Spatial and temporal characteristics of the daily precipitation concentration index over China from 1979 to 2015

Irregular precipitation has a nontrivial influence on hydrological processes and regional agriculture. The precipitation concentration index provides convenient quantitative characterizations of precipitation variability. To explore the spatial and temporal distribution of the precipitation concentration index, the long-term concentration index (LCI) and the annual concentration index (ACI) during 1979–2015 were calculated based on the China Meteorological Forcing Dataset. The results are as follows: (1) The LCI in China ranged from 0.4571 to 0.9197, and the values between 0.6 and 0.7 accounted for 61.61% of the dataset. The highest and lowest LCI values were both recorded in Northwest China, which features low precipitation levels. Additionally, there are high LCI values (greater than 0.6) in Southeast China, which features high precipitation levels. (2) Application of the Mann-Kendall test (M-K test) and Sen’s slope revealed that more than 88% of the grids exhibited nonsignificant positive or negative ACI trends and that more than 10% of the grid ACI values exhibited positive trends, with approximately 2.8% showing significant changes at the 0.1 significance level. (3) Application of the Pettitt test revealed that approximately 11.9% of the grid ACI values exhibited an abrupt change at the 0.5 significance level, with abrupt changes occurring in 1991, 1992 and 1993, together accounting for 45.89% of all grids with abrupt changes. 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.2020.149 om http://iwaponline.com/hr/article-pdf/51/3/562/698590/nh0510562.pdf er 2020 Huihua Du Yimin Wang State Key Laboratory of Eco-hydraulics in Northwest Arid Region of China, Xi’an University of Technology, Jinhua Road 5, Xi’an 710048, Shaanxi, China Zongzhi Wang (corresponding author) Kelin Liu Liang Cheng State Key Laboratory of Hydrology-Water Resources and Hydraulic Engineering, Nanjing Hydraulic Research Institute, Nanjing 210029, China E-mail: wangzz77@163.com


INTRODUCTION
Climate change is on an upward trend in the context of global warming and has a significant effect on the global natural ecosystem and socioeconomic system (IPCC ; Chang et al. ). One of the most noticeable consequences of climate change is changes in the water cycle, with precipitation being a key aspect of this process. Changes in precipitation patterns, including intensity, amount, duration, timing and rate, may lead to anomalous weather events, such as floods and droughts (Zhang et al. b, ; Parajka et al. ; Chang et al. ). In this context, the analysis of precipitation concentration is a subject of great interest and can be used to identify irregular temporal patterns of precipitation (Vyshkvarkova et al. ; Wang et al. ). In general, a higher precipitation concentration, represented by a higher percentage of the yearly total precipitation in a few rainy days, has the potential to cause flood and drought phenomena (Li et al. ).
In recent years, the analysis of daily precipitation concentration has been a focus of research. The essence of analyzing precipitation concentration is describing the percentage of precipitation on very rainy days relative to the total precipitation over a period of time. The precipitation concentration index was presented (Martin-Vide ) based on the Gini concentration index, which relates the magnitude of precipitation events to the time period in which they occur (Serrano-Notivoli et al. ). Due to its considerable scientific and practical merits, the precipitation concentration index has been widely employed to analyze spatio-temporal patterns in many regions, such as Considering that the characteristics of precipitation concentration play a key role in watershed management, the purpose of this study is (1) to explore the spatial distribution of the precipitation concentration index and (2) to investigate the temporal variation and abrupt changes in the precipitation concentration index. In this paper, the daily precipitation concentration index in China was analyzed based on a gridded daily precipitation dataset for China with a resolution of 0.1 for the period from 1979 to 2015. Then, the Mann-Kendall test (M-K test) and Sen's slope estimator were used to determine the trend in the precipitation concentration index. Finally, the nonparametric Pettitt test was employed to identify the change point in the precipitation data.

METHODOLOGY Study area and data
China is located in East Asia and has a climate dominated by monsoon winds (Yihui & Chan ). The remarkable regional diversity of natural conditions results in spatial characteristics in the meteorological factors across China.

Precipitation exhibits spatial and temporal variability in
China because of differences in climate and complex topography (Xu et al. ; Zhang et al. ). To clarify the characteristics of precipitation concentration in association with various regional climates and geographical features, the study area is divided into ten major basins, as defined  the precipitation amount (Y) and the cumulative percentages of rainy days (X). The relationship can be expressed as follows: where a and b are estimated using the least-squares method (Martin-Vide ).
3. Based on estimates of a and b, the area S under the Lorenz curve is the definite integral of the exponential curve between 0 and 1: 4. The precipitation concentration index (CI in the following equation) can be calculated as follows: If the period is 1 year, the final calculated result is called the annual concentration index (ACI). If the period is 1979-2015, the final calculated result is called the long-term concentration index (LCI) in this study.

Mann-Kendall test and Sen's slope estimator
The M-K test is a nonparametric trend test that is applicable for analyzing nonnormally distributed data, and the sample data are not necessarily compliant with a specific distribution (Hirsch & Slack ). Using the M-K test, the concentration trends can be quantified, and the comparison of trends among regions is more intuitive.
According to this test, the null hypothesis H 0 states that the deseasonalized data (x 1 , x 2 , . . . , x n ) is a sample of n independent and identically distributed random variables (Partal & Kahya ). The procedures to calculate the test statistics, the variance of the test statistics, and the standardized test statistics are as follows: where S is the test statistic, VAR is the variance of S, Z is the standardized test statistic, n is the total number of sample data, x j and x i are the yearly mean values of years i and j, and t is the extent at any given time. If jZj > Z 1Àα=2 , the null hypothesis is rejected with a given confidence level α; namely, there is a significant trend in the time series data {X}. When α is equal to 0.05 and 0.01, Z 1Àα is equal to 1.960 and 2.576, respectively. If Z > 0, there is an increasing trend of S, which means that the sample data {X} exhibit the same trend. In contrast, if Z < 0, {X} exhibits a decreasing trend.
The slope of the trend of the data is evaluated by the nonparametric Sen's slope estimator (Sen ) and is calculated as follows: where x j and x i are data values at times j and i. The median of the ascending sequence of values Sen's slope estimator, and the confidence interval of the slope can be calculated as follows: where Z 1Àα and VAR(S) are defined as mentioned above. N is the length of the ascending series {(x j À x i )/( j À i)}, written as X, and X (M1) represents the M 1 th values in series X. Sen's slope estimator has been widely used in trend analysis of

Pettitt test
The nonparametric Pettitt test (Pettitt ) was employed to identify the change point in the precipitation data. It is a rank-based and distribution-free test used to detect signifi-

A time series of
index U τ is defined as follows: The most likely change point can be found as follows: The significance probability p associated with K τ can be calculated as follows:

Spatial distribution of annual precipitation
Based on the daily precipitation dataset, the spatial distribution of the mean annual precipitation in China during the period of 1979-2015 is shown in Figure 2. The regional average annual precipitation indices of the ten major basins were calculated from all grids, and the statistical results are shown in

Spatial distribution of LCI values
The data with daily precipitation values of !0.  Based on the ACI trend in each grid, the numbers of grids with different trends are shown in Figure 6. The results show that there are more grids with positive ACI trends than negative ones. More than 88% of the grids exhibited nonsignificant positive or negative ACI trends. More than 10% of the grid ACI values exhibited positive trends, and approximately 2.8% showed significant changes at the 0.1   were calculated in the ten major basins at the 5% and 1% significance levels and are shown in

Oscillation characteristics of ACI
The Pettitt test was applied to the 37-year ACI data to assess oscillations in the daily precipitation concentration. The distribution of the statistical parameter P of ACI derived from the Pettitt test results was calculated in each grid and is presented in Figure 8.  Based on ACIs for ten major basins, the Pettitt test was employed to assess the oscillations of daily precipitation concentration, and the results are shown in Table 4. The results showed that there are no significant abrupt changes in the LIA, HUA and PEA basins and that there are

Relations between ACI and annual precipitation parameters
Pearson's correlation coefficient (r) was used to analyze the statistical relations between the ACI values and the annual precipitation and the annual number of rainy days (daily precipitation !0.1 mm). The distribution of r at a 0.05 significance level is shown in Figure 10, and the statistical results   The differences between these ranges can be explained by the period of the climatic data. The ACI values in the Lancang River basin, located in the eastern SWR basin, range   Figure 11, and the results indicate that there is three regional flood management.
1. Flood management in the soil erosion area (red rectangular box in Figure 11). In this study, the ACI values in the YEL basin are dominated by a significant increasing trend, and grids with increasing trends are mainly located in the middle reach, corresponding to the Loess Plateau.
A similar phenomenon can be found in existing studies (Ran et al. ). In regions with severe soil erosion increase. The prevention and control of water-driven soil erosion may be the main work in the regional flood management, which can reduce the sediment input. Furthermore, based on the forecast of flood, water and sediment regulation should be developed to improve the relation between runoff and sediment.
2. Flood management in the urban area (blue rectangular box in Figure 11). If flooding occurs, there will be considerable damage in the urban area, as it is the place with a dense population, accumulation of poverty and gathering of human activities. This becomes a serious challenge for urban drainage and flood control due to the extreme precipitation events which have occurred more frequently. Therefore, it is necessary to develop an urban drainage project and to enhance non-engineering measures including flood forecasting and emergency rescue capacity building in the urban area.
3. Flood management in a remote mountainous area (yellow rectangular box in Figure 11). Floods and debris floods will happen more frequently in the remote mountainous area with the increasing trend of ACI. The to 2010, and more than 45% of the total grids had abrupt changes in 1991, 1992 and 1993.
5. Pearson's correlation coefficient was used for the analysis of the statistical relations between the ACI values and the annual precipitation and the annual number of rainy days. At the 0.05 significance level, the results indicated that the correlation between the ACI and the annual number of rainy days was greater than that between the ACI and the annual precipitation. The number of grids with negative correlations was larger than that with positive correlations, in terms of both the correlation between the ACI and the annual precipitation and the correlation between the ACI and the annual number of rainy days.
6. The increasing trend of ACI reveals that there will be more heavy precipitation events in the future. Regional flood management should be carried out according to topography, landscape characteristics and human activities, such as in soil erosion areas, urban areas and remote mountainous areas.