Temporal and Spatial Changes of Extreme Precipitation Indices in Jilin Province During 1960–2019 and Future Projections Under CMIP6 Scenarios

Abstract

Extreme precipitation constitutes one of the most devastating climatic resulting from global climate change. Jilin Province, a significant commodities grain base in China by a temperate monsoon climate, is particularly susceptible to flood disasters caused by extreme precipitation, usually occurring from late July to early August. The 2010 flood impacted moreover 5.12 million individuals and resulted in direct economic damages amounting to 45.1 billion CNY. However, research on the spatiotemporal characteristics and future trends of extreme precipitation in Jilin Province is still quite inadequate. This study examined the spatiotemporal distribution and future forecasts of extreme precipitation utilizing daily meteorological data from 31 stations (1960–2019) and three CMIP6 models (CanESM5, MPI-ESM1-2-HR, FGOALS-g3) under SSP2-4.5 and SSP5-8.5 scenarios. Eleven extreme precipitation indices, as specified by the WMO, were analyzed utilizing linear regression, the Mann–Kendall test, wavelet analysis, and inverse distance weighting interpolation. The findings indicated that from 1960 to 2019, extreme precipitation demonstrated traits of “increased frequency and total amount, decreased intensity”, with a significant decline in CDD (−2.184 d·(10a)⁻¹, p < 0.001), a notable increase in PRCPTOT (1.493 mm·(10a)⁻¹, p < 0.05), and a significant reduction in SD II (−0.016 mm·d⁻¹·(10a)⁻¹, p < 0.01). The majority of indicators had a predominant cycle of 30 to 50 years. A significant northwest-to-southeast gradient characterized most indicators, with PRCPTOT varying from 327.5 mm in Baicheng to 824.3 mm in Tonghua. Future projections (2025–2100) suggested scenario-dependent intensification. Under SSP5-8.5, all three models forecast substantial increases in precipitation amount indices (PRCPTOT: 2.071–2.457 mm·(10a)⁻¹) and SD II (0.010–0.013 mm·d⁻¹·(10a)⁻¹), reversing the past downward trend in intensity. The anticipated alterations exhibited a northwest-to-southeast gradient, with PRCPTOT increases above 230 mm in the central and southeastern regions. These findings offer a scientific basis for flood management and climate change adaptation in Jilin Province and analogous areas.

1. Introduction

The Sixth Assessment Report of the Intergovernmental Panel on Climate Change (IPCC) unequivocally states that the rate and magnitude of changes in the current global climate system have surpassed historical records. The rise in the global average surface temperature, primarily driven by human activities (with a rise of 0.8–1.3 °C from 2010 to 2019 compared to pre-industrial levels), stands as the principal factor contributing to the frequent occurrence of extreme weather events [1].

Extreme precipitation events, among the most catastrophic meteorological disasters related to global climate change, are occurring with greater frequency and intensity. In recent years, secondary disasters including floods, landslides, and debris flows induced by excessive precipitation have occurred often worldwide. These disasters have resulted in significant mortality and extensive property destruction, while also imposing irrevocable effects on regional water resource security, agricultural practices, and ecological equilibrium [2,3]. In light of this severe scenario, rigorous study on extreme precipitation has become an urgent necessity to tackle climate change and foster regional sustainable development.

Recent years have witnessed substantial progress in studies on extreme precipitation at global, national, and regional scales. Studies have revealed a notable upward trend in the frequency and intensity of global extreme precipitation from 1980 to 2020, with distinct regional differences [4]. China is one of the countries most severely affected by extreme precipitation. In southern China, the frequency of hourly extreme precipitation the frequency increased in southern China, intensity mainly increased in northern China has increased; in northern China, the intensity has mainly risen; and in the southeastern region, the total daily precipitation has continued to increase [5]. In the arid northwestern part of China, despite an increase in the number of consecutive drought days, both average and extreme precipitation have shown an upward trend [6,7]. The frequency of extreme weather and climate events in Northeast China has increased significantly, which is considered a direct consequence of global warming has had impacts on regional food production [8]. Existing studies suggest that extreme precipitation in Northeast China is likely to become more frequent and intense in the future [9].

To date, research on extreme precipitation has evolved into a comprehensive multi-scale and multi-method system. The academic community extensively uses the 11 extreme precipitation indices endorsed by the World Meteorological Organization (WMO) as analytical parameters [10,11], thereby establishing these indices as standard instruments for the analysis of extreme precipitation [12].

The Mann–Kendall trend test, a non-parametric method first introduced by Mann [13] and later refined by Kendall [14], has been extensively utilized for identifying monotonic trends in hydroclimatic time series owing to its robustness, and lack of assumptions regarding data normality. Wavelet analysis has been progressively utilized to discern localized periodicities, long-term trends, and non-stationary characteristics in climate series, particularly for managing noisy hydrological data during climate change [15,16]. The empirical orthogonal function (EOF) method effectively identifies predominant spatial modes by decomposing multivariate fields into principal components that optimize variance contribution [17]. Meanwhile, inverse distance weighting (IDW) interpolation effectively reconstructs the spatial distribution of precipitation indices from discrete station observations, thereby elucidating regional heterogeneity [18].

Accurate prediction of future trends in extreme precipitation forms the basis for developing medium- and long-term climate adaptation strategies. The Coupled Model Intercomparison Project (CMIP) has entered its sixth phase (CMIP6). Compared to previous generations, CMIP6 demonstrates improvements in the number of models, data accuracy, and simulation ability [19,20,21].

Jilin Province is located at the geographical center of Northeast Asia and serves as a major commercial grain production base in China [22]. Studies conducted in Jilin Province and neighboring regions have shown that the Northeast Asian region experienced a decadal abrupt increase in summer water vapor deficit in the late 1990s [23]; extreme precipitation in southern Eastern Siberia and Mongolia exhibits a northwestward shift [24]; summer precipitation anomalies in the Selenga River basin are the primary cause of low-flow periods [25]; extreme precipitation events in Liaoning Province showed a decreasing trend from 1991 to 2021, although some regions still experienced an increase frequency and intensity [21].

Due to its geographical location, with the western part bordering the arid Inner Mongolia Plateau and being far from the ocean, the western region of Jilin Province from water shortages. The southeastern region is also prone to drought because of the barrier formed by the Changbai Mountains. However, the eastern mountainous areas are adjacent to the Songhua River and the Sea of Japan. Warm, moist air currents from the southwest influence, northeastward movement of water vapor by the southwest monsoon, leading to significant moisture in the region [26,27]. Consequently, the region is also susceptible to extreme precipitation events such as floods and waterlogging. Given Jilin Province’s climate sensitivity and its heavy reliance on agriculture, it holds a unique position in the study of extreme precipitation [28,29].

Nonetheless, there has been relatively limited research on spatiotemporal characteristics and future trends of extreme precipitation in Jilin Province. In light of this, the specific objectives of this study are as follows: (1) Based on daily meteorological observation data from 31 weather stations in Jilin Province from 1960 to 2019, this study will employ the 11 extreme precipitation indices recommended by the WMO, combined with methods such as simple linear regression, the Mann–Kendall test, IDW interpolation, and wavelet analysis, to systematically describe the spatiotemporal evolution of extreme precipitation during this period. (2) This study will utilize data from the CanESM5, MPI-ESM1-2-HR, and FGOALS-g3 models within the CMIP6 project to extract daily grid data for 31 meteorological stations from 2025 to 2100, aiming to analyze future spatiotemporal trends of various extreme precipitation indicators under distinct climate scenarios. This work establishes a scientific foundation for disaster prevention programs in Jilin Province and analogous climatic regions as well as for forecasting future alterations in extreme precipitation events.

2. Materials and Methods

2.1. Study Area

Jilin Province is located at the geometric center of Northeast Asia (40°50′–46°19′ N, 121°38′–131°19′ E). Northeast Asia includes Japan, Russia, North Korea, South Korea, Mongolia, and northeastern China. The province spans 769.62 km from east to west and 606.57 km from north to south, covering a total land area of 187,400 km², approximately 2% of China’s territorial expanse. Geomorphologically, Jilin Province exhibits a pronounced southeast-to-northwest gradient, characterized by elevated terrains in the southeast and lower elevations in the northwest. It largely comprises two principal geomorphic units: the eastern mountainous region and the central-western plains. The Changbai Mountains in the southeast and the low-lying plains in the west create notable topographic variations throughout the province, affecting the spatial distribution of precipitation. located on the eastern side of the Eurasian continent in the mid-latitudes, Jilin Province has a moderate continental monsoon climate characterized by four distinct seasons with simultaneous periods of rainfall and warmth. Annual precipitation generally varies from 400 to 600 mm, exhibiting significant seasonal and geographical disparities, with approximately 80% occurring in summer and the eastern section of the province having the highest levels of precipitation [30].

2.2. Data

This study employed daily precipitation, maximum temperature, and minimum temperature data gathered from 31 national-level ground observation stations in Jilin Province (Appendix A 

Table A1. Information for the 31 meteorological stations in Jilin Province and their corresponding CMIP6 grid points.

Number	Station Name	Station Latitude (°N)	Station Longitude (°E)	Grid Latitude (°N)	Grid Longitude (°E)
50936	Baicheng	45.37	122.49	45.375	122.375
50945	Daan	45.3	124.16	45.375	124.125
50948	Qianan	45	124.1	44.875	124.125
50949	Qianguo	45.05	124.52	45.125	124.625
54041	Tongyu	44.48	123.4	44.375	123.375
54049	Changling	44.15	123.58	44.125	123.625
54063	Fuyu	44.58	126	44.625	125.875
54064	Nongan	44.23	125.09	44.125	125.125
54142	Shuangliao	43.3	123.31	43.375	123.375
54157	Siping	43.1	124.19	43.125	124.125
54161	Changchun	43.54	125.13	43.625	125.125
54165	Shuangyang	43.33	125.38	43.375	125.375
54169	Yantongshan	43.18	126.01	43.125	126.125
54172	Jilinchengjiao	43.52	126.32	43.625	126.375
54181	Jiaohe	43.42	127.2	43.375	127.125
54186	Dunhua	43.22	128.12	43.125	128.125
54192	Luozigou	43.42	130.16	43.375	130.125
54195	Wangqing	43.18	129.47	43.125	129.375
54260	Liaoyuan	42.55	125.05	42.625	125.125
54263	Panshi	42.58	126.05	42.625	126.125
54266	Meihekou	42.32	125.38	42.375	125.375
54273	Huadian	42.59	126.45	42.625	126.375
54276	Jingyu	42.24	126.48	42.125	126.375
54284	Donggang	42.09	127.3	42.125	127.375
54285	Erdao	42.25	128.07	42.125	128.125
54286	Helong	42.32	129	42.375	128.875
54292	Yanji	42.52	129.3	42.625	129.375
54363	Tonghua	41.41	125.54	41.375	125.625
54374	Linjiang	41.48	126.53	41.375	126.375
54377	Jian	41.09	126.13	41.125	126.125
54386	Changbai	41.25	128.11	41.375	128.125

), covering the years from 1960 to 2019. Precipitation data were obtained from the China Meteorological Science Data Sharing Service Platform (http://data.cma.cn/, accessed on 26 February 2026) and subjected to stringent quality control measures. Figure 1 depicts the geographical distribution of Jilin Province and its meteorological stations. The areal mean precipitation for Jilin Province was calculated utilizing the inverse distance weighting (IDW) approach.

This research, building on prior studies, selected three commonly utilized and validated CMIP6 climate models for northern China: CanESM5, MPI-ESM1-2-HR, and FGOALS-g3 [31,32]. The SSP2-4.5 and SSP5-8.5 scenarios were selected for examination. Daily precipitation and temperature data from 2025 to 2100 were obtained from grid points associated with 31 meteorological stations or their nearest grid cells (Appendix A Table A1). These data were sourced from the China Regional CMIP6 Downscaled Precipitation [33], Temperature, and Wind Speed Dataset maintained by the National Qinghai-Tibet Plateau Science Data Center (http://data.tpdc.ac.cn, accessed on 26 February 2026). This dataset underwent bias-correction at a resolution of 0.25°, utilizing the equidistant cumulative distribution function (EDCDF) approach, referencing the China Meteorological Driving Force Dataset (CMDF) as the observational standard. Upon rectification, the correlation between model outputs and observations for monthly precipitation nears unity, validating the dependability of the downscaled data for regional climate study. In accordance with the official CMIP6 recommendation, the period from 1995 to 2014 was established as the baseline for quantifying future alterations in extreme precipitation indices.

2.3. Methods

2.3.1. Extreme Precipitation Indicators

The definition and criteria for extreme precipitation occurrences differ among research, typically involve the development of specified thresholds for precipitation intensity, duration, and regional distribution. This study utilized 11 extreme precipitation indices as specified by the Expert Group on Climate Change Detection and Indices (EGCCDI) of the World Meteorological Organization [34].

Table 1. Definition of extreme precipitation index.

Indicator Name	Index	Definition	Units
Number of heavy precipitation days	R10	Annual count of days with RR ≥ 10 mm	days
Number of very heavy precipitation days	R20	Annual count of days with RR ≥ 20 mm	days
Number of extreme heavy precipitation days	R50	Annual count of days with RR ≥ 50 mm	days
Consecutive dry days	CDD	Maximum number of consecutive days with RR < 1 mm	days
Consecutive wet days	CWD	Maximum number of consecutive days with RR ≥ 1 mm	days
Very wet days	R95p	Total annual precipitation from days with RR > 95th percentile	mm
Extreme wet days	R99p	Total annual precipitation from days with RR > 99th percentile	mm
Max 1-day precipitation amount	Rx1day	Annual maximum 1-day precipitation amount	mm
Max 5-day precipitation amount	Rx5day	Annual maximum 5-day precipitation amount	mm
Simple precipitation intensity index	SD II	The ratio annual total wet-day precipitation to the number of wet days	mm/day
Annual total wet-day precipitation	PRCPTOT	Total annual precipitation from days with RR ≥ 1 mm	mm

provides comprehensive information regarding the names, abbreviations, and definitions of these indexes. These indices can be classified into three categories [35]: (1) extreme precipitation day indices (R10, R20, R50, CDD, CWD), which measure the occurrence of extreme precipitation occurrences. R10, R20, and R50 quantify the number of days with precipitation at specified intensities, while CDD and CWD signify the frequencies of dry and wet spells, respectively; (2) extreme precipitation quantity indices (R95p, R99p, Rx1day, Rx5day, PRCPTOT) represent the total precipitation volume and the overall characteristics of extreme heavy precipitation events; and (3) the extreme precipitation intensity index (SD II) denotes the concentration of precipitation.

2.3.2. Inverse Distance Weighting Interpolation

Spatial interpolation was conducted using the inverse distance weighting (IDW) method, wherein the estimated value at an unsampled location is computed as a weighted average of surrounding observations, with weights diminishing as the distance from the prediction point increases. The interpolated value at location ${\mathit{x}}_{\mathbf{0}}$ is expressed as:

$$\mathit{Z}({\mathit{x}}_{\mathbf{0}})={\displaystyle \frac{{\sum }_{\mathit{i}=\mathbf{1}}^{\mathit{n}}{\mathit{w}}_{\mathit{i}}\cdot \mathit{Z}\left({\mathit{x}}_{\mathit{i}}\right)}{{\sum }_{\mathit{i}=\mathbf{1}}^{\mathit{n}}{\mathit{w}}_{\mathit{i}}}}$$

(1)

where $\mathit{Z}\left({\mathit{x}}_{\mathit{i}}\right)$ is the observed value at location ${\mathit{x}}_{\mathit{i}}$ and ${\mathit{w}}_{\mathit{i}}$ is the weight assigned to point ${\mathit{x}}_{\mathit{i}}$. The weight is defined as:

$${\mathit{w}}_{\mathit{i}}={\displaystyle \frac{\mathbf{1}}{\mathit{d}{\left({\mathit{x}}_{\mathit{i}},{\mathit{x}}_{\mathbf{0}}\right)}^{\mathit{p}}}}$$

(2)

where $\mathit{d}({\mathit{x}}_{\mathit{i}},{\mathit{x}}_{\mathbf{0}})$ is the distance between the known point ${\mathit{x}}_{\mathit{i}}$ and the prediction location ${\mathit{x}}_{\mathbf{0}}$, and $\mathit{p}$ is the power parameter controlling the rate of distance decay. In this study, the optimal value of this parameter was determined to be 2.75 using leave-one-out cross-validation based on the minimum interpolation error criterion [18].

2.3.3. Mann–Kendall Trend Test and Change-Point Detection

The Mann–Kendall (MK) test is a non-parametric method widely used to detect monotonic trends in time series [36]. For a time, series $\mathit{X}=({\mathit{x}}_{\mathbf{1}},{\mathit{x}}_{\mathbf{2}},\dots ,{\mathit{x}}_{\mathit{n}})$, the MK statistic $\mathit{S}$ is calculated as:

$$\mathit{S}=\sum _{\mathit{i}=\mathbf{1}}^{\mathit{n}-\mathbf{1}}\sum _{\mathit{j}=\mathit{i}+\mathbf{1}}^{\mathit{n}}\mathbf{s}\mathbf{g}\mathbf{n}({\mathit{x}}_{\mathit{j}}-{\mathit{x}}_{\mathit{i}})$$

(3)

$$\mathit{S}\mathit{g}\mathit{n}({\mathit{x}}_{\mathit{j}}-{\mathit{x}}_{\mathit{i}})=\left\{\begin{array}{ll}\mathbf{1},& {\mathit{x}}_{\mathit{j}}>{\mathit{x}}_{\mathit{i}}\\ \mathbf{0},& {\mathit{x}}_{\mathit{j}}={\mathit{x}}_{\mathit{i}}\\ -\mathbf{1},& {\mathit{x}}_{\mathit{j}}<{\mathit{x}}_{\mathit{i}}\end{array}\right.$$

(4)

A positive value of $\mathit{S}$ indicates an increasing trend, whereas a negative value indicates a decreasing trend. For $\mathit{n}>\mathbf{8}$, $\mathit{S}$ approximately follows a normal distribution with variance:

$$\mathit{Var}\left(\mathit{S}\right)={\displaystyle \frac{\mathit{n}\left(\mathit{n}-\mathbf{1}\right)\left(\mathbf{2}\mathit{n}+\mathbf{5}\right)-{\sum }_{\mathit{k}=\mathbf{1}}^{\mathit{n}}{\mathit{t}}_{\mathit{k}}\left({\mathit{t}}_{\mathit{k}}-\mathbf{1}\right)\left(\mathbf{2}{\mathit{t}}_{\mathit{k}}+\mathbf{5}\right)}{\mathbf{18}}}$$

(5)

where $\mathit{n}$ is the number of tied groups and ${\mathit{t}}_{\mathit{k}}$ is the number of data points in the $\mathit{k}$-th tied group. The standardized test statistic $\mathit{Z}$ is given by

$${\mathit{Z}}_{\mathit{M}\mathit{K}}=\left\{\begin{array}{ll}{\displaystyle \frac{\mathit{S}-\mathbf{1}}{\sqrt{\mathit{V}\mathit{a}\mathit{r}\left(\mathit{S}\right)}}},& \mathit{S}>\mathbf{0}\\ \mathbf{0},& \mathit{S}=\mathbf{0}\\ {\displaystyle \frac{\mathit{S}+\mathbf{1}}{\sqrt{\mathit{V}\mathit{a}\mathit{r}\left(\mathit{S}\right)}}},& \mathit{S}<\mathbf{0}\end{array}\right.$$

(6)

An upward trend is identified when $\mathit{Z}>\mathbf{0}$, and a downward trend is identified when $\mathit{Z}<\mathbf{0}$. The trend is considered statistically significant at the 95% confidence level when $\mid \mathit{Z}\mid >\mathbf{1.96}$.

Potential abrupt changes in the time series were further examined using the sequential Mann–Kendall (SQMK) test and the Bai–Perron structural breakpoint test. The SQMK test was utilized to initially identify potential change sites by examining the intersection of the forward ($\mathit{U}{\mathit{F}}_{\mathit{k}}$) and backward ($\mathit{U}{\mathit{B}}_{\mathit{k}}$) standardized statistics. Intersections occurring within the critical boundaries of $\pm \mathbf{1.96}$ were considered significant at the 0.05 level. The Bai–Perron test was utilized to detect and confirm structural changes in the series by identifying one or more statistically significant breakpoints and establishing the best segmentation of the time series. By integrating the results of the SQMK and Bai–Perron tests, the abrupt change years were determined [37].

2.3.4. Morlet Wavelet Analysis

Wavelet analysis was utilized to discern localized temporal characteristics of the precipitation series, encompassing multi-scale variability and periodic oscillations. The discrete wavelet representation is as follows [38]:

$${\mathit{W}}_{\mathit{f}}\left(\mathit{a},\mathit{b}\right)={\displaystyle \frac{\mathbf{1}}{\sqrt{\left|\mathit{a}\right|}}}\sum _{\mathit{k}=\mathbf{1}}^{\mathit{N}}\mathit{f}\left(\mathit{k}\mathsf{\Delta }\mathit{t}\right)\mathsf{\Delta }\mathit{t}\overline{\mathit{\Psi }}\left({\displaystyle \frac{\mathit{k}\mathsf{\Delta }\mathit{t}-\mathit{b}}{\mathit{a}}}\right)$$

(7)

where $\mathit{a}$ and $\mathit{b}$ denote the scale and translation factors; $\mathit{\psi }\left(\cdot \right)$ is the mother wavelet function; $\mathit{f}\left(\cdot \right)$ is the variable time sequence. Positive and negative real parts of the wavelet coefficients indicate relatively wet and dry conditions, respectively, at a given time and scale.

To identify the dominant periodicities, the wavelet variance was calculated as:

$${\mathit{w}}_{\mathit{f}}\left(\mathit{a}\right)={\int }_{-\mathit{\infty }}^{\mathit{\infty }}{\left|{\mathit{w}}_{\mathit{f}}\left(\mathit{a},\mathit{b}\right)\right|}^{\mathbf{2}}\mathit{d}\mathit{b}$$

(8)

where ${\mathit{w}}_{\mathit{f}}\left(\mathit{a}\right)$ represents the distribution of signal energy across scales. Wavelet variances were employed to discern the predominant periodicities in the precipitation series.

2.3.5. Spatial Modal Analysis

Empirical orthogonal function (EOF) decomposition was used to identify the dominant spatial patterns of extreme precipitation and their associated temporal variations [17]. The precipitation indices were arranged into a matrix $\mathit{X}$ with $\mathit{m}$ spatial points and $\mathit{n}$ temporal observations. EOF decomposition of the covariance matrix yielded eigenvalues and eigenvectors, and the eigenvector of the $\mathit{j}$-th mode is expressed as:

$${\mathit{W}}_{\mathit{j}}={\left({\mathit{W}}_{\mathbf{1}\mathit{j}},{\mathit{W}}_{\mathbf{2}\mathit{j}},\dots {,\mathit{W}}_{\mathit{m}\mathit{j}}\right)}^{\mathit{T}}$$

(9)

where ${\mathit{W}}_{\mathit{j}}$ represents the spatial pattern of the $\mathit{j}$-th EOF mode. The corresponding temporal coefficients were calculated as:

$$\mathit{P}={\mathit{W}}^{\mathit{T}}\mathit{X}$$

(10)

where $\mathit{P}$ is the matrix of temporal coefficients.

3. Results

3.1. Characteristics of Temporal Change Trend of Extreme Precipitation

From 1960 to 2019, the extreme precipitation indices in Jilin Province exhibited notable temporal patterns. R10, R20, and R50 (Figure 2a–c) demonstrated minor, non-significant rising trends (0.003–0.03 d·(10a)⁻¹), whereas CWD (Figure 2e) saw a tiny reduction (−0.001 d·(10a)⁻¹). CDD (Figure 2d) exhibited a significant reduction of −2.184 d·(10a)⁻¹ (p < 0.001), signifying a considerable decrease in the duration of consecutive dry spells. Among the indices of precipitation quantity, R95p, R99p, and Rx1day (Figure 2f–h) exhibited a non-significantly rise, whereas Rx5day (Figure 2i) experienced a minor fall. PRCPTOT (Figure 2k) increased considerably at a rate of 1.493 mm·(10a)⁻¹ (p < 0.05). The intensity index SD II (Figure 2j) exhibited a substantial decrease of −0.016 mm·d⁻¹·(10a)⁻¹ (p < 0.01). The 5-year moving average indicated that CDD and SD II were comparatively elevated throughout the 1960s and 1970s; R10, R20, R50, CWD, R95p, and PRCPTOT attained their zenith the 1980s–1990s, whereas R99p, Rx1day, and Rx5day peaked in the 2010s. Overall, only CDD and SD II exhibited significant decreases, whereas PRCPTOT demonstrated a substantial rise, indicating that the frequency and total amount of extreme precipitation usually grew, although precipitation intensity diminished over the study period.

The Mann–Kendall mutation test, in conjunction with the Bai–Perron structural breakpoint test, was utilized to identify abrupt change points (Figure 3 and

Table 2. Statistical values of extreme precipitation index characteristics in Jilin Province from 1960 to 2019.

Index	Regression Scope	T-Test p-Value	MK z-Value	MK Trends	MK p-Value	Point of Change	Year
R10	0.03	0.124	0.874	↑	0.382
R20	0.013	0.242	0.938	↑	0.348
R50	0.003	0.295	1.078	↑	0.281
CDD	−2.184 ***	<0.001	−6.002 ***	↓	<0.001	1	1979
CWD	−0.001	0.748	−0.529	↓	0.597
R95p	0.308	0.377	0.644	↑	0.519
R99p	0.168	0.374	1.193	↑	0.233
Rx1day	0.021	0.765	0.721	↑	0.471
Rx5day	−0.013	0.911	−0.733	↓	0.463
SD II	−0.016 **	<0.01	−2.417 *	↓	<0.05	1	1975
PRCPTOT	1.493 *	<0.05	1.818	↑	0.0691	1	1982

Notes: “*” indicates that passed the significance level of 95%; “**” the significance level of 99%; “***” the significance level of 99.99%. ↑ indicates an increasing trend, and ↓ indicates a decreasing trend.

). The UF curves for the majority of indices exhibited fluctuations devoid of distinct turning points. Discernible rapid changes were detected for CDD (1979), SDII (1975), and PRCPTOT (1982), with CDD and SD II exhibiting statistically significant z-values that corroborate their long-term declining patterns, whereas the z-value for PRCPTOT lacked significance. For the remaining indices, although the UF and UB curves overlap multiple times within the 95% confidence interval, the trends preceding and after these intersections rarely achieve significance, suggesting an absence of meaningful abrupt change.

3.2. Characterization of Extreme Precipitation Cycle Change Trends

Wavelet analysis (Figure 4) revealed multi-scale rhythmic features across all 11 indices, demonstrating a common short-term cycle of 10–20 years. R10, R50, CWD, R99p, Rx1day, and Rx5day (Figure 4a,c,e,g–i) exhibited a three-tier periodic structure comprising a brief cycle of 10–20 years, a predominant cycle of 35–50 years, and an extended cycle of 50–60 years. R20, R95p, and PRCPTOT (Figure 4b,f,k) exhibited a similar structure, but with a reduced dominant cycle of 30–45 years. CDD (Figure 4d) exhibited a predominant cycle of 35–45 years and a secondary cycle of 45–60 years. SD II (Figure 4j) had a brief cycle of 10–20 years and an interdecadal cycle of 30–45 years, with feeble signals over extended time scales. Dominant cycles were primarily seen within the 30–50-year range, with the intensity of the periodic signal fluctuating with time, indicating the quasi-periodic characteristics of intense precipitation variability.

3.3. Characterization of Spatial Trends in Extreme Precipitation

All extreme precipitation indices, with the exception of SD II, exhibited a notable northwest-to-southeast increasing gradient (Figure 5a–i,k). Stations in the northwest, exemplified by Baicheng, consistently displayed the lowest metrics (e.g., R10: 11.2 d; R20: 5.4 d; R95p: 96.4 mm; PRCPTOT: 327.5 mm), while southeastern stations such as Ji’an and Tonghua exhibited the highest values (e.g., R10: 26.1 d at Ji’an; R20: 12.2 d at Tonghua; R95p: 244.1 mm at Ji’an; PRCPTOT: 824.3 mm at Tonghua). Rx1day and Rx5day displayed slight discrepancies, with eastern locations, such as Yanji (Rx1day: 51 mm) and Luozigou (Rx5day: 75.8 mm), recording relatively low values, indicating a more complex spatial arrangement. In contrast, SD II (Figure 5j) demonstrated an increase from east to west, with heightened intensity at western and southwestern stations, including Ji’an (11.1 mm·d⁻¹), Changchun (9.4 mm·d⁻¹), and Baicheng (9.3 mm·d⁻¹), while diminished intensity was noted at eastern stations such as Wangqing (8.0 mm·d⁻¹) and Luozigou (7.5 mm·d⁻¹).

An EOF decomposition of R50, R99p, and Rx5day further analyzed the principal spatial modes (Appendix A 

Table A2. Cumulative variance contribution rates of the first three principal component eigenvalues for R50, R99p, and Rx5day.

Extreme Precipitation Index	Modal	Eigenvalue	Cumulative Variance Contribution Rate/%
R50	1	5.898	19.027
	2	3.175	29.268
	3	2.366	36.9
R99p	1	5.681	18.325
	2	3.255	28.824
	3	2.376	36.488
Rx5day	1	6.407	20.668
	2	4.263	34.421
	3	2.729	43.225

). The cumulative variance contributions of the first three principal components were 36.9%, 36.488%, and 43.225%, respectively. The preliminary mode of all three indices (Figure 6a,d,g) demonstrated heightened precipitation in the central-southern and southern regions, particularly near Liaoyuan, Jilin, and Tonghua. The correlated temporal coefficients (Figure 7a) were predominantly negative from 1960 to 1980 and shifted to positive values after 1990, signifying a prolonged intensification of extreme precipitation in these regions. The second mode (Figure 6b,e,h) exhibited a north–south spatial disparity, with temporal coefficients (Figure 7b) fluctuating consistently between positive and negative values during the study period, indicating alternating dry and wet conditions. The third mode (Figure 6c,f,i) was centered in the southeastern region, with significant fluctuations in time coefficients (Figure 7c) that indicated intense precipitation activity in this area. All three indices exhibited largely consistent spatial patterns across the three modes, implying cohesive large-scale influences on extreme precipitation.

3.4. Analysis of Changes in Extreme Precipitation Indices Under Future Scenarios

3.4.1. Projected Temporal Trends

The extreme precipitation indices forecasted by the CanESM5, MPI-ESM1-2-HR, and FGOALS-g3 models displayed diverse patterns under the SSP2-4.5 and SSP5-8.5 scenarios (Appendix A 

Table A3. Statistical Values of the CanESM5 Future Extreme Precipitation Index for Jilin Province.

Index	Baseline		SSP245		SSP585
	Regression Scope	T-Test p-Value	Regression Scope	T-Test p-Value	Regression Scope	T-Test p-Value
R10	0.078	0.542	0.044	0.062	0.043	0.068
R20	0.029	0.655	0.016	0.118	0.016	0.175
R50	0.005	0.495	0.008 ***	<0.001	0.011 ***	<0.001
CDD	0.491	0.129	−0.0007	0.990	−0.060	0.253
CWD	−0.163	0.480	0.063	0.047	−0.010	0.727
R95p	1.060	0.635	0.874	0.024	1.090	0.011
R99p	1.109	0.357	0.746 ***	<0.001	1.048 ***	<0.001
Rx1day	0.385	0.177	0.216 ***	<0.001	0.260 ***	<0.001
Rx5day	0.353	0.554	0.180	0.077	0.459 ***	<0.001
SD II	0.012	0.544	0.0095 *	<0.05	0.011 **	<0.01
PRCPTOT	−0.796	0.797	2.360 ***	<0.001	2.071 **	<0.01

Notes: “*” indicates that passed the significance level of 95%; “**” the significance level of 99%; “***” the significance level of 99.99%.

,

Table A4. Statistical Values of the MPI-ESM1-2-HR Future Extreme Precipitation Index for Jilin Province.

Index	Baseline		SSP245		SSP585
	Regression Scope	T-Test p-Value	Regression Scope	T-Test p-Value	Regression Scope	T-Test p-Value
R10	−0.016	0.888	0.030	0.083	0.101 ***	<0.001
R20	−0.034	0.360	0.008	0.191	0.022 **	<0.01
R50	−0.0003	0.785	0.0004	0.212	0.0009	0.119
CDD	−0.250	0.139	0.0006	0.981	−0.042	0.117
CWD	−0.018	0.872	−0.011	0.564	0.051 *	<0.05
R95p	−1.145	0.464	0.388	0.101	0.845 ***	<0.001
R99p	−0.877	0.251	0.121	0.341	0.325 *	<0.05
Rx1day	−0.230	0.081	0.023	0.320	0.071 **	<0.01
Rx5day	−0.395	0.264	0.013	0.790	0.129 *	<0.05
SD II	−0.006	0.651	0.002	0.193	0.010 ***	<0.001
PRCPTOT	−1.626	0.601	0.673	0.132	2.457 ***	<0.001

Notes: “*” indicates that passed the significance level of 95%; “**” the significance level of 99%; “***” the significance level of 99.99%.

and

Table A5. Statistical Values of the FGO-ALS-g3 Future Extreme Precipitation Index for Jilin Province.

Index	Baseline		SSP245		SSP585
	Regression Scope	T-Test p-Value	Regression Scope	T-Test p-Value	Regression Scope	T-Test p-Value
R10	0.057	0.700	0.026	0.182	0.057 *	<0.05
R20	0.033	0.701	0.014	0.240	0.026 *	<0.05
R50	0.042 *	<0.05	0.001	0.956	0.012 ***	<0.001
CDD	−0.164	0.637	0.013	0.845	−0.139	0.052
CWD	0.042	0.676	0.010	0.280	−0.006	0.620
R95p	3.129	0.258	0.168	0.738	1.272 **	<0.01
R99p	2.807	0.066	−0.037	0.735	0.780 ***	<0.001
Rx1day	0.761	0.070	−0.004	0.875	0.221 **	<0.01
Rx5day	0.953	0.089	0.133	0.328	0.333 **	<0.01
SD II	0.022	0.531	0.002	0.682	0.013 *	<0.05
PRCPTOT	3.181	0.504	0.693	0.273	2.088 **	<0.01

Notes: “*” indicates that passed the significance level of 95%; “**” the significance level of 99%; “***” the significance level of 99.99%.

).

Under SSP2-4.5, the majority of indicators exhibited mild, non-significant trends across all three models. The exception was CanESM5 (Figure 8), where R99p (0.746 mm·(10a)⁻¹, p < 0.001; Figure 8g), Rx1day (0.216 mm·(10a)⁻¹, p < 0.001; Figure 8h), PRCPTOT (2.360 mm·(10a)⁻¹, p < 0.001; Figure 8k), and SD II (0.0095 mm·d⁻¹·(10a)⁻¹, p < 0.05; Figure 8j) exhibited substantial increases. MPI-ESM1-2-HR (Figure 9) and FGOALS-g3 (Figure 10) exhibited inconclusive albeit statistically insignificant trends.

Under the SSP5-8.5 scenario, all three models projected notable increases in precipitation amount indices and SD II. R99p, Rx1day, Rx5day, and PRCPTOT demonstrated statistically significant increases across all models (p < 0.05 to p < 0.001; Figure 8g–i,k, Figure 9g–i,k and Figure 10g–i,k). The trends of PRCPTOT varied from 2.071 to 2.457 mm·(10a)⁻¹, while those of R99p ranged from 0.325 to 1.048 mm·(10a)⁻¹. R95p attained significance in MPI-ESM1-2-HR (0.845 mm·(10a)⁻¹, p < 0.001; Figure 9f) and FGOALS-g3 (1.272 mm·(10a)⁻¹, p < 0.01; Figure 10f) but not in CanESM5. SD II exhibited a substantial rise across all models, varying from 0.010 to 0.013 mm·d⁻¹·(10a)⁻¹ (Figure 8j, Figure 9j and Figure 10j). The frequency indices exhibited model-dependent variations: R10, R20, and R50 generally increased, with notable trends in MPI-ESM1-2-HR (e.g., R10: 0.101 d·(10a)⁻¹, p < 0.001; Figure 9a) and FGOALS-g3 (e.g., R50: 0.012 d·(10a)⁻¹, p < 0.001; Figure 10c), whereas only R50 attained significance in CanESM5 (0.011 d·(10a)⁻¹, p < 0.001; Figure 8c). CWD exhibited a substantial rise solely in MPI-ESM1-2-HR (0.051 d·(10a)⁻¹, p < 0.05; Figure 9e). CDD diminished across all models although did not attain statistical significance (Figure 8d, Figure 9d and Figure 10d).

3.4.2. Projected Spatial Distribution

Projected changes relative to the baseline period (1995–2014) were analyzed at grid points corresponding to or nearest to observational stations (Figure 11, Figure 12, Figure 13, Figure 14, Figure 15 and Figure 16).

Under SSP2-4.5, the majority of indices exhibited minimal variation with restricted spatial coherence. CanESM5 projected the most significant increases in PRCPTOT, varying from 119.37 mm in Da’an to 183.97 mm in Ji’an (Figure 11k). CWD exhibited notable inter-model divergence: CanESM5 and MPI-ESM1-2-HR projected overall increases, but FGOALS-g3 projected extensive decreases (e.g., Jingyu: −16.37 d) (Figure 11e, Figure 13e and Figure 15e).

Under SSP5-8.5, all models projected more extensive and spatially consistent alterations. The indices of precipitation amounts demonstrated a steady gradient of increase from northwest to southeast. PRCPTOT rises above 230 mm at central and southeastern grid points in CanESM5 (e.g., Huadian: 244.08 mm), 190 mm in MPI-ESM1-2-HR (e.g., Tonghua: 195.78 mm), and 200 mm in FGOALS-g3 (e.g., Erdao: 200.47 mm), while northwestern grid points exhibited lesser increases (Figure 12k, Figure 14k and Figure 16k). CDD decreased more significantly than in SSP2-4.5, with the most pronounced reductions noted in CanESM5 (e.g., Fuyu: −6.28 d; Changchun: −5.37 d) (Figure 12d, Figure 14d and Figure 16d). SD II escalated throughout the province, with elevated values in the southwestern and southeastern regions (e.g., Shuangliao: 1.16 mm·d⁻¹ in CanESM5; Ji’an: 1.18 mm·d⁻¹ in MPI-ESM1-2-HR) (Figure 12j, Figure 14j and Figure 16j). The inter-model disparity in CWD continued: FGOALS-g3 projected extensive reductions (e.g., Jingyu: −14.78 d), whereas CanESM5 and MPI-ESM1-2-HR projected overall increases (Figure 12e, Figure 14e and Figure 16e).

Overall, the precipitation amount indices exhibited a northwest-to-southeast increasing gradient under both scenarios, with the most significant increases occurring in the middle and southeastern regions. SD II exhibited the most significant growth in the southwest and southeast regions. CDD transitioned from model-dependent patterns under SSP2-4.5 to extensive reductions under SSP5-8.5. The most significant inter-model divergence was CWD, with FGOALS-g3 projecting declines, whereas the other two models projected increases.

4. Discussion

4.1. Temporal Evolution Patterns of Extreme Precipitation

From 1960 to 2019, extreme precipitation events in Jilin Province demonstrated a complex and heterogeneous trend. Specifically, both the frequency and total volume of extreme precipitation generally exhibited an upward trend, whereas the intensity of precipitation showed a declining tendency. This pattern was in line with the relationship between inter-decadal variation of summer precipitation in East Asia and the intensity of the Asian summer monsoon [39,40]. The increasing trends were consistent with the Clausius–Clapeyron equation, as global warming enhanced the atmospheric water-holding capacity and the water vapor transport capacity of the East Asian summer monsoon [41,42,43]. The concurrent decreases in SD II and CDD were in accordance with observations across China. After the 1970s, the inter-decadal weakening of the East Asian Summer Monsoon reduced its northward transport capacity, resulting in precipitation becoming more dispersed rather than concentrated [44,45,46]. Comparable trends observed in Heilongjiang Province (1958–2017) reflected common characteristics across Northeast China [47].

The abrupt change in CDD around 1979 was consistent with the inter-decadal transition of the East Asian monsoon in the late 1970s, where enhanced northward movement led to shorter dry spells [45,48]. The transition of SD II around 1975 may have been associated with the AO phase shift from negative to positive. During the positive phase, stable mid-to high latitude westerly circulation inhibited severe convective weather [49,50]. Moreover, AO-driven circulation adjustments could also indirectly affect regional precipitation by modulating the East Asian winter monsoon [51]. The turning point of PRCPTOT around 1982 was attributed to the PDO phase change from cold to warm [52]. Only CDD, SD II, and PRCPTOT showed significant changes, indicating that these indices were more sensitive to climate forcing, while others were more influenced by local factors [45,53].

Wavelet analysis revealed a multi-cycle superposition pattern. The dominant 30–50-year cycles corresponded to PDO transitions [52,54]. The 10–20-year short cycles reflected the effects of the ENSO and solar activity, as warm ENSO events tended to increase summer precipitation through air–sea coupling processes [55,56,57]. The 50–60-year-long cycles were likely related to the AMO, which triggered Rossby wave trains to enhance monsoon water vapor transport. Additionally, the out-of-phase effect between AMO and PDO could amplify inter-decadal precipitation fluctuations [58,59]. SD II exhibited a simpler periodic structure, possibly because precipitation intensity was more sensitive to underlying surface changes, such as grassland degradation in western Jilin and forest coverage changes in eastern Jilin [60].

Furthermore, Antokhina et al. [24] identified a regime shift in circulation patterns associated with extreme precipitation over Eastern Siberia and Mongolia around the late 1990s. This shift was characterized by a transition from subtropical-only PV gradient inversion to simultaneous wave breaking at both subtropical and mid-latitude levels. It was accompanied by enhanced East Asian summer monsoon moisture transport, providing a large-scale dynamic context for the intensification of extreme precipitation observed in Jilin Province after 1990.

4.2. Spatial Distribution Mechanism of Extreme Precipitation

The growing gradient of extreme precipitation from northwest to southeast in Jilin Province is primarily influenced by the combined effects of Changbai Mountains’ morphology and the East Asian Summer Monsoon [61]. Long-term observational data have validated this trend, attributing it to the water vapor movement by monsoon and the topography blocking effect [62]. Numerical simulations indicate that the Changbai Mountains obstruct southerly airflow during summer, resulting in a 26% augmentation of precipitation on the southern windward slopes, such as Tonghua and Ji’an, while producing a rain shadow on the leeward side [63,64]. The anisotropic topography facilitates the ascension of water vapor, leading to frequent orographic precipitation in the southeast and a reduced number of precipitation days in the northwest [65]. This aligns with prior observations in Jilin Province from 1961 to 2015 [66]. The piedmont region functions as a high-frequency zone for summer convective clouds, induced by orographic uplift [67]. A case study of Super Typhoon Maysak demonstrated that the Changbai Mountains augment precipitation via forced uplift, resulting in an increase in severe rainfall by 6.8 mm/h, which constitutes 41% of the total rainfall [67].

The comparatively low values of Rx1day and Rx5day at specific eastern locations are ascribed to inadequate water vapor delivery. This is attributable to their distance from the Sea of Japan and the obstructive influence of local topography, namely the Laoyeling Mountains [53]. The east-to-west gradient of SD II is associated with the thermal disparities of the underlying surface. The western plain grasslands, characterized by little vegetation, underwent accelerated surface warming, leading to brief but intense precipitation events. Conversely, the eastern woodlands demonstrated diminished convective activity. The elevated precipitation intensity in Ji’an is attributable to the synergistic effects of orographic uplift and localized thermal conditions, as the intricate piedmont topography amplifies dynamic uplift and thermal instability [68].

The first mode derived from EOF analysis designates the southern region as the principal area for extreme precipitation variability. The temporal coefficients of this mode turn positive post-1990, signifying an escalation of intense precipitation. Tonghua and Ji’an are classified as high-risk zones [26,69]. The second mode encapsulates the north–south anti-phase variation, illustrating the interannual discrepancies in the northward progression of the summer monsoon [70]. This alternating pattern aligns with the precipitation seesaw mode observed between Mongolia and Eastern Siberia, influenced by the upper-tropospheric Rossby wave configurations and related wave-breaking occurrences. These occurrences result in a significant redistribution of moisture throughout the region [24]. The third mode is centered in the southeast, with increased variations, underscoring the necessity to monitor flood hazards in this region [71].

4.3. Trends and Limitations of Extreme Precipitation in Future Scenarios

All three models exhibited a comparable trend: an increased emission intensity led to a more significant amplification of extreme precipitation. This discovery aligns with studies performed in other areas of China [72,73]. The phenomenon adheres to the Clausius-Clapeyron connection, wherein the augmented water vapor transfer by the East Asian Summer Monsoon [74], in conjunction with the orographic uplift of the Changbai Mountains, collectively exacerbated exceptional precipitation episodes. In the SSP5-8.5 scenario, the extreme precipitation indices were markedly elevated compared to those in the SSP2-4.5 scenario, aligning with the scenario-driven patterns noted in other basins [75,76]. The declining trend in CDD under SSP5-8.5 indicates an intensified water vapor cycle, marked by reduced dry intervals and more intense precipitation events [10,77].

The anticipated regional alterations in extreme precipitation were predominantly aligned with historical trends. Significant increases were noted in grid sites in the central and southeastern regions, reflecting the persistent impact of topography and the monsoon. The highest significant inter-model variation among the extreme precipitation indices was noted in CWD. The FGOALS-g3 model forecasted a broad reduction in CWD, but the CanESM5 and MPI-ESM1-2-HR models anticipated increases. The disparities amongst the models were mainly primarily evident in simulated conservatism under the SSP2-4.5 scenario [78,79]. The CanESM5 model had an earlier response, with multiple indices attaining statistical significance, while only a limited number of indices were significant for the other two models. These inconsistencies align with the inter-model variations observed in the CMIP6 assessments for China [80,81].

While the models identified the principal patterns in extreme precipitation, uncertainties remain. Disparities in feedback mechanisms between CMIP6 models influence forecasts, and multi-model ensemble evaluations can offer more reliable support [10]. Furthermore, elements such as greenhouse gas emissions, urbanization, and alterations in land cover have substantially influenced extreme weather phenomena [82,83,84,85]. Consequently, subsequent study should concentrate on clarifying the impact of human activities on regional extreme climate change.

5. Conclusions

This study comprehensively examined the spatiotemporal characteristics and future projections of extreme precipitation in Jilin Province using observational data (1960–2019) and three CMIP6 models under SSP2-4.5 and SSP5-8.5 scenarios. Over the period from 1960 to 2019, extreme precipitation in Jilin Province exhibited the following characteristics: an increase in frequency and total amount, along with a decrease in intensity. The CDD showed a significant decreasing trend at a rate of −2.184 (d·(10a)⁻¹, p < 0.001), the PRCPTOT increased significantly at a rate of 1.493 (mm·(10a)⁻¹, p < 0.05), while the SD II decreased significantly at a rate of −0.016 (mm·d⁻¹·(10a)⁻¹, p < 0.01). Most of the extreme precipitation indices shared a dominant cycle of 30–50 years. Spatially, a prominent northwest-to-southeast increasing gradient was observed for most of the indices. For instance, the PRCPTOT ranged from 327.5 mm at Baicheng to 824.3 mm at Tonghua.

The future projections for the period from 2025 to 2100 indicated a scenario-dependent intensification of extreme precipitation. Under the SSP5-8.5 scenario, all three models projected significant increases in the precipitation amount indices, PRCPTOT: increasing at a rate of 2.071–2.457 mm·(10a)⁻¹, and the SD II increasing at a rate of 0.010–0.013 mm·d⁻¹·(10a)⁻¹, reversing the historical decreasing trend in intensity. Spatially, the projected changes maintained the northwest-to-southeast gradient, with PRCPTOT increases exceeding 230 mm in the southeast.

However, this study possesses multiple drawbacks. Initially, there exist inter-model discrepancies among the CMIP6 models employed, which may generate errors in the projections. Secondly, potential biases may emerge from the alignment between grid cells and in the models and observation stations. Thirdly, the study did not specifically address the impacts of urbanization. To mitigate these constraints, subsequent study should utilize multi-model ensemble techniques to enhance the reliability of projections and examine the driving mechanisms of extreme precipitation, including the effects of land-use alterations.