Spatio-Temporal Variations in Precipitation Extremes in the Endorheic Hongjian Lake Basin in the Ordos Plateau, China

: Precipitation extremes have important implications for regional water resources and ecological environment in endorheic (landlocked) basins. The Hongjian Lake Basin (HJLB), as the representative inﬂow area in the Ordos Plateau in China, is su ﬀ ering from water scarcity and an ecosystem crisis; however, previous studies have paid little attention to changes in precipitation extremes in the HJLB. In this study, we investigated the spatio-temporal variations of the core extreme precipitation indices (i.e., PRCTOT , R99p , Rx1day , Rx5day , SDII , R1 , R10 , CWD , and CDD ) recommended by the Expert Team on Climate Change Detection and Indices (ETCCDI), and analyzed the climatic dry–wet regime indicated by these extreme indices during 1960–2014 in the HJLB. The results show that the nine extreme indices had large di ﬀ erences in temporal and spatial variation characteristics. All the nine extreme precipitation indices showed a large ﬂuctuation, both in the whole period and in the three detected di ﬀ erent sub-periods, with variation magnitudes of 13%–52%. Most extreme indices had non-signiﬁcant downward trends, while only the consecutive wet days ( CWD ) had a signiﬁcant upward trend. The eight extreme wet indices increased from northwest to southeast, while the consecutive dry days ( CDD ) had the opposite change direction. Each index had a di ﬀ erent trend with di ﬀ erent spatial distribution locations and areas. The nine extreme indices revealed that the climate in the HJLB has become a drought since the early 1980s. This was speciﬁcally indicated by all four extreme precipitation quantity indices ( PRCTOT , R99p , Rx1day , Rx5day ) and the extreme intensity index ( SDII ) declining, as well as the number of heavy precipitation days ( R10 ) decreasing. When the dry–wet variations was divided into the di ﬀ erent sub-periods, the climatic dry–wet changes of each index demonstrated more inconsistency and complexity, but most indices in the ﬁrst sub-period from 1960 to the late 1970s could be regarded as a wet high-oscillation phase, the second sub-period after the early 1980s was a relatively dry low-oscillation phase, and the third sub-period after the late 1990s or early 21st century was a dry medium-oscillation phase. It is worth noting that most extreme indices had an obvious positive linear trend in the third sub-period, which means that in the last 20 years, the precipitation extremes showed an increasing trend. This study could provide a certain scientiﬁc reference for regional climate change detection, water resources management, and disaster prevention in the HJLB and similar endorheic basins or inland arid regions.


Introduction
An endorheic (hydrologically landlocked) basin is a closed drainage basin lacking any water outflow to other external bodies of water, such as rivers or the ocean [1,2]. Endorheic basins, spatially concurring with arid and semi-arid climate regions, mostly with basin outlets ending in endorheic lakes [3], cover approximately 20% of the Earth's land surface area [4]. According to

Study Area
The Hongjian Lake Basin (HJLB), with a longitude of 109 • 30 38" E-110 • 5 46" E and a latitude of 38 • 51 10" N-39 • 20 20" N, is located in the Ordos Plateau, China [14,42]. The HJLB is also located in the middle regions of the Yellow River Basin, and closest to the Tuwei River, which is one of the main outflow tributaries in the middle reaches of the Yellow River ( Figure 1). The HJLB spans the two provinces, and its western and northwestern parts are in the Inner Mongolia Autonomous Region, where the eastern and southeastern parts belong to Shaanxi Province. The HJLB is sparsely populated, and there are only four counties within the HJLB (Figure 1).
The HJLB is an inflow zone, and its surface water does not flow into the Yellow River system. Seven seasonal rivers flow into the central Hongjian Lake [14]. There are four seasonal rivers in the Inner Mongolia Autonomous Region (i.e., the Maogaitu River, the Zhashake River, the Mudushili River, and the Songdao River) flowing into the lake from the west and northwest, and three seasonal rivers in Shaanxi Province (i.e., the Qibosu River, the Erlintu River, and the Haolai River) flowing into the lake from the south, southeast, and north ( Figure 1). In this study, based on hydrological analysis, the HJLB covers an area of ≈1513.10 km 2 under the projected coordinate system projection of Beijing_1954_GK_Zone_19N, with an elevation from 1337 m to 1640 m.

Study Area
The Hongjian Lake Basin (HJLB), with a longitude of 109°30′38'' E-110°5′46'' E and a latitude of 38°51′10'' N-39°20′20'' N, is located in the Ordos Plateau, China [14,42]. The HJLB is also located in the middle regions of the Yellow River Basin, and closest to the Tuwei River, which is one of the main outflow tributaries in the middle reaches of the Yellow River ( Figure 1). The HJLB spans the two provinces, and its western and northwestern parts are in the Inner Mongolia Autonomous Region, where the eastern and southeastern parts belong to Shaanxi Province. The HJLB is sparsely populated, and there are only four counties within the HJLB (Figure 1).
The HJLB is an inflow zone, and its surface water does not flow into the Yellow River system. Seven seasonal rivers flow into the central Hongjian Lake [14]. There are four seasonal rivers in the Inner Mongolia Autonomous Region (i.e., the Maogaitu River, the Zhashake River, the Mudushili River, and the Songdao River) flowing into the lake from the west and northwest, and three seasonal rivers in Shaanxi Province (i.e., the Qibosu River, the Erlintu River, and the Haolai River) flowing into the lake from the south, southeast, and north ( Figure 1). In this study, based on hydrological analysis, the HJLB covers an area of ≈1513.10 km 2 under the projected coordinate system projection of Beijing_1954_GK_Zone_19N, with an elevation from 1337 m to 1640 m. The HJLB is a typical inland semi-arid and arid region. According to the daily meteorological data collected in this study (see Section 2.2.1), the average annual precipitation during 1960-2014 is 393.0 mm for the whole basin. The rainy season is concentrated from June to September, and precipitation in the rainy season accounts for 75.9% of the annual precipitation. The average annual The HJLB is a typical inland semi-arid and arid region. According to the daily meteorological data collected in this study (see Section 2.2.1), the average annual precipitation during 1960-2014 is 393.0 mm for the whole basin. The rainy season is concentrated from June to September, and precipitation in the rainy season accounts for 75.9% of the annual precipitation. The average annual surface water evaporation is generally 964-1124 mm, increasing from southeast to northwest. The average annual temperature is 7.4 • C. The coldest month is January, with an average temperature of −10.0 • C, while the hottest month is July, with an average temperature of 22.4 • C. Grassland, sand land, woodland, cropland, and lake water surface are the main land-use types, and grassland and shrubs are the primary vegetation types in the HJLB [14,39]. The HJLB has an extremely fragile ecological environment. The central Hongjian Lake area has shrunk severely since 1999 due to climate change and intensified human activity [14].

Daily Precipitation
The time series of daily precipitation records are the necessary basic data for calculating extreme indices. The China Meteorological Data Service Center (CMDC), an upgraded system of the meteorological data-sharing platform developed by the Climate Data Center of the National Meteorological Information Center of the China Meteorological Administration (http://data.cma.cn/en/ ?r=site/index), makes a dataset of daily climate data, including such daily data as pressure, temperature, precipitation amount, evaporation amount, relative humidity, wind direction, wind speed, and sunshine duration from more than 2000 national weather stations in China, and has done so since the 1950s. After strict repeated verifications and quality control by the CMDC, all suspicious and incorrect data have been rechecked and rectified manually, the missing digitalized data are supplemented, and eventually all element data are flagged with quality control codes. The availability of various weather elements in the daily dataset is generally above 99%, and the correct data rate approaches nearly 100%. However, in this study, the study area (HJLB) belongs to a typical scarce data region, with no national weather stations within the HJLB. There are only 10 national weather stations around the HJLB (Figure 1), and their basic information is shown in Table 1. From the CMDC's daily climate dataset, we obtained the daily precipitation data during 1960-2014 at the 10 national weather stations around the HJLB. On the basis of this daily precipitation data, we further applied the thin plate spline method [43,44] in ANUSPLIN version 4.4 software (https://fennerschool.anu.edu.au/research/products/anusplin, Australian National University, Canberra, Australia) [45] to interpolate the site's daily data into grid data with a resolution of 0.05 • × 0.05 • to better obtain the basic daily precipitation and perform the spatial analysis of extreme precipitation within the HJLB.

Extreme Precipitation Indices
In the study of climate change detection and extreme events, the 27 core climate change indices recommended by the Expert Team on Climate Change Detection and Indices (ETCCDI) (http://etccdi. pacificclimate.org) [26,46,47] are most widely used around the world. Because not all the indices are meaningful in the HJLB, we finally chose the nine extreme precipitation indices from the core indicators, and their exact definitions are shown in Table 2. Based on the definitions of indices and the obtained daily precipitation data, we calculated the nine extremes precipitation indices using the MATLAB R2017b (9.3.0.713579) software (The MathWorks, Inc., Natick, MA, USA). According to the attribute and unit of these extreme indices, the nine indices can be further divided into three classes: the quantity indices expressed in mm units (i.e., PRCTOT, R99p, Rx1day, and Rx5day), the intensity index expressed in mm/day (i.e., SDII), and the time indices expressed in days (i.e., R1, R10, CWD, and CDD).

Basic Statistical Analysis
In this study, we selected the mean value (M), standard deviation (Sd), coefficient of variation (Cv, Cv = Sd/M), and slope of the linear trend (S) as basic statistics in order to help analyze the changes in each extreme precipitation index during the study period of 1960-2014 and the sub-periods. These basic statistics were calculated using IBM SPSS Statistics Version 19 software (https://www.ibm.com/ products/spss-statistics).

Trend Analysis
The trends of extreme precipitation indices in the period of 1960-2014 and in the sub-periods were determined using the linear regression method. Then, the nonparametric Mann-Kendall test [48,49], as the widely used method for trend detection due to its robustness for non-normally distributed data [20,50,51], was applied to detect the significance of trends in the indices. For a given time series X (X 1 , X 2 . . . X n ); their ranks are R 1 , R 2 . . . R n ; and the Mann-Kendall rank statistic (S) is calculated as: where the symbolic function is defined as: An increasing or decreasing trend in the time series is indicated by a positive S and a negative S, respectively. If the null hypothesis H 0 (i.e., no trend in the data) is true, then S can be assumed to be approximately normally distributed with: The Z score of S is calculated as: In this study, in order to statistically distinguish the differences in the trends of the indices, as well as the differences in the spatial trends of each indicator, and to concurrently ensure that there is at least one indicator with a significant change trend, the corresponding significant level was set as p = 0.1, which means that a positive Z score larger than 1.64 indicated a significant increasing trend, while a negative Z score lower than −1.64 indicated a significant decreasing trend. When the absolute value of the Z score was less than 1.64, it meant that the trend was not significant.

Turning Points Detection
The non-parametric Pettitt test [52] is commonly used to detect the turning points in the extreme precipitation indices [53][54][55]. For a given time series composed of X 1 , X 2 , . . . , X n , the Pettitt test considers the series as two samples X 1 , X 2 . . . , X t and X t+1 , X t+2 , . . . , X n , and the Mann-Whitney statistic U t,n is expressed as: where the symbolic function sign(X t − X j ) is the same as mentioned in Equation (2), and n is the data set length. Then the statistic K t,n and the corresponding significance level are defined as: p 2 exp −6(K t,n ) 2 / n 3 + n 2 If p < 0.05, a significant turning point exists.

Spatial Analysis
Based on every grid's daily precipitation data with a resolution of 0.05 • × 0.05 • , as well as the definition of extreme precipitation indices (Table 2), we first calculated the nine extreme precipitation indices and the change trends of the indices for every grid, using the MATLAB R2017b software. Second, using the georasterref function and the geotiffwrite function in MATLAB software (https://ww2. mathworks.cn/help/map/ref/geotiffwrite.html?s_tid=srchtitle), we converted the gridded calculation results (i.e., extreme precipitation indices and their change trends) into data in tiff format using the coordinate system of GCS_WGS_1984. Third, we imported the targeted data in tiff format into ArcGIS Desktop 10.5 software (https://desktop.arcgis.com/en/). We selected the format of a stretched color ramp map (Layer Properties\Symbology\Stretched\Color Ramp) to symbolize and show the spatial results. To display the numerical distribution of the results more intuitively, using the Contour geoprocessing tool integrated into the Surface of ArcToolbox (Toolboxes\System Toolboxes\Spatial Analyst Tools.tbx\Surface\Contour), we further applied the format of contours to jointly show the spatial distributions of grid extreme indices and change trends. Lastly, as for the spatial change trends of indices, we needed to find the area of positive trends and negative trends. We used the Extract by Attributes tool integrated into Extraction within ArcToolbox (Toolboxes\System Toolboxes\Spatial Analyst Tools.tbx\Extraction\Extract by Attributes) to extract the targeted raster with positive trends (where clause is "VALUE > 0"), as well as the targeted raster with negative trends (where clause is "VALUE < 0"). When we got the positive and negative rasters, since each raster size was the same, we counted the number of rasters separately (Layer Properties\Symbology\Classified\Classify\Classification Statistics\Count) to get the required raster area. positive and negative rasters, since each raster size was the same, we counted the number of rasters separately (Layer Properties\Symbology\Classified\Classify\Classification Statistics\Count) to get the required raster area.

Temporal Variation Characteristics of Extreme Precipitation Indices
The    Figure 3 shows the three different change sub-periods of the nine extreme indices divided by the detected turning points. Although the nine indices had different turning points at different years, their variations could be divided into three sub-stages. Table 3 listed the corresponding basic statistical results in the three sub-periods. Taking the first sub-period as the benchmark, the percentage of relative change in the mean value in the second or third sub-periods (i.e., Rc2 or Rc3 in Table 3) Table 3). Generally, most indices had the largest linear slope in the third and first sub-periods, and the lowest S in the second sub-period.
Water 2019, 11, x FOR PEER REVIEW 8 of 19 Figure 3 shows the three different change sub-periods of the nine extreme indices divided by the detected turning points. Although the nine indices had different turning points at different years, their variations could be divided into three sub-stages. Table 3 listed the corresponding basic statistical results in the three sub-periods. Taking the first sub-period as the benchmark, the percentage of relative change in the mean value in the second or third sub-periods (i.e., Rc2 or Rc3 in Table 3 Table 3). Generally, most indices had the largest linear slope in the third and first sub-periods, and the lowest S in the second sub-period.

Spatial Variation Characteristics of Extreme Precipitation Indices
The spatial distribution of the nine extreme precipitation indices during 1960-2014 in the HJLB is shown in Figure 4. Except for the consecutive dry days (CDD), the other eight indices (PRCTOT, R99p, Rx1day, Rx5day, SDII, R1, R10, CWD) had similar spatial variation characteristics, namely, they increased from northwest to southeast. Meanwhile, the CDD had the opposite change direction, namely, they declined from northwest to southeast. In detail, as for the four quantity indices, the range of variations (i.e., minimum-maximum) in the PRCTOT, R99p, Rx1day, Rx5day was 351.9-392.4 mm, 110.1-116.5 mm, 41.0-43.0 mm, and 63.5-70.4 mm, respectively. For the intensity index, the SDII increased from 6.2 mm/day in the northwest to 6.5 mm/day in the southeast of the study area. For the four time indices, the min-max values of R1, R10, CWD, and CDD were 55.1-61.2 days, 9.5-10.7 days, 5.1-5.6 days, and 61.1-64.9 days, respectively.
Water 2019, 11, x FOR PEER REVIEW 10 of 19

Spatial Variation Characteristics of Extreme Precipitation Indices
The spatial distribution of the nine extreme precipitation indices during 1960-2014 in the HJLB is shown in Figure 4. Except for the consecutive dry days (CDD), the other eight indices (PRCTOT, R99p, Rx1day, Rx5day, SDII, R1, R10, CWD) had similar spatial variation characteristics, namely, they increased from northwest to southeast. Meanwhile, the CDD had the opposite change direction, namely, they declined from northwest to southeast. In detail, as for the four quantity indices, the range of variations (i.e., minimum-maximum) in the PRCTOT, R99p, Rx1day, Rx5day was 351.9-392.    Figure 5 shows the spatial distribution of the change trends of the nine extreme precipitation indices during 1960-2014 in the HJLB. Table 4 shows the area and area ratio of positive trends and negative trends. The spatial variation in the change trends of indices was more complicated and variable than the spatial distribution of the indices themselves. Each index had different spatial trends with different area and area ratios, and the specific spatial locations of the highest and lowest values were also different. For example, the changing trend of PRCTOT ranged from −0.23 mm/year to 0.04 mm/year. The negative trend of PRCTOT was distributed over most regions, while the positive trend was located in the eastern regions ( Figure 5). The area of negative trends and positive trends was 1507.53 km 2 and 141.22 km 2 , accounting for 90.67% and 9.33% of the total area, respectively ( Table 4). Trends of Rx5day were negative, ranging from −0.40 mm/year in the west region to −0.19 days/year in the eastern region. All the trends of R1 were positive, ranging from 0.02 days/year in the west region to 0.08 days/year in the north and northeast regions. In addition, there was a difference in the significance of the spatial variation trend of each index, indicated by the value of the Mann-Kendall test's Z statistic at a significance level of p = 0.1 (i.e., Z 0.1 stands for the significant positive trends with the Z larger than 1.64). Only consecutive wet days (CWD) had significant positive change trend points or sub-regions in the south, southeast, and east regions ( Figure 5). Although the spatial change trend of the remaining eight indices was different, the trend was not significant.

Precipitation Extremes and Climatic Dry-Wet Regime Compared with Other Studies
Although there are few studies on precipitation extremes in our study area (HJLB), in the neighborhood or in some typical wider regions in China (e.g., Loess Plateau, Yellow River basin, the northwest arid region, and Tibetan Plateau), there are still some related studies that can be used for comparison regarding precipitation extremes indices, as recommended by the ETCCDI and widely used worldwide. Table 5 shows the trends of precipitation extremes from this study and other studies in some typical regions in China. As for the change trends of the same index, a large difference did exist in the magnitude, direction, and significance of trends in these different regions. These differences, to a large extent, could reveal the complexity and spatial-temporal differences of changes in precipitation extremes. The possible causes of these differences include data sources, data sequence length, regional size, regional climate and geographic conditions, regional spatial heterogeneity, and trend analysis methods. It is very meaningful to further analyze the differences and underlying mechanisms of precipitation extremes in different regions. Furthermore, endorheic basins and related lakes are sensitive to climate change, responding rapidly to environmental changes [4,13]. Variations in the climatic dry-wet regime can directly affect the state of water deficit in these water-scarce endorheic basins. Therefore, it is necessary to conduct further discussions on the climatic dry-wet regime in the endorheic HJLB and adjacent regions. The nine selected extreme precipitation indices can be used to reflect the changes in the climatic dry-wet regime in the HJLB. The results of this study (Figures 2 and 3, Table 3) show that, overall, the climate in the Hongjian Lake Basin during 1960-2014 developed toward aridification. The specific performance was that all four extreme precipitation quantity indices (PRCTOT, R99p, Rx1day, Rx5day) declined, and the extreme intensity index (SDII) also slightly decreased. Although the number of wet days (R1) and the consecutive wet days (CWD) slightly increased, the number of heavy precipitation days (R10) decreased. Some studies on annual precipitation and precipitation extremes in the neighborhood or a wider region, for example, Li et al. [59] used daily precipitation data during 1957-2007 at Shenmu meteorological station near to HJLB (i.e., the station labeled as number 10 in Figure 1) to analyze climate change, showing that the precipitation in the HJLB in the past 50 years experienced a slight increase and then decreased to become a relatively dry climate. Sun et al. [56] studied changes in precipitation extreme indices recommended by the ETCCDI in the Loess Plateau during 1960-2013, showing that the changes of the rapid warming and consecutive drying in the Loess Plateau is indicated by most extreme indices (exclude PRCTOT) ( Table 5). Although there are differences relative to these findings, to some extent, it could reflect that the climate in HJLB and adjacent regions is overall changing towards aridification. However, when the climatic dry-wet variations was divided into the three sub-periods, it became more complicated and varied than that during the whole period of 1960-2014. Generally, most indices in the first sub-period were a wet high-oscillation phase with a relatively high mean value and having the largest variation coefficient. Compared with that in the first sub-period, the dry-wet state in the second and third sub-periods had four kinds of combinations, namely drier-driest (Rx5day, SDII, R10), wetter-drier (R99p, Rx1day), drier-wetter (PRCTOT, R1), and wetter-wettest (CWD, CDD). It could be seen that the climatic dry-wet changes of each index in the different three sub-periods demonstrated more inconsistency, and this inconsistency was masked in the overall trend. Therefore, special attention needs to be paid to the inconsistency of climate change in sub-phases.

Implications in the Study
As the main source of water resources and the key factor for the ecological environment in the closed HJLB, variations in precipitation and its extremes have some important implications for water resource management, ecological protection, and disaster prevention.
First, our results showed that the extreme precipitation indices fluctuated greatly over time, and heterogeneity was clear in the spatial distribution. The central Hongjian Lake's four main seasonal recharging rivers are mainly distributed in the northwestern part of the closed basin (Figure 1), while the northwest region was the drier area of the basin, as indicated by these extreme indices. These obvious differences in time and space are very detrimental to the replenishment and effective use of precipitation. Second, the climate in the HJLB, as indicated by the nine extreme precipitation indices, became drier after the late 1970s and early 1980s, which means that the water resources in the HJLB have decreased. This overall drying climate will inevitably exacerbate the already existing water stress, which puts more stringent requirements and challenges on local water resources management. Third, the central Hongjian Lake in the HJLB plays an important role in protecting the biological and ecological environment. However, some studies [9,14] have revealed that the shrinkage of Hongjian Lake due to climate change and intensified human activities has seriously threatened the survival of rare species. How the gradual drying of the climate affects the water supply of lakes and the survival of the species is an important scientific issue that deserves further study. Lastly, our results revealed that although the climate was generally dry throughout the study period , when the linear trends of the indices were examined separately in the three sub-periods, it was found that: the index value in the third sub-period was smaller than that in the first sub-period, the linear change trends of the six indices in the third sub-period was greater than the value in the first sub-period, and the linear trends of most extreme indices (except for CDD) clearly increased in the third sub-period after the late 1990s ( Figure 3 and Table 3); this means that the climate change trends were generally enhanced and have created humidification in the last 20 years. On the one hand, these kinds of climate change trends have a certain positive effect on the precipitation recharge into the endorheic basin, and on the other hand, it will increase the risk of potential rainstorm damage. There are still great risks and challenges for the extreme climate disaster prevention in the HJLB.

Uncertainty in this Study
First, there is a certain degree of uncertainty in the division of sub-periods of variations in the indices. The turning points were detected using the Pettitt test [52]. Some indices had the other possible turning points at the year 1966 (SDII in Figure 3) or year 2011 (PRCTOT, Rx5day in Figure 3); however, considering that the length of the sub-period should not be too short, we did not take the 1966 or 2011 years as turning points. The further combined use of multiple abrupt point detection methods, such as the cumulative anomaly method [9,60], will be more conducive to accurate detection of the actual turning point.
Second, due to the large fluctuation of the indices, only a linear trend was used to represent the overall trends of the extreme indices, which will cause a certain degree of uncertainty in the quantitative description of the trend size. Subsequent attempts will be made to more accurately quantify the fine trend of the index using a non-linear approach, such as polynomial fitting.
Lastly, the Hongjian Lake Basin (HJLB) is located in a sparsely populated region, lacking enough meteorological stations within the basin, which will cause a certain degree of fine mapping of the spatial distribution characteristics inside the study area. We suggest that it is necessary to add some rainfall stations inside the HJLB in the future.

Conclusions
In this study, we used the nine extreme precipitation indices recommended by the Expert Team on Climate Change Detection and Indices (ETCCDI) to quantify the spatio-temporal extreme climate changes during 1960-2014 in the endorheic Hongjian Lake Basin (HJLB), as well as investigate the corresponding climatic dry-wet regime. We found that, both in the whole period of 1960-2014 and in the three different sub-periods, all nine extreme precipitation indices (i.e., PRCTOT, R99p, Rx1day, Rx5day, SDII, R1, R10, CWD, and CDD) had different temporal variation characteristics, with a total fluctuation magnitude of 13%-52%. Most indices had a non-significant change trend, and only the CWD had a significant positive trend. The large fluctuation, the different change direction, and the magnitude of the change trends jointly demonstrated the complexity of precipitation extremes change in the HJLB. Eight indices (PRCTOT, R99p, Rx1day, Rx5day, SDII, R1, R10, CWD) had similar spatial distribution characteristics, namely, they increased from northwest to southeast, while the CDD had the opposite change direction. Each index had a different trend in some different regions with different areas and area ratios. The nine extreme indices revealed that the climate in the HJLB has become a drought since the early 1980s. This was specifically indicated by all four extreme precipitation quantity indices (PRCTOT, R99p, Rx1day, Rx5day) and the extreme intensity index (SDII) declining, as well as the number of heavy precipitation days (R10) decreasing. When the dry-wet variations disintegrated into the different sub-periods, the climatic dry-wet changes of each index demonstrated more inconsistency and complexity, but most indices in the first sub-period from 1960 to the late 1970s could be regarded as a wet high-oscillation phase, the second sub-period after the early 1980s belonged a relatively dry low-oscillation phase, and the third sub-period after the late 1990s or early 21st century was a dry medium-oscillation phase. However, it is worth noting that most extreme indices had an obvious positive linear trend in the third sub-period, which means that in the last 20 years, the precipitation extremes showed an increasing trend. This changing trend will not only have a certain effect on the precipitation recharge into the endorheic basin but also increase the risk of rainstorm disaster damage.