbai Song Institute of Water Saving Agriculture in Arid Areas of China, Northwest Agriculture and Forestry University, Yangling, Shaanxi 712100, China Yi Li Key Lab of Agricultural Water and Soil Engineering of Education Ministry, Northwest Agriculture and Forestry University, Yangling 712100, China De Li Liu Bin Wang NSW Department of Primary Industries, Wagga Wagga Agricultural Institute, Wagga Wagga, NSW 2650, Australia De Li Liu Climate Change Research Centre and ARC Centre of Excellence for Climate Extremes, University of New South Wales, Sydney, NSW 2052, Australia Linchao Li and Yufeng Zou contributed equally.


INTRODUCTION
Climate changes are expected to influence the occurrence of extreme precipitation events, which have attracted considerable attention (Croitoru et al. ). Extreme precipitation indices (EPIs) have been extensively used to quantitatively characterize extreme precipitation events. Some frequently used EPIs are the maximum 1-day (Rx1day) or consecutive 5-day (Rx5day) precipitation amounts, simple daily intensity index (SDII), the number of heavy (R10) or very heavy precipitation days (R20), consecutive wet days (CWDs), consecutive dry days (CDDs), very (R95p) or extremely wet days (R99p) and annual total wet-day precipitation (PRCPTOT) ( Table 1). Each index has an interior meaning. For example, the 'day-count' type has fixed thresholds (R10, R20, …,   Figure 1) (Zhao ). Detailed sub-region information is given in Table 2. Detailed descriptions of multi-year mean climatic variables in each sub-region and China are listed in Table 1 in Yao et al. (a). We excluded the site with the missing or abnormal data longer than 1 month in our analysis. In northwestern China and Qinghai-Tibet Plateau, where the density of stations is very sparse, the missing data are interpolated by the nearest site when the data missing ratio is not higher than 1% (data (n 1 À 1)r jj for jj > 1 1 þ 2 r n1þ1 1 À n 1 r 2 1 þ (n 1 À 1)r 1 n 1 (r 1 À 1) 2 for jj ¼ 1 where r jj is the sample self-correlation coefficient.
The slope of the trend (b) is estimated by Sen () as where x i and x j are the values in the ith and jth year, respectively.
A sequential Mann-Kendall analysis containing sequential progressive u(t) and backward u 0 (t) analyses was applied

Empirical frequencies
The annual PRCPTOT from 1961 to 2017 (a total of 57 years) were ranked in the descending order to compute empirical frequencies (i.e., m/(n þ 1), where m is the order and n is 57) for the seven sub-regions and mainland China.
The empirical frequencies were used to select represent year of extreme precipitation.

Wavelet analysis
A cluster of wavelet functions was used to show signal The key function was written as follows: where t is the time (year), and ψ(t) is a wavelet function that forms a cluster of functions on the timeline: where ψ a,b (t) is a sub-wavelet, a is a scale factor reflecting the wavelet length, and b is a translation factor that shows the translation of time. In this study, the multi-Morlet-wavelet was selected as a basic function.

RESULTS
Temporal variations and trends of the 10 selected EPIs

Annual variations
The annual variations of the selected 10 EPIs over 1961-2017 in seven different sub-regions and EMC (averaged from the sites) are illustrated in Figure 2. First, regional variability in all EPIs was observed, and nine EPI curves (except CDD which indicated dry conditions) generally decreased in sub-regions VII to VI, V, IV, II, III and I. The This result showed that the spatial patterns of the nine EPIs were generally similar but differed with that of CDD. The annual Rx1day ranged between 13. in EPIs were not only region-specific but also had random principles. Second, the pairs of R10 vs. R20, Rx1day vs.

Trends and significance
Six out of nine wet EPIs in sub-region I had significantly increasing trends, indicating strong wetter signals ( Due to the complexity of climatic systems, variations or  The '*' and significant at the 95% confidence level (|Zm| ! 1.96).
CDD, the average values of EPI before the change points were less than the values after the change points, and they consistently indicated the intensified EPI values of the recent two or three decades. This non-stationary feature in the EPIs intensified the complexity of extreme precipitation event identification.

Spatial variations in the 10 EPIs
Spatial distributions of long-term mean EPIs

Trends and significance
From the trend test results of the 10 EPIs ( Figure 6 and  in Figure 6(f))) were distributed in more than half of mainland China; this finding implied a generally wetter However, considering the larger PRCPTOT ( Figure 5(j)), sub-region VI may have a much higher risk of extreme wet events than sub-region III.
In general, the spatial distribution of trends in the 10 EPIs implied that sub-regions I and VII as well as the eastern part of sub-region VI would receive more Prs, be subject to larger Pr intensities, and have longer Pr days,    (Table S1).
Although the values varied for each sub-region, the three    showed general increasing areas of EPIs, which implied a higher risk of flooding in 2016.

The relationship between EPIs and atmospheric circulation indices
We analyzed the correlations between EPIs and three atmospheric circulation indices, i.e., Pacific Decadal Oscillation (PDO), Arctic Oscillation (AO) and NINO3. Figure 7 shows the spatial distribution of correlation coefficients between 10 EPIs and NINO3. The range of correlation coefficients was from À0.09 to 0.36. The correlation coefficients had a significant spatial pattern and decreased from southern to northern China for nine EPIs expect CDD which showed generally reverse pattern.
Although the correlation coefficients of 10 EPIs and other two climate indices were lower than NINO3, the spatial pattern had significant regional differences ( Figures S7 and S8).
The wavelet coherence relationships between 10 EPIs and NINO3 were analyzed using the cross-wavelet for EMC ( Figure 8)  The selected representative years were similar by different methods. In this study, we also calculated the low (10%) and high percentiles (90%) of 57 years' EPIs as the extreme dry/wet years and then selected the years when the EPIs were below and above thresholds (Table S2). Figure S4 also showed that the extreme wet and dry EPI values can clearly reflect the differences between wet and dry years.
In 2016, the area of agricultural land covered by floods and direct economic damage was larger than those in 1998, but the number of deaths from floods, the number of people affected and the number of houses destroyed were much less than during 1998 (Table 6)    The spatial pattern of EPI was similar with Figure 7, which implied that EPI was influenced by NINO3, especially in wet regions. The correlation between 10 EPIs and other two climate indices also have regional differences ( Figures S7 and S8), but the value of the correlation coefficient was lower than NINO3 (Figure 7). It implied that the NINO3 is the main factor influencing extreme precipitation events among the three selected atmospheric circulation indices. The wavelet coherence relationships between EPI and NINO3 were also stronger than between EPI and PDO/AO ( Figures S9 and S10).
Trend and abrupt change tests are useful tools in the time series analysis. In this study, a large number of time series were analyzed by the modified Mann-Kendall (MMK) test.
Although this method considers the influences of self-correlation, it is still insufficient or is applied only to some tests  In general, NINO3 was the major factor influencing extreme precipitation events among PDO, AO and NINO3. The NINO3 impacted EPI especially in southern China (sub-regions VI and VII). This study provided important references for the prevention of extreme precipitation events and heavy rains in China.