Spatiotemporal Characteristics of Extreme Precipitation Events in Central Asia: Insights from an Event-Based Analysis

Abstract

Extreme precipitation events, increasingly driven by climate change, are becoming more frequent and pose significant challenges to both the ecological environment and human society. Using the MSWEP data, this study constructed eight event-based extreme precipitation indicators so as to systematically analyze the spatiotemporal characteristics and dominant types of extreme precipitation across Central Asia and its three sub-regions from 1979 to 2023. The results revealed the following: (1) Extreme precipitation events exhibit a pronounced spatial preference for high-altitude areas, with the total number of events reaching up to 698 in these regions. (2) From 1979 to 1991, the frequency of extreme precipitation events has decreased in Central Asia (by 1.742 events per 13 years), while their duration has however increased (by 0.52 days per 13 years). The period from 1992 to 2009 experienced the most significant and widespread decline in the magnitude of extreme precipitation indicators. In contrast, from 2010 to 2023, all indicators—except for the event frequency (EF) and event intensity (EI)—have shown rising tendencies across the region. (3) Regarding the dominant event types, based on the proportion of extreme precipitation frequency across areas, the Southwestern Desert (SD) and northern Kazakhstan (NK) regions are characterized by a more prominent combination of rear-peak (TDP2) and front-peak (TDP1) events, whereas the southeastern mountains (SM) region is rather dominated by a combination of rear-peak (TDP2) and balanced-type (TDP3) events. (4) The EF and event duration (ED) are strongly associated with the Digital Elevation Model (DEM) and Aridity Index (AI). The spatial patterns of EF and ED are closely linked, with the sub-humid and mountainous regions demonstrating the highest frequency and longest duration of extreme precipitation events.

1. Introduction

Due to their strong suddenness, short duration, and large destructive power, extreme precipitation events (EPEs) have always been the key focus of disaster meteorology research against the background of climate change. Secondary disasters such as mountain floods and debris flows caused by extreme precipitation frequently happen, which have a serious effect on the ecological environment, infrastructure, and social and economic systems. Therefore, systematic depictions of the temporal and spatial distribution patterns and change characteristics are of great significance for regional disaster prevention and mitigation and adaptation to climate change [1,2,3].

At present, the majority of the research on extreme precipitation at home and abroad is carried out based on a series of precipitation indices proposed by the Expert Team on Climate Change Detection and Indices (ETCCDI). It mainly covers aspects such as the intensity, frequency, and duration of the precipitation. Relevant studies have already explored the spatiotemporal distribution patterns and trend evolution of indicators like R95pTOT, SDII, and CWD in different regions by calculating those [4,5,6]. A further analysis of its relationship with potential driving factors such as the atmospheric circulation and surface conditions will also prove to be essential [7,8,9]. For example, different scholars analyzed this from different time scales. Darwish et al. [10], Maria Douka et al. [11], and Zhang et al. [12] investigated the capture of hourly EPEs in several regions. Shang et al. [13] analyzed the spatial distribution and temporal variation characteristics of the EPEs in the study area on a daily scale. These studies provide important references in revealing the characteristics of the extreme precipitation and its climate responses. However, most traditional index methods are based on statistical methods with fixed time scales (such as days, months, and years), and it is rather complicated to reflect the process continuity and dynamic changes in the precipitation. We encountered considerable limitations especially when depicting the development stages, continuous structure, and intensity evolution of the extreme events [14].

In response to this issue, some scholars have begun to attempt to conduct research on the spatiotemporal characteristics of the EPEs during recent years from the perspective of the event process, analyzing precipitation as a continuously evolving physical process and extracting the full-process characteristics of the precipitation events based on the identification threshold. This assumption is in line with the objective situation of the actual continuity of the precipitation. For instance, She et al. [14] considered the continuity and threshold of extreme precipitation, proposed the concept of event-based extreme precipitation and the statistics of the characteristics of the EPEs, and also verified the superiority of the event-based EPEs using the Han River Basin in China. Wu et al. [15] examined the spatiotemporal distribution characteristics of the EPEs in China based on the definition of the EPEs; Li et al. [16] oriented towards the processes of the EPEs; and this paper explores the spatiotemporal characteristics of EPEs in the Haihe River Basin and reveals their circulation characteristics. Malla et al. [17] analyzed the multi-day EPEs in the Hindu Kush–Himalayas region, identified the temporal distribution patterns, and looked into their characteristic trends. This type of research has substantially enhanced the ability to characterize the temporal structure and regional heterogeneity of extreme precipitation and offers more process-aware tool support for risk assessments.

Located in the hinterland of the Eurasian continent, Central Asia is one of the largest inland arid regions in the world, with a dry regional climate, scarce precipitation, highly stressed water resources, and fragile ecosystems, and seems extremely vulnerable to extreme climate events [18]. Historical data show that since the 1980s, the annual precipitation has generally been on the rise in Central Asia, especially during the 1990s, which was the wettest period in nearly half a century [19]. Against the background of global warming, the frequency of precipitation anomalies and extreme events has been considerably augmented at the regional scale [20,21,22,23,24]. Over the past few decades, severe flood events have frequently taken place in Central Asia, causing huge economic losses and serious ecological damage [25]. In the spring of 2017, for instance, Kazakhstan suffered from severe floods due to heavy rainfall, forcing thousands of people to evacuate and inducing multiple rivers to overflow, affecting many areas in northern Kazakhstan [26]. In June 2024, extreme precipitation happened in the northern mountainous areas of Kyrgyzstan, triggering flash floods and mudslides. This generated severe damage to the infrastructure and had a large influence on the agricultural production and the residents’ lives [27]. Studies denote that the increasing trend of extreme precipitation in Central Asia might even further intensify in the future [28].

Compared with climate variables (e.g., extreme temperatures), however, systematic research on the extreme precipitation in Central Asia still remains largely insufficient. The existing literature mainly focuses on simulation and prediction based on the index method [29], an analysis of the response relation of climate factors [30,31], and the applicability evaluation of various precipitation datasets in the monitoring of extreme events observed in the Central Asian region [32,33], despite the fact that these studies have already achieved certain outcomes in revealing the changing trends, driving mechanisms, and data availability of extreme precipitation, the majority of which still rely on the observation data from the ground meteorological stations. The number of stations in the Central Asian countries is limited and the spatial distribution is rather sparse. For example, Zhang et al. [34] examined the modifications in the extreme precipitation indicators based on 22 weather stations in five Central Asian countries and Zhang et al. [23] further expanded to include 55 stations in Xinjiang, but the overall coverage and resolution still did not suffice to comprehensively describe the precipitation process at the regional scales. The Xinjiang region possesses relatively complete and high-quality rain gauge data, while the Central Asian countries have serious data gaps, resulting in a limited spatiotemporal representativeness.

Furthermore, Schiemann et al. [35] found that the overall variation range of extreme precipitation in Central Asia and South Asia has been relatively small from 1961 to 2000. However, Peng et al. [36] predicted—based on scenario simulation—that the frequency and intensity of extreme precipitation would significantly increase under the RCP8.5 path in Central Asia in the future. Overall, the current research mostly remains at the level of statistics and simulation based on fixed indices and there is still a lack of systematic work in order to deeply explore the structural characteristics and evolution processes of extreme precipitation from the perspective of continuous processes. Especially in areas with sparse observation data, introducing the perspective of event continuity would not only help to compensate for the problem of an insufficient spatial resolution but it might also reduce the sensitivity influence on the threshold setting, thereby reflecting the entire process of the EPEs more truly [14].

Therefore, it is necessary to introduce an analysis framework based on the event process urgently. With the support of data with a higher spatiotemporal resolution, the system could identify and analyze the spatiotemporal heterogeneity and dominant types of EPEs in Central Asia, making up for the deficiencies in the existing studies. This study employs the Multi-Source Weighted-Ensemble Precipitation (MSWEP) dataset—a high-quality precipitation product that integrates multiple sources of observational and reanalysis data—for the period 1979–2023. The EPEs were defined from the perspective of event continuity and their characteristics including frequency, duration, intensity, and severity were quantitatively analyzed. The spatial distribution patterns and interdecadal evolution trends of these events were examined under varying topographic and hydroclimatic conditions. Moreover, the study has identified dominant types of the EPEs and their distribution patterns. These findings will contribute to enhancing the understanding of extreme precipitation processes in arid regions and might offer a scientific basis and data support for regional water resource management, disaster risk assessment, and the development of adaptation strategies.

2. Materials and Methods

2.1. Study Area

Central Asia is situated in the heart of the Eurasian continent, encompassing five countries: Kazakhstan, Kyrgyzstan, Tajikistan, Uzbekistan, and Turkmenistan. The region spans approximately 4 × 10⁶ km², within the geographical extent of 45° E–88° E and 34° N–56° N. Topographically, Central Asia is characterized by a general elevation decline from east to west, with plains and hills dominating the landscape, while the highlands appear to be mainly concentrated in the eastern part. The region is strongly influenced by the mid-latitude westerlies and the continental stationary front, resulting in a low annual precipitation and high evaporation rates. It exhibits a typical continental arid climate and has been recognized as the largest inland dryland region in the world [33,37,38]. The precipitation mainly occurs during winter and the winter precipitation in the southern part of Central Asia accounts for more than 40% of the total annual precipitation [24,39,40]. Many rivers exist in Central Asia, which are primarily supplied by the melting snow and glaciers from the upper reaches of the Pamir Plateau and the Tianshan Mountains. Apart from the ice and snow, water bodies, construction land, and forests, the landscape of Central Asia is mainly covered by grasslands, farmlands, and wastelands [41] (Figure 1a,b).

According to the Food and Agriculture Organization of the United Nations (FAO) classification, using the Aridity Index (AI) with thresholds of 0.05, 0.2, 0.5, and 0.65, climates can be categorized into five types, including hyper-arid, arid, semi-arid, sub-humid, and humid [42]. As shown in Figure 1c, Central Asia is mainly composed of the arid, semi-arid, and sub-humid areas. The arid areas are mainly concentrated in the regions including Uzbekistan, Turkmenistan, and the southwestern part of Kazakhstan. The semi-arid areas are especially distributed in most parts of Kazakhstan. The sub-humid area is sporadically distributed in Kyrgyzstan and Tajikistan, as well as the northern and eastern parts of Kazakhstan. The humid area is situated in the central region of Tajikistan (Figure 1c). Combining precipitation, AI distribution, and topography in Central Asia, this study divides Central Asia into three sub-regions, comprising northern Kazakhstan (NK), the Southwestern Desert (SD), and the southeastern mountains (SM) (Figure 1d).

Central Asia is marked by a noteworthy arid climate, with annual mean precipitation ranging from approximately 98.9 mm to 994.5 mm. The precipitation illustrates clear spatial gradients, with higher amounts in the eastern and peripheral regions and lower values in the central and western parts of the domain (Figure 1d).

Among the three defined sub-regions, the southeastern mountainous area (SM) demonstrates the largest precipitation, with an average annual precipitation measuring up to as high as 481.8 mm. The northern Kazakhstan region (NK) demonstrates a moderate precipitation, particularly in its northeastern areas. In contrast, the Southwestern Desert Region (SD), which includes most areas of Uzbekistan and Turkmenistan, is very dry, with the lowest annual precipitation only amounting to 180.5 mm (Figure 1e).

These spatial patterns denote the combined influences of topography, westerly circulation, and continentality on the regional hydrological regime.

2.2. Datasets

2.2.1. CATPD

The observed precipitation data from meteorological stations were obtained from the Central Asia Temperature and Precipitation Data (CATPD), released by the National Snow and Ice Data Center (NSIDC). The CATPD covers stations in Kazakhstan, Kyrgyzstan, Tajikistan, Turkmenistan, and Uzbekistan. The period of record varies among stations; however, most stations have nearly a century of observations, with the earliest records dating back to 1879 and the most recent extending to 2003 [43]. The station data used in this study underwent rigorous quality control and inspection procedures, including the removal of invalid values, extreme value tests, internal consistency checks, and spatial consistency checks. A total of 142 meteorological stations with complete data for the period 1979–1999 were ultimately retained, including 16 stations in the NK region, 28 stations in the SD region, and 98 stations in the SM region.

2.2.2. MSWEP

This study has adopted the daily scale data of the MSWEP from 1979 to 2023 as basic data to monitor the EPEs [44]. MSWEP is a multi-source ensemble precipitation product that integrates the advantages of the site observation data, remote sensing precipitation products, and reanalysis products and enables the provision of precipitation grid data with a global coverage of a 3 h scale. The MSWEP incorporates various precipitation data and conducts quality control, resampling, weighted ensemble, deviation correction, and spatial smoothing on them so as to generate high-resolution and spatiotemporal consistent global precipitation data. The precipitation data mainly include remote sensing precipitation data, reanalysis data, and surface precipitation data. The estimates of the artificial satellites mostly encompass the TRMM 3B42RTV7, GSMaP, GriSMat B1 infrared data, and the CMORPH estimation data. The reanalysis data comprise the interim reanalysis data of the European Centre for Medium-Range Meteorological Forecasting, the 55-year reanalysis data of Japan, the GPCC data, and the WorldClimV2.0 month-scale meteorological data. On this basis, by means of various observation data, such as GHCN-D and GSOD, and the estimation results of MSWEP are corrected daily. Secondly, the MSWEP shows a relatively high accuracy in Central Asia where the data are scarce. The research concludes that the MSWEP data denote a higher accuracy than the data of other remote sensing precipitation products [45,46,47]. In Central Asia, the monitoring accuracy of the MSWEP data seems to be the greatest and the MSWEP has a strong ability to capture the precipitation and to characterize the dry–wet changes in Central Asia [48] (Figure A1).

2.3. The Definition and Characteristics of the EPEs

This study has adopted the double threshold method to identify EPEs. To minimize potential biases arising from invalid precipitation records, a threshold of 1 mm/day was established to define effective precipitation. Building upon this, the 90th percentile of the daily effective precipitation series at each grid cell was determined as the extreme precipitation threshold (ExtP90). This method has been extensively employed in extreme precipitation identification studies [14,49] and has demonstrated robust applicability in arid and semi-arid climate zones [25]. The spatial distribution of the ExtP90 showed an obvious unevenness (Figure A2). It was lower than 6.5 mm/day in the central and southeastern humid areas, while higher in NK and southwest of the SM.

Based on this framework, an EPE has been defined as a continuous sequence of the daily precipitation in a given grid cell, in which the precipitation equals to the ExtP90 during at least one day and also when all days within the sequence—including the days before and after the peak—record a precipitation amount that is no less than 1 mm/day. Such a sequence has been identified as an EPE [14].

In order to comprehensively depict the evolution process and intensity structure of the EPEs, this paper introduced eight event-level quantitative indicators (Figure 2), including event duration (ED, the total number of days from the beginning to the end of the event, Formula (1)), event severity (ES, the precipitation throughout the event process, Formula (2)), event intensity (EI, the average precipitation intensity within a unit of time, Formula (3)), event peak (EP, the maximum single-day precipitation during the event period, Formula (4)), and event frequency (EF, the total number of times EPEs have occurred in this grid during the study period). In addition, to further highlight the extremity of the precipitation intensity above the threshold, three enhanced indicators were introduced, including the ExtED (the number of days above the ExtP90), ExtES (the total precipitation above ExtP90), and the ExtEI (the intensity above ExtP90, calculated as ExtES/ExtED). The eight indicators mentioned above jointly created a multi-dimensional quantitative description of the EPEs in terms of persistence and intensity, which could effectively reveal their heterogeneous characteristics at the spatiotemporal scale.

The formulas are given as follows:

$$ED=TE-TS$$

(1)

$$ES={\int }_{TS}^{TE}f\left(t\right)$$

(2)

$$EI={\displaystyle \frac{ES}{ED}}={\displaystyle \frac{{\int }_{TS}^{TE}f\left(t\right)}{TE-TS}}$$

(3)

$$EP=\underset{\mathit{TS}\le t\le \mathit{TE}}{\max }f\left(t\right)$$

(4)

where $TE$ and $TS$ are the start and end times, respectively, and $f\left(t\right)$ denotes the function of the extreme precipitation events.

2.4. Types of EPEs

Based on the event duration and the temporal position of the peak intensity, the EPEs in this study have been classified into three distinct types. The events that last more than one day are stipulated as multi-day continuous events, while those confined to one single day are considered as single-day events. Given the daily resolution of the used dataset, the single-day events are incorporated into the balanced-type category.

Regarding the multi-day events, the classification is based on the timing of the peak precipitation relative to the event duration, specifically when the precipitation exceeds the ExtP90. The three types are determined as follows:

Front-peak events (TDP1): The peak precipitation intensity takes place during the early stage of the event, followed by a rapid decline with little or no extreme rainfall later.

Rear-peak events (TDP2): The initial phase is relatively dry and the peak intensity is in the latter part of the event.

Balanced-type events (TDP3): The extreme precipitation is distributed more symmetrically, with peak or significant rainfall present in both the early and late stages. All single-day events are also categorized as balanced-type events.

This classification framework effectively captures the temporal structure of the extreme precipitation processes and supports the analysis of their spatial heterogeneity and the dominant types across the study region.

2.5. Trend Analysis Method

In order to examine the temporal trends in the EPEs, this study employed a combination of the Sen slope estimator, the Mann–Kendall [50] trend test, and the Locally Estimated Scatterplot Smoothing method (LOESS).

The Sen slope estimator, originally proposed by Theil and then later refined by Sen [51,52], is a robust non-parametric method that was used to estimate the magnitude of the monotonic trends in the time series data. It calculates the median of all possible pairwise slopes between the time points, making it resistant to outliers and applicable to abnormally distributed data. This method alone is however not suitable to evaluate the statistical significance of the trend [53]. The slope is calculated as:

$$Sen=Median\left[{\displaystyle \frac{x_j-x_i}{j-i}}\right],j>i$$

(5)

where $x_i$ and $x_j$ represent the observed values at the time steps i and j, respectively, and n stands for the length of the time series.

The Mann–Kendall test, initially introduced by Mann and further developed by Kendall [54,55], is a widely used non-parametric method so as to detect the presence of monotonic trends in the time series data. It is particularly effective for the data that are abnormally spread, contain outliers, or comprise missing values, making it more appropriate than the traditional parametric tests in such contexts [56,57].

Aiming to capture the localized and non-linear trends in the data, we further applied the LOESS method [58]. LOESS is a non-parametric local regression technique that performs a weighted polynomial fitting over a subset of neighboring data points. For each target point, closer observations are assigned higher weights, allowing for a flexible fitting and an accurate characterization of the complex and evolving precipitation trends.

3. Results

3.1. Analysis of Spatiotemporal Characteristics of Annual Average Extreme Precipitation in Central Asia

From 1979 to 2023, spatial differences have been noticed in the characteristic indicators of the eight EPEs in Central Asia (Figure 3). In general, the spatial patterns of the event-level quantitative indices for extreme precipitation (ED, EI, ES) closely resemble those of their corresponding enhanced indices (ExtED, ExtEI, ExtES) (Figure 3c–h). The values of the eight indicators were significantly larger in the SM region than those in the other two regions.

Specifically, the frequency of the EPEs exhibits a distinct spatial pattern characterized by a decrease from the northeast to the southwest. In the northern parts of the Kazakhskiy Melkosopochnik and the Turgay Plateau, the number of EPEs reaches as high as 500–600. The greatest total frequencies have been seen in the high-altitude regions like the Altai Mountains, Tianshan Mountains, and the Pamir Plateau, where up to 698 events have been recorded—considerably exceeding those in the lower-altitude areas (Figure 3a). In the SD region, the EPEs seem relatively infrequent, with 99.73% of the area experiencing fewer than 500 events. In contrast, the duration of the EPEs is notably longer in the SM region, particularly in the Tianshan Mountains of Kyrgyzstan, in which the mean event duration ranges from 8 to 12.76 days (Figure 3c,d). The spatial distributions of the EP, ExtEI, EI, ES, and ExtES are generally similar, demonstrating a decreasing gradient from the southeast to the northwest. The higher values of these indicators are concentrated in the Tianshan Mountains and Pamir Plateau within the SM region, the high-altitude areas in the southern part of the SD region and the Altai Mountains in eastern NK (Figure 3b,e–h). This conforms to the feature ‘the higher the altitude, the more prominent the extreme precipitation characteristics’. The latter might be attributed to the fact that high mountains serve as barriers to the atmospheric moisture transport, thereby reducing precipitation on the leeward slopes [59]. In the SM region, the Pamir Plateau denotes markedly higher values of the ES and ExtES when compared to other areas, with exceptionally large values noticeable in Tajikistan, reaching 62.15 mm and 36.20 mm, respectively. The cumulative precipitation associated with the EPEs is a key indicator of the event severity; the bigger the cumulative amount, the more severe the event and the higher the likelihood of triggering precipitation-induced disasters [60]. The potential risk of extreme precipitation-related disasters is therefore substantially larger in Tajikistan within the SM region.

During the period 1979–2023, the extreme precipitation characteristics in Central Asia demonstrates obvious patterns. Except for the EI and EF, the SD region shows the lowest average values for all other indicators, followed by the NK region, while the SM region illustrates the highest averages. Moreover, the extreme precipitation characteristics appear to be relatively concentrated in the SD and NK areas, whereas those are rather highly dispersed in the SM region (Figure 4).

Compared with the ED and ExtED, the average ED across Central Asia is 3.1 days longer than the ExtED. Furthermore, the ExtED denotes a smaller spatial variability than the ED across the entire region and its sub-regions (Figure 4a,b). Similar patterns have also been observed for the ES and ExtES, with the ExtES showing reduced spatial differences compared to the ES (Figure 4e,f).

At a sub-regional scale, the difference between the ED and ExtED is most pronounced in the SM region, reaching 4.9 days, while the smallest difference is detected in the SD region at 2.3 days (Figure 4a,b). Although the average EI in the SD region is 0.4 mm/day higher than the one in the NK region, the average ExtEI in SD measures 0.7 mm/day lower than in the NK (Figure 4c,d).

Although the EF varies greatly among the three sub-regions, the EP differences are relatively small (Figure 4g,h). Over 45 years, the average number of the EPEs in the SD region has been lower than in other regions (EF = 345) and the EP seems to be the smallest (averaging 9.0 mm/day). It is worth noting that the average number of EPEs amounts to 516 in the NK region, which is slightly bigger than the average EF in the SM region (EF = 509). Moreover, the EF fluctuates little in various places within the NK area, while the latter fluctuates greatly vary in several locations within the SM area.

From 1979 to 2023, the trends in the extreme precipitation characteristics have largely varied across several sub-regions in Central Asia. The annual frequency of the EPEs exhibit a substantial interannual variability in all three sub-regions, with an overall significant falling trend (Figure 5a). Over 45 years, the event duration has shown a rising tendency in the SM region (2835 days/45 years), with a notably sharper increase visible after 2020. In contrast, the ED in the SD and NK regions fluctuates over time but remains relatively stable in terms of the long-term trends (Figure 5c). Both the ES and ExtED demonstrate climbing trends in the SM region, with the duration of the extreme events augmenting by 3.555 days over 45 years (Figure 5d,g). Although slightly decreasing trends have been recorded for the EI, EP, ExtEI, and ExtES, the overall change magnitudes are instead relatively small (Figure 5b,e,f,h).

3.2. Analysis of the Characteristics of EPEs at the Decadal Scale in Central Asia

The eight indicators have demonstrated consistent spatial patterns across Central Asia during the periods 1979–1991, 1992–2009, and 2010–2023. This is generally consistent with the spatial heterogeneity of the EPEs in the region, whereby the high-altitude areas generally experienced more intense extreme precipitation compared to the low-altitude areas. Except for the EF, the spatial characteristics of the other EPE indicators exhibited minimal variation during these three periods.

The frequency of the EPEs has significantly increased in Central Asia during 1992–2009 compared to 1979–1991, with an average rise of 36 events. During 1992–2009, an increase in the frequency of extreme precipitation events was observed across 97.9% of Central Asia, with markedly greater increases occurring in the high-altitude regions of NK and SM (Figure A4a). In particular, the NK sub-region underwent up to 270 events and 63.44% of the combined area of the NK and SM registered more than 180 EPEs (Figure 6a,b). From 1979 to 2023, the duration of the EPEs (ED and ExtED) markedly increased in the SM region, while a slight rise in duration was noticed in the northern part of the NK sub-region, specifically in the West Siberian Plain (Figure 6d–i). In contrast, in most of the central SD, a decline in the accumulated precipitation was noted (ES and ExtES) during 1992–2009, with maximum decreases amounting to 4.65 mm and 5.80 mm, respectively (Figure A4d). In the Aral Sea region, both the duration and accumulated precipitation of extreme events increased, which may be attributed to the humidifying effect of the large lake surface [59]. After 1992, the accumulated extreme precipitation increased in the SM region, while a similar increase was observed in the SD region after 2010. Over the past 45 years, the Tianshan Mountains have exhibited a notable increase in extreme precipitation, with the maximum increase reaching up to 16.3 mm (Figure 6j–o and Figure A4j–o). Compared to the period before 1992, 61.90% of Central Asia experienced a weakening of EPEs’ intensity during 1992–2009, with decreases ranging from 0 to 3.19 mm/day. Similarly, compared to the period before 2010, 67.82% of the region showed a reduced intensity during 2010–2023, with declines ranging from 0 to 3.60 mm/day (Figure 6p–u and Figure A4p–u). The areas with the most pronounced reductions in event intensity were mainly located in the northern NK region, the central desert areas of SD, and the high-altitude zones of SM. Furthermore, 78% of Central Asia experienced a decrease in peak precipitation (EP) during 1979–2023, with reductions ranging from 0 to 9.61 mm/day. These declining EP areas were scattered across the NK and SM, with 89.2% of the SM region exhibiting a weakening trend in extreme precipitation (Figure 6v–x and Figure A4v–x).

Based on the LOESS of the precipitation in Central Asia from 1979 to 2023, three distinct sub-periods have been recognized according to the trend changes in the smoothed curve: 1979–1991, 1992–2009, and 2010–2023 (Figure A3). As a whole, the period from 1992 to 2009 has been the most dramatic and substantial period of extreme precipitation characteristics in Central Asia (

Table 1. Analysis and fitting of the extreme precipitation characteristics in Central Asia during the periods 1979–1991, 1992–2009, and 2010–2023, as well as the significance testing. * indicates p < 0.05, ** indicates p < 0.01, and *** indicates p < 0.001.

Year	Feature	CA_Slope	NK_Slope	SD_Slope	SM_Slope
1979–1991	EF	−0.134	−0.130	−0.158	−0.080
	EP	−0.024	−0.020	−0.028	−0.021
	ED	0.040 ***	0.036 *	0.031 **	0.081 **
	ExtED	0.002	−0.001	0.004 *	0.005
	EI	−0.025 **	−0.015	−0.030 *	−0.037 *
	ExtEI	−0.033	−0.017	−0.045	−0.053
	ES	0.059	0.022	0.040	0.217
	ExtES	−0.006	−0.032	−0.003	0.051
1992–2009	EF	−0.040	−0.024	−0.049	−0.082
	EP	−0.081 ***	−0.070 **	−0.089 ***	−0.088 **
	ED	−0.003	−0.021	0.004	0.026
	ExtED	−0.001	−0.002	0.001	−0.001
	EI	−0.032 ***	−0.023 ***	−0.039 **	−0.033 **
	ExtEI	−0.105 ***	−0.095 **	−0.109 ***	−0.120 **
	ES	−0.118 **	−0.152 **	−0.096	−0.101
	ExtES	−0.122 ***	−0.121 **	−0.113 **	−0.161 *
2010–2023	EF	0.017	0.094	−0.015	−0.061
	EP	0.018	0.022	0.005	0.063 *
	ED	0.038	0.029	0.032	0.057
	ExtED	0.001	−0.001	0.001	0.006
	EI	−0.009	−0.005	−0.018	0.019
	ExtEI	0.028	0.015	0.223	0.092 *
	ES	0.085	0.029	0.060	0.283
	ExtES	0.245	0.006	0.037	0.245

).

During 1979–1991, the EF has demonstrated a falling trend across Central Asia in the NK, SD, and SM sub-regions, with declines of 1.742 (95% CI: ±0.04), 1.690 (±0.03), and 2.054 (±0.08) per 13-year period, respectively (Figure 7a and Table 1). The EP also shows a downward trend in Central Asia (Figure 7b). In contrast, the event duration was augmented rather significantly, at a rate of 0.52 days per 13 years (95% CI: ±0.01) (Figure 7c). Both EI and ExtEI demonstrated declining tendencies across the region, with noticeable decreases of 0.325 and 0.429 mm/day per 13 years, respectively. Among these, the 95% confidence intervals for the SD and NK sub-regions are relatively narrow, suggesting that the trends in these regions are more stable and reliable (Figure 7e,f and Table 1). The ExtES remained, however, relatively stable during this period, only exhibiting minor fluctuations (Figure 7h).

The period 1992–2009 has represented the most essential phase of change in the extreme precipitation characteristics across Central Asia, with notable variations seen in the EP, EI, ExtEI, ES, and ExtES across all sub-regions. The EF recorded a pronounced decreasing trend, with a regional reduction rate of 0.72 events per 18 years (Figure 7a and Table 1). Overall, the EP declined during this time, with the most substantial decrease taking place in the SD region (−1.602 mm/day per 18 years), (Figure 7b and Table 1). Both the EI and ExtEI considerably declined across Central Asia and all sub-regions, indicating a marked weakening in the precipitation intensity. Among those, the SD region has experienced the largest decrease in EI (−0.705 ± 0.02 mm/day per 18 years), while the SM region listed the greatest reduction in the ExtEI (−2.16 ± 0.01 mm/day per 18 years) (Figure 7e,f and Table 1). In addition, both ES and ExtES illustrated major decreasing tendencies throughout Central Asia, with the most pronounced changes noticeable in the NK region (Figure 7g,h).

During 2010–2023, all extreme precipitation indicators—except for EF and EI—demonstrated rising trends. The EF increased more rapidly in the NK region than in other sub-regions at a rate of 1.316 events per 14 years [60] (Figure 7a and Table 1). However, the ExtED in the NK region and the EI in the SM region proved to have trends opposite to those observed in other sub-regions. The SM region demonstrated a notable upward tendency in the extreme precipitation characteristics. It is expected that the degree of extreme precipitation in high-altitude areas such as the SM region will further intensify in the future. This expectation is supported by the statistically significant increasing tendencies observed in 2010–2023, with 95% confidence intervals excluding zero [61,62].

3.3. Analysis of the Dominant Types of EPEs in Central Asia

The dominant types of EPEs were identified based on two criteria: the total event frequency and the accumulated precipitation amount (Figure 8 and Figure 9). A specific type of EPE is considered prevalent in a given region if it accounts for the largest proportion among the three event types, either in terms of frequency or cumulative precipitation.

The proportions of TDP1 and TDP2 were consistently higher in the SD and NK regions when compared to the SM region, indicating that the TDP1 and TDP2 occurred more frequently in the western and northern parts of Central Asia. A further analysis of the dominant EPE types revealed that during 1979–2023, the TDP2 remained the prevailing type across 90.04% of the Central Asia region (

Table 2. Based on the frequency of the events and the cumulative precipitation amounts, the proportion of the dominant EPE types was determined in the Central Asia region.

Year	Frequency			Cumulative Precipitation Amounts
	TDP1 (%)	TDP2 (%)	TDP3 (%)	TDP1 (%)	TDP2 (%)	TDP3 (%)
1979–2023	3.29	90.04	6.67	1.52	44.97	53.51
1979–1991	4.25	90.09	5.66	2.23	43.64	54.13
1992–2009	3.16	90.02	6.82	1.44	45.45	53.11
2010–2023	4.32	89.11	6.57	1.78	35.74	62.48

). During the period 1992–2009, 61.1% of Central Asia experienced an increase in the frequency of TDP2 compared with the first period, while after 2010, 57.5% of the region showed a higher frequency of TDP2 relative to the first period. These increases were spatially distributed across the entire region and were particularly pronounced in Kazakhstan (Figure A5d–f). Although the TDP1 accounted for a relatively larger part of the SD and NK regions, its frequency was lower than that of TDP2. The spatial dominance of the TDP1 reached 4.25%, 3.16%, and 4.32% during the periods 1979–1991, 1992–2009, and 2010–2023, respectively (Figure 8d,g,j and Table 2). Compared with the period 1979–1991, the frequency of TDP1 events in the central desert areas of the SD region increased significantly after 1992, with the magnitude of increase reaching up to 30% (Figure A5a–c). The TDP3 indicates a significantly high value in the SM region, especially during 1979–1991 and 2010–2023. However, the dominance of the TDP3 was relatively weak and its proportion of the dominant area was consistently lower than 6.9% for the entire study area (Figure 8c,f,i,l and Table 2). From 1979 to 2023, based on the prevailing event types determined by the relative frequency, the TDP1 and TDP3 exposed a fluctuating increasing tendency, while the TDP2 rather showed a slight decline (Figure 8a–c). The frequency of TDP3 demonstrated a declining tendency across Central Asia. Specifically, during the period 1992–2009, the SD region underwent a pronounced reduction, with 76.1% of its area exhibiting decreases in TDP3 frequency ranging from 0% to 21% (Figure A5g–i).

From 1979 to 2023, the spatial distribution patterns of the accumulated precipitation proportions for the TDP1, TDP2, and TDP3 closely resembled those of their corresponding frequency proportions (Figure 8 and Figure 9).

Among the EPEs in Central Asia, the accumulated precipitation proportion of the TDP1 seemed relatively low and its dominance has been consistently less than the other two types across all periods, reaching a maximum of only 2.23% during 1979–1991 (Table 2). Spatially, the TDP1 demonstrated larger amounts in the SD region and lower proportions in the SM region (Figure 9a,d,g,j). During 1992–2009, the accumulated precipitation associated with the TDP1 has generally risen across Central Asia, with an average increase of 1.5% (Table 2). The TDP2 type has maintained a big part of the accumulated precipitation throughout the entire study period, particularly between 1979 and 2023, when its dominant spatial coverage reached 75.12%, which is most widespread in the NK and SD sub-regions (Figure 9b,e,h,k). The dominance of TDP2 exhibited an increasing trend followed by a subsequent decline, as its spatial coverage declined from 43.64% during 1979–1991 to 35.74% in 2010–2023, indicating a contraction in the ruling area (Table 2). From 1992 to 2009 relative to 1979–1991, 59.6% of Central Asia exhibited an increase in the proportion of accumulated precipitation associated with TDP1, while 66.0% of the region showed a corresponding increase in TDP2. Particularly pronounced increases were observed in the desert areas of central SD. Nevertheless, the dominant influence of both TDP1 and TDP2 weakened after 2010. In contrast, the interdecadal variations in TDP3 were opposite to those of the other two types. During 1992–2009, the proportion of accumulated precipitation attributable to TDP3 reached its minimum in the Kazakh Hills, the desert areas of SD, and the high-altitude mountainous regions of SM (Figure 9c,f,i,l and Figure A6). However, when viewed over the long-term time series, the prevalent spatial coverage of the TDP3 expanded from 54.13% during 1979–1991 to 62.48% in 2010–2023 (Table 2). This tendency might be attributed to the fact that the individual TDP3 tends to show a stronger precipitation intensity or longer duration, resulting in the largest and increasing share of the accumulated precipitation.

3.4. Analysis of the Correlation of Extreme Precipitation Characteristics in Central Asia

The frequency of EPEs in Central Asia shows a strong correlation with the event duration, and the uncertainty of the fitted curve initially increases and subsequently decreases. The high-value areas of the extreme precipitation frequency are mainly situated where the event duration ranges between 6 and 7 days, predominantly distributed along the margins of the Tianshan Mountains in the SM region and the Altai Mountains in the NK region (Figure 10a). The ED exerts a certain influence on ES and ExtES (Figure 10e,f), while its impact on the other three characteristics appears to be relatively limited (Figure 10). The EF increases with longer durations but the rate of increase gradually slows down. When the ED exceeds 7 days, the EF clearly denotes a declining trend in the increase rate (Figure 10a). Both the ES and ExtES demonstrate a consistently rising tendency with a longer ED, with the growth rate decelerating as the duration augments (Figure 10e,f). However, no vital or consistent correlations have been noticed between ED and EP, EI, or ExtEI across Central Asia (Figure 10b–d).

Generally, areas with a Digital Elevation Model (DEM) value below 500 m are defined as plains, those with elevations between 500 and 1500 m are classified as hilly, and low-lying mountainous regions and those above 1500 m are treated as mountainous areas. The mountainous regions with elevations between 2000 and 3000 m are often associated with the peak frequencies of the EPEs (Figure 11a). In the hilly and mountainous areas (DEM value ranging from 0.5 km to 2.5 km), the extreme precipitation characteristics illustrate a considerable variability (ρ = 0.47 ***) (Figure 11). In Central Asia, the DEM shows a strong correlation with the EPE duration. The mountainous area with a DEM of 4000–5000 m is mostly the peak area of the extreme precipitation duration (Figure 11c,d). When the DEM measures below 4.75 km, the ED increases with the elevation, indicating a positive correlation (ρ = 0.57 ***) (Figure 11c). The DEM also exerts a certain influence on ExtED, ES, and ExtES, all of which display an increasing-then-decreasing trend with elevation (Figure 11d,g,h). Nevertheless, the relationship between DEM and ExtED is characterized by relatively high fitting uncertainty (95% CI: ±0.06). On the other hand, we conclude that the DEM has a rather limited impact on the other three characteristics (Figure 11b,e,f).

In Central Asia, the AI exhibits a strong correlation with the EF, ED, EP, ExtED, ES, and ExtES of the extreme precipitation. The greatest EF values are concentrated in the semi-humid zone, where the AI ranges between 0.7 and 0.8, and the latter are predominantly spread across the Tianshan Mountains and the Pamir Plateau in the SM region (Figure 3a and Figure 12a). The majority of the AI values are found within the range of 0.05 to 0.6 (Figure 12). Taking the AI approximately equal to 1.3 as the dividing line for the sub-humid zone, the ED exhibits the following tendency: firstly increasing and then decreasing with the rising AI (Figure 12c). The intensity of the EPEs augments with the rising AI and the rate of increase becomes more pronounced at higher AI values (Figure 12e,f). The EP, ES, and ExtES all denote positive correlations with AI across its value range (Figure 12b,g,h).

In summary, as the annual precipitation increases or the potential evapotranspiration decreases, the characteristics of the extreme precipitation become more prominent in the arid, semi-arid, and semi-humid regions of Central Asia.

4. Discussion

4.1. Spatial Heterogeneity of the TDPs in Central Asia

Based on the proportion of the EPEs’ frequency and the cumulative extreme precipitation amount, the dominant types of EPEs in Central Asia expose a pronounced spatial heterogeneity. From 1979 to 2023, the classification was based on the proportion of the extreme precipitation frequency, which indicates that TDP2 is the dominant event type across 90.04% of the region. In contrast, the TDP1- and TDP3-dominated events only account for a small fraction of the domain and are mainly scattered across the desert areas in the northwest of the SD region and the northern plains of the NK region. When classified by the cumulative precipitation amount over the same period (1979–2023), the TDP3-dominated regions seem to be slightly more extensive than those ruled by TDP2. These TDP3-dominated areas are primarily located in the deserts of the northwestern SD, the plains and hilly areas of northern NK, and high-altitude zones covering 57.17% of the SM region. In terms of decadal variability, while the frequency of TDP3 occurrences has declined, the cumulative precipitation associated with TDP3-dominated areas has increased, particularly across the deserts of the SD region and the high-altitude zones of the SM region. Defining the dominant type of EPEs is of great significance for the regions with a highly extreme precipitation frequency, as it offers insights for the decision-makers when managing the extreme precipitation risks [15]. For example, in regions where the TDP2 seems to be the prevailing event type, the severity of the extreme precipitation tends to intensify during the later stages of the event, highlighting the need for enhanced preparedness and mitigation efforts focused on the latter part of the event.

4.2. Analysis of the Changing Trend of EPEs’ Characteristics in Central Asia

Central Asia is a typical inland arid region, influenced by the westerly circulation in summer and the Siberian high in winter. Specifically, the variations in the westerly circulation are the primary factors affecting precipitation in the region [63]. With the increase in the Arctic Zonal Index (AZI), the westerly circulation becomes stronger, resulting in an accelerated airflow over the 500 hPa isobaric surface between 45° N and 60° N, thereby enhancing the moisture transport into Central Asia [64]. Aizen et al. [65] found that the North Atlantic Oscillation (NAO) and the Western Pacific Oscillation (WPO) indices exhibit a negative correlation with the mean precipitation over the mountainous and plain regions of Central Asia. The combined effects of the enhanced pressure gradients and the anomalous westerly winds likely facilitate increased moisture transport into Central Asia [66]. With respect to the climatic drivers of extreme precipitation characteristics, the findings of this study indicate that EF is strongly associated with sea surface temperature anomalies. In particular, EF shows significant positive correlations with the Niño 3.4, Niño 4, and Pacific Decadal Oscillation (PDO), and these relationships remain robust at a high level of statistical significance. Moreover, EF, EP, and EI are significantly negatively correlated with the Southern Oscillation Index (SOI), underscoring the suppressive role of ENSO-related ocean–atmosphere coupling processes in modulating regional extreme precipitation anomalies. Additionally, ED is significantly negatively correlated with the PDO, while EP and EI exhibit significant positive correlations with the PDO, suggesting that different phases of the PDO may alter moisture transport and circulation patterns, thereby exerting distinct influences on the persistence and intensity of extreme precipitation events. In contrast, ES, ExtED, ExtEI, and ExtES display comparatively weak associations with the major atmospheric circulation indices (Figure 13).

The precipitation in the Tianshan Mountains is substantially higher than that of the surrounding lowlands, with a higher frequency of intense precipitation events [35]. It has been confirmed that the period from 1973 to 1984 corresponded to a drought phase in the Tianshan region. Starting in 1985, the annual precipitation has indicated a rising tendency across the majority of the Tianshan area within Central Asia. In particular, the high-altitude regions of Kyrgyzstan and Tajikistan experienced an elevated likelihood of extreme precipitation occurrences [22,31]. These findings corroborate the increasing trend of extreme precipitation observed in the SM region in this study. In order to forecast future changes in extreme precipitation trends, Peng et al. [36] assessed the performance of the CMIP5 models in simulating the extreme precipitation in Central Asia. They also predicted a major augmentation in the extreme precipitation under the RCP8.5 scenario, which aligns with the outcomes of our study.

4.3. Uncertainties

Meteorological station observations are widely recognized as the most reliable representation of actual precipitation conditions and have been extensively applied in previous research. In this study, CATPD station data were employed to evaluate the applicability of MSWEP. Nevertheless, station-based observations are not without limitations, including uneven spatial distribution, sparse coverage in complex terrain, and difficulties in data acquisition, all of which may introduce potential uncertainties into error assessments. Although the stations used in this study are national meteorological stations subject to strict quality control, several issues remain noteworthy: (1) The spatial distribution of stations across Central Asia is highly uneven. While relatively dense observations are available in the southeastern mountainous regions (98 stations), northern Kazakhstan and the Southwestern Desert areas are characterized by sparse and scattered stations, thereby limiting their spatial representativeness. (2) Station observations provide point-scale precipitation measurements, whereas MSWEP data have a spatial resolution of 0.1° × 0.1°. This inherent scale mismatch may introduce uncertainties in the evaluation. Consequently, the validation of MSWEP against CATPD observations may be subject to potential errors in regions lacking station coverage. Nonetheless, considering the unique conditions of Central Asia—such as the scarcity and limited accessibility of station records—the CATPD dataset remains advantageous, offering longer temporal coverage and a comparatively larger number of stations than alternative sources. (3) As a multi-source ensemble precipitation product, MSWEP integrates gauge, remote sensing, and reanalysis data to produce high-resolution precipitation estimates. Inevitably, this approach is subject to estimation errors, which may propagate into the spatiotemporal characterization of extreme precipitation events. For instance, slight underestimations have been observed in arid regions. However, both the present study and previous investigations demonstrate that MSWEP exhibits higher accuracy relative to other satellite-based precipitation products, effectively capturing precipitation processes and characterizing regional dry–wet variability.

5. Conclusions

This study introduces an event-based conceptual framework for extreme precipitation and quantitatively classifies different types of EPEs. Eight characteristic indicators related to extreme precipitation and its events have been applied to analyze the spatiotemporal characteristics of the extreme precipitation in Central Asia and its three sub-regions from 1979 to 2023. Furthermore, the relations between the EPE characteristics and the ED, DEM, and AI have been examined. In addition, the dominant types of EPEs and their contributions to the cumulative precipitation were evaluated for Central Asia and its sub-regions. The main conclusions are as follows:

(1)

Extreme precipitation is more prevalent in high-altitude regions. The characteristic indicators of EPEs are significantly higher in the SM region, particularly in the Tianshan Mountains and the Pamir Plateau. Within the SM region, the event duration reaches up to 12.76 days. The highest values of the ES and ExtES are recorded in Tajikistan, measuring 62.15 mm and 36.20 mm, respectively. The cumulative frequency of EPEs generally appears to be higher in the high-altitude areas of the SM than in other regions of Central Asia, amounting to 698 events. Due to the smaller spatial variability in the cumulative frequency of the EPEs in the NK region, the average event frequency is higher than for the SM region, with the NK region experiencing approximately seven more EPEs on average.

(2)

On an interdecadal scale, the extreme precipitation characteristics illustrate rather pronounced phase changes, with substantial regional trend disparities. During 1979–1991, the frequency of the EPEs across Central Asia declined (−1.742 events per 13 years), while the event duration instead increased (0.52 days per 13 years). The period from 1992 to 2009 has undergone the most significant and pronounced a decline in the magnitude of extreme precipitation indicators. Compared to 1979–1991, it increased by a factor of 36 during the time frame 1992–2009, with 63.44% of the region experiencing EF values exceeding 180. Except for ED and ExtED, all other EPE characteristics showed significant downward trends. From 2010 to 2023, apart from EF and EI, most other EPE characteristics displayed augmenting tendencies across Central Asia.

(3)

The dominant types of EPEs exhibit regional differences. Using the proportion of the extreme precipitation frequency as a criterion, the regions dominated by the TDP2 account for a large part of Central Asia spatially. The SD and NK regions are characterized by a more prominent combination of the “TDP2 + TDP1”, while the SM region is dominated by the “TDP2 + TDP3” combination. Based on the total amount of extreme precipitation, 99.48% of Central Asia primarily features the “TDP2 + TDP3” combination. Temporally, the frequency of the TDP3 events demonstrates a decreasing tendency across the region.

(4)

The frequency and duration of the EPEs are closely related to DEM and AI. From the perspective of the elevation, the mountainous areas at 2000–3000 m are identified as peak zones for the EPEs’ frequency, while the regions at 4000–5000 m correspond to the peak values of the EPEs’ duration. In terms of AI, the peak values of the EF and ED are detected in the sub-humid zone. Additionally, there is a strong correlation visible between the EF and ED across Central Asia. The high-frequency areas of the EPEs are predominantly concentrated in the areas where the ED ranges between 6 and 7 days, mainly spread along the fringes of the Tianshan Mountains in the SM region and the Altai Mountains in the NK region.

Overall, this study demonstrates that an event-based framework can effectively characterize the complexity of extreme precipitation in Central Asia, revealing its correlations with elevation and the Aridity Index, as well as its non-uniform evolution on interdecadal timescales. By quantitatively classifying the dominant event types, we translate complex extreme precipitation processes into actionable categories, providing both historical insights and a methodological foundation for future monitoring, risk assessment, and adaptation strategies in the region.