Preamble

ARID LAND GEOGRAPHY
Vol. 48 No. 8 Aug. 2025

Spatio-Temporal Characteristics of Extreme Precipitation in the Ili River Basin Based on CMIP6

LIU Jinghui, YUAN Xushan, LI Yanmin, LI Xinxu
School of Emergency Technology and Management, Institute of Disaster Prevention, Sanhe 065201, Hebei, China

Abstract: Under the influence of global warming, disaster risks associated with extreme precipitation events have intensified, posing severe threats to regional socioeconomic development and public safety. This study analyzes the spatio-temporal characteristics of eight extreme precipitation indices in the Ili River Basin from 1981 to 2024 using historical meteorological data. Furthermore, employing multi-model datasets from the Coupled Model Intercomparison Project Phase 6 (CMIP6) under different scenarios, we project changes in extreme precipitation indices from 2025 to 2050 using multi-model ensemble averaging and Sen's slope estimation. Results indicate: (1) During 1981–2024, most extreme precipitation indices exhibited upward trends, particularly pronounced in the eastern and southwestern mountainous regions. (2) Under both SSP245 and SSP585 scenarios, extreme precipitation shows large interannual fluctuations but an overall upward trajectory. The SSP585 high-emission scenario projects more frequent and intense extreme precipitation, with significant increases in annual precipitation and heavy precipitation events in the eastern and southern mountainous areas, demonstrating stronger trends and higher risk. Conversely, the northern and central plains are projected to experience relatively fewer extreme precipitation events. This spatial heterogeneity may differentially impact regional disaster frequency, agricultural production, and livestock husbandry. These findings provide a scientific basis for local governments to formulate prevention and mitigation strategies for extreme precipitation events.

Keywords: CMIP6; extreme precipitation; spatio-temporal variation characteristics; Ili River Basin

Introduction

Climate change intensifies extreme precipitation events by increasing their frequency, intensity, and duration, causing substantial damage to socioeconomic development and ecosystems. When precipitation intensity reaches certain thresholds, it can trigger flood disasters that seriously threaten lives and property. Therefore, investigating the spatio-temporal patterns of extreme precipitation and projecting future scenarios are crucial for enhancing regional disaster resilience and developing adaptive strategies. Climate change has led to divergent trends in extreme precipitation across regions, drawing widespread attention due to associated hazards. Previous studies have employed various methodologies: Zhang et al. analyzed three-stage spatio-temporal evolution of extreme precipitation in the Qinling region using extreme value distribution and statistical methods; Zou et al. examined the Wei River Basin using Kendall trend tests and wavelet transforms; Zhao et al. applied mutation tests to study extreme precipitation in Northwest China; and Yang et al. used percentile thresholds to analyze summer extreme precipitation in Xinjiang. These studies reveal increasing extreme precipitation events in Northwest China's arid regions, particularly in Northern Xinjiang and the Tianshan Mountains, leading to more flood disasters and geological hazards such as landslides and debris flows.

The Ili River Basin, located in the hinterland of the Eurasian continent and surrounded by mountains on three sides, provides a unique geographical setting where warm, moist airflows from the west directly converge in the valley, significantly increasing the probability of extreme precipitation events. The basin's large topographic variations cause most precipitation to concentrate in the Ili River, posing severe flash flood threats to riverside communities. Given the basin's high susceptibility to extreme precipitation and its significant impacts on regional ecology, livelihoods, and economic exchanges, comprehensive analysis of extreme precipitation patterns is urgently needed.

CMIP6, the sixth phase of the Coupled Model Intercomparison Project led by the World Climate Research Programme, serves as a primary tool for understanding climate system complexity and projecting future changes. Previous CMIP6 studies have evaluated model performance for various applications: Wang et al. assessed temperature simulation capabilities; Hu et al. evaluated precipitation characteristics using high-resolution models; Yang and Xiang analyzed climate state distributions and extreme temperature/precipitation; and Zhang projected precipitation and temperature changes in Xinjiang. While some research has addressed precipitation patterns in the Ili River Basin, in-depth analysis of extreme precipitation and future projections remains limited. This study addresses this gap by systematically analyzing eight extreme precipitation indices from 1981 to 2024 and projecting their evolution under different CMIP6 scenarios through 2050.

1 Study Area Overview

The Ili River Basin is situated in Xinjiang Uygur Autonomous Region, China, between 80°09′–84°56′E and 42°14′–44°50′N, serving as a critical corridor connecting China with Central Asia, West Asia, and Europe. The region features diverse landforms with a spatial pattern characterized as "high in the east and low in the west, narrow in the east and wide in the west." Its location in the Eurasian interior, surrounded by mountains on three sides, allows western warm-moist air masses to directly converge in the Ili Valley, substantially increasing the probability of extreme precipitation events. Additionally, significant topographic relief causes most precipitation to drain into the Ili River, exposing riverside populations to severe flash flood risks. These unique geographical and climatic conditions make the Ili River Basin a high-risk area for extreme precipitation events with significant regional impacts.

2 Data Sources and Model Selection

This study utilizes two categories of meteorological data: historical observations and future simulations. Historical precipitation data comprise daily datasets from nine national meteorological stations in the Ili River Basin for 1981–2024, obtained from the China Meteorological Administration (http://data.cma.cn) with quality control applied. Drawing from previous CMIP6 model selection experience for China and Xinjiang, and considering data availability, we selected 20 climate models from CMIP6 (https://esgf-node.llnl.gov/search/cmip6/). These models were interpolated to a uniform 0.25°×0.25° resolution using bilinear interpolation and downscaled to each meteorological station. Delta bias correction was then applied to obtain corrected data for each model.

To identify suitable climate models for the Ili River Basin, we compared corrected model data against observed daily data using Taylor diagrams (Figure [FIGURE:2]). The Taylor diagram evaluates model performance through three metrics: standard deviation (optimal when接近 observed), correlation coefficient (higher values indicate better fit), and root-mean-square error (smaller radius indicates higher accuracy). Models ACCESS and CESM2-WACCM showed negative or weak correlations with observations and were therefore excluded. The final selection retained seven models: CanESM5, CESM2, Earth3, FGOALS, NESM3, NorESM2, and UKESM1, all demonstrating correlation coefficients greater than 0.6 with comparable performance. To minimize individual model biases, all seven models were retained for analysis.

3 Methods

3.1 Extreme Precipitation Indices

This study selected eight extreme precipitation indices from the 27 core indices defined by the WMO Commission for Climatology and Climate Variability and Predictability (CLIVAR) program. Table [TABLE:2] provides detailed definitions. Following the RclimDex 1.0 framework, these indices are categorized into intensity, relative, and absolute groups. Historical indices were calculated using RclimDex 1.0 software, while future indices were derived from bias-corrected CMIP6 multi-model ensemble daily precipitation data.

3.2 Delta Method

The Delta method is a widely used bias correction technique that calculates deviations between model and observed data, then applies these corrections to future model outputs. This approach adjusts future CMIP6 precipitation data to produce more realistic projections by removing systematic biases identified over the historical period.

3.3 Multi-Model Ensemble Mean Method

Multi-model ensemble averaging is a crucial technique in climate simulation research, comprising equal-weight and unequal-weight approaches. This study employs the equal-weight method calculated as:

$$EE = \frac{1}{N}\sum_{i=1}^{N}F_i$$

where $EE$ is the ensemble mean, $N$ is the number of models, and $F_i$ represents the $i$-th model simulation.

3.4 Mann-Kendall (M-K) Trend Test

The Mann-Kendall trend test is a non-parametric method for detecting monotonic trends in time series data. Its advantage lies in independence from data distribution assumptions and robustness to outliers, making it ideal for climatological and hydrological analyses. A Z-value greater than 0 indicates an upward trend.

3.5 Sen’s Slope Estimation

Sen’s slope estimator is a non-parametric statistical method that calculates the median slope among all possible point pairs to estimate trend magnitude. Insensitive to outliers and suitable for non-normal distributions, it is widely applied in trend analysis and anomaly detection. A slope $\beta > 0$ indicates an upward trend, while $\beta < 0$ indicates a downward trend.

4 Results

4.1 Bias Correction Results

Applying the Delta bias correction method to the seven climate models significantly improved simulation accuracy. Prior to correction, simulated values showed large deviations from observations with unclear seasonal patterns. After correction, daily precipitation simulations closely matched observations, with markedly reduced bias and enhanced seasonal representation. Summer months (June–August) exhibit substantially higher precipitation than winter (December–February), consistent with observed seasonal cycles. These seasonal differences are well captured in the bias-corrected data, demonstrating the method's effectiveness.

4.2 CMIP6 Climate Model Simulation Performance

Taylor diagram analysis compares the performance of seven individual models and the multi-model ensemble (MME) against observations (Figure [FIGURE:4]). The MME demonstrates optimal performance with the lowest root-mean-square error (0.112) and highest correlation coefficient (0.85), outperforming all single models. Individual models show varying strengths: Earth3 exhibits good correlation (0.82) but higher standard error and RMSE (1.043, 0.133); CanESM5 shows moderate correlation (0.78) with relatively high standard error and RMSE (1.164, 0.174), indicating substantial uncertainty; FGOALS performs poorly across all metrics with weak correlation (0.65); NESM3 shows moderate performance (standard error 1.115, RMSE 0.128, correlation 0.71). Overall, the MME provides superior simulation performance, justifying its use for projecting extreme precipitation indices under SSP245 and SSP585 scenarios.

4.3 Historical Spatio-Temporal Characteristics of Extreme Precipitation Indices (1981–2024)

Temporal analysis reveals significant interannual variability in extreme precipitation indices from 1981 to 2024. The total precipitation index (PRCPTOT) shows a slow upward trend overall, peaking at 789.2 mm in 1998 and reaching its minimum at 514.5 mm in 1985. Heavy precipitation (R95p) and very heavy precipitation (R99p) exhibit large fluctuations but a general upward trend, particularly increasing after 2010. Maximum 1-day precipitation (RX1day) and maximum 5-day precipitation (RX5day) display strong interannual variability with slow upward trends, reaching notably high values in 1998 and 2010, indicating severe extreme precipitation events during those years. Precipitation intensity (SDII) remains relatively stable in most years but increases during high-precipitation years, suggesting more concentrated and intense rainfall. The consecutive dry days (CDD) index shows a gradual overall decline, peaking at 54.1 days in 1985 and reaching its lowest value of 22.9 days in 2010, while consecutive wet days (CWD) remain relatively stable, fluctuating between 3–6 days but reaching 8 days in 1993.

Spatial distribution of four extreme precipitation indices (Figure [FIGURE:5]) shows higher values in eastern and southwestern mountainous areas and lower values in northwestern mountains and central plains. PRCPTOT reaches 502.95 mm in Xinyuan County and 482.3 mm in Zhaosu County, indicating these as the most precipitation-abundant regions. RX1day and RX5day are particularly elevated in Xinyuan County, with RX1day reaching 6.24 mm·d⁻¹, far exceeding other counties and demonstrating both high volume and intensity. The consecutive dry days (CDD) index shows a west-to-east decreasing gradient, with Holgos City (34.28 days) and Qapqal Xibe Autonomous County (33.15 days) experiencing longer dry spells, while Nilka County (25.75 days) receives more abundant precipitation. Consecutive wet days (CWD) are higher in eastern and southwestern mountains, with Zhaosu County reaching 6.01 days, significantly above the basin average of 3.6–4.8 days, indicating more persistent precipitation periods.

4.4 Temporal Distribution Characteristics of Future Extreme Precipitation Indices (2025–2050)

Projected extreme precipitation indices under SSP245 and SSP585 scenarios reveal distinct temporal patterns with significant fluctuations (Figure [FIGURE:6]). All eight indices show upward trends, with SSP585 exhibiting greater amplification and volatility. PRCPTOT increases under both scenarios, but SSP585 shows more dramatic fluctuations, ranging from 314.9 mm (2029) to 435.5 mm (2045), compared to SSP245's range of 384.5–404.5 mm. R95p and R99p display similar patterns, with SSP585 showing larger fluctuations and higher intensities, particularly peaking in 2045. RX1day fluctuates between 13.28–55.08 mm under SSP245 and 6.17–55.29 mm under SSP585, with SSP585 reaching extreme peaks in 2035 and 2045. RX5day shows minimal differences between scenarios, varying between 13.0–21.9 mm under SSP245 and 14.5–21.8 mm under SSP585, though SSP585 exhibits greater variability. Precipitation intensity (SDII) ranges from 1.65–1.91 mm·d⁻¹ under SSP245 and 1.74–1.98 mm·d⁻¹ under SSP585, with the high-emission scenario showing consistently higher values and greater extreme event probability. CDD shows no clear long-term trend but fluctuates significantly, while CWD increases slightly under both scenarios, particularly under SSP585.

4.5 Mann-Kendall Trend Characteristics of Future Extreme Precipitation Indices

Mann-Kendall trend analysis for 2025–2050 (Figure [FIGURE:7]) shows Z-values greater than 0 for most indices under both scenarios, indicating upward trends. Under SSP245, all indices except CDD show increasing trends, with PRCPTOT, RX1day, and RX5day demonstrating significant increases (Z-values 0.64–1.67). Under SSP585, trend significance strengthens, with Z-values ranging 0.90–1.97, suggesting that increased greenhouse gas emissions will intensify extreme precipitation. Spatial analysis dividing the basin into northern, southern, central, and eastern regions reveals that eastern and southern mountainous areas exhibit the strongest increasing trends under both scenarios, likely due to topographic and climatic sensitivity. The central region shows smaller increases under SSP245, while the SSP585 scenario produces more pronounced amplification across all stations.

4.6 Spatial Distribution of Future Extreme Precipitation Trends

Comparing projected precipitation with historical data reveals similar spatial patterns. Multi-year average daily precipitation (Figure [FIGURE:8]) shows higher values in southern regions under SSP245, while SSP585 expands high-value areas to include both southern and eastern regions. The spatial trend distribution (Figure [FIGURE:9]) demonstrates that under SSP245, PRCPTOT increases in eastern and southern mountains, with R95p and R99p also increasing in northeastern and southwestern mountains, elevating flood and geological disaster risks. The central plains show fewer extreme events. Under SSP585, trend magnitudes intensify, with Sen's slope values increasing across all stations. RX1day and RX5day show significant increases in eastern and southern mountains, with Zhaosu County exhibiting the highest slope value (0.21), while northern and central stations show weaker increases (0.11–0.13). CDD decreases in most stations under SSP245 but increases in Holgos City and Qapqal under SSP585, indicating longer dry spells. CWD increases under both scenarios, with more consistent amplification under SSP585, particularly in Zhaosu County, suggesting more frequent consecutive precipitation events.

5 Conclusions

This study analyzed spatio-temporal characteristics of extreme precipitation events in the Ili River Basin from 1981 to 2024 and projected their evolution under different CMIP6 scenarios through 2050. Key findings are:

1. During 1981–2024, most extreme precipitation indices exhibited increasing trends, particularly significant in eastern and southwestern mountainous areas where precipitation is concentrated and intense. Northwestern mountains and central plains showed relatively lower precipitation.

2. For 2025–2050, both SSP245 and SSP585 scenarios project upward trends in extreme precipitation indices, with SSP585 showing more frequent, intense, and volatile events. Eastern and southern mountainous regions are projected to experience significant increases in annual precipitation and heavy precipitation events, indicating stronger trends and higher risk. Northern and central plains will likely experience fewer extreme events. This spatial heterogeneity may differentially affect regional disaster frequency and agricultural/pastoral production.

3. As carbon emissions increase, extreme precipitation events in the Ili River Basin show clear upward trends, particularly in intensity indices (PRCPTOT, R95p, R99p, RX1day, RX5day). Statistical significance (Z-values) increases with emission levels, reflecting intensifying trends. Under high-emission SSP585, the basin may face more extreme precipitation events, increased annual precipitation, and higher precipitation intensity.

These results provide a scientific foundation for local government agencies to develop targeted prevention and mitigation strategies for extreme precipitation events, supporting regional sustainable development and disaster risk reduction.

References

[1] IPCC. Climate change the physical science basis: Contribution of working group I to the fifth assessment report of the Intergovernmental Panel on Climate Change[M]. Cambridge: Cambridge University Press, 2013.

[2] Ou T H, Chen D L, Linderholm W H, et al. Evaluation of global climate models in simulating extreme precipitation in China[J]. Tellus: Series A, Dynamic Meteorology and Oceanography, 2013, 65(1): 1-16.

[3] Xu H W, Chen H P, Wang H J. Future changes in precipitation extremes across China based on CMIP6 models[J]. International Journal of Climatology, 2022, 42(1): 635-651.

[4] Zhang Hongfang, Pan Liujie, Lu Shan, et al. Variation characteristics of extreme precipitation in Qinling and surrounding areas over the past 40 years[J]. Arid Land Geography, 2024, 47(3): 380-390.

[5] Zou Lei, Yu Jiangyou, Wang Feiyu, et al. Spatial temporal variations of extreme precipitation indices and their response to atmospheric circulation factors in the Weihe River Basin[J]. Arid Zone Research, 2021, 38(3): 764-774.

[6] Zhao Li, Han Xueyun, Yang Qing. Spatial and temporal variability of the extreme precipitation in the arid region of northwest China during recent 50 years[J]. Desert and Oasis Meteorology, 2016, 10(1): 19-26.

[7] Yang Xia, Zhou Hongkui, Zhao Keming, et al. Variation features of hourly precipitation in Xinjiang Province during 1991—2018[J]. Plateau Meteorology, 2020, 39(4): 762-773.

[8] Velazquez J A, Troin M, Caya D, et al. Evaluating the time invariance hypothesis of climate model bias correction: Implications for hydrological impact studies[J]. Journal of Hydrometeorology, 2015, 16(5): 2013-2026.

[9] Yang Xinxin. Analysis on spatial and temporal distribution of precipitation in Xinjiang Ili River Basin for more than 50 years[J]. Water Resources Development and Management, 2018(6): 65-68.

[10] Zhao Li, Yang Qing, Han Xueyun. Analysis of temporal and spatial distribution and variation characteristics of precipitation index in Ili during 1961—2009[J]. Journal of Arid Land Resources and Environment, 2024, 28(10): 82-89.

[11] Chen Shilong, Meng Qingkai, Dai Yong, et al. Geological disaster hazard assessment and prediction in the Ili River Basin based on CMIP6 future scenarios[J]. Arid Land Geography, 2025, 48(4): 599-611.

[12] Gong Xiao, Dai Changlei, Li Shuling, et al. Spatio temporal distribution characteristics and future trend analysis of extreme climate indices in Nenjiang River Basin over the past 60 years[J]. Pearl River. [2025-03-19]. http://kns.cnki.net/kcms/detail/44.1037.TV.20250120.1855.004.html.

[13] Wang Shuangshuang, Xie Wenqiang, Yan Xiaodong. Evalution on CMIP6 model simulation of the diurnal temperature range over China[J]. Climatic and Environmental Research, 2022, 27(1): 79-93.

[14] Hu Yiyang, Xu Ying, Li Jinjian, et al. Evaluation on the performance of CMIP6 global climate models with different horizontal resolution in simulating the precipitation over China[J]. Climate Change Research, 2021, 17(6): 730-743.

[15] Yang Xiaoling. Evaluation and projection of temperature and precipitation in China using CMIP6 model ensembles[D]. Nanjing: Nanjing University of Information Science and Technology, 2022.

[16] Xiang Junwen, Zhang Liping, Deng Yao, et al. Projection and evaluation of extreme temperature and precipitation in major regions of China by CMIP6 models[J]. Engineering Journal of Wuhan University, 2021, 54(1): 46-57, 81.

[17] Feng Anlan, Zhang Qiang, Song Jinbo, et al. Spatiotemporal characteristics of extreme precipitation in the Yellow River Basin based on CMIP6[J]. Journal of Beijing Normal University (Natural Science Edition), 2024, 60(2): 270-284.

[18] Zhang Xiaolu, Wang Xiaoxin, Hua Lijuan, et al. Projections of temperature and precipitation over Xinjiang based on CMIP6 models[J]. Chinese Journal of Atmospheric Sciences, 2023, 47(2): 387-398.

[19] Song J B, Zhang Q, Wu W H, et al. Amplifying flood risk across the lower Yellow River Basin, China, under shared socioeconomic pathways[J]. Frontiers in Earth Science, 2022, 10: 900866, doi: 10.3389/feart.2022.900866.

[20] Qi Bin. Temporal and spatial changes of extreme precipitation and flood disaster risk projection in the Wei River Basin based on CMIP6 models[D]. Xi'an University, 2023.

[21] Li Shuanglin, Han Leqiong, Bian Jie. Projecting heavy rainfall events in the middle and lower reach of the Yangtze River valley in the 21st century based on IPCC AR4 simulations[J]. Torrential Rain and Disasters, 2012, 31(3): 193-200.

[22] Wang L Y, Chen S F, Zhu W B, et al. Spatiotemporal variations of extreme precipitation and its potential driving factors in China North South Transition Zone during 1960—2017[J]. Atmospheric Research, 2021, 252: 105429, doi: 10.1016/j.atmosres.2020.105429.

[23] Liu Jing, Zhou Yushu, Yang Lianmei, et al. A diagnostic study of water vapor during an extreme precipitation event in the Yili River Valley[J]. Chinese Journal of Atmospheric Sciences, 2019, 43(5): 959-974.

[24] Yang Xia, An Dawei, Zhou Hongkui, et al. Daily variation of winter precipitation in Ili River valley of Xinjiang from 2012 to 2017[J]. Journal of Glaciology and Geocryology, 2020, 42(2): 609-619.

[25] Momin Gulzar, Yang Lianmei, Liu Yan, et al. Spatial and temporal distribution of flood disasters and characteristics of disaster prone environment in Ili River Basin in recent 30 years[J]. Arid Land Geography, 2024, 47(8): 1314-1326.

[26] Guan Yinghui. Extreme climate change and its trend prediction in the Yangtze River Basin[D]. Northwest A&F University, 2015.

[27] Li Huihui, Luan Chengmei, Xia Xu, et al. Prediction of future extreme precipitation scenarios in China based on CMIP6 climate model[J]. Water Resources and Hydropower Engineering, 2023, 54(8): 16-29.