Introduction

Extreme climate conditions can exert strong impacts on vegetation productivity1,2,3. Mounting evidence shows that the rising frequency and intensity of extreme heatwaves and droughts have critically reduced terrestrial carbon sinks over the past few decades, as seen in events like the European summers of 2003, 2010, and 20224,5,6 and in Eastern China in 2013 and 20227,8. Persistent droughts can also heighten wildfire risks, triggering widespread tree mortality and CO2 emissions into the atmosphere2,9,10, significantly affecting regional and global carbon cycles1,11. Furthermore, increased atmospheric CO2 concentrations exacerbate global warming through biosphere feedback12,13,14. These events are likely to offset increases in carbon uptake due to elevated temperatures and CO2 concentrations in a warmer world1,15,16, exerting strong impacts on global and regional carbon balances.

Southwest China (SWC, mainly including the Yunnan, Guizhou, Guangxi, and Sichuan provinces, Fig. 1) is renowned for its abundant vegetation biomass and serves as a predominant carbon sink area in China17,18,19. However, vegetation growth in this region is also susceptible to drought suppression due to its unique and widespread karst geomorphology20,21. Precipitation amounts here have decreased remarkably in the past few decades, with an increasing frequency of extreme dry events22,23. In particular, a compound drought–heatwave event with extreme severity impacted this region in the spring of 2023. Widespread drought occurred in the preceding winter (Fig. S1), followed by a persistent heatwave, with record-breaking temperatures of 40 °C observed in Yunnan even in April24. These events triggered both the lowest recorded area-mean precipitation and surface soil moisture (SM) for the same period since 195025. Similarly, a “once-in-a-century” extreme drought of 2010 in SWC also originated from a drought in the preceding winter20, which led to significant reductions in regional vegetation productivity, especially from March to May26. It took approximately three months for the vegetation growth to recover to pre-drought conditions20. Several studies have suggested that such extreme events might become the “new normal” in the future under climate change27,28,29.

Fig. 1: Study region and corresponding land-cover types.
figure 1

a Geographical location of Southwest China (SWC) in East Asia, represented by a red box. Shadings represent elevations inland and the ocean depth. b Distribution of the four provinces, including Yunnan, Guangxi, Guizhou, and Sichuan in SWC. Shadings represent the major land-cover types.

Full size image

Low SM and a high atmospheric vapor pressure deficit (VPD) are considered two major abiotic stressors on vegetation growth during droughts30,31. Both of these can restrict terrestrial carbon uptake by limiting plant water availability and driving vegetation mortality2,4,32. On the one hand, low SM is unfavorable for hydraulic transfer from soil to leaves, causing plant xylem embolism1,32. On the other hand, a high VPD can contribute to declining carbon sequestration by limiting leaf stomatal conductance to minimize water loss, which is expected to restrict plant photosynthesis33,34,35. Along with global temperature increases and oceanic evaporation decreases, a rapidly increasing trend in the global mean VPD has been observed since the late 1990s34. Under these circumstances, the effects of high VPD on global vegetation growth have received growing attention recently35,36. Most previous analyses on extreme droughts in the SWC up to this point have primarily focused on the impacts of SM deficiency on vegetation growth20,26,37, but they have neglected the considerable contribution from high VPD anomalies.

In this study, we first explored the extent to which the extreme drought–heatwave concurrence in the spring of 2023 suppressed vegetation productivity in SWC. Then, the influencing mechanisms were systematically investigated, mainly from the perspective of contributions from both the SM and VPD anomalies. Finally, the relative roles of SM and VPD in suppressing vegetation productivity were assessed across different ecosystem types.

Results

Characterization of the 2023 spring heatwave–drought concurrence in SWC

An extremely severe compound drought–heatwave event struck SWC in late spring 2023, following the preceding winter–spring drought. The mean temperature and the number of high-temperature days in April and May ranked second and first since 1979, respectively24. In April, SWC saw domain-wide warming anomalies and decreasing precipitation amounts, with temperatures rising by up to 2 °C in Eastern Yunnan (Fig. 2a, d). The combination of a heatwave and a precipitation shortage exacerbated the drought conditions and triggered a northwest–southeast-oriented severe dry zone from southern Sichuan to eastern Yunnan, where the three-month Standardized Precipitation Evapotranspiration Index (SPEI-3, see Methods) was less than −1.5 (Fig. 2g). Warming anomalies and precipitation deficits persisted into May but shifted to eastern SWC, resulting in the eastward expansion of the dry zone. In contrast, cooling anomalies emerged over the western region with slight precipitation increases, which alleviated the local drought stress (Fig. 2b, e, and h). In June, cooling anomalies further expanded eastwards with positive precipitation anomalies occupying a larger domain (Fig. 2c, f), easing the drought in most parts of SWC, except the border areas of Yunnan, Guizhou, and Guangxi (Fig. 2i). Several similarities exist in the characteristics of the extreme drought of spring 2010, which also developed from the preceding winter drought (Fig. S1). The main difference is that the severe 2010 drought began to alleviate in April, whereas the 2023 extreme case was sustained and strengthened until May under the combined forcings from extreme warming and precipitation decreases. These seasonal differences between the two events suggest that in the 2023 drought events, the heatwave may induce a more severe SM shortage via the land-atmosphere feedback effect. Therefore, the SPEI reached its lowest level since 195025, while the drought range in 2023 is smaller than that in 2010. In the following sections, we will focus on the potential impacts of these extreme climate conditions on vegetation growth in SWC during late spring 2023.

Fig. 2: Overview of the SWC extreme compound heatwave–drought event in 2023.
figure 2

Detrended anomalies in T2m a–c; shading, unit: °C) and precipitation d–f; shading, unit: mm d−1) from April to June in 2023 relative to the climatology of 2001–2022. The hatching in a–f indicates values exceeding one standard deviation of those for the entire research period, and the red box spanning from 23–29°N and 102–107°E denotes the study region (similar to Figs. 3, 5 and 6). The SPEI-3 (g–i; shading) from April to June in 2023, and the solid gray lines in (g–i denote values less than −1.5 (i.e., severe drought).

Full size image

Responses of vegetation productivity and greenness to extreme climate conditions

It is widely recognized that extreme heatwaves and dry spells suppress vegetation carbon uptake due to the responses of photosynthesis to high temperatures and water availability limitations1,4,6. The gross primary productivity (GPP) showed slight changes before April 2023 (Fig. S2), suggesting that droughts do not always suppress vegetation growth because of the resistance of terrestrial ecosystems. In April, the GPP exhibited a significant decline across the southern part of SWC (Fig. 3a). As warming and drying anomalies persisted into May, significant GPP decreases expanded further northwards, with the magnitude greater than that in April (Fig. 3b). Although warming and drought conditions were alleviated in most areas by June, GPP decreases were more serious than in the preceding two months, even exceeding 4 gC m–2d–1 (half of the climatological mean for June) in the border areas of Yunnan and Guizhou (Fig. 3c). The evolution of the enhanced vegetation index (EVI) showed similar characteristics, but with relatively smaller changing amplitudes (Fig. 3d–f). Over the most-affected area within SWC (red boxes shown in Fig. 3), the regional mean of both the GPP and EVI during late spring (April–June) sharply decreased to the minimum value for the past two decades in 2023 (Fig. 3g). Compared to the 2010 extreme event, drought-induced declines in GPP and EVI in 2023 showed a remarkably larger scope and magnitude (Fig. S2). For both cases, the largest declines in GPP and EVI occurred one month later than the peak of the drought intensity (Fig. 4). The drought intensity reached its peak in May (March) for 2023 (2010), while the largest declines in GPP and EVI occurred in June (April).

Fig. 3: Changes in vegetation growth during the 2023 extreme events.
figure 3

The detrended anomalies in GPP (a–c; shading, unit: gC·m–2d–1) and EVI (d–f; shading, unitless) from April to June in 2023. The anomalies are relative to the climatology of 2001–2022. Hatching indicates anomalies exceeding the 90% confidence level, and the red box spanning from 23–29°N and 102–107°E denotes the study region (similar to Figs. 4 and 5). g Time series of the detrended anomalies of AMJ (i.e., April, May, and June) for the mean GPP and EVI in the red box region from 2001 to 2023. h Monthly evolutions of the GPP and EVI anomalies for 2023. The green and orange shading denotes the 10% and 90% quartiles of the GPP and EVI anomalies, respectively, in the red box region, and the yellow shading denotes the period with significant GPP and EVI anomalies.

Full size image
Fig. 4: Violin plots of the GPP and EVI anomalies during 2010 and 2023 extreme events.
figure 4

The frequency distribution of GPP (a–e) and EVI (f–j) anomalies from March to July in 2010 and 2023. The anomalies are relative to the climatology of 2001–2022. White dots indicate the median values, with gray boxes and thin gray lines indicative of the interquartile range and interdecile range, respectively.

Full size image

In general, it takes a certain amount of time for vegetation growth to recover to its pre-drought functional state after severe heatwaves and droughts, which is a crucial aspect of ecosystem resilience1,3. The most significant declines for both the GPP and EVI in 2023 occurred in June, with magnitudes of approximately 2 gC m–2d–1 and 0.07, respectively (Fig. 3h). Subsequently, the negative anomalies began to diminish gradually, but they did not fully recover from the drought suppression until September. Similar lagged impacts have been well-documented for previous extreme climate events8,20,38. Under severe drought stress, hydraulic imbalance and prolonged temperatures above the optimal temperature can damage the physiological structure of plants. The photosynthetic rate cannot immediately return to normal even after the drought lifted39. It is worth noting that the sensitivity of vegetation growth to extreme events depends highly on ecosystem types. The reductions in both the GPP and EVI in forests were less pronounced than those in the grasslands, especially in June (Fig. S3), as those areas may benefit from the high resistance and resilience of forest ecosystems1,40.

Possible mechanisms of vegetation growth suppression

Vegetation growth strongly relies on climate conditions, which are generally dominated by large-scale atmospheric circulations31,41. During late spring, significant anticyclonic circulation anomalies were observed in the upper troposphere over the SWC (Fig. 5a–c). These anomalies may be linked to decreased snow depth in North America, which in turn stimulates eastward-propagating Rossby waves24. These anomalous anticyclonic circulations can lead to anomalous descending motions that showed stronger intensities in April and May (Fig. 5d–f), which further favored total cloud cover (TCC) reductions (Fig. 5g–i). Lower cloud cover allowed for more incoming solar radiation to reach the land surface, and triggered heatwave events. They also suppressed the precipitation amounts and thus exacerbated the SM shortages. In late spring 2023, significant SM shortages mainly occupied southern Sichuan, eastern Yunnan, and western Guizhou and Guangxi, which is highly consistent with the spatial distribution corresponding to drought development (Fig. 6a–c). It should be noted that, when the drought tension alleviated over most areas in June, SM remained significantly low, which may result from the failure of deep soil water compensation in a timely manner (Fig. S4). Low SM directly limits evaporative cooling, and thus contributes to surface warming anomalies31,42. Although increased incoming solar radiation and warming anomalies could enhance the photosynthetic capacity, especially for the forest ecosystems43, these positive effects may be offset by the negative effects of a severe SM deficiency44. In addition, significant horizontal divergence anomalies appeared throughout the SWC in the lower troposphere, accompanied by anomalous winds blowing from land to sea (Fig. 5j–l). This caused reductions in the water vapor and actual vapor pressure (AVP)34,45. The combination of warming and low AVPs led to a significant increase in the VPD in April and May, with strong signals concentrated over the areas significantly affected by heatwaves and droughts (Fig. 6d, e). Warming anomalies weakened in June, returning VPD conditions to normal levels except in the border area between Yunnan and Guangxi (Fig. 6f). In general, low SM and high VPD anomalies induced by anomalous atmospheric circulations together provided favorable conditions for widespread reductions in both GPP and EVI across SWC.

Fig. 5: Anomalies in atmospheric circulations during the 2023 extreme events.
figure 5

Anomalies in the 200-hPa geopotential height (a–c; shading, unit: gpm), vertical velocity at 500 hPa (d–f; shading, unit: 10−2 Pa s–1), total cloud cover (TCC, g–i; shading, unit: %), and 700-hPa horizontal divergence (j–l; shading, unit:10–6 s–1) overlaid by the horizontal wind anomalies (blue vector, unit: 3 m s–1). The anomalies are relative to the climatology of 2001–2022. Hatching indicates values exceeding one standard deviation of those for the whole research period.

Full size image
Fig. 6: Changes in soil and atmospheric water during the 2023 extreme events and their relative roles on vegetation growth.
figure 6

Anomalies in the volumetric soil water content throughout the whole layer (a–c; shading, unit: m3 m–3) and VPD (d–f; shading, unit: kPa). The anomalies are relative to the climatology of 2001–2022. Hatching indicates anomalies exceeding the 90% confidence levels. Partial correlation coefficients of SM and VPD with GPP g and EVI h over different vegetation types over the SWC. The marker ‘**’ indicates a p-value less than 0.01.

Full size image

Dominant driver for vegetation growth suppression—SM or VPD?

Using partial correlation analysis (see Methods), we further examined whether differences existed in the sensitivity of various vegetation types to low SM and high VPD anomalies during late spring 2023. Though GPP and EVI in the forests, grasslands, and croplands were all positively correlated with the SM and showed negative correlations to the VPD (Fig. 6g, h), the relative importance of SM and VPD varied among different vegetation types. Compared to the VPD, the SM showed stronger relationships to both the GPP and EVI in forests, with correlation coefficients of 0.36 and 0.34, respectively. Both correlations are significant at the 99% confidence level. Forests with deep rooting systems can access soil water in deeper layers, thus maintaining strong resilience and resistance to droughts1,46,47. In June, although the precipitation increases compensated for the surface soil water shortage (Fig. 2f), the water did not penetrate deeply enough, resulting in overall significantly negative SM anomalies (Fig. S4). Thus, the SM deficiency might be the predominant reason for the declines in both the GPP and EVI in forest ecosystems during the heatwave–drought concurrence over SWC in spring 2023. In contrast, the GPP and EVI in the grasslands and croplands were more sensitive to the variations in the VPD. A dry atmosphere is more likely to demand water vapor from the vegetation in these circumstances48. Meanwhile, grasslands with shallower roots have difficulties absorbing soil water in deeper layers49, thus displaying closer relationships to the VPD50,51 than to the SM. Although the VPD in most parts of SWC returned to normal levels in June (Fig. 6f), the vegetation required time to return to its pre-drought functional state from extreme climates3,52. Similarly, the GPP and EVI in croplands were also more sensitive to the VPD variations. In this case, weaker correlations of the SM with both the GPP and VPD may be attributed to improved SM conditions by human management.

Previous studies have presented divergent assessments on the impact of dryness stressors on terrestrial carbon uptake30,35. It is worth noting that the dominant factors for vegetation responses to extreme droughts can vary among different cases even within the same areas6, likely due to discrepancies in the climatic context under which these drought events occurred. Regarding the SWC cases, most extreme and persistent droughts like the 2010 event took place outside the main growing season20,26,37. In this context, although slight VPD increases during drought may initially support vegetation growth, prolonged SM shortages are likely to play a more substantial role in suppressing growth43,44. To further substantiate our findings regarding the 2023 extreme case, we analyzed a similar extensive high-temperature event in SWC during 2019, which also occurred within the main growing season (April to June)53,54. Unlike the extreme compound drought-heatwave events of 2023, the 2019 high-temperature conditions were not preceded by drought in SWC (Fig. S5). The same partial correlation suggested that the GPP in forest ecosystems was still dominated by SM, while the GPP in grassland ecosystems was primarily influenced by VPD (Fig. S6). These findings align closely with those from the 2023 event, reinforcing the reliability of our conclusions.

Discussion and conclusions

Vegetation productivity in SWC experienced the most significant reductions for the past two decades during late spring 2023. This is mainly attributed to the preceding compound heatwave–drought events, with the mean temperature and the number of high-temperature days in April and May ranked the second and first highest since 1979, respectively. Studies have widely noted that extreme drought-related water deficiencies can damage the physiological structure of plants, preventing vegetation from restoring quickly to its pre-drought functional state1,3. The lagged effects of these extreme climate conditions led to the most substantial decreases in GPP and EVI occurring one month after the peak of drought intensity, with the vegetation growth not fully recovering from the drought’ s suppressive impact until September. Similar lagged impacts have been well-documented for the extreme events in 2010 (Fig. S2). However, we suggest that not every extreme dry event will have a lagged impact on vegetation growth. For instance, both the GPP reductions and drought intensity peaked in June during the extreme heatwave events over SWC in 2019 (Fig. S6). As the droughts got alleviated in most areas in July, the GPP reductions did not continue to develop. Therefore, it is essential to consider such impacts comprehensively, taking into account the climatic background of the event, as well as its duration and intensity.

For the underlying physical mechanisms, we found that anomalous anticyclonic circulations emerged in the upper troposphere over SWC in spring 2023, which triggered descending motion anomalies and further resulted in elevated temperatures and suppressed precipitation. These circumstances exacerbated SM shortages and high VPD—two major abiotic stressors regulating the terrestrial carbon uptake. The relative roles of the SM and VPD in driving vegetation productivity declines varied through different vegetation types in SWC. Forests with deep rooting systems can access soil water in deeper layers, maintaining strong resilience and resistance to surface soil and atmospheric dryness1,46,47. Thus, low SM conditions played a dominant role in forests, showing similar characteristics to those for the widespread European extreme drought events in the summers of 2003 and 20184,5. In contrast, grasslands and croplands with shallower roots have difficulties in absorbing soil water in deeper layers. Additionally, a dry atmosphere is more likely to demand water from the vegetation. Thus, both the grasslands and croplands were more sensitive to VPD changes than to SM conditions. Compared to previous studies on the extreme drought in 201020,26,37 that mainly concentrated on the SM shortage–related effects, we took into consideration the combined influences of changes in the SM and VPD on vegetation growth, providing new insights into drought-induced vegetation productivity anomalies at regional scales.

Thus far, the relative roles of SM and VPD in determining vegetation responses to dry conditions have remained unclear. Reichstein et al.1 suggested that the definition of extremes should consider a combination of variables rather than a single variable. From this perspective, the characterization of extreme events that affect vegetation growth should also take into account the climatic context. For instance, an increased VPD may initially promote vegetation photosynthesis to a certain extent, but the photosynthesis rate should be expected to decrease as the VPD continues to rise55,56. Therefore, the responses of vegetation productivity to VPD conditions may be complex and nonlinear. Although our study provided insights into the relative roles of the SM and VPD in the responses of different vegetation types to the 2023 extreme event in SWC, it is essential to acknowledge the joint contributions of other factors. More comprehensive research efforts are needed to understand the intricate responses of vegetation productivity to SM and VPD anomalies in a changing climate. Particularly, the occurrences of extreme drought events are projected to increase significantly, leading to a more extreme climate in SWC in the future28,29,57. Attention should be paid to the inhibitory effect of high VPD on vegetation growth. In summary, our study provides valuable guidance for future research endeavors aimed at understanding and mitigating the impacts of extreme drought events on ecosystems and human society at regional scales.

Methods

Climate variables data

We utilized atmospheric data from the ERA-558, including the monthly 2-m air temperature (T2m), 2-m dewpoint temperature, total precipitation, TCC, and the geopotential height and winds at different pressure levels. To characterize the SM conditions, we employed volumetric soil water content from the ERA5-Land reanalysis, which is defined as the ratio of the water volume to the unit soil volume. This dataset offers volumetric soil water content at four different depths (0–7 cm, 7–28 cm, 28–100 cm, and 100–289 cm). All datasets from the ERA-5 have a horizontal resolution of 0.25° × 0.25°, and this study uses the period from 2001 to 2023.

Drought severity was assessed using monthly SPEI datasets acquired from the National Center for Atmospheric Research with a horizontal resolution of 0.5° × 0.5°. The SPEI, an extension of the widely used Standardized Precipitation Index (SPI), incorporates both precipitation and potential evapotranspiration (PET) in determining drought59. Unlike the SPI, the SPEI not only depicts the impacts of precipitation changes on water demand but also accounts for the significant contribution from changing temperatures. Consistent with previous studies60,61, drought grades were categorized as mild drought (−0.5 to −1), moderate drought (−1.5 to −1), severe drought (−2 to −1.5), and extreme drought (≤−2). In this study, the SPEI at a time scale of three months was employed to depict the seasonal drought development.

Vegetation productivity and greenness data

The GPP dataset utilized in this study is derived from the FluxSat GPP version 2.2 product62. This dataset is generated through machine learning, employing the collection 6.1 MCD43C Bidirectional Reflectance Distribution Function (BRDF)–Adjusted Reflectance (NBAR) from the Moderate-Resolution Imaging Spectroradiometer (MODIS) instruments on the NASA Terra and Aqua satellites, which are trained on FLUXNET2015 and OneFlux eddy covariance data. The FluxSat GPP v2.2 is provided on the MERRA-2 grid with a horizontal resolution of 0.625° × 0.5° and is available from March 2000 to the present.

We utilized the monthly EVI as a canopy greenness product sourced from the Terra MODIS Vegetation Indices 16-Day (MOD13C1) Version 6.1 product63. In contrast to the Normalized Difference Vegetation Index (NDVI), the EVI is corrected for certain atmospheric conditions to the canopy background noise, making it more sensitive in areas with dense vegetation. The global MOD13C1 data represent cloud-free spatial composites derived from the gridded 16-day 1-kilometer MOD13A2 data and are presented as a Level-3 product projected at the 0.05° Climate Modeling Grid (CMG) resolution. The dataset is available from February 2000 to the present.

Land-cover type map

The land-cover type data utilized in this study were obtained from the Terra and Aqua combined MODIS Land Cover CMG (MCD12C1) Version 6.1 data product64. Our study primarily focused on SWC (mainly including the Yunnan, Guizhou, Guangxi, and Sichuan provinces), spanning from 95–110°E and from 20–30°N—a typical subtropical region. The land cover in the region predominantly consists of forests (including shrublands), grasslands, and croplands.

Calculations of VPD

The VPD is calculated as the difference between the saturation vapor pressure (SVP) and actual vapor pressure (AVP)65:

$${\rm{VPD}}={\rm{SVP}}-{\rm{AVP}}$$
(1)

The SVP is calculated by Tetens formula for temperatures above 0 °C:

$${\rm{SVP}}=0.611\times {{\rm{e}}}^{\frac{17.27{{\rm{T}}}_{{\rm{as}}}}{{{\rm{T}}}_{{\rm{as}}}+237.29}}$$
(2)

Here, the Tas represents the T2m in degrees Celsius. The calculated SVP is expressed in kilopascals (kPa). The AVP is determined by the 2-m dewpoint temperature (Td):

$${\rm{AVP}}=0.611\times {{\rm{e}}}^{\frac{17.27{{\rm{T}}}_{{\rm{d}}}}{{{\rm{T}}}_{{\rm{d}}}+237.29}}$$
(3)

Removing the linear trend

To remove the CO2 fertilization effects due to increased CO2 concentrations, the time series of the monthly GPP and EVI were linearly detrended. Similarly, the monthly temperature and precipitation time series were also linearly detrended to eliminate the warming trend. This detrending process helps to ensure that the analysis focuses on the variability and impacts of extreme events rather than on the long-term trends.

Partial correlation analysis

To quantify individual contributions of the VPD and SM to GPP/EVI anomalies over different vegetation types, we utilized a simple partial correlation analysis. We primarily focused on the areas in which the vegetation growth was significantly affected in the 2023 during the extreme events from April to June 2023 (as shown in Fig. 3). Combining the land-cover type data from MODIS, we extracted the anomalies of GPP and EVI of the three vegetation types, including forests, grasslands, and croplands, and calculated their partial correlation coefficients with the VPD and SM anomalies.