A 3.5-fold increase in the synchrony of extreme precipitation and temperature events across China from 1930 to 2022

Abstract

Global warming has led to an increasing frequency of extreme climate events; however, existing studies have largely focused on individual extremes or selected pairs occurring at specific locations over relatively short time frames, with limited attention to the long-term spatial synchrony of multiple extremes across different geographic regions. This study addresses this gap by examining the synchrony of eight precipitation and temperature extreme climate events across China from 1930 to 2022, using monthly-scale data. We quantify synchrony using two indicators: the proportions of land area (LA_p) and population (Pop_p) affected by these events. Our results reveal that over the 93-year period, the annual LA_p increased approximately 3.5-fold, while Pop_p doubled, suggesting that land exposure is expanding faster than population exposure. Although underpopulated regions northwest of the Hu Huangyong Line exhibit generally lower exposure levels, they experience a higher rate of increase in land exposure compared to densely populated areas. Extreme wet and hot events drive more than 60% of the total increase in land area exposure. We also identify a significant shift in ecosystem exposure: forest lands were the most affected before 1956, but since then, grasslands have become the dominant ecosystem exposed, especially during the warmer months. This study enhances understanding of the long-term spatial synchrony of multiple extreme climate events in China and emphasizes the critical need to integrate climatic, ecological, and social vulnerabilities into adaptation strategies to effectively manage growing risks.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-13415-3.

Introduction

The Intergovernmental Panel on Climate Change (IPCC) reported that global surface temperatures have risen faster since 1970 than in any other half-century in at least the last two millennia¹. This accelerated warming has exacerbated the occurrence of extreme climate events^2,3, including heatwaves⁴, droughts⁵, heavy precipitation⁶, wildfires⁷, tropical cyclones⁸, and others. The impacts of these extreme events are far-reaching, disrupting agricultural production^9–11, causing major economic losses^12,13, threatening public health¹⁴, and severely affecting biodiversity and ecosystem stability^15,16. In addition, the chances of these events occurring at the same time are increasing (e.g., a heatwave and a drought event occurring simultaneously in the same region), leading to more severe and widespread social and environmental impacts¹⁷.

Multiple extreme climate events occurring simultaneously or sequentially in time or space are defined as compound extreme events^18,19 and are of particular concern because of their cumulative and intensified impacts²⁰. Although significant progress has been made in describing the frequency, intensity and socio-economic impacts of compound extreme events^21,22, they have been mainly focused on the context of multivariate compound events occurring at the same location, such as simultaneous or sequential occurrences of heatwaves and heavy precipitation^17,23, with less attention to their spatial synchrony. Spatial synchrony refers to the co-occurrence of extreme climate events across different geographic regions during the same period of time (which can be the same week, month, and year) and have profound implications for risk assessment and disaster management. Recent examples, such as the record-breaking heatwaves in Europe²⁴ and the severe drought in the Yangtze River basin in China¹⁵, which co-occur in the summer of 2022, highlight the interconnectedness of extreme climate events around the globe and underscore the challenges posed by spatial compound extreme events.

While recent studies have advanced detection of spatially co-occurring extremes globally^25,26, their long-term evolution in China—a region dominated by monsoonal dynamics and complex topography—remains insufficiently understood. This gap limits actionable insights for regional adaptation planning. China, one of the most climate-sensitive regions in the world, is warming faster than the global average²⁷. Its wide-ranging climate patterns²⁸, large and growing population²⁹, extensive agricultural lands³⁰, and rapidly urbanizing society³¹ render it particularly vulnerable to extreme climate events. For instance, the severe flooding caused by heavy precipitation in Zhengzhou in 2021 resulted in substantial economic losses³², while the drought at Poyang Lake in 2022 significantly impacted agriculture and ecosystems^33,34. Although numerous studies have examined multivariate extreme events in China^23,28—including analyses of the changing frequency of compound hot-dry and hot-wet extremes at single locations³⁵, and investigations into the spatiotemporal variability of heatwaves, droughts, and floods in specific regions and timescales³⁶—these efforts have largely focused on isolated or regional manifestations. A subset of studies has also conducted attribution analyses of typical compound events, most of which emphasize heat-related extremes, with limited consideration of cold extremes^37,38. However, these research efforts have rarely addressed the spatial synchrony of a diverse array of extreme events—encompassing both individual extremes and temporally compounding events at the same location—from a long-term perspective. In particular, few studies have assessed the extent to which land area and population are jointly exposed to synchronous extreme events over time. Such simultaneous extreme events across different regions can result in compound socio-economic impacts, especially on agricultural systems. For example, concurrent flooding in grain-producing areas of South China and drought in North China would pose a critical threat to national food security. Accordingly, a comprehensive understanding of the evolving spatial synchrony of extreme climate events is critical for informing risk assessments and enhancing adaptive capacity in China.

To address existing knowledge gaps, this study systematically investigates the long-term evolution of the spatial synchrony of precipitation- and temperature-related extreme events across China from 1930 to 2022. Distinct from prior studies that have mainly concentrated on the frequency of extreme event occurrences, we quantify spatial synchrony using two complementary indicators: the annual proportions of land area and population affected by extreme climate events. This dual-metric framework allows us to jointly evaluate physical and societal exposure. In addition to single extremes, we incorporate temporally compounding events occurring at the same location, providing a more realistic picture of hazard complexity. By examining patterns across different land use types—such as croplands, forest lands, grasslands, and urban/rural residential areas—we further explore how spatial synchrony interacts with ecosystems exposure. Our findings highlight not only long-term trends in exposure, but also shifting ecological sensitivities (e.g., a post-1956 shift exposure from forest lands to grasslands), offering novel insights into how climate change reshapes risk landscapes. This work adds depth to the current understanding of climate extremes in China and offers an integrated perspective essential for designing future adaptation strategies.

Data and methods

Data

This study uses multiple high-resolution datasets to investigate the spatial synchrony of extreme precipitation- and temperature-related events in China. These datasets include gridded monthly precipitation and temperature data, as well as gridded annual population distribution and land use data.

Gridded monthly precipitation and temperature data are sourced from the National Environmental Data Center for the Tibetan Plateau/Third Pole (http://data.tpdc.ac.cn). They have a spatial resolution of 1 km × 1 km and span the period from 1901 to 2022³⁹.

Population distribution data are obtained from WorldPop (https://hub.worldpop.org), with the same spatial resolution of 1 km × 1 km as the precipitation and temperature data.

Land use data are provided by the Resource and Environmental Science Data Platform (https://www.resdc.cn) and also have a spatial resolution of 1 km × 1 km. The land use categories include cropland, forest land, grassland, and urban/rural residential areas, among others.

To isolate the impacts of population growth and land use changes in the study period, population distribution and land use data from a single year (2015) are utilized in the analysis. Since population distribution and land use data are annual in scale, it is assumed that they remain consistent at the monthly scale throughout the year.

Definition of extreme climate events

To define extreme climate events, this study adopts the threshold method based on prior research^35,40 , using the 5th and 95th percentiles of monthly precipitation and temperature over the study period (1901–2022) as critical thresholds. We calculate the 5th and 95th percentiles of monthly precipitation and temperature based on the entire study period (1901–2022), without defining a separate base period. These thresholds are then consistently applied throughout the full time series to identify exceedances. For each month, the extreme climate events are defined as follows: (1) An extreme wet (WW) event occurs when precipitation exceeds the 95th percentile (P95); (2) An extreme dry (DD) event occurs when precipitation falls below the 5th percentile (P5); (3) An extreme hot event (HH) occurs when temperature exceeds the 95th percentile (T95); (4) An extreme cold event (CC) occurs when temperature fall below the 5th percentile (T5). Additionally, a multivariate extreme event is identified within the same grid cell when both precipitation and temperature extremes occur in the same month.

Extreme climate events in this study are classified into eight types, including four individual extremes defined above (WW, DD, HH, and CC), as well as the two-by-two combinations of HH/CC with WW/DD, i.e., compound cold and dry (CD), compound hot and dry (HD), compound cold and wet (CW), and compound hot and wet (HW). Importantly, for any given grid cell, only one of the eight extreme events can occur in a particular month. This is because each event type is defined by distinct thresholds for precipitation and temperature extremes, with no overlap in the conditions that define the individual and compound events. As such, each grid cell is classified into only one of these event categories based on the specific exceedance of the defined thresholds for temperature and precipitation. This approach ensures that there is no overlap or double-counting of extreme events within a grid cell.

To evaluate the robustness of our results, we also conduct a sensitivity analysis using the 10th and 90th percentiles as alternative thresholds. The main analyses presented in the main text are based on the 5th and 95th percentile thresholds, while the sensitivity results using the 10th and 90th percentiles are provided in the Supplementary Information (Figs. S1–S6).

Statistical analysis

To investigate the spatial synchrony of extreme climate events, a multi-step statistical framework is employed.

1. National- and Provincial-level synchrony rates.

   At both the national and provincial scales, the monthly synchrony rate is calculated as the proportion of land area affected by the eight extreme climate events defined above (LA_p). aThe annual synchrony rate is derived by calculating the average of the monthly synchrony rates, which is then moving averaged over 30 years to analyze its long-term trends from 1930 to 2022 using Sen’s slope estimator and the Mann-Kendall trend test^41,42. Specifically, each 30-year average represents the value for the final year of the period, such as the 1901–1930 average representing the value for 1930. This approach is applied at the national scale to assess overall trends and at the provincial scale to provide a more localized understanding of spatial synchrony dynamics. Since the annual data reflect 30-year moving averages, the analysis focuses on the period from 1930 to 2022, rather than extending back to 1901.

2. Dominance of extreme event types.

   The relative contribution of each extreme event type to the overall synchrony rate is determined by decomposing the synchrony rates into the eight event categories. The dominant event type is identified based on its contribution to the overall synchrony rate and its trend over time. Specifically, for each year, we calculate the LA_p separately for each of the eight types of extreme events. By summing the annual LA_p values of all event types, we obtain the total land area proportion (LA_p) affected by any extreme precipitation or temperature event. Similarly, by aggregating the trends (Sen’s slope) of the individual LA_p series, we determine the overall trend of extreme climate event synchrony. The contribution of each extreme event type to the total LA_p and its long-term trend is then quantified based on its proportion relative to the total. This approach allows us to identify the dominant extreme event types that influence spatial synchrony, both in magnitude and in their trend over time.

3. Population-weighted synchrony rate.

   A population-based synchrony rate is calculated by replacing grid cell area with grid cell population, reflecting the proportion of the population exposed to extreme climate events (Pop_p). The trends in population-weighted synchrony rates are analyzed at both national and provincial scales from 1930 to 2022, following the same trend calculation method as for LA_p. The correlations between population-weighted and area-weighted synchrony rates are also examined to compare these two indicators.

4. Land use contributions.

   To assess the influence of land use on extreme event synchrony, grid cells are categorized by land use type (e.g., cropland, forest, grassland, urban/rural residential areas). The proportion of affected grid cells within each land use type is calculated relative to the total number of affected grid cells across the country. Temporal trends in these proportions from 1930 to 2022 are analyzed to examine how the contributions of different land use types to the overall synchrony of extreme climate events have evolved. Monthly comparisons are also conducted to explore seasonal variations in land use contributions, providing insights into how land use patterns shape the spatial and temporal dynamics of extreme events in different ecological contexts.

Results

An increase of 3.5-fold in annual LA_p from 1930 to 2022

From 1930 to 2022, the annual proportion of land area (LA_p) affected by multiple extreme climate events increases significantly at the national scale. Based on 30-year moving averages, the national annual LA_p rises from an average of 7.7% ± 1.8% in 1930 to 27.2% ± 13.4% in 2022, reflecting a 3.5-fold increase. This represents an average annual increase of 0.16% (p < 0.05), as determined by Sen’s slope estimator and the Mann-Kendall trend test (Fig. 1a). To evaluate the robustness of this trend, we conducted a sensitivity analysis using the 10th and 90th percentile thresholds. Under these less stringent thresholds, the national annual LA_p increases from 16.8% ± 2.7% in 1930 to 46.5% ± 11.6% in 2022, representing approximately a 2.8-fold increase. While the thresholds yield higher absolute LA_p values due to their broader definition of extremes, the persistent upward trend remains consistent with that based on the 5th and 95th percentiles. This confirms the robustness of the observed increase in spatial synchrony across China (see Supplementary Information, Fig. S1).

Fig. 1.

Spatial and temporal patterns of the annual proportion of land area (LA_p) affected by multiple extreme climate events based on 5th and 95th percentile thresholds. (a) Time series of annual LA_p from 1930 to 2022. The gray and blue lines represent the national annual LA_p and 30-year average LA_p, respectively, while the blue shaded area indicates the standard deviation of the 30-year LA_p. The red dashed line denotes the Sen’s slope for the national LA_p from 1930 to 2022 based on the 30-year moving averages, with ** indicating statistical significance (p < 0.05) according to the Mann-Kendall trend test. (b) Spatial distribution of the multi-year average annual LA_p from 1930 to 2022 at the provincial scale. (c) Spatial patterns of trends in provincial LA_p, calculated from the 30-year average over the period from 1930 to 2022. The histograms in (b) and (c) display the distribution of the multi-year average annual LA_p and the trends in LA_p across provinces, respectively, where the y-axis represents the number of provinces. In both panels, the black dashed line represents the Hu Huanyong Line. (d, e) Scatter plots illustrating the relationship between provincial LA_p and its population density, as well as the trend of provincial LA_p and population density, respectively, from 1930 to 2022.

At the provincial level, there are clear regional differences in LA_p values, with higher and lower LA_p in regions to the southeast and northwest of the Hu Huanyong Line (Chen et al., 2016; Hu, 1935), a key line for population distribution in China. The population density and LA_p are both higher in southeastern of the Hu Huanyong Line, and vice versa. Specifically, provinces in the northeastern, eastern coastal, and southwestern China, such as Heilongjiang, Shandong and Yunnan, exhibit higher annual LA_p levels, generally exceeding 17.5%. In contrast, provinces in the western China, such as Xinjiang, Qinghai and Tibet, have lower annual LA_p levels, typically below 16.5% (Fig. 1b). The relationship between LA_p and population density by province (Fig. 1d) shows a significant positive correlation with a correlation coefficient of 0.74. This result quantitatively supports the conclusion that provinces with higher population densities also have higher LA_p values.

Trend analysis of provincial annual LA_p reveals that regions with initially lower LA_p values have experienced more substantial increases. For instance, Tibet and Qinghai with smaller LA_p demonstrate significant increases in LA_p, with average annual growth rates exceeding 0.25%, suggesting a rise in both the frequency and spatial extent of extreme climate events in these areas. In contrast, eastern provinces, which already have higher initial LA_p levels (Fig. 1b), show more modest increases, with annual growth rates under 0.15% (Fig. 1c). Furthermore, the correlation analysis between provincial LA_p trends and population densities reveals a significant negative correlation, indicating that provinces with higher population densities tend to experience slower LA_p growth rates (Fig. 1e). These regional variations emphasize that historically less vulnerable areas (i.e., provinces with lower population densities) are becoming increasingly susceptible to extreme weather events, whereas regions already prone to such events are undergoing more gradual changes. These regional patterns and trends remain broadly consistent when using alternative thresholds (10th and 90th percentiles), further confirming the robustness of our findings (see Supplementary Information, Fig. S1).

Extreme wet and hot events are the primary contributors to the increase in LA_p

From Fig. 2a, it is evident that the four single-factor extreme climate events—extreme drought (DD), extreme heat (HH), extreme cold (CC), and extreme wet (WW)—all contribute significantly to the overall land area affected, with each accounting for more than 20% of the total annual LA_p. Among these, WW event has the largest share, representing 24.4% of the annual LA_p, followed closely by CC and HH events, which account for 23.6% and 23.1%, respectively. In contrast, the four compound extreme events—compound cold and dry (CD), compound hot and dry (HD), compound cold and wet (CW), and compound hot and wet (HW)—all constitute a smaller proportion, with a total contribution of less than 10% of the total annual LA_p. A similar pattern is observed when applying the 10th and 90th percentile thresholds, with WW, CC, and HH events remaining the dominant contributors, accounting for 22.6%, 22.4%, and 21.3% of the annual LAp, respectively (see Supplementary Information Fig. S2a).

Fig. 2.

Contributions of eight extreme climate events based on 5th and 95th percentile thresholds to the change in the annual proportion of land area (LA_p) affected by extreme climate events at the national scale. (a) Multi-year average annual LA_p affected by the eight extreme climate events shown with different colored boxplots: compound cold and dry (CD), extreme dry (DD), compound hot and dry (HD), extreme hot (HH), extreme cold (CC), compound cold and wet (CW), extreme wet (WW), and compound hot and wet (HW), along with the overall total (labeled as “TOL”), from 1930 to 2022. The LA_p values are accumulated from CD to HW, with the box heights and error bars representing the averages and standard deviations of the annual LA_p over the years. (b) Trends in annual LA_p affected by the eight extreme climate events and their combined total (“TOL”) from 1930 to 2022, with the box height indicating trend (slope) of each extreme event from CD to HW or their total and ** indicating statistical significance (p < 0.05) based on the Mann–Kendall trend test. The final column represents the aggregated trend for all eight extreme events combined.

When analyzing the contribution of each extreme event type to the overall increase in annual LA_p, WW emerges as the dominant factor, accounting for 34.8% of the total increase. This is followed by HH (29.5%) and DD (19.7%) (Fig. 2b). Despite the widespread occurrence of CC event (Fig. 2a), the trend calculated using Sen’s slope estimator shows a decline of -0.04% per year. These trends are consistent at both the national and provincial levels, with extreme wet and extreme hot events becoming increasingly prevalent across China. Similar results are obtained when using the 10th and 90th percentile thresholds, where WW, HH, and DD remain the primary contributors, accounting for 31.8%, 31.5%, and 20.1% of the total increase, respectively, while CC shows a notable decline of − 7.6% per year (Fig. S2b).

The increasing prominence and seasonal variation of extreme wet and hot events are underscored by the monthly distribution and trends of LA_p by eight extreme weather events (Fig. 3). For WW event, the affected land area is more evenly distributed across months throughout the year, with slightly lower proportions in November and December compared to other months (Fig. 3a). Notably, the growth rate of monthly LA_p affected by the WW event from 1930 to 2022 is higher in these two months (Fig. 3b). For HH event, the affected land area is significantly lower in July and August than in other months (Fig. 3a), and the growth rate of monthly LA_p affected by the HH event during these months is also the lowest (Fig. 3b). In contrast, the growth rate of monthly LA_p affected by the HH event is higher during the cooler months, particularly in November, where it reaches 0.09% per year (Fig. 3b). Meanwhile, the monthly LA_p affected by the CC event shows a significant decline in the cooler months, especially in December and January (Fig. 3b). These findings suggest that attention should be given to the increasing occurrences of extreme wet and hot events, particularly during the cooler months. Similar monthly distributions and increasing trends of WW and HH events, as well as the declining trend of CC events during cooler months, are also evident when using the 10th and 90th percentile thresholds (Fig. S3), reinforcing the robustness of these seasonal patterns.

Fig. 3.

Monthly distribution and trends of the proportion of land area (LA_p) affected by multiple extreme climate events from 1930 to 2022 based on 5th and 95th percentile thresholds. (a) The heatmap displays the monthly distribution of multi-year average LA_p affected by the eight extreme climate events from 1930 to 2022. Different colors highlight the monthly variation in LA_p affected by various extreme climate event types (CD, DD, HD, HH, CC, CW, WW, HW). (b) The heatmap shows the trends in monthly LA_p based on Sen’s slope estimator from 1930 to 2022, calculated using 30-year moving averages. Significant trends identified by the Mann–Kendall trend test are marked with **.

Twofold increase in annual Pop_p from 1930 to 2022

The annual proportion of the population affected by multiple extreme climate events (Pop_p) increases significantly at the national level from 1930 to 2022, independent of population growth trends. According to 30-year moving averages, the national annual Pop_p rises from 12.0 ± 3.6% in 1930 to 24.6 ± 9.0% in 2022, reflecting a twofold increase over the study period. This represents to an average annual increase of 0.11% (p < 0.05), as assessed using Sen’s slope estimator and the Mann-Kendall trend test (Fig. 4a). While the absolute values of Pop_p are higher when using the 10th and 90th percentile thresholds—rising from 26.1% ± 5.2% in 1930 to 42.5% ± 9.4% in 2022—the overall increasing trend remains consistent. However, the relative increase is smaller, at approximately 1.6-fold over the study period, reflecting the less stringent nature of these alternative thresholds (see Supplementary Information Fig. S4).

Fig. 4.

Spatial and temporal patterns of the annual proportion of population (Pop_p) affected by multiple extreme climate events based on 5th and 95th percentile thresholds. (a) Time series of annual Pop_p from 1930 to 2022, the gray and green lines represent the national annual Pop_p and its 30-year moving average, while the green fill area illustrates the standard deviation of the 30-year Pop_p. The pink dashed line indicates the Sen’s slope for the national Pop_p from 1930 to 2022 based on the 30-year moving averages, with ** denoting statistical significance (p < 0.05) according to the Mann-Kendall trend test. (b) Spatial distribution of the multi-year average annual Pop_p from 1930 to 2022 at the provincial scale. (c) Spatial trends in provincial Pop_p, calculated using the 30-year average over the period from 1930 to 2022. The histograms in panels b and c show the distribution of the multi-year average annual Pop_p and the trends in Pop_p across provinces, respectively, where the y-axis represents the number of provinces. In both panels, the black dashed line represents the Hu Huanyong Line. (d, e) Scatter plots illustrating the relationship between provincial Pop_p and its population density, as well as the trend of provincial Pop_p and population density, respectively, from 1930 to 2022.

At the provincial level, annual Pop_p values exhibit notably regional differences. Similar to LA_p, higher and lower Pop_p are found in regions to the southeast and northwest of the Hu Huanyong Line (Chen et al., 2016; Hu, 1935). The population density, LA_p, and Pop_p are all higher in southeastern of the Hu Huanyong Line, and vice versa. Specifically, provinces in the northeastern, eastern coastal, and southwestern regions—such as Liaoning, Shandong, and Yunnan—tend to have higher annual Pop_p values, generally exceeding 17.5%. In contrast, provinces in western China, including Tibet and Qinghai, show lower annual Pop_p levels, typically below 17% (Fig. 4b). The relationship between Pop_p and population density by province (Fig. 4d) exhibits a strong positive correlation, with a correlation coefficient of 0.67. This quantitative evidence reinforces the conclusion that provinces with higher population densities tend to have elevated LA_p values as well as higher Pop_p values.

Trend analysis of annual Pop_p by province indicates that regions with lower initial Pop_p levels (Fig. 4b) experience the most substantial growth (Fig. 4c). For instance, Tibet and Qinghai with smaller Pop_p have shown substantial increases, with average annual growth rates surpassing 0.2%, indicating a larger increase in the proportions of their population being exposed to extreme climate events. In contrast, provinces in eastern China, which already have high initial Pop_p levels (Fig. 4b), exhibit slower growth, with average annual rates below 0.1% (Fig. 4c). In addition, the correlation analysis between Pop_p trends and population density across provinces reveals a weak negative, albeit non-significant, correlation. This suggests that provinces with higher population densities tend to experience slower Pop_p growth rates (Fig. 4e). This pattern highlights a critical shift, while some regions remain consistently more vulnerable, areas historically less exposed to extreme events (i.e., provinces with lower population densities) are now experiencing the fastest increase in population exposure, which could intensify challenges related to local adaptation and disaster management. These spatial patterns and trends of provincial Pop_p remain consistent when using the 10th and 90th percentile thresholds, with the correlation coefficient between Popp and population density even slightly increasing, further confirming the robustness of our findings (Supplementary Information Fig. S4).

The annual Pop_p and LA_p affected by multiple extreme climate events exhibit a significant positive correlation at both national and provincial levels (Fig. 5). At the national level, the scatterplot of annual Pop_p versus LA_p for the period from 1930 to 2022 clearly shows a positive relationship, with higher Pop_p values corresponding to increased LA_p values (Fig. 5a). This strong association suggests that as the spatial extent of extreme climate events increases, a larger proportion of the population is affected, highlighting the interconnectedness of environmental and societal vulnerabilities. At the provincial scale, the spatial distribution of correlation coefficients between annual Pop_p and LA_p confirms a significant positive relationship across all provinces (Fig. 5b). Provinces with higher correlations between LA_p and Pop_p are generally located in regions where the LA_p and Pop_p growth rates are similar, i.e., either both high or both low. These regions include some eastern coastal and southwestern provinces as shown in Figs. 2c and 4c, and 5c. On the other hand, the correlation is weaker in provinces where the LA_p and Pop_p have different growth rates, i.e., one high and the other low, which can be seen in Qinghai and Sichuan provinces (Figs. 2b and 4b, and 5b). The results based on the 10th and 90th percentile thresholds are fully consistent with those shown here, reinforcing the robustness of the positive correlation between annual Pop_p and LA_p at both national and provincial levels (Supplementary Information Fig. S5).

Fig. 5.

Correlation between the annual proportion of land area (LA_p) and population (Pop_p) affected by multiple extreme climate events based on 5th and 95th percentile thresholds. (a) Scatter plot showing the relationship between national LA_p and Pop_p from 1930 to 2022. Each point represents the 30-year average annual value, illustrating the correlation between LA_p and Pop_p. (b) Spatial distribution of correlation coefficients between provincial annual LA_p and annual Pop_p over the same period (1930 to 2022). Provinces with statistically significant correlations (p < 0.05) are indicated with **. The histogram in the panel b display the distribution of the correlation coefficients across provinces, with the y-axis representing the number of provinces.

Increasing share of grasslands in annual LA_p from 1930 to 2022

Based on 30-year moving averages, the proportion of grasslands within the annual LA_p affected by extreme climate events has increased significantly from 1930 to 2022, with an average annual growth rate of 0.13%, as estimated by Sen’s slope estimator. This share rises from 22.1 ± 3.4% in 1930 to 33.7 ± 2.6% in 2022 (Fig. 6a). In contrast, the proportions of forest lands and croplands in the annual LA_p have decreased notably, at rates of 0.13% and 0.09% per year, respectively. Grasslands surpassed forests as the ecosystem most exposed to extreme climate events starting in 1956 (Fig. 6a). Meanwhile, the proportion of urban/rural residential areas in the annual LA_p has remained relatively stable, with a slight annual decline of 0.01% (Fig. 6a).

Fig. 6.

Temporal changes in the share of annual LA_p relative to total LA_p for different land uses and their monthly distribution patterns based on 5th and 95th percentile thresholds. (a) Temporal changes in the share of annual LA_p affected by eight extreme climate events relative to total annual LA_p for croplands, forest lands, grasslands, and urban/rural residential areas from 1930 to 2022 at the national scale. The dashed lines represent the respective trends over this period, the shaded area indicates the standard deviation of the 30-year values. ** indicates a p-value less than 0.05 based on the Mann–Kendall trend test. (b) Monthly distribution patterns of the multi-year average share of LA_p for croplands, forest lands, grasslands, and urban/rural residential areas, respectively.

Seasonal analysis shows that grasslands consistently account for a larger proportion of LA_p during the warmer months from May to September, particularly in July, compared to other land use types such as forest lands, croplands, and urban/rural residential areas (Fig. 6b). In contrast, both forest lands and croplands accounted for a higher proportion of LA_p in the cooler months of November and December (Fig. 6b). This seasonal pattern underscores the heightened exposure of grasslands to extreme climate events, especially during periods of higher temperatures. For forest lands and croplands, attention should be focused on the impacts of extreme climate events likely to occur during the cooler months. These results, based on the proportion of each ecosystem type affected, are fully consistent with the findings using the 10th and 90th percentile thresholds (Supplementary Information Fig. S6), further confirming the robustness of the observed temporal trends and seasonal patterns across different land use types.

Discussion

By focusing on the synchrony of multiple climate extremes, this study advances our understanding of their spatiotemporal dynamics across China. While previous research has primarily concentrated on individual types of extreme events^42,43 or on multivariate events occurring either simultaneously or sequentially at the specific locations over time^21,44, these studies have largely emphasized changes in the frequency of such events. In recent years, a growing number of studies³⁸ have begun to address spatially compounding climate extremes; however, these efforts have mostly been limited to specific sub-regions—such as arid North China or Eastern China—or focused on a narrow range of hazard pairs, such as drought–pluvial and flood–nocturnal heat combinations. In contrast, our study adopts a more comprehensive and integrative perspective by examining not only multiple climate extremes—including both precipitation- and temperature-related extremes as well as their combinations—but also spatial synchrony in their evolving process across regions over a 93-year period (1930–2022). We introduce two exposure-based indicators—annual land area proportion (LA_p) and population proportion (Pop_p) affected by multiple extreme events—to capture both the physical extent and societal relevance of this synchrony. Our results reveal a significant increase in both LA_p and Pop_p, which have grown approximately 3.5-fold and twofold, respectively. Importantly, the more rapid expansion of land exposure compared to population exposure highlights a critical but often overlooked aspect of climate risk: the growing exposure of sparsely populated regions.

Underpopulated areas northwest of the Hu Huanyong Line are experiencing a disproportionately high growth rate in land exposure to extreme events, underscoring the need to account for spatial disparities in future climate adaptation and risk management strategies. While these regions are witnessing faster increases in exposure, it is important to note that densely populated areas currently bear a higher absolute burden of climate extremes. To further explore these spatial patterns, we examine the relationship between LA_p, Pop_p, and population density at the provincial scale. Analysis of annual data reveals strong positive correlations between both LA_p and Pop_p and population density, with correlation coefficients of 0.74 and 0.67, respectively (Figs. 1d and 4d). These findings indicate that provinces with higher population densities generally experience greater exposure to extreme events, particularly in the densely populated southeastern regions beyond the Hu Huanyong Line, in contrast to the more rapid growth of exposure in the sparsely populated northwest. This suggests a dual dimension of risk: while sparsely populated areas face rapidly increasing exposure, densely populated regions are already subject to high levels of hazards exposure, which—when combined with concentrated human and infrastructural assets—may further amplify their overall climate risk.

In addition to providing an innovative analysis of the temporal evolution of synchrony among multiple extreme climate events, our findings highlight the dominant role of extreme wet and hot events in driving the increase in LA_p. Together, these two types of extremes account for over 60% of the overall rise in LA_p (Fig. 2b), far exceeding the contributions from other extreme event combinations. This dominance can be traced to fundamental climatic and environmental drivers. Rising global temperatures increase the atmosphere’s moisture-holding capacity via the Clausius–Clapeyron relationship, intensifying the hydrological cycle and increasing the likelihood of extreme precipitation⁴⁵—particularly under the modulation of the East Asian summer monsoon^46,47. At the same time, growing thermal stress has led to more frequent and intense heatwaves across the country⁴⁸. Regionally, intensified summer monsoons in southern and eastern China^47,49 and greater interannual variability in northern rainfall patterns⁵⁰ further reinforce spatial disparities in wet extremes. In addition, widespread land-use change—most notably urban expansion—has exacerbated urban heat island effects and modified land–atmosphere energy and moisture exchanges, amplifying both temperature and precipitation extremes^51,52. In contrast, compound extreme events such as dry/hot or wet/cold extremes contribute relatively little to the LA_p increase. This is partly due to the strict thresholds (5th and 95th percentiles) used in our analysis, which emphasize the most severe extremes. These stringent criteria inherently limit the frequency occurrence of compound events in our dataset, especially those requiring concurrent exceedance of both thresholds^35,40.

Beyond population and land exposure, our study reveals significant temporal shifts in ecosystem exposure to extreme climate events across China. Notably, until 1956, forest lands bore the greast burden of exposure. However, in subsequent decades, grasslands have emerged as the most exposed ecosystem type, driven primarily by an increased frequency and intensity of extreme hot and drought events. This transition is particularly alarming given grasslands’ intrinsic sensitivity to thermal and hydrological stresses, which can undermine key ecological functions such as carbon sequestration, soil stabilization, and biodiversity support. The heightened exposure of grasslands poses risks of degradation, reduced productivity, and shifts in species composition, potentially triggering cascading effects on regional ecosystem services and local livelihoods dependent on these landscapes. Moreover, grasslands’ vulnerability to prolonged drought and heat suggests potential feedbacks that may exacerbate climate change impacts locally, such as increased soil erosion and desertification. Our findings highlight an urgent need for adaptive ecosystem management and conservation strategies tailored to these shifting exposure patterns. This includes enhancing monitoring systems to dynamically track ecosystem responses, implementing drought and heat-resilience practices (e.g., restoration with drought-tolerant species), and integrating ecosystem-based approaches into broader climate adaptation frameworks. While prior studies have addressed the exposure of individual ecosystems to extreme climate events^53–55, our research uniquely captures the evolving spatial-temporal patterns of ecosystem exposure over nearly a century, emphasizing the importance of flexible, ecosystem-specific adaptation planning to effectively mitigate climate risks.

Despite the valuable insights gained from this study, there are still some limitations that should be acknowledged. First, the use of relatively strict thresholds (5th and 95th percentiles) to define extreme events may have overlooked the role of moderately intense compound events. While these capture the most severe events, less extreme but more frequent events—such as those defined by the 10th and 90th percentiles—can still have significant cumulative effects on ecosystems and populations. We performe a systematic sensitivity analysis using these alternative thresholds, which demonstrate that the overall temporal trends, spatial patterns, and dominant drivers remain highly consistent with those derived from the 5th/95th percentiles. For example, although the absolute magnitudes of exposure metrics (e.g., LA_p and Pop_p) are somewhat larger with the 10th/90th percentiles due to inclusion of more moderate events, the key conclusions about increasing exposure and the primary role of extreme wet and hot events held robustly. This finding underscores the reliability of our results while suggesting that future studies should incorporate a wider range of event intensities to capture the full range of extreme event impacts. Second, while our analysis captures exposure in terms of affected land and population, it does not account for the underlying socioeconomic determinants of vulnerability, such as adaptive capacity, infrastructure resilience, or social inequality, which strongly influence actual climate risk. Incorporating these factors in future studies would yield a more holistic understanding of vulnerability and help identify the regions and populations most at risk. Finally, observational uncertainties—particularly in precipitation data—may affect the results. In this study, we used a single dataset, which could introduce potential limitations, such as biases or inaccuracies inherent in that dataset. While the threshold method applied in this study helps mitigate some of these uncertainties by focusing on extreme deviations, it does not fully eliminate them. Future work could benefit from incorporating multiple datasets and cross-validation techniques to improve robustness and minimize potential data-related uncertainties.

Although this study investigates changes in exposure to multiple extreme climate events, it does not directly assess actual damages or impacts. The potential impacts we refer to here are based on the integration of hazard changes with land area and population exposure. However, our analysis does not incorporate critical vulnerability factors—such as adaptive capacity, infrastructure quality, social inequality, and institutional preparedness—which fundamentally shape how exposure translates into actual impacts. Therefore, future research should move beyond exposure metrics to incorporate multi-dimensional vulnerability assessments. Integrating social, economic, and institutional factors alongside environmental exposures would provide a more comprehensive understanding of climate risk and resilience^56–59.

Conclusions

This study provides a comprehensive analysis of the long-term evolution of the spatial synchrony of eight extreme climate events related to precipitation and temperature across China from 1930 to 2022. Using Sen’s slope estimator and the Mann-Kendall trend test, we find that the proportion of land area (LA_p) affected by these events has increased 3.5-fold, while the proportion of the population (Pop_p) exposed has doubled, with land exposure growing more rapidly than population exposure. This divergence highlights an important but often overlooked aspect of climate risk: increasing hazards in sparsely populated regions, particularly northwest of the Hu Huanyong Line, alongside sustained high exposure in densely populated southeastern provinces. Our analysis identifies extreme wet and hot events as the primary drivers of this increase, jointly accounting for over 60% of the growth in land area exposure. Additionally, a significant ecological shift is observed, with grasslands surpassing forest lands as the most exposed ecosystem type after 1956, raising concerns about ecosystem degradation and cascading impacts on regional services and livelihoods.

These findings underscore the intensifying synchrony of extreme precipitation- and temperature-related extremes in China and highlight the need for targeted adaptation strategies that consider both physical exposure and societal vulnerability. In particular, adaptive planning should address the dual challenge posed by rapid exposure growth in less populated areas and the already high risk in populous regions. While this study advances understanding by integrating multiple extreme event types, spatial synchrony, and ecosystem impacts over nearly a century, it also highlights key areas for future research. Specifically, relaxing the strict event thresholds and incorporating socioeconomic and vulnerability factors would enable a more holistic assessment of climate risks and their impacts.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

B.Y. drafted the main manuscript, created visualizations, conducted validation, investigation, and formal analysis, and contributed to the methodology. J.Y.L. assisted with validation, methodology, and formal analysis. S.S.H. reviewed and edited the manuscript, supervised the research, contributed to methodology and formal analysis, and led the study’s conceptualization. All authors reviewed and approved the final manuscript.

Data availability

All data supporting the analysis in this study are publicly available from open sources: Gridded monthly precipitation and temperature data are available from the National Tibetan Plateau Data Center at http://data.tpdc.ac.cn.Population distribution data are obtained from WorldPop Hub at https://hub.worldpop.org.Land use data can be accessed via the Resource and Environment Data Cloud Platform at https://www.resdc.cn.

Declarations

Competing interests

The authors declare no competing interests.