Intensification of summer mesoscale convective systems in East China under global warming Authors: Yutong Lu, John H. Marsham, Douglas J. Parker, Juan Fang, Sinan Gu, Jianping Tang, Deliang Chen Published: 2026 ABSTRACT Mesoscale Convective Systems (MCSs) are the primary drivers of extreme rainfall and flood hazards. Understanding their response to global warming is therefore important for improving climate projections and disaster risk reduction. In East China, however, evidence remains largely limited to case studies, and long-term projections of MCSs are scarce due to their poor representation in coarse-resolution climate models. Using convection-permitting model simulations with a pseudo–global warming approach, this study presents 22-summer projections of future MCSs over East China under the SSP5-8.5 scenario. Results show that future MCSs exhibit higher precipitation intensity and wider convective areas. Convective precipitation intensifies more than stratiform precipitation, indicating a transition toward more convection-dominated systems, accompanied by enhanced mid-level updrafts and super–Clausius–Clapeyron scaling of extreme rainfall. Increases in convective available potential energy (CAPE) and total column water vapor (TCWV) largely explain the intensification of maximum convective precipitation, with vertical zonal wind shear acting as a key dynamic modulator. The product of TCWV and vertical velocity is identified as an effective predictor of MCS peak precipitation. These findings imply heightened risks of intense rainfall from MCSs in East China and highlight the combined roles of thermodynamic and dynamic processes in shaping future MCSs. Introduction Anthropogenic greenhouse gas emissions have continued to increase with atmospheric CO2 concentrations increasing by ~50% above pre-industrial (1850–1900) levels, resulting in a rise in global mean surface temperature of approximately 1.3 °C in 2025, with widespread impacts on the climate system1. Ongoing climate warming has led to more frequent and intense weather hazards2, particularly extreme rainfall3,4, posing growing challenges to regional climate system stability1. The Clausius–Clapeyron (CC) relationship describes an increase of saturation water vapor pressure increase by approximately 7% per degree of warming, providing a thermodynamic constraint on potential precipitation intensification. However, the response of extreme precipitation to temperature can be substantially stronger, reaching up to twice the CC scaling5–7, especially at mean temperatures between 12 °C and 22 °C8. Mesoscale convective systems (MCSs), which are organized deep-convection systems composed of convective and stratiform clouds with spatial scales exceeding 100 km, are major contributors to extreme rainfall. Typically producing widespread precipitation over several hours, MCSs account for more than 40% of summer extreme precipitation in East China9. Beyond thermodynamic control, MCSs are strongly influenced by dynamic processes and multiscale internal interactions10, making their response to global warming difficult to explain using the CC relationship alone. A deeper understanding of MCS behavior under warming is therefore essential for assessing future flood and extreme rainfall risks. Observational and reanalysis-based studies indicate that the frequency and intensity of MCSs have increased under global warming in regions including the Sahel in West Africa11, North America12, and East Asia13. Correspondingly, both total and extreme MCS-related precipitation show increasing trends11–16. Over eastern China, MCS frequency and precipitation intensity increases significantly over the past two decades13, primarily driven by enhanced moisture availability16. Although regional climate regimes differ, key controlling factors, such as vertical wind shear, moisture content, and atmospheric instability, consistently shape MCS behavior. Increases in MCS activity are generally associated with rising total column water vapor (TCWV), enhanced instability, strengthened temperature gradients due to uneven surface warming, and warmer, drier mid-level conditions12,17,18. Convection-permitting models (CPMs), with horizontal grid spacings of approximately 1–4 km, explicitly resolve deep convection without relying on cumulus parameterization. This allows for much improved simulations of key MCS characteristics, including initiation, propagation, diurnal cycle, lifecycle, and precipitation structure9,19–21, thereby enabling quantitative projections of future MCS changes. In addition to dynamical downscaling driven by global climate model projections, the pseudo–global warming (PGW) approach has been widely applied in regional climate studies. PGW imposes climate change signals associated with future warming onto historical simulations, enabling the isolation of warming effects while largely preserving the present-day circulation patterns22,23. PGW-based CPM projections suggest MCS responses that differ from historical trends. Overall, MCS precipitation amount and intensity are projected to increase under future warming scenarios18,24. Changes in MCS frequency, however, are more complex. In some regions, such as North America, only modest increases or even decreases in total MCS frequency are projected, mainly due to reductions in weak to moderate systems25,26. In contrast, the frequency of extreme MCSs is projected to increase by more than threefold27, implying a future climate characterized by fewer but more intense convective events and elevated flood risk. GCM projections presenting convections with parametrization have also reported a decrease in total MCS frequency accompanied by increasing precipitation intensity, implying a reduction in weaker systems and a greater contribution from intense events28,29. Many studies attribute these changes primarily to thermodynamic processes. In a warming climate, low-level and column-integrated moisture increase substantially, while convective available potential energy (CAPE) and convective inhibition (CIN) both intensify, suppressing weak convection. Once CIN is overcome, strong CAPE combined with abundant moisture favors deeper convection and larger convective systems17,18,30. Warming also enhances MCS maintenance through microphysical changes, including a higher freezing level, increased liquid and ice water content, and a deepened warm-cloud layer27. Nevertheless, the role of vertical wind shear, a well-established control on MCS organization and rainfall31,32, remains uncertain, and even CPMs struggle to fully represent storm responses to wind shear33,34. Research on MCS responses to climate warming in eastern China remains limited. A PGW-based sensitivity experiment on the extreme MCS associated with the 20 July 2021 Henan rainstorm, which developed under favorable synoptic-scale conditions influenced by nearby typhoons35, showed that removing anthropogenic warming reduced MCS precipitation intensity and weakened the system structure36. These changes were mainly attributed to reduced CAPE and weakened vertical ascent, underscoring the strong thermodynamic sensitivity of MCSs to warming. However, it is still unclear how thermodynamic and dynamic factors regulate changes in MCS occurrence, intensity, and structure under climate warming. The primary objective of this study is to examine future changes in MCS characteristics over eastern China and to disentangle the roles of thermodynamic and dynamic drivers using PGW experiments under the SSP5-8.5 scenario. A convection-permitting regional climate simulation covering the summers of 2000–2021 is employed. The analysis systematically investigates future changes in the spatiotemporal distribution, structure, evolution, and precipitation characteristics of MCSs, quantifies the relationship between extreme MCS precipitation and temperature, and identifies the dominant mechanisms governing these changes. Results Projected changes of circulation and precipitation Under climate warming, the large-scale summer circulation over East China undergoes pronounced changes. Near-surface air temperature increases across the entire study region, with more pronounced warming exceeding 5 °C north of 35°N (Fig. 1a). As a result, the meridional temperature gradient weakens compared to WRF-CTL, which is unfavorable for the formation and maintenance of frontal zones. The overall warming leads to a thickening and upward expansion of the troposphere, as indicated by increased 500 hPa geopotential heights (Fig. 1c). The projected warming reflects the combined climate response under SSP5-8.5, which includes both greenhouse-gas and aerosol forcings that may influence regional precipitation through different mechanisms37. Meanwhile, low-level moisture transport from southeastern regions toward northern China is enhanced. Combined with higher temperatures, this leads to an increase in TCWV of up to 10 kg m⁻², thereby increasing the potential for extreme precipitation (Fig. 1b). However, because warming raises saturation vapor pressure, total column relative humidity (TCRH) decreases overall, with pronounced reductions along the eastern coastal regions (~−3%) and over the eastern Tibetan Plateau (~−7%) (Fig. 1c).Fig. 1Projected changes in background circulation, shown as differences between WRF-PGW and WRF-CTL.a Near-surface (2 m) air temperature (shading; °C) and 500–850 hPa vertical wind shear (vectors; m s⁻¹); b total column water vapor (TCWV; shading; kg m⁻²) and 925 hPa wind fields (vectors; m s⁻¹); c total column relative humidity (TCRH; shading; %), integrated vapor transport (IVT; vectors; kg m–1 s–1), and 500 hPa geopotential height (contours; dam) from WRF-PGW (purple) and WRF-CTL (green). Geopotential height contours are plotted at 2-dam intervals. All fields are averaged over the 22 summers (June, July, and August) from 2000 to 2021. The black box indicates analysis domain. Projected changes in background circulation, shown as differences between WRF-PGW and WRF-CTL. a Near-surface (2 m) air temperature (shading; °C) and 500–850 hPa vertical wind shear (vectors; m s⁻¹); b total column water vapor (TCWV; shading; kg m⁻²) and 925 hPa wind fields (vectors; m s⁻¹); c total column relative humidity (TCRH; shading; %), integrated vapor transport (IVT; vectors; kg m–1 s–1), and 500 hPa geopotential height (contours; dam) from WRF-PGW (purple) and WRF-CTL (green). Geopotential height contours are plotted at 2-dam intervals. All fields are averaged over the 22 summers (June, July, and August) from 2000 to 2021. The black box indicates analysis domain. Vertical wind shear between the 500 and 850 hPa levels is projected to weaken across most of China during future summers, with particularly strong reductions in regions dominated by westerly shear (Fig. 1a). According to the thermal wind relationship, this weakening is consistent with the reduced near-surface meridional temperature gradient. This interpretation is further supported by the smoother and more widely spaced 500 hPa geopotential height contours in WRF-PGW (Fig. 1c). Because the development and maintenance of MCSs depend on sufficient vertical wind shear to sustain tilted updrafts and to prevent convective cells from being suppressed by cold pool outflows38, the widespread reduction in wind shear suggests that the future background dynamical environment will be less favorable for MCS initiation and persistence. Figure 2 presents the projected changes in summer precipitation amount (PA), precipitation frequency (PF), and precipitation intensity (PI) over eastern China in the WRF-PGW simulations, shown as both absolute and relative changes. Regions with statistically significant changes at the 99.5% confidence level are highlighted. The results indicate an overall increase in total summer precipitation across eastern China under warming conditions, with the largest increases occurring over the southwestern coast, the Yangtze River basin, North China, and Northeast China. In contrast, decreases are projected over the southeastern and eastern coastal regions, the eastern Tibetan Plateau, and parts of northwestern China. The absolute magnitude of precipitation amount changes generally remains within ±0.1 mm h⁻¹, and most regions do not pass the 99.5% significance threshold (Fig. 2a). Relative changes, however, show that northern China experiences the largest precipitation increases, reaching up to 40% (Fig. 2d).Fig. 2Projected changes in total precipitation, shown as differences between WRF-PGW and WRF-CTL.a, d Precipitation amount (PA; mm h⁻¹), b, e precipitation frequency (PF; %), and c, f precipitation intensity (PI; mm h⁻¹). Panels (a–c) show absolute changes, while panels (d–f) show relative changes. All fields are averaged over the 22 summers (June, July, August) from 2000 to 2021. Hatching indicates regions where the changes are statistically significant at the 99.5% confidence level. Projected changes in total precipitation, shown as differences between WRF-PGW and WRF-CTL. a, d Precipitation amount (PA; mm h⁻¹), b, e precipitation frequency (PF; %), and c, f precipitation intensity (PI; mm h⁻¹). Panels (a–c) show absolute changes, while panels (d–f) show relative changes. All fields are averaged over the 22 summers (June, July, August) from 2000 to 2021. Hatching indicates regions where the changes are statistically significant at the 99.5% confidence level. Precipitation frequency decreases across most of eastern China, except along the southwestern coast (Fig. 2b). The largest reductions—up to 36% relative to WRF-CTL—occur over the southeastern coast, the eastern Tibetan Plateau, and northwestern China (Fig. 2e). In contrast, precipitation intensity increases significantly throughout eastern China. The largest absolute increases are concentrated over the Yangtze River basin and the North China Plain, reaching up to 1.6 mm h⁻¹ (Fig. 2c), while the strongest relative increases occur over the Loess Plateau and northwestern regions, exceeding 42% (Fig. 2f). These results indicate that, under future warming, precipitation events are expected to become less frequent but substantially more intense. The spatial correspondence between reduced TCRH (Fig. 1c) and decreased precipitation frequency (Fig. 2b, e) suggests that relative humidity, rather than absolute moisture content alone, plays a critical role in triggering rainfall39. Overall, the PGW projections point to a shift toward less frequent yet more intense precipitation across most of eastern China, implying heightened risks of both drought and flood hazards under future climate conditions. Projected spatial variations of MCS and their associated precipitation Projected changes in MCS frequency exhibit pronounced spatial heterogeneity across eastern China (Fig. 3a). MCS frequency is projected to decrease substantially over Northeast China, the North China Plain, the Yangtze–Huaihe River basin, and the southeastern coastal region, with reductions of approximately −15%, −15%, and −8%, respectively. In contrast, increases are projected over the southwestern region and the arid areas of Northwest China, reaching about 5%, 25%, and 10%, respectively. Notably, the WRF-PGW simulations indicate a general decline in total precipitation frequency across the entire study region, with particularly strong reductions in the arid northwest (SI Appendix, Fig. S1). This pattern suggests a shift toward less frequent but more extreme precipitation events under future warming, indicating a transition in precipitation characteristics whereby events become more intense when they occur.Fig. 3Spatial distribution of projected changes in mesoscale convective systems (MCSs), shown as differences between WRF-PGW and WRF-CTL.Projected changes in (a) MCS frequency (yr⁻¹), b MCS occurrence hours (h yr⁻¹), c MCS precipitation amount (mm yr⁻¹), and d the contribution of MCS precipitation to total precipitation (% yr⁻¹). All fields are averaged over the 22 summers (June, July, and August) from 2000 to 2021. The subregions, including southeast China (SEC), Yangtze-Huaihe River Basin (YRB), Sichuan Basin (SCB), north China (NC), northeast China (NEC), Loess Plateau (LP), and northwest China (NWC), are indicated by black boxes in panel a. Spatial distribution of projected changes in mesoscale convective systems (MCSs), shown as differences between WRF-PGW and WRF-CTL. Projected changes in (a) MCS frequency (yr⁻¹), b MCS occurrence hours (h yr⁻¹), c MCS precipitation amount (mm yr⁻¹), and d the contribution of MCS precipitation to total precipitation (% yr⁻¹). All fields are averaged over the 22 summers (June, July, and August) from 2000 to 2021. The subregions, including southeast China (SEC), Yangtze-Huaihe River Basin (YRB), Sichuan Basin (SCB), north China (NC), northeast China (NEC), Loess Plateau (LP), and northwest China (NWC), are indicated by black boxes in panel a. Despite increasing CAPE (Fig. 4a), MCS frequency along the eastern coast declines. This reduction is likely associated with enhanced moisture divergence driven by weakened onshore flow and increased export of moisture away from the coastline under warming conditions (Fig. 1b, c). In contrast, the total number of hours influenced by MCSs increases across most regions (Fig. 3b), with particularly pronounced increases over the Sichuan Basin, the lower Yangtze River basin, and the southern coastal areas. This increase is consistent with longer MCS lifetimes and an expansion of their horizontal scales in a warmer climate.Fig. 4Projected changes in thermodynamic factors, shown as differences between WRF-PGW and WRF-CTL.a Convective available potential energy (CAPE; J kg⁻¹) and b convective inhibition (CIN; J kg⁻¹). All fields are averaged over the 22 summers (June, July, and August) from 2000 to 2021. Projected changes in thermodynamic factors, shown as differences between WRF-PGW and WRF-CTL. a Convective available potential energy (CAPE; J kg⁻¹) and b convective inhibition (CIN; J kg⁻¹). All fields are averaged over the 22 summers (June, July, and August) from 2000 to 2021. Further analysis reveals that both the total amount of MCS-related precipitation (Fig. 3c) and its contribution to total precipitation (Fig. 3d) are projected to increase across most regions of China, with the strongest relative increases occurring in the arid northwest, reaching up to 55%. These results point to a substantial intensification of MCS-related precipitation under future warming scenarios. In summary, MCSs in a warmer climate exhibit clear intensification and enhanced precipitation impacts, implying increased risks of extreme rainfall events. This enhancement is particularly pronounced in the arid northwest and southwestern regions, where both MCS frequency and precipitation intensity are projected to rise, posing heightened challenges for water resource management and disaster risk reduction. Projected changes in convection instability Under climate warming, both CAPE and CIN are projected to increase (Fig. 4). Enhanced CIN suppresses the initiation of convection, making MCS triggering more difficult and allowing convective energy and moisture to accumulate in the atmosphere (Fig. 4b). Once convection is initiated, this accumulated energy favors stronger convective outbreaks and the development of larger and more intense systems. The projected reductions in MCS frequency over Northeast China and the Yangtze–Huaihe River basin are closely associated with increased CIN in these regions (Fig. 4b). Despite increases in TCWV in these areas (Fig. 1b), TCRH decreases (Fig. 1c), indicating that although moisture availability is enhanced, atmospheric moisture recharge times are prolonged39. This combination provides a physical explanation for the coexistence of reduced MCS frequency and enhanced precipitation intensity per event. In contrast, the projected increases in both MCS frequency and intensity over southwestern China are associated with negligible changes in CIN and substantial increases in CAPE, reaching approximately 700 J kg⁻¹ (Fig. 4a). These changes occur alongside elevated TCWV (Fig. 1b) and nearly unchanged with TCRH (Fig. 1c), creating a thermodynamically favorable environment for frequent and intense convective activity. Over the arid northwest, CIN increases markedly by more than 30 J kg⁻¹ and is accompanied by decreases in relative humidity. Nevertheless, MCS frequency and intensity are still projected to increase in this region. This apparent inconsistency suggests that, in addition to thermodynamic forcing, strengthened land–atmosphere interactions under future warming may play an important role in modulating MCS activity in arid environments40,41 Extreme MCS precipitation response under the Clausius-Clapeyron relationship Figure 5a, d illustrates the temperature sensitivity of total extreme precipitation and MCS-related extreme precipitation, respectively, based on comparisons between the WRF-CTL and WRF-PGW simulations. On average, extreme precipitation exhibits super–CC scaling with increasing temperature, with mean scaling rates (SRs) of 1.09 and 1.15 times the CC rate for total and MCS-related extremes, respectively. This super-CC behavior is particularly evident over the arid northwest, the eastern Tibetan Plateau, and the Yangtze River basin, indicating heightened sensitivity to warming and elevated risks of extreme precipitation in these regions. In contrast, sub-CC scaling is found over the North China Plain, Northeast China, and coastal regions, suggesting a weaker response of extreme precipitation to temperature increases.Fig. 5Clausius–Clapeyron (CC) scaling of extreme precipitation and its decomposition.a, d CC scaling response of extreme precipitation, decomposed into contributions from (b, e) 2 m specific humidity changes and c, f) residual components. Panels (a–c) show results for total extreme precipitation, while panels (d–f) show results for MCS-associated extreme precipitation. The spatially averaged scaling rate (SR) is indicated in the lower-right corner of each panel. Clausius–Clapeyron (CC) scaling of extreme precipitation and its decomposition. a, d CC scaling response of extreme precipitation, decomposed into contributions from (b, e) 2 m specific humidity changes and c, f) residual components. Panels (a–c) show results for total extreme precipitation, while panels (d–f) show results for MCS-associated extreme precipitation. The spatially averaged scaling rate (SR) is indicated in the lower-right corner of each panel. Further analysis shows that the increase in 2 m specific humidity with temperature is smaller than the theoretical increase in saturation vapor pressure predicted by the CC relationship. The average SRs of specific humidity with temperature are approximately 0.72 and 0.88 for total and MCS-related extremes, respectively (Fig. 5b, e), reflecting the overall decline in atmospheric relative humidity under warming conditions. Consequently, changes in moisture availability alone cannot fully explain the observed super-CC scaling of extreme precipitation. Instead, the relationship between extreme precipitation and specific humidity (Fig. 5c, f) suggests that enhanced precipitation efficiency plays a key role. Here, the residual term represents processes not directly explained by thermodynamic moisture increases, and therefore implicitly includes dynamical contributions such as strengthened vertical updrafts, enhanced moisture convergence, and more efficient conversion of condensate into surface precipitation. In a warming climate, even modest increases in specific humidity can lead to disproportionate increases in extreme precipitation. This indicates that in regions exhibiting super-CC scaling, the intensification of extreme precipitation is not solely thermodynamically driven, but likely involves stronger vertical moisture transport and enhanced land–atmosphere interactions42. The mechanisms underlying sub-CC scaling vary regionally. Over the North China Plain and Northeast China (Fig. 5a, d), sub-CC behavior is primarily attributable to a less-than-CC increase in atmospheric moisture with warming (Fig. 5b, e). Along coastal regions, however, sub-CC scaling arises from a weakened sensitivity of precipitation intensity to moisture availability (Fig. 5c, f), pointing to dynamic or microphysical constraints on rainfall efficiency rather than moisture limitation. These contrasting responses highlight the coexistence of thermodynamic constraints on moisture supply over inland regions and dynamic or microphysical controls on precipitation efficiency in coastal environments. The temperature scaling of MCS-related extreme precipitation is similar to that of total extreme precipitation (Fig. 5d). Notably, the scaling rate of pre-MCS specific humidity with temperature is approximately 0.3 higher than that for total extremes, approaching the theoretical CC scaling across northern China, including Northeast China, the Loess Plateau, and the arid northwest (Fig. 5e). In these regions, pre-storm moisture increases more closely follow saturation vapor pressure, favoring intense convective development and increasing the likelihood of MCS occurrence, likely driven by enhanced moisture transport ahead of MCSs11. In contrast, the sensitivity of extreme precipitation to specific humidity is weaker for MCS-related extremes than for all extremes (Fig. 5f), suggesting a comparatively weaker intensification of updrafts in MCS events. Taken together, MCS-related extremes are thermodynamically better primed than all extreme precipitation events, as pre-storm humidity increases with temperature at rates closer to CC scaling, consistent with enhanced moisture convergence associated with mesoscale circulations11. Conversely, for all extremes, precipitation intensity increases more rapidly relative to available moisture, implying stronger dynamical intensification of updrafts in non-MCS extreme events. Changes in MCS features including lifecycle and structure Figure 6a presents the statistical characteristics of key MCS features based on Z-score standardization. The simulations indicate that under future warming scenarios, the horizontal extent of convective clouds expands significantly, with the mean convective area increasing from 8.2 × 10⁴ km² to 9.0 × 10⁴ km². In contrast, the total MCS area shows no systematic change, exhibiting only minor regional variability. The lack of a consistent expansion in stratiform cloud coverage suggests that the projected increase in MCS size is primarily attributable to the broadening of convective clouds rather than to changes in stratiform regions.Fig. 6Key structural and precipitation characteristics of mesoscale convective systems (MCSs) and their evolution throughout the life cycle.a Boxplots of standardized MCS properties (Z-scores), including duration, propagation speed, propagation distance, convective area, total MCS area, convective precipitation rate, and total MCS precipitation rate. Mean values for each variable are shown below the boxplots, with units of hours, 10 km h⁻¹, km, 10⁴ km², 10⁵ km², mm h⁻¹, and mm h⁻¹, respectively. Features with statistically significant changes at the 95% confidence level are indicated by stars next to the tick labels. Normalized evolution of MCS structure and precipitation over the life cycle, including the evolution of (b) convective cloud-top brightness temperature (K) and convective precipitation rate (mm h⁻¹); c convective precipitation (10³ mm) and convective cloud area (10⁴ km²); d the fraction of convective cloud area within the total MCS area (%) and convective cloud-top temperature (K); and e total MCS precipitation (10³ mm) and total MCS area (10⁴ km²). Key structural and precipitation characteristics of mesoscale convective systems (MCSs) and their evolution throughout the life cycle. a Boxplots of standardized MCS properties (Z-scores), including duration, propagation speed, propagation distance, convective area, total MCS area, convective precipitation rate, and total MCS precipitation rate. Mean values for each variable are shown below the boxplots, with units of hours, 10 km h⁻¹, km, 10⁴ km², 10⁵ km², mm h⁻¹, and mm h⁻¹, respectively. Features with statistically significant changes at the 95% confidence level are indicated by stars next to the tick labels. Normalized evolution of MCS structure and precipitation over the life cycle, including the evolution of (b) convective cloud-top brightness temperature (K) and convective precipitation rate (mm h⁻¹); c convective precipitation (10³ mm) and convective cloud area (10⁴ km²); d the fraction of convective cloud area within the total MCS area (%) and convective cloud-top temperature (K); and e total MCS precipitation (10³ mm) and total MCS area (10⁴ km²). MCS precipitation intensity also increases under warming. The mean convective precipitation rate rises from 3.1 mm h⁻¹ to 3.5 mm h⁻¹, while the total MCS precipitation rate increases from 2.3 mm h⁻¹ to 2.6 mm h⁻¹. Overall, future MCSs are characterized by expanded convective areas, and stronger precipitation intensity, while the overall system size remains largely unchanged. These changes indicate a clear structural shift toward more convection-dominated MCSs. Figure 6b–e compare the evolution of precipitation structure and cloud characteristics throughout the normalized MCS life cycle. Convective intensity peaks during the initial stage of MCS development (Stage 1), which is marked by the lowest cloud-top brightness temperatures and the strongest convective rainfall rates (Fig. 6b). As the system evolves, precipitation intensity gradually weakens, accompanied by a progressive warming of cloud-top temperatures. The convective cloud area expands during the early development stage, reaching a maximum at Stage 5 before rapidly declining (Fig. 6c). The evolution of total convective precipitation closely follows this pattern, reflecting the combined influence of precipitation intensity and convective area. Figure 6d shows that the proportion of convective clouds within the total MCS area varies systematically throughout the life cycle. Prior to Stage 3, convective cloud growth outpaces that of surrounding stratiform clouds, leading to a rapid increase in convective dominance. During Stages 4–5, this proportion peaks and remains relatively stable, followed by a gradual decline as convective clouds contract rapidly while stratiform clouds expand more slowly, indicating a transition toward stratiform dominance. Meanwhile, the total cold cloud area expands rapidly after Stage 1, reaching its maximum at Stage 7 before gradually decreasing, reflecting the lagged outward expansion of the MCS (Fig. 6e). The WRF-PGW projections further reveal an enhancement of convective characteristics throughout the MCS life cycle. Compared with the present climate, MCSs in a warming climate exhibit stronger convective precipitation intensity (Fig. 6b) and larger convective areas (Fig. 6c) already at Stage 1, along with a faster growth rate. Although the total MCS area slightly decreases (Fig. 6e), the proportion of convective clouds within the system increases markedly (Fig. 6d), further confirming a structural shift toward convectively dominated systems with reduced stratiform contributions. This structural evolution suggests that future MCSs are more likely to manifest as intense convective rainfall events, increasing uncertainty in hydrological responses and posing greater challenges for flood risk management, water resource allocation, and urban drainage system adaptation. To further examine changes in MCS precipitation structure in a warmer world, Fig. 7 presents composite mean precipitation fields centered on the convective centroid, including total, convective, and stratiform precipitation components. Under the current climate, total MCS precipitation exhibits an elliptical southwest–northeast orientation (Fig. 7a), consistent with the climatological mean southeastward propagation of summer MCSs in China43. The maximum precipitation center is located slightly north of the convective core, and the area enclosed by the 10 mm contour spans approximately 400 km. Convective precipitation displays a similar but more compact structure, with the 10 mm contour enclosing an area of approximately 350 km (Fig. 7b). In contrast, stratiform precipitation appears as a broader crescent-shaped rainfall shield surrounding the convective core, extending primarily northward and slightly eastward of the convective core, covering a broader area but with much weaker intensity, with maxima below 4 mm (Fig. 7c).Fig. 7Composite mean precipitation fields centered on MCS convective centroids.a, d Total MCS precipitation (mm), b, e convective precipitation (mm), and c, f stratiform precipitation (mm). Panels (a–c) show results from the WRF-CTL simulation, while panels (d–f) show results from the WRF-PGW simulation. Contours in panels (d–f) denote the corresponding precipitation fields from WRF-CTL for reference. Dashed boxes in panels (a–c) indicate the compositing domains used for the vertical hydrometeor profiles in Fig. 9. Composite mean precipitation fields centered on MCS convective centroids. a, d Total MCS precipitation (mm), b, e convective precipitation (mm), and c, f stratiform precipitation (mm). Panels (a–c) show results from the WRF-CTL simulation, while panels (d–f) show results from the WRF-PGW simulation. Contours in panels (d–f) denote the corresponding precipitation fields from WRF-CTL for reference. Dashed boxes in panels (a–c) indicate the compositing domains used for the vertical hydrometeor profiles in Fig. 9. Under the warming scenario, the southwest–northeast orientation of MCS precipitation is preserved, but the area of heavy rainfall becomes larger, with maximum precipitation exceeding 30 mm (Fig. 7d). Relative to the current climate, the areas enclosed by the 20 mm, 15 mm, 10 mm, and 5 mm contours increase by 113%, 54%, 37%, and 22%, respectively. The enhancement is particularly pronounced for convective precipitation, with corresponding increases of 136%, 62%, 42%, and 31% (Fig. 7e). In contrast, stratiform precipitation shows only weak changes under warming. Its spatial distribution shifts slightly southwestward, but both intensity and spatial extent remain largely unchanged (Fig. 7f), with areas exceeding 3 mm, 2 mm, and 1 mm increasing by only 18%, 14%, and 6%, respectively. These results indicate that the projected enhancement of MCS precipitation under warming is dominated by convective precipitation, while stratiform precipitation exhibits no systematic intensification or expansion. This behavior is consistent with the structural evolution identified in Fig. 6 and aligns with previous findings over North America24,26,44. Further analysis of composite maximum 500 hPa vertical velocity (ω₅₀₀) reveals a strong spatial correspondence between mid-tropospheric ascent and precipitation distribution, with regions of intense upward motion closely aligned with heavy rainfall (Fig. 8a–c). In the WRF-PGW simulations, both the intensity and spatial extent of strong mid-level ascent increase substantially, with maximum ω₅₀₀ strengthening from −6.3 Pa s⁻¹ in the current climate to −7.1 Pa s⁻¹ under warming. Areas with ω₅₀₀ exceeding −6 Pa s⁻¹ and −5 Pa s⁻¹ expand by 587% and 69%, respectively (Fig. 8d). Focusing on convective precipitation regions, maximum ω₅₀₀ intensifies from −6.22 Pa s⁻¹ to −7.00 Pa s⁻¹, while areas exceeding −6 Pa s⁻¹ and −5 Pa s⁻¹ increase by 1767% and 78%, respectively (Fig. 8e). These results indicate substantially strengthened vertical ascent within convective regions, facilitating more efficient upward transport of low-level moisture and promoting enhanced condensation and precipitation. This is consistent with CPM evidence that warming shifts vertical motion toward stronger updrafts and strengthens coupling between ascent, humidity, and rainfall45.Fig. 8Composite mean of the maximum 500 hPa vertical velocity (ω500, Pa s–1) during the MCS life cycle centered on MCS convective centroids.ω500 in the range of (a, d) total MCS precipitation, b, e convective precipitation, and c, f stratiform precipitation. Panels (a–c) show results from the WRF-CTL simulation, while panels (d–f) show results from the WRF-PGW simulation. Contours in panels (d–f) denote the corresponding ω500 fields from WRF-CTL for reference. Composite mean of the maximum 500 hPa vertical velocity (ω500, Pa s–1) during the MCS life cycle centered on MCS convective centroids. ω500 in the range of (a, d) total MCS precipitation, b, e convective precipitation, and c, f stratiform precipitation. Panels (a–c) show results from the WRF-CTL simulation, while panels (d–f) show results from the WRF-PGW simulation. Contours in panels (d–f) denote the corresponding ω500 fields from WRF-CTL for reference. In contrast, the stratiform precipitation region is characterized by weaker, rearward mid-level ascent typical of mature MCSs10 (Fig. 8c, f). Under the WRF-PGW projection, ω₅₀₀ in stratiform regions shows minimal enhancement, with areas exceeding −2 Pa s⁻¹ and −1 Pa s⁻¹ increasing by only 9% and 7%, respectively. This weak response is consistent with the limited changes in stratiform precipitation intensity and spatial extent shown in Fig. 7. Hydrometeor profiles (Fig. 9) provide further insight into the microphysical processes underlying these structural changes. The mean MCS profile is dominated by convective regions (Fig. 9a, b). Within convective core (Fig. 9b), ice- and snow-phase mixing ratios (QICE + QSNOW) increase substantially in the mid-to-upper troposphere, accompanied by enhanced graupel content (QGRAUPEL) above the elevated melting level (Fig. 8). This reflects intensified moist updrafts and amplified latent heat release from deposition and freezing. The resulting diabatic heating enhances buoyancy and further strengthens core updrafts, although this effect may be offset by the enhanced drag associated with falling hydrometeors. The net response is consistent with the projected increase in mid-level vertical velocity (Fig. 8). The increased ice-phase hydrometeor production ultimately leads to enhanced surface convective rainfall through melting processes, as reflected by increased rainwater mixing ratios (QRAIN, Fig. 9b).Fig. 9Projected changes in MCS-related hydrometeor vertical profiles.Shown are mean differences in mixing ratios of rain, ice and snow, and graupel between WRF-PGW and WRF-CTL for (a) entire MCSs, b MCS convective regions, and c MCS stratiform regions. Spatial averaging was performed over a 320 × 320 km box centered on the MCS centroid for the entire MCS, a 160 × 160 km box centered on the centroid for convective regions, and a 320 × 160 km box located 160 km north of the centroid for stratiform regions. The compositing domains were defined according to the spatial distribution and scales of convective and stratiform regions derived from Fig. 7. The selected dimensions therefore reflect typical structural features of simulated MCSs. Projected changes in MCS-related hydrometeor vertical profiles. Shown are mean differences in mixing ratios of rain, ice and snow, and graupel between WRF-PGW and WRF-CTL for (a) entire MCSs, b MCS convective regions, and c MCS stratiform regions. Spatial averaging was performed over a 320 × 320 km box centered on the MCS centroid for the entire MCS, a 160 × 160 km box centered on the centroid for convective regions, and a 320 × 160 km box located 160 km north of the centroid for stratiform regions. The compositing domains were defined according to the spatial distribution and scales of convective and stratiform regions derived from Fig. 7. The selected dimensions therefore reflect typical structural features of simulated MCSs. In contrast, stratiform regions exhibit a markedly weaker response (Fig. 9c). Ice-phase hydrometeor increases are modest and confined to lower altitudes than in convective regions, suggesting limited enhancement of stratiform precipitation sources. Below the melting level, decreases in snow, ice, and graupel mixing ratios are approximately balanced by increases aloft, while surface stratiform rainfall slightly decreases. This reduction is likely associated with enhanced evaporation in a warmer and potentially drier environment (Fig. 1). Given that stratiform precipitation primarily originates from detrainment and lateral spreading of upper-level anvils10,46, these results suggest that enhanced updrafts and elevated diabatic heating preferentially support deeper, more concentrated convection rather than expansion of stratiform outflow. Enhanced buoyancy and vertical mass flux47 thus favor the development of taller and more intense convective cores. In summary, the WRF-PGW projections indicate that under climate warming, precipitation enhancement and intensified vertical motion within MCSs are concentrated primarily in convective cloud regions, reflecting strong thermodynamic and dynamic amplification of convection. In contrast, stratiform cloud regions exhibit a much weaker response, highlighting fundamentally different physical controls on convective and stratiform precipitation under warming. These contrasting responses warrant further investigation to better constrain future changes in MCS-related extreme precipitation. Thermodynamic and dynamic drivers of MCSs To examine changes in the environmental conditions associated with MCS initiation under climate warming, Fig. 10a, b present probability density distributions of a dynamic factor, 500–850 hPa zonal vertical wind shear, and a thermodynamic factor, TCWV, averaged within a 200 km radius of the convective centroid during the 3 hours preceding MCS initiation. Here, MCS initiation is defined as the first time a convective system satisfies the MCS identification criteria. Although the WRF-PGW simulations project a weaker mean-state low-level zonal vertical wind shear relative to WRF-CTL, the shear distribution associated with MCS initiation shifts slightly toward stronger values (Fig. 10a). This indicates that, in a warming climate, MCS initiation preferentially occurs in environments with relatively stronger vertical wind shear, suggesting that enhanced shear is increasingly required to organize and sustain deep convection. As expected, pre-initiation moisture conditions are substantially enhanced in WRF-PGW, with peak TCWV increasing from approximately 60 kg m⁻² to about 75 kg m⁻² (Fig. 10b), reflecting a marked increase in background moisture availability that favors deep convective development.Fig. 10Projected changes in vertical wind shear and moisture conditions prior to MCS initiation.Probability density distributions of (a) 500–850 hPa zonal vertical wind shear (m s⁻¹) and b total column water vapor (TCWV; kg m⁻²), averaged within a 200 km radius of the convective centroid during the 3 h preceding MCS initiation. Blue curves denote WRF-CTL simulations, and red curves denote WRF-PGW projections. Black dotted (dashed) lines indicate the 10th (25th) and 90th (75th) percentiles in WRF-CTL, while the solid black line marks the median (50th percentile). c Projected changes in the composite mean 500–850 hPa zonal vertical wind shear and d projected changes in the composite mean TCWV. Stippling indicates regions where changes are statistically significant at the 99.5% confidence level. Projected changes in vertical wind shear and moisture conditions prior to MCS initiation. Probability density distributions of (a) 500–850 hPa zonal vertical wind shear (m s⁻¹) and b total column water vapor (TCWV; kg m⁻²), averaged within a 200 km radius of the convective centroid during the 3 h preceding MCS initiation. Blue curves denote WRF-CTL simulations, and red curves denote WRF-PGW projections. Black dotted (dashed) lines indicate the 10th (25th) and 90th (75th) percentiles in WRF-CTL, while the solid black line marks the median (50th percentile). c Projected changes in the composite mean 500–850 hPa zonal vertical wind shear and d projected changes in the composite mean TCWV. Stippling indicates regions where changes are statistically significant at the 99.5% confidence level. Figure 10c, d further shows composite mean differences in zonal vertical wind shear and TCWV between WRF-PGW and WRF-CTL, averaged hourly within a 200 km box centered on the convective centroid at the time of MCS initiation. A modest enhancement of zonal vertical wind shear is evident in WRF-PGW, with the strongest increases (~0.8 m s⁻¹) located primarily south of the convective centroid (Fig. 10c). In contrast, the increase in TCWV is much more pronounced, with moisture content near the convective core increasing by at least 15 kg m⁻² and peaking at approximately 18.5 kg m⁻² relative to WRF-CTL (Fig. 10d). Figure 11 further elucidates the dynamic–thermodynamic controls on MCS intensity and structure by examining the response of maximum convective precipitation intensity and maximum convective area to environmental conditions. Two-dimensional histograms reveal that maximum convective precipitation intensity is sensitive to both zonal vertical wind shear and TCWV in both climate scenarios (Fig. 11a, b). Increases in either variable are associated with stronger MCS precipitation. Notably, for comparable wind shear and TCWV conditions, maximum precipitation intensity is consistently higher in WRF-PGW than in WRF-CTL, indicating an upward shift in the upper bound of MCS precipitation under warming. This enhancement likely reflects stronger vertical ascent in a warmer atmosphere. Univariate analyses further show that although the response slopes of maximum convective precipitation intensity to vertical wind shear (Fig. 11e) and TCWV (Fig. 11f) are broadly similar between the two simulations, the increased frequency and magnitude of extreme TCWV and wind shear values in WRF-PGW substantially amplify extreme MCS precipitation.Fig. 11Response of MCS intensity and spatial extent to wind shear, moisture, and instability under present-day and warming climates.a, c WRF-CTL and b, d WRF-PGW two-dimensional histograms illustrating the relationships between (a, b) maximum convective precipitation rate and c, d maximum convective area of MCSs with 500–850 hPa zonal vertical wind shear and total column water vapor (TCWV). Numbers in the subtitles indicate the total number of MCS samples in each panel, and values shown at each grid point represent the mean maximum convective precipitation rate or convective area within that bin. Response of maximum convective precipitation rate to (e) zonal vertical wind shear, (f) TCWV, (i) convective available potential energy (CAPE), and (j) TCWV×\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\sqrt{{\rm{CAPE}}}$$\end{document}CAPE. Response of maximum convective area to (g) zonal vertical wind shear, (h) TCWV, (k) CAPE, and l TCWV × \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\sqrt{{\rm{CAPE}}}.$$\end{document}CAPE. Dashed lines denote linear regression fits, and error bars represent the standard error within each bin. Response of MCS intensity and spatial extent to wind shear, moisture, and instability under present-day and warming climates. a, c WRF-CTL and b, d WRF-PGW two-dimensional histograms illustrating the relationships between (a, b) maximum convective precipitation rate and c, d maximum convective area of MCSs with 500–850 hPa zonal vertical wind shear and total column water vapor (TCWV). Numbers in the subtitles indicate the total number of MCS samples in each panel, and values shown at each grid point represent the mean maximum convective precipitation rate or convective area within that bin. Response of maximum convective precipitation rate to (e) zonal vertical wind shear, (f) TCWV, (i) convective available potential energy (CAPE), and (j) TCWV×\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\sqrt{{\rm{CAPE}}}$$\end{document}CAPE. Response of maximum convective area to (g) zonal vertical wind shear, (h) TCWV, (k) CAPE, and l TCWV × \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\sqrt{{\rm{CAPE}}}.$$\end{document}CAPE. Dashed lines denote linear regression fits, and error bars represent the standard error within each bin. To assess whether enhanced vertical ascent associated with increased convective instability contributes to this intensification, Fig. 11i, j examine the relationship between maximum convective precipitation intensity and CAPE, as well as TCWV×\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\sqrt{{\rm{CAPE}}}$$\end{document}CAPE. Maximum precipitation increases with CAPE up to approximately 2000 J kg⁻¹, beyond which further increases are limited, indicating that extreme precipitation is not governed by instability alone but is also constrained by moisture availability. Consistent with theoretical expectations that vertical velocity scales with the square root of CAPE, a more linear and coherent relationship emerges between maximum convective precipitation intensity and TCWV×\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\sqrt{{\rm{CAPE}}}$$\end{document}CAPE (Fig. 11j). Minor differences between WRF-CTL and WRF-PGW are likely modulated by variations in vertical wind shear. Overall, TCWV×\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\sqrt{{\rm{CAPE}}}$$\end{document}CAPE emerges as a robust predictor of extreme MCS precipitation intensity, consistent with observations of Sahelian MCS rainfall17. The maximum convective area of MCSs also exhibits positive relationships with moisture, vertical wind shear, and CAPE. In WRF-CTL, convective area increases with low-level zonal vertical wind shear (Fig. 11c). In contrast, WRF-PGW shows a weaker and non-monotonic response, with convective area increasing at low to moderate shear values but decreasing at higher shear (Fig. 11d, g). The response to TCWV remains similar across both simulations, with convective area increasing steadily with moisture content (Fig. 11c, d, h). CAPE exerts a strong and consistent positive influence on convective area in both scenarios (Fig. 11k), underscoring its dominant role in regulating the horizontal extent of convection. Given the generally higher CAPE values in WRF-PGW, maximum convective area tends to increase accordingly. For a given MCS translation speed, such expansion directly enhances rainfall accumulation even without increases in rain rate. A near-linear relationship is also observed between maximum convective area and TCWV×\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\sqrt{{\rm{CAPE}}}$$\end{document}CAPE. However, for equivalent TCWV×\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\sqrt{{\rm{CAPE}}}$$\end{document}CAPE values, WRF-PGW yields smaller convective areas than WRF-CTL (Fig. 11l). This difference can be attributed to the reduced sensitivity of convective area to TCWV under warming, as evidenced by the weaker TCWV–area relationship in WRF-PGW (Fig. 11h). In summary, maximum convective precipitation intensity in MCSs is primarily governed by the combined effects of TCWV and CAPE, with additional modulation by zonal vertical wind shear, whereas maximum convective area is mainly controlled by CAPE and moisture availability. Additional analysis (SI Appendix, Fig. S2) further suggests the existence of an optimal regime, characterized by high TCWV, large CAPE, and moderate vertical wind shear. The intensity of extreme MCS precipitation increases even though the climatological background vertical wind shear weakens in the projection. Under climate warming, the increased co-occurrence of extreme TCWV and CAPE substantially enhances extreme MCS precipitation. In contrast, changes in the horizontal expansion of convective structures are more complex and reflect evolving dynamical controls. While low-level vertical wind shear strongly influences convective area in the present climate, its regulating role appears to weaken under warming in the WRF-PGW projections. The combined effect of intensified precipitation rates and expanded convective areas leads to larger event-integrated rainfall totals, thereby substantially amplifying future flood risks. Discussion In summary, CP WRF simulations using a PGW approach project that summer MCSs over East China under the SSP5-8.5 scenario shift toward fewer but more intense events in historically active regions. Warming enhances atmospheric moisture availability and instability, leading to a 13% increase in convective rain rate (from 3.1 to 3.5 mm h–1) and an expansion of convective areas from 8.2 × 104 to 9.0 × 104 km2. In contrast, stratiform precipitation changes remain comparatively small. In arid northwestern regions, the concurrent increases in MCS occurrence and intensity suggest heightened risks of extreme precipitation despite the region’s overall dryness. These changes collectively increase the contribution of MCS rainfall to total precipitation, with a relative increase of up to 55%. The results indicate a systematic shift toward more convectively dominated MCS structure. Enhanced thermodynamic conditions promote stronger updrafts and intensified convective precipitation. The combined effects of TCWV and instability largely contribute to precipitation intensification, while vertical wind shear acts primarily as a secondary modulator. The product of TCWV and square root of CAPE is identified as the most effective predictor of MCS maximum precipitation. Our results are consistent with PGW-based projections from other regions, which generally indicate that MCSs intensify and become increasingly convectively dominated under warming. For example, simulations over North America show that projected increases in total MCS rainfall are primarily driven by a stronger and more extensive convective core rather than by changes in event frequency26, while composite analysis of flood-producing MCSs similarly attribute future rainfall intensification to strengthened convective structures24. Comparable responses have also been reported in tropical regions22,30,42. Despite regional differences in large-scale circulation, these studies collectively suggest a robust thermodynamic control whereby warming enhances deep convection and shifts MCS precipitation toward a more convectively dominated regime. These findings suggest that rainstorm risk assessments should shift from focusing solely on event frequency or total rainfall to changes in storm structure and intensity under climate change. Even if the overall frequency of MCSs decrease, the risk of hazardous rainfall may still increase due to intensified convective precipitation. Consequently, future disaster risk assessments and urban drainage design should place greater emphasis on the strengthening of convective rainfall and short-duration precipitation extremes. Moreover, as MCS represents a major contributor to East Asian monsoon rainfall, structural changes in these systems may also influence the regional hydrological cycle and the spatial distribution of flood risk. These changes therefore have important implications for water resource management and flood mitigation. Despite its advantages in isolating the thermodynamic response of MCSs to warming, the PGW approach has inherent limitations. PGW experiments cannot fully capture potential systematic changes in large-scale circulation, such as shifts in monsoon or storm tracks27. Dynamical downscaling studies suggest that such circulation changes, together with moisture transport variation, can substantially influence MCS frequency and seasonality48. Nevertheless, circulation responses associated with thermodynamic changes may still be partially assessed within the PGW framework, such as warming-induced modifications in large-scale temperature gradients49. In addition, PGW experiments typically apply mean climate perturbations and therefore do not represent changes in internal climate variability. Although our multi-decadal simulations could permit an analysis of interannual variability in MCS activity, this aspect is beyond the scope of this study. Alternative approaches have therefore been explored. For instance, the thermodynamic global-warming (TGW) method introduces time-varying climate perturbations, allowing the evolution of warming signals to be represented50,51. With the development of global CP models, combining PGW-type thermodynamic perturbations with such simulations may benefit quantifying the joint roles of thermodynamic and dynamical changes in shaping future extreme precipitation and MCS activity. Methods Model configuration The Weather Research and Forecasting (WRF) model52 version 4.3.3 was employed to conduct CP scale simulation. The model was initialized and forced at the lateral boundaries using ERA5 reanalysis data53 at 31 km resolution, which was dynamically downscaled to CP scale for regional climate simulations. The simulation domain encompassed most of eastern China, configured on an 840 × 840 grid with 4 km horizontal spacing under a Lambert conformal projection. A sponge layer with 15 vertical levels was applied near the lateral boundaries to minimize reflection. The model used 50 vertical levels, extending up to 50 hPa. Simulations were conducted for 22 boreal summers (2000–2021), each initialized on May 21 and integrated through September 1, with the first 10 days designated as spin-up. The physical parameterizations used include WRF Thompson microphysics scheme54, Yonsei University scheme (YSU) as planetary boundary physics scheme55, CAM shortwave and longwave schemes56, and Noal-MP land surface model57. Cumulus parameterization was disabled to explicitly model moist convective processes. More information on the model design and configuration can be found in Lu et al. (2025)43. This simulation is hereafter referred to as WRF-CTL. Simulation and analysis domain is presented in supporting information (SI Appendix, Fig. S3). The performance of the WRF-CTL simulation in reproducing the climatological characteristics of MCSs and their associated precipitation over eastern China was evaluated. The model effectively captures key features such as MCS propagation, life cycle, and diurnal cycle, though it tends to overestimate extreme MCS precipitation. Additionally, WRF-CTL demonstrates skill in simulating the sensitivity of MCS-related rainfall to key environmental drivers, including zonal wind shear and TCWV43. To assess the applicability of the PGW framework, we evaluated the performance of WRF-CTL in representing large-scale circulation. The model realistically reproduces the spatial distribution of TCWV and CAPE compared with ERA5, although underestimation is found43. The simulated moisture flux divergence and wind fields at 850 hPa closely resemble ERA5, capturing the major convergence zone over southern and eastern China, associated with moisture transport from the South China Sea and western Pacific (SI Appendix, Fig. S4). Overall, the model reproduces the key circulation and moisture features, providing a physically consistent baseline for PGW-based climate change analysis. Pseudo global warming (PGW) experiment The Pseudo Global Warming (PGW) approach has been widely adopted in regional climate change projection studies58–60. In this study, the PGW experiment was designed to be consistent with the WRF-CTL in terms of model configuration, spatial domain, simulation period, and resolution. The initial and boundary conditions consist of ERA5 reanalysis data superimposed with global warming perturbations constructed based on the SSP5-8.5 greenhouse gas emission scenario. The SSP5-8.5 scenario represents a fossil-fueled development pathway characterized by rapid economic growth, high energy demand, and continued reliance on fossil fuels, resulting in very high greenhouse gas emissions and radiative forcing of 8.5 W m-2 by 210061. Hereafter, this PGW experiment is referred to as WRF-PGW. The global warming perturbation fields were derived from the ensemble mean climate change signals of 20 models (Table 1) participating in the Coupled Model Intercomparison Project Phase 6 (CMIP6). The adoption of a multi-model ensemble helps reduce model uncertainty and internal variability. Specifically, the forcing fields for the WRF-PGW projections can be described as:1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{WRF}}_{{forcing}}={ERA}5+\Delta {{CMIP}6}_{{SSP}}$$\end{document}WRFforcing=ERA5+ΔCMIP6SSPwhere ERA5 represents the ERA5 reanalysis data, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\triangle {{CMIP}6}_{{SSP}}$$\end{document}∆CMIP6SSP denotes the monthly mean differences between the future period (2070–2099) under the SSP5-8.5 scenario and the historical period (1971–2000), calculated as:2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\Delta {{CMIP}6}_{{SSP}}={{CMIP}6}_{{SSP}\,}(2070-2099)-{{CMIP}6}_{{historical}}(1971-2000)$$\end{document}ΔCMIP6SSP=CMIP6SSP(2070−2099)−CMIP6historical(1971−2000)Table 1The 20 climate models from the Coupled Model Intercomparison Project Phase 6 (CMIP6) used in this study, including their modeling institutions, horizontal resolutions (longitude grid points [nlat, 180°]×latitude grid points [nlon, 360°]), and variant idNumberModelInstitutionnlat×nlonVariant id1ACCESS-CM2CSIRO-ARCCSS(Australia)144144×192192r1i1p1f12ACCESS-ESM1-5CSIRO(Australia)145145×192192r1i1p1f13BCC-CSM2-MRBCC (China)160160×320320r1i1p1f14CanESM5CCCma (Canada)6464×128128r1i1p1f15CESM2NCAR (USA)192192×288288r1i1p1f16CMCC-ESM2CMCC(Italy)192192×288288r1i1p1f27CNRM-CM6CNRM-CERFACS(France)256256×256256r1i1p1f28CNRM-ESM2-1CNRM-CERFACS(France)256256×128128r1i1p1f19EC-Earth3EC-Earth-Consortium (EU)256256×512512r1i1p1f110FGOALS-F3-LCAS (China)180180×360360r1i1p1f111FGOALS-g3CAS (China)8080×180180r1i1p1f112GFDL-ESM4NOAA-GFDL (USA)180180×360360r1i1p1f113HadGEM3-GC31-LLMOHC NERC(UK)144144×192192r1i1p1f314INM-CM5-0INM(Russia)120120×180180r1i1p1f115IPSL-CM6A-LRIPSL (France)143143×144144r1i1p1f116MIROC6MIROC (Japan)128128×256256r1i1p1f117MPI-ESM1-2-HRMPI-M DWD DKRZ (Germany)192192×384384r1i1p1f118MPI-ESM1-2-LRMPI-M DWD DKRZ (Germany)9696×192192r1i1p1fl19MRI-ESM2-0MRI(Japan)160160×320320r1i1p1f120UKESM1-0-LLMOHC NERC NIMS-KMA NIWA(UK)144144×192192r1i1p1f1 Table 1The 20 climate models from the Coupled Model Intercomparison Project Phase 6 (CMIP6) used in this study, including their modeling institutions, horizontal resolutions (longitude grid points [nlat, 180°]×latitude grid points [nlon, 360°]), and variant idNumberModelInstitutionnlat×nlonVariant id1ACCESS-CM2CSIRO-ARCCSS(Australia)144144×192192r1i1p1f12ACCESS-ESM1-5CSIRO(Australia)145145×192192r1i1p1f13BCC-CSM2-MRBCC (China)160160×320320r1i1p1f14CanESM5CCCma (Canada)6464×128128r1i1p1f15CESM2NCAR (USA)192192×288288r1i1p1f16CMCC-ESM2CMCC(Italy)192192×288288r1i1p1f27CNRM-CM6CNRM-CERFACS(France)256256×256256r1i1p1f28CNRM-ESM2-1CNRM-CERFACS(France)256256×128128r1i1p1f19EC-Earth3EC-Earth-Consortium (EU)256256×512512r1i1p1f110FGOALS-F3-LCAS (China)180180×360360r1i1p1f111FGOALS-g3CAS (China)8080×180180r1i1p1f112GFDL-ESM4NOAA-GFDL (USA)180180×360360r1i1p1f113HadGEM3-GC31-LLMOHC NERC(UK)144144×192192r1i1p1f314INM-CM5-0INM(Russia)120120×180180r1i1p1f115IPSL-CM6A-LRIPSL (France)143143×144144r1i1p1f116MIROC6MIROC (Japan)128128×256256r1i1p1f117MPI-ESM1-2-HRMPI-M DWD DKRZ (Germany)192192×384384r1i1p1f118MPI-ESM1-2-LRMPI-M DWD DKRZ (Germany)9696×192192r1i1p1fl19MRI-ESM2-0MRI(Japan)160160×320320r1i1p1f120UKESM1-0-LLMOHC NERC NIMS-KMA NIWA(UK)144144×192192r1i1p1f1 The 20 climate models from the Coupled Model Intercomparison Project Phase 6 (CMIP6) used in this study, including their modeling institutions, horizontal resolutions (longitude grid points [nlat, 180°]×latitude grid points [nlon, 360°]), and variant id Monthly mean perturbations preserve the year-to-year variability inherited from ERA5 at the lateral boundaries5. The forcing variables used in the WRF-PGW experiments include horizontal wind fields, geopotential height, temperature, specific humidity at each vertical model level, as well as sea surface temperature and sea level pressure at the surface. Soil moisture and temperature were not perturbed in this study. This method also enables a direct comparison between present-day and future climates22. Cloud-top brightness temperatures The Community Radiative Transfer Model (CRTM) version 2.3.062 was employed to calculate cloud-top brightness temperature (Tb) corresponding to the WRF simulation using the 11.2 μm near-infrared band of the Himawari-8 geostationary meteorological satellite. The input to CRTM consisted of three-dimensional atmospheric output variables from the WRF-CTL and WRF-PGW simulations, including pressure, temperature, water vapor, and mass mixing ratios of cloud water, rainwater, graupel, snow, and ice. Additionally, two-dimensional surface fields, including longitude, latitude, and surface temperature, were included. These inputs enable CRTM to generate the corresponding the cloud-top brightness temperature. MCS tracking To robustly identify MCSs, we applied the following criteria10,63,64 with adaptations for eastern China MCSs65: (1) A cold cloud system (CCS) must exhibit a Tb ≤ -32 °C and an area ≥40,000 km2; (2) within the CCS, a convective core must be present with a Tb ≤ –52 °C and an area ≥30,000 km2; (3) the convective core must contain a precipitation feature (PF) with rainfall intensity ≥ 2 mm/h and a major axis length ≥100 km; (4) conditions (1) and (2) must persist for at least 3 consecutive hours, while condition (3) must persist for no less than 2 h. Under warming scenarios, the troposphere becomes generally thicker and warmer, raising the question of whether the Tb thresholds used for MCSs remain applicable in WRF-PGW simulations. We conducted statistical analyses on troposphere-top height, troposphere-top temperature, and the difference between troposphere-top temperature and Tb in both WRF-CTL and WRF-PGW. Under the SSP5-8.5 warming scenario, results show that both the tropopause-top height and temperature increase (SI Appendix, Fig. S5a), indicating a deepening and warming troposphere. However, the distribution of the temperature difference between the tropopause-top temperature and Tb remains largely unchanged (SI Appendix, Fig. S5b). Therefore, it can be concluded that under the SSP5-8.5 warming scenario, the Tb thresholds used for MCS identification under historical climate conditions remain applicable under warming scenarios despite the overall thickening of the troposphere. An object-based Tracking (OT) algorithm was applied to enable the identification and tracking of MCSs. This method is highly flexible and can be adapted to various datasets by adjusting threshold values and processing procedures. For the identification process, the dataset was spatially resampled to an 8 km resolution by averaging the values of all grid points within an 8 km radius around each grid point66. This smoothing effectively reduces over-segmentation and filters out small-scale noise, thereby enhancing the detection of MCSs. The MCS identification and tracking procedure consists of five main steps. First, during the identification stage, convective clouds are identified based on predefined thresholds of Tb, along with a minimum area criterion at each time step. Second, a pixel-wise expansion is performed to include surrounding stratiform cloud regions with higher Tb threshold. The combined convective and stratiform regions form a cold cloud system (CCS). If multiple convective cores exist within the same time step, the expansion process is prioritized for the largest core, followed by others in descending order. If the boundaries of two CCSs come into contact during expansion, the expansion is terminated. Third, precipitation features (PFs) within each CCS are identified, and their characteristics are extracted, including precipitation intensity, major axis length, and length-width ration (LWR). Fourth, CCSs are tracked over time using an area-overlap method. If two CCSs in consecutive time steps overlap more than 50%, they are considered to belong to the same potential MCS. In case of multiple overlaps, the largest overlapping system is retained, while others are terminated. If a CCS splits, the largest fragment continues to be tracked as part of the MCS case, while the other portions are treated as new potential MCS cases. Tracking is performed hourly and continues until the 50% overlap criterion is no longer met. Fifth, all identified potential MCS cases are filtered based on the duration of the system and the PF. Only those meeting the temporal thresholds are retained as final MCS cases and labeled as MCS1, MCS2, …, MCSN, where N represents the total number of tracked MCSs. Typhoon-related MCSs were excluded from the analysis based on the Joint Typhoon Warning Center (JTWC) Tropical Cyclone Best-Tracks67. Analyses methods In this study, MCSs were first identified, tracked, and labeled at hourly intervals within the defined study region. Using the mask files generated for each identified MCS, a corresponding MCS-specific precipitation dataset was constructed. For each time step, the mass centroid of the convective core was calculated and tracked over time to quantify propagation characteristics, including speed, track length, and direction. All MCSs affecting the analysis region were included for spatial patterns (Figs. 2–5), whereas analysis of MCS properties (Figs. 6–11) were restricted to systems with complete life cycles within the analysis domain. The lifetime of each individual MCS is divided into ten normalized life stages68, with stage 1 representing initiation and stage 10 indicating system dissipation. This normalization is achieved by linearly interpolating the original time series of each variable across the MCS lifecycle. Previous evaluations have shown that WRF-CTL realistically reproduces the observed temporal evolution of key MCS properties, including the timing of coldest cloud-top temperature during early development, peak convective area during the mature stage, and subsequent stratiform expansion during later stages43. We analyzed the Clausius-Clapeyron scaling relationship of extreme precipitation and MCS-related extremes over eastern China by decomposition42 as:3\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\frac{{\Delta }_{ \% }P}{{\Delta }_{ \% }{e}_{{sat}2m}}=\frac{{\Delta }_{ \% }{q}_{2m}}{{\Delta }_{ \% }{e}_{{sat}2m}}\times \frac{{\Delta }_{ \% }P}{{\Delta }_{ \% }{q}_{2m}}$$\end{document}Δ%PΔ%esat2m=Δ%q2mΔ%esat2m×Δ%PΔ%q2mwhere \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\Delta }_{ \% }$$\end{document}Δ% represents relative change, P is extreme precipitation (or extreme MCS precipitation), \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${e}_{{sat}2m}$$\end{document}esat2m is saturation vapor pressure at 2 m and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${q}_{2m}$$\end{document}q2m is specific humidity at 2 m. The first term represents ratio of changes in extreme precipitation to changes in 2 m saturation vapor pressure, which assesses whether the increase in extreme precipitation aligns with theoretical CC scaling. The second term is ratio of changes in 2 m specific humidity relative to changes in saturation vapor pressure, reflecting the extent to which atmospheric moisture content responds to temperature increases. The third term represents the rate of change in extreme precipitation relative to the change in specific humidity, often referred to as the residual term. Saturation vapor pressure change is defined as the climatological mean change in 2 m saturation vapor pressure. Extreme precipitation change is calculated as the average changes in the hourly precipitation exceeding 99th percentile threshold. Specific humidity change was computed as the mean changes in 2 m specific humidity within three hours preceding the extreme precipitation. All changes are expressed as relative changes. To minimize the cooling and moisture removal effect of ongoing precipitation on temperature and specific humidity, only extreme precipitation with hourly rainfall below 1 mm h-1 in the preceding hour is considered. A ratio of 1 indicates consistency with CC scaling, values below 1 suggest sub-CC scaling, and values above 1 indicate super-CC scaling. Maximum convective available potential energy (CAPE) and maximum convective inhibition (CIN) was calculated using python library wrf-python. The maximum equivalent potential temperature (theta-e) height level in the lowest 3000 m within each column was firstly found. A parcel with a depth of 500 m was then calculated and centered over this maximum theta-e height level. The parcel’s moisture and temperature characteristics are calculated by averaging over the depth of this 500 m parcel. CAPE and CIN were calculated for the parcel with the maximum theta-e within the full atmospheric column. The distributions of MCS features, including MCS duration, propagation speed, propagation distance, convective area, CCS area, convective precipitation, and CCS precipitation, were each rescaled via Z-score standardization so that4\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${Z}_{i}=\frac{{X}_{i}-{\mu }_{i}}{{\sigma }_{i}},$$\end{document}Zi=Xi−μiσi,where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${X}_{i}$$\end{document}Xi denotes an individual variable ii, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mu }_{i}$$\end{document}μi and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\sigma }_{i}$$\end{document}σi are its sample mean and standard deviation. This affine transformation aligns all variables onto a common zero-mean, unit-variance scale without altering their inherent distributional shapes (including skewness, kurtosis and tail behavior). As a result, features of vastly different magnitudes can be overlaid in a single plot for direct visual comparison, while preserving each parameter’s unique variability and probabilistic structure. Supplementary information Supplementary information Supplementary information