
==== Front
Natl Sci Rev
Natl Sci Rev
nsr
National Science Review
2095-5138
2053-714X
Oxford University Press

10.1093/nsr/nwae274
nwae274
Research Article
Earth Sciences
Nsr/9
AcademicSubjects/MED00010
AcademicSubjects/SCI00010
Compound hot–dry events greatly prolong the recovery time of dryland ecosystems
Yao Ying Conceptualization Data curation Formal analysis Investigation Methodology Writing - original draft State Key Laboratory of Earth Surface Processes and Resource Ecology, Faculty of Geographical Science, Beijing Normal University, Beijing 100875, China

Fu Bojie Conceptualization Writing - review & editing State Key Laboratory of Urban and Regional Ecology, Research Center for Eco-Environmental Sciences, Chinese Academy of Sciences, Beijing 100085, China

Liu Yanxu Conceptualization Writing - review & editing State Key Laboratory of Earth Surface Processes and Resource Ecology, Faculty of Geographical Science, Beijing Normal University, Beijing 100875, China

Zhang Yao Writing - review & editing Sino-French Institute for Earth System Science, College of Urban and Environmental Sciences, Peking University, Beijing 100871, China

Ding Jingyi Writing - review & editing State Key Laboratory of Earth Surface Processes and Resource Ecology, Faculty of Geographical Science, Beijing Normal University, Beijing 100875, China

https://orcid.org/0000-0002-6336-0981
Li Yan Writing - review & editing State Key Laboratory of Earth Surface Processes and Resource Ecology, Faculty of Geographical Science, Beijing Normal University, Beijing 100875, China

Zhou Sha Writing - review & editing State Key Laboratory of Earth Surface Processes and Resource Ecology, Faculty of Geographical Science, Beijing Normal University, Beijing 100875, China

Song Jiaxi Writing - review & editing State Key Laboratory of Earth Surface Processes and Resource Ecology, Faculty of Geographical Science, Beijing Normal University, Beijing 100875, China

Wang Shuai Writing - review & editing State Key Laboratory of Earth Surface Processes and Resource Ecology, Faculty of Geographical Science, Beijing Normal University, Beijing 100875, China

https://orcid.org/0000-0001-7297-9658
Li Changjia Writing - review & editing State Key Laboratory of Earth Surface Processes and Resource Ecology, Faculty of Geographical Science, Beijing Normal University, Beijing 100875, China

https://orcid.org/0000-0001-5342-354X
Zhao Wenwu Writing - review & editing State Key Laboratory of Earth Surface Processes and Resource Ecology, Faculty of Geographical Science, Beijing Normal University, Beijing 100875, China

Corresponding author. E-mail: bfu@rcees.ac.cn
10 2024
09 8 2024
09 8 2024
11 10 nwae27401 12 2023
23 5 2024
09 7 2024
10 9 2024
© The Author(s) 2024. Published by Oxford University Press on behalf of China Science Publishing & Media Ltd.
2024
https://creativecommons.org/licenses/by/4.0/ This is an Open Access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited.

ABSTRACT

Compound hot–dry events cause more severe impacts on terrestrial ecosystems than dry events, while the differences in recovery time (ΔRT) between hot–dry and dry events and their contributing factors remain unclear. Both remote sensing observations and eddy covariance measurements reveal that hot–dry events prolong the recovery time compared with dry events, with greater prolongation of recovery time in drylands than in humid regions. Random forest regression modeling demonstrates that the difference in vapor pressure deficit between hot–dry and dry events, with an importance score of 35%, is the major factor contributing to ΔRT. The severity of stomatal restriction exceeds that of non-stomatal limitation, which restricts the vegetation productivity that is necessary for the recovery process. These results emphasize the negative effect of vapor pressure deficit on vegetation recovery during hot–dry events and project an extension of drought recovery time considering elevated vapor pressure deficit in a warming world.

A large increase in VPD during compound hot–dry events dominated the greater prolongation of recovery time in dryland ecosystems.

recovery time
drought severity
drought duration
vegetation response
vapor pressure deficit
high temperature
National Natural Science Foundation of China 10.13039/501100001809 41991235 42171088 Science and Technology Project of Inner Mongolia Autonomous Region NMKJXM202109 Fundamental Research Funds for the Central Universities 10.13039/501100012226
==== Body
pmcINTRODUCTION

Under global warming, the probability and severity of compound hot–dry events have increased substantially [1–3]. It is widely reported that hot–dry events cause more serious losses than droughts without heat (precipitation-deficit droughts) [4,5]. For example, in Spain, moderate compound dry and hot conditions increased the likelihood of productivity loss by 8%–11% compared with that under moderate dry conditions [6]. In addition to the negative impacts on productivity, hot–dry events may also increase the mortality of vegetation and induce detrimental and sudden changes in the structure or function of ecosystems [7], which reduce the structural and functional stability of ecosystems. Given the importance of recovery capacity in maintaining ecosystem stability [8], it is critical to explore the recovery of ecosystems after hot–dry events.

Recovery time—the length of time it takes for ecosystems to return to normal states—is one of the important indicators used to measure the recovery of ecosystems from droughts [9,10]. Previous research has reported that the recovery time for most ecosystems is between 2 and 8 months [11,12], but these drought recovery time assessments did not distinguish between hot–dry events and dry events. Considering that hot–dry events usually caused more severe vegetation losses than dry events, ecosystems are expected to require a longer time to recover from hot–dry events. However, a quantitative assessment of the recovery time of ecosystems from hot–dry events is lacking. With the increasing frequency of hot–dry events [13], ecosystems with long recovery times may suffer another hot–dry event without full recovery, resulting in superimposed damage to ecosystems. Therefore, it is crucial to reveal the recovery dynamics of ecosystems from hot–dry events, contributing to the management and protection of ecosystems.

Post-drought recovery largely depends on the vegetation productivity for root growth, rebuilding of lost leaf area and regeneration of non-structural carbohydrate [14]. These processes are inseparable from efficient carbon and water cycles [15]. When vegetation is in water deficit (high vapor pressure deficit and low soil moisture), it regulates stomatal conductance to maximize carbon gains while reducing water loss [16]. Previous studies have reported that soil moisture effects dominated dryness stress on ecosystems and determined the recovery time of ecosystems from droughts [11,16]. However, when it comes to the recovery time of hot–dry events, there is no evidence proving the relative importance of low soil moisture and high vapor pressure deficit, which inhibits our understanding of ecosystem recovery from these two types of droughts. Given the significant increase in vapor pressure deficit under hot–dry events, is the relative importance of vapor pressure deficit for recovery time greater under such conditions? In addition, under hot–dry events, high temperature could reduce ribulose-1,5-bisphosphate (RuBP) content and Rubisco activity, and such non-stomatal limitations inhibit various physiological activities and ultimately impact vegetation recovery [17]. However, the performances of stomatal and non-stomatal limitations under these two types of droughts are unclear.

To address the abovementioned research gaps, this study explored the difference in the recovery time between hot–dry and dry events, and analysed the factors contributing to the difference in recovery time. We conducted this research in the following three parts. First, a hot–dry event refers to a dry event accompanied by a hot event and a dry event is not accompanied by hot event (see ‘Methods’). We determined the difference in recovery time between hot–dry and dry events based on the normalized difference vegetation index (NDVI), leaf area index (LAI), gross primary productivity (GPP) and vegetation optical depth (VOD). Second, we applied a random forest regression model to quantify the relative importance of contributing factors to the difference in recovery time. Third, we comparatively analysed the stomatal activities (canopy conductance, Gc) and non-stomatal activities (maximum photosynthetic assimilation rate, Amax) under hot–dry and hot conditions based on eddy covariance measurements to understand the recovery process from vegetation physiology. Our results indicate that the difference in the recovery time between hot–dry and dry events was more protracted in drylands than in humid regions, and the main factor was the difference in vapor pressure deficit. This study highlights that the negative effects of a high vapor pressure deficit on ecosystem recovery outweigh those of low soil moisture under hot–dry events. With global warming, the vapor pressure deficit is expected to increase [18], causing damage to a wider range of ecosystems, especially in drylands.

RESULTS AND DISCUSSION

Difference in recovery time of ecosystems between hot–dry and dry events

Compared with dry events, hot–dry events extend the recovery time by >1 month (Fig. 1). The global average recovery time of ecosystems from hot–dry and dry events was 5.43 and 4.23 months, respectively, based on the NDVI (Fig. 1a and c), consistently with the result obtained from LAI with average recovery time of 5.18 and 4.1 months, respectively (Fig. 1b and d). More than 80% of ecosystems recover from dry events within 6 months and ∼10% of ecosystems take >8 months to recover (Fig. 1c–f). Ecosystem recovery from hot–dry events tends to be more difficult—specifically, 20% of ecosystems need ≥8 months to recover (Fig. 1a, b, e and f), including those in southern South America, the Mediterranean region, southern Africa and Australia (Fig. 1a and b). This result suggests that declined moisture content (from humid region to dryland) reduces ecosystem resilience (Fig. S1) [9,19]. In addition to moisture availability, biological factors, such as canopy height, can also affect recovery after drought. Using global forest canopy height data [20], we discussed the correlation coefficients between recovery time and canopy height at the dryland, humid region and global scales, respectively. There are significant positive correlation coefficients between recovery time and canopy height (Fig. S2b). These results suggest that the recovery time increases with canopy height, which is supported by previous studies of drought-related tree mortality which have shown that large trees find it more difficult to recover and have higher mortality rates than small trees [21,22]. This could be explained by the fact that larger trees tend to take longer to produce sufficient material to repair the productivity loss (growth decline) caused by droughts [21].

Figure 1. Recovery time of ecosystems from hot–dry and dry events. (a, b) Recovery time of ecosystems from hot–dry events (RThot–dry) based on the NDVI and LAI, respectively. (c, d) Recovery time of ecosystems to dry events (RTdry) based on the NDVI and LAI, respectively. (e, f) Percentage of different recovery times based on the NDVI and LAI, respectively. Regions with sparse vegetation or no droughts are masked with white. Review drawing number: GS京(2024)1579.

We use the difference in recovery time between hot–dry and dry events (ΔRT) to characterize the prolongation of recovery time caused by hot–dry events compared with dry events, in which a larger ΔRT means a greater prolongation of recovery time during hot–dry events. Along different aridity gradients, ΔRT is always >0 (Fig. 2c and d), meaning that it is more difficult for ecosystems to recover from hot–dry events than from dry events. There is no doubt that there is longer time for ecosystems to recover from hot–dry than from dry events, given that the hot–dry events have greater drought severity and more severe moisture deficits (Fig. S3). Notably, ΔRT shortens with an increase in the aridity index, and the slope of the linear regression of ΔRT and the aridity index is –1.18 and –0.65 based on the NDVI and LAI, respectively (Fig. 2c and d), indicating that the prolongation of recovery time caused by hot–dry events is more severe in drier regions. In humid regions, ΔRT based on the NDVI and LAI is 0.93 and 0.90 months, respectively; however, ΔRT reaches 1.54 and 1.20 months in drylands (Fig. 2c and d, and Fig. S1) and even exceeds 4 months in Australia, Southern Africa, Southern South America and the Mediterranean region (Fig. 2a and b).

Figure 2. Difference in recovery time between hot–dry and dry events. (a, b) Spatial patterns of the difference in recovery time between hot–dry and dry events (ΔRT = RThot–dry – RTdry) based on the NDVI and LAI, respectively. (c, d) ΔRT under different aridity levels (aridity index, the ratio of precipitation to potential evapotranspiration) based on the NDVI and LAI, respectively. Asterisks indicate statistically significant differences (**P < 0.05). Review drawing number: GS京(2024)1579.

Consistently with NDVI and LAI, microwave VOD data reveal that hot–dry events lead to longer recovery time than dry events (Fig. S4a–c) and greater prolongation of recovery time in drylands than in humid regions (Fig. S4d). This suggests that hot–dry events cause more serious damage to biomass and carbon stock in dryland than in humid regions [23]. Notably, recovery time calculated using VOD (carbon stock) data is the longest among these three data sets (Fig. 1 and Fig. S4). Specifically, for hot–dry events, the recovery times based on NDVI, LAI and VOD are 5.43, 5.18 and 5.66 months, respectively (Fig. 1a and b, and Fig. S4c). For dry events, the recovery times based on NDVI, LAI and VOD are 4.23, 4.1 and 4.26 months, respectively (Fig. 1c and d, and Fig. S4c). This result implies that biomass and carbon stock recover more slowly after disturbance than leaf-scale area or greenness [24,25].

Considering that human activities may affect recovery time, we supplemented the recovery time of ecosystems from hot–dry and dry events under different human footprints. Under a low human footprint (human footprint is below the global average), the recovery time of ecosystems from hot–dry and dry events is 5.20 and 4.04 months (Fig. S5a), respectively, and ΔRT in dryland and humid regions is 1.22 and 1.07 months (Fig. S5b), respectively. Under a high human footprint (human footprint is above the global average), the recovery time of ecosystems from hot–dry and dry events is 5.15 and 4.18 months (Fig. S5c), respectively, and ΔRT in dryland and humid regions is 1.14 and 0.84 months (Fig. S5d), respectively. This result indicates that the finding of greater prolongation of recovery time under hot–dry events in drylands than in humid regions is robust under different human footprints.

Contributing factors of ΔRT

We attribute the ΔRT between drylands and humid regions to the differences in climatic factors during the recovery period, drought severity and vegetation loss (see ‘Methods’). The average ΔVPD (difference in vapor pressure deficit) anomaly between hot–dry and dry events is 2.8 and 1.85 in drylands and humid regions, respectively (Fig. 3a). This could be explained by the high temperature in the compound hot–dry environment increasing the atmospheric moisture demand and exacerbating soil evaporation and vegetation transpiration [26,27]. A previous study revealed that the evaporation rate could increase by 0.07 mm/day for each 1°C increase in soil temperature [28], leading to a decline in soil moisture [28,29]. When the soil water drops to relatively low levels, evaporation and transpiration are water-limited [30], resulting in less moisture transported to the atmosphere, especially in drylands [31]. Thus, there are substantial increases in vapor pressure deficit in drylands [32].

Figure 3. Contributing factors to the difference in recovery time between hot–dry and dry events. (a–g) ΔVPD, ΔDS, Δloss, ΔSM, ΔPRE, ΔSrad and ΔTEM. (h) Variable importance score based on the random forest regression model. (i–k) Partial dependence plots of ΔRT with the three most important variables (ΔVPD, ΔDS and Δloss). ΔVPD, ΔSM, ΔPRE, ΔSrad and ΔTEM represent the difference in vapor pressure deficit, soil moisture, precipitation, shortwave radiation and temperature during the recovery period between hot–dry and dry events. ΔDS represents drought severity indicated by the sum of SPI during drought duration. Δloss represents the difference in vegetation loss. Asterisks indicate statistically significant differences (**P < 0.05).

The average ΔDS (difference in drought severity) is 1.55 and 1.31 in drylands and humid regions, respectively, indicating that hot–dry events cause greater drought severity than dry events, and drylands are more severely impacted than humid regions (Fig. 3b). Vegetation losses caused by hot–dry events are greater than those caused by dry events, and the average Δloss is 0.08 and 0.04 in drylands and humid regions, respectively (Fig. 3c). Random forest regression modeling showed that ΔVPD is the most important contributing factor to ΔRT, followed by ΔDS and Δloss (Fig. 3h). The partial dependence plot reveals that ΔRT increases with increasing ΔVPD, ΔDS and Δloss (Fig. 3i–k) and these results suggest that the greater ΔVPD, ΔDS and Δloss in drylands than in humid regions contribute to the variation in ΔRT along the aridity gradient.

Although the contribution of ΔSM (difference in soil moisture) to ΔRT is lower than that of ΔVPD, ΔDS and Δloss (Fig. 3h), the significant negative correlation and partial correlation coefficients between ΔRT and ΔSM show that a lower ΔSM prolongs recovery time (Fig. 3d and Fig. S6). The ΔTEM (difference in temperature) in drylands is significantly higher than that in humid regions (Fig. 3g), but its contribution to ΔRT is lower than that of ΔVPD and ΔSM (Fig. 3h and Fig. S6). This could be explained by the fact that soil moisture and vapor pressure deficit have a direct impact on the vegetation recovery process by affecting the carbon–water cycle (vegetation roots absorb water from soil, and stomata regulate carbon uptake and water loss) [15], while temperature indirectly affects the recovery time by affecting soil moisture evaporation, vegetation transpiration, etc. [33].

Stomatal and non-stomatal limitations caused by hot–dry and dry events

Post-drought recovery is largely dependent on the vegetation productivity for reconstructing vegetation loss caused by drought [14] and it is crucial to eliminate the stomatal and non-stomatal limitations for the efficient progress of photosynthesis. We evaluate the recovery time of ecosystems from hot–dry and dry events based on GPP, temperature and precipitation from the FLUXNET2015 data set, and analyse the stomatal and non-stomatal anomalies under hot–dry and dry events based on canopy conductance (Gc) and the maximum photosynthetic assimilation rate (Amax) (see ‘Methods’). Consistently with remote sensing, eddy covariance measurements show that hot–dry events cause longer recovery time than dry events (Fig. 4a) and greater prolongation of recovery time in drylands than in humid regions (Fig. 4b). Compared with dry events, hot–dry events cause more severe stomatal and non-stomatal limitations (Fig. 4c and d). This result means greater damage to photosynthesis, explaining the longer recovery time from hot–dry than from dry events.

Figure 4. Response of vegetation to hot–dry and dry events based on eddy covariance measurements from the FLUXNET2015 dataset. (a) Comparison between recovery time of ecosystems from hot–dry and dry events. (b) Difference in recovery time between ecosystems from hot–dry and dry events. (c) Gc anomalies under hot–dry and dry events. (d) Amax anomalies under hot–dry and dry events. (e) Response curves of Gc anomalies and Amax anomalies related to increased vapor pressure deficit. (f) Response curves of Gc anomalies and Amax anomalies related to decreased soil moisture.

Notably, the Gc anomalies of hot–dry events are 2.55 times those of dry events, and the Amax anomalies of hot–dry events are 1.52 times those of dry events, meaning that hot–dry events lead to more severe stomatal restriction than non-stomatal restriction (Fig. 4c and d). This result is supported by an analysis of stomatal and non-stomatal limitations in mature deciduous tree species. It was found that, when considered independently from leaf age, the response of trees to drought was dominated by stomatal limitations, accounting for ∼75% of the total limitation [34]. Under moisture pressure, plants close their stomata to avoid the collapse caused by hydraulic failure due to xylem embolism [35]. Previous studies have reported that vapor pressure deficit directly affects vegetation transpiration through water potential [35,36] and the response of stomatal closure to the water potential of leaves or the canopy is timely and sensitive [37–40]. Stomatal limitations are more sensitive to both vapor pressure deficit and soil moisture than are non-stomatal limitations (Fig. 4e and f). Therefore, there are more severe stomatal limitations due to lower soil moisture and higher vapor pressure deficit under hot–dry events (Fig. S3).

CONCLUSION AND IMPLICATIONS

Considering the increasing hot–dry events under global warming (Figs S7 and S8), a wider range of ecosystems may suffer from repeated hot–dry events when they have not fully recovered. Clarifying the process of ecosystem recovery from hot–dry events is conducive to ecosystem management and conservation. Here, our study presents comprehensive evidence of the prolonged recovery time of vegetation in response to hot–dry events compared with dry events and the driving mechanisms based on global evidence from both remote sensing observations and eddy covariance measurements. We found that the recovery time from hot–dry events exceeds 5 months, which is more than that from dry events (∼4 months), and the difference in recovery time between hot–dry and dry events is exacerbated with intensified dryness. Spatially, ΔRT is >1.5 months in drylands, which is significantly longer than that in humid regions. The intensification of vapor pressure deficit in drylands under hot–dry events is the main factor contributing to the greater prolongation of recovery time in drylands.

The longer recovery time in drylands is indicative of high fragility during compound hot–dry events and the delayed recovery may bring about risks of degradation due to the inability to recover from repeated droughts [9,41]. It should be noted that glacier meltwater is one of the sources of water in arid regions [42] and precipitation alone could not adequately reflect local water deficits. This may lead to uncertainty in the recovery time assessment in the arid areas. High-temperature environments could exacerbate the glacial meltwater to compensate for the lack of precipitation [43]. More glacial meltwater and soil water monitoring in arid areas would facilitate regional drought assessment.

The intensification of vapor pressure deficit in drylands under hot–dry events dominates greater prolongation of recovery time in drylands than in humid regions. This finding updates the previous notion that soil moisture determines dryness stress and dominates drought recovery time [11,16], providing new insights for the vegetation recovery process of dryland ecosystems during compound hot–dry events. The significant increase in vapor pressure deficit in drylands due to high temperatures leads to severe stomatal limitations [36,38], which restrict the vegetation productivity necessary for the recovery process, resulting in longer recovery times. Considering global warming and decreasing terrestrial relative humidity [44], terrestrial ecosystems are expected to face increasing drought risk and recovery pressure, which may pose risks to the carbon sinks of ecosystems and challenge climate mitigation based on natural climate solutions.

METHODS

Drought index and vegetation index

We used the standardized precipitation index (SPI), which is a universal drought index, to identify dry conditions [45]. Notably, we did not use the standardized precipitation evapotranspiration index calculated based on monthly precipitation and evapotranspiration impacted by temperature because we needed to define compound hot–dry events and droughts without hot events separately [46]. SPI at 3 months—a short- and medium-term drought indicator—was selected to explore the impact of drought on vegetation in this study [47,48]. We selected the NDVI and LAI from the Global Inventory Monitoring and Modeling System 3g as proxies for characterizing the dynamics of vegetation growth [47,49,50].

Recovery time of ecosystems in response to hot–dry events and dry events

To accurately quantify ecosystem recovery time [51,52], the vegetation data used needed to be devoid of seasonal cycles and long-term trends (Fig. S9). We defined hot–dry events as the co-occurrence of the following three criteria (Fig. S10): (i) SPI was below –1 and lasted for ≥2 months [52,53]; (ii) a hot event was defined as an average temperature for two consecutive months exceeding the 90th percentile of temperature over the same period [2]; (iii) detrended vegetation data were below –0.1 SD (Fig. S10). Meanwhile, we determined the dry events according to three criteria: (i) SPI was below –1 and lasted for ≥2 months; (ii) there were no two consecutive months with temperatures above the 90th percentile of the same period; (iii) detrended vegetation data were below –0.1 SD. Recovery time was defined as the time it took for vegetation to return to its normal state from the maximum loss (Fig. S10). According to the aridity index (the ratio of precipitation to potential evapotranspiration) [54], we further determined the differences in recovery time (ΔRT) between hot–dry and dry events under different aridity levels. We supplemented the assessment of recovery time based on a negative vegetation anomaly threshold of –0.5 SD (Fig. S11).

Contributing factors of ΔRT between hot–dry and dry events

We explained the ΔRT between hot–dry and dry events according to the following three aspects: DS, vegetation loss and climatic factors during the recovery period caused by drought [12,55,56]. More details are available in the Supplementary material.

Stomatal and non-stomatal limitations caused by hot–dry and dry events derived from eddy covariance measurements

Productivity loss or greenness decline of vegetation results from stomatal and non-stomatal (maximum photosynthetic rate) limitations caused by droughts [38]. Stomatal limitation refers to the decline in photosynthesis caused by the partial closure of stomata during drought to save water [57] and non-stomatal restriction mainly refers to the decline in photosynthesis caused by non-stomatal factors such as the degradation of chloroplasts [58] and the declines in RuBP content and Rubisco activity [17]. We characterized the stomatal and non-stomatal limitation under hot–dry and dry events by using canopy conductance (Gc) and the maximum photosynthetic assimilation rate (Amax) derived from eddy covariance measurements from the FLUXNET2015 Tier 1 data set (Figs S12 and S13) [34,38,59–63].

DATA AND CODE AVAILABILITY

The LAI data set is available at http://www.mdpi.com/2072-4292/5/2/927. NDVI 3gv1 is available at http://poles.tpdc.ac.cn/en/data/9775f2b4-7370-4e5e-a537-3482c9a83d88/. The FLUXNET2015 data set is available at https://fluxnet.org/data/fluxnet2015-dataset/. The microwave-based VOD data are available at http://files.ntsg.umt.edu/data/LPDR_v2/. Precipitation, potential evapotranspiration, actual vapor pressure and temperature data are from the Climatic Research Unit gridded Time Series (CRU TS 4.05), available at https://crudata.uea.ac.uk/cru/data/hrg/. Root soil moisture data are available at https://www.gleam.eu/. Shortwave radiation data are available at https://cds.climate.copernicus.eu/cdsapp#!/dataset/reanalysis-era5-land-monthly-means?tab=overview. Human footprint data are available at https://figshare.com/articles/figure/An_annual_global_terrestrial_Human_Footprint_dataset_from_2000_to_2018/16571064. Global forest canopy height data are available at https://glad.umd.edu/dataset/gedi/.

Data preprocessing and the calculation of recovery time and Gc were performed in MATLAB (R2020b), Amax was calculated in R.4.0.2, random forest regression was performed in Python 3.9.0 and the figures were produced in Origin 2023.

Supplementary Material

nwae274_Supplemental_File

FUNDING

This work was supported by the National Natural Science Foundation of China (41991235 and 42171088), the Science and Technology Project of Inner Mongolia Autonomous Region, China (NMKJXM202109) and the Fundamental Research Funds for the Central Universities of China.

AUTHOR CONTRIBUTIONS

B. Fu, Y. Yao and Y. Liu designed the research. Y. Yao performed the analysis, created all figures and drafted the manuscript. Y. Zhang, J. Ding, Y. Li, S. Zhou and J. Song reviewed the manuscript. All authors contributed to the interpretation of the results and to the text.

Conflict of interest statement. None declared.
==== Refs
REFERENCES

1. Alizadeh  MR, Adamowski  J, Nikoo  MR  et al.  A century of observations reveals increasing likelihood of continental-scale compound dry-hot extremes. Sci Adv  2020; 6 : eaaz4571.10.1126/sciadv.aaz4571 32967839
2. Bevacqua  E, Zappa  G, Lehner  F  et al.  Precipitation trends determine future occurrences of compound hot–dry events. Nat Clim Chang  2022; 12 : 350–5.10.1038/s41558-022-01309-5
3. Sarhadi  A, Ausín  MC, Wiper  MP  et al.  Multidimensional risk in a nonstationary climate: joint probability of increasingly severe warm and dry conditions. Sci Adv  2018; 4 : eaau3487.10.1126/sciadv.aau3487 30498780
4. Byrne  B, Liu  J, Lee  M  et al.  The carbon cycle of Southeast Australia during 2019–2020: drought, fires, and subsequent recovery. AGU Adv  2021; 2 : e2021AV00046.10.1029/2021AV000469
5. Feng  SF, Wu  XY, Hao  ZC  et al.  A database for characteristics and variations of global compound dry and hot events. Weather Clim Extreme  2020; 30 : 100299.10.1016/j.wace.2020.100299
6. Ribeiro  AFS, Russo  A, Gouveia  CM  et al.  Risk of crop failure due to compound dry and hot extremes estimated with nested copulas. Biogeosciences  2020; 17 : 4815–30.10.5194/bg-17-4815-2020
7. Anderegg  WRL, Kane  JM, Anderegg  LDL. Consequences of widespread tree mortality triggered by drought and temperature stress. Nat Clim Chang  2013; 3 : 30–6.10.1038/nclimate1635
8. Hillebrand  H, Langenheder  S, Lebret  K  et al.  Decomposing multiple dimensions of stability in global change experiments. Ecol Lett  2018; 21 : 21–30.10.1111/ele.12867 29106075
9. Liu  LB, Gudmundsson  L, Hauser  M  et al.  Revisiting assessments of ecosystem drought recovery. Environ Res Lett  2019; 14 : 114028.10.1088/1748-9326/ab4c61
10. Zhang  SL, Yang  YT, Wu  XC  et al.  Postdrought recovery time across global terrestrial ecosystems. J Geophys Res-Biogeosci  2021; 126 : e2020JG005699.10.1029/2020JG005699
11. Yao  Y, Liu  YX, Zhou  S  et al.  Soil moisture determines the recovery time of ecosystems from drought. Glob Change Biol  2023; 29 : 3562–74.10.1111/gcb.16620
12. Schwalm  CR, Anderegg  WRL, Michalak  AM  et al.  Global patterns of drought recovery. Nature  2017; 548 : 202–5.10.1038/nature23021 28796213
13. Zhang  Y, Hao  ZC, Zhang  X  et al.  Anthropogenically forced increases in compound dry and hot events at the global and continental scales. Environ Res Lett  2022; 17 : 024018.10.1088/1748-9326/ac43e0
14. Kannenberg  SA, Schwalm  CR, Anderegg  WRL  et al.  Ghosts of the past: how drought legacy effects shape forest functioning and carbon cycling. Ecol Lett  2020; 23 : 891–901.10.1111/ele.13485 32157766
15. Joshi  J, Stocker  BD, Hofhansl  F  et al.  Towards a unified theory of plant photosynthesis and hydraulics. Nat Plants  2022; 8 : 1304–16.10.1038/s41477-022-01244-5 36303010
16. Liu  LB, Gudmundsson  L, Hauser  M  et al.  Soil moisture dominates dryness stress on ecosystem production globally. Nat Commun  2020; 11 : 4892.10.1038/s41467-020-18631-1 32994398
17. Bota  J, Medrano  H, Flexas  J. Is photosynthesis limited by decreased Rubisco activity and RuBP content under progressive water stress?  New Phytol  2004; 162 : 671–81.10.1111/j.1469-8137.2004.01056.x 33873761
18. Fang  ZX, Zhang  WM, Brandt  M  et al.  Globally increasing atmospheric aridity over the 21st century. Earth Future  2022; 10 : e2022EF003019.10.1029/2022EF003019
19. Yao  Y, Fu  BJ, Liu  YX  et al.  Evaluation of ecosystem resilience to drought based on drought intensity and recovery time. Agric For Meteorol  2022; 314 : 108809.10.1016/j.agrformet.2022.108809
20. Potapov  P, Li  XY, Hernandez-Serna  A  et al.  Mapping global forest canopy height through integration of GEDI and Landsat data. Remote Sens Environ  2021; 253 : 112165.10.1016/j.rse.2020.112165
21. Trugman  AT, Detto  M, Bartlett  MK  et al.  Tree carbon allocation explains forest drought-kill and recovery patterns. Ecol Lett  2018; 21 : 1552–60.10.1111/ele.13136 30125446
22. Stovall  AEL, Shugart  H, Yang  X. Tree height explains mortality risk during an intense drought. Nat Commun  2029; 10 : 4385.10.1038/s41467-019-12380-6
23. Du  JY, Kimball  JS, Jones  LA  et al.  A global satellite environmental data record derived from AMSR-E and AMSR2 microwave Earth observations. Earth Syst Sci Data  2017; 9 : 791–808.10.5194/essd-9-791-2017
24. Fan  L, Wigneron  J-P, Ciais  P  et al.  Siberian carbon sink reduced by forest disturbances. Nat Geosci  2023; 16 : 56–62.10.1038/s41561-022-01087-x
25. Yang  J, Tian  HQ, Pan  SF  et al.  Amazon drought and forest response: largely reduced forest photosynthesis but slightly increased canopy greenness during the extreme drought of 2015/2016. Glob Change Biol  2018; 24 : 1919–34.10.1111/gcb.14056
26. Buras  A, Rammig  A, Zang  CS. Quantifying impacts of the 2018 drought on European ecosystems in comparison to 2003. Biogeosciences  2020; 17 : 1655–72.10.5194/bg-17-1655-2020
27. Drake  JE, Tjoelker  MG, Vårhammar  A  et al.  Trees tolerate an extreme heatwave via sustained transpirational cooling and increased leaf thermal tolerance. Glob Change Biol  2018; 24 : 2390–402.10.1111/gcb.14037
28. Kidron  GJ, Kronenfeld  R. Temperature rise severely affects pan and soil evaporation in the Negev Desert: temperature rise affects pan and soil evaporation. Ecohydrology  2016; 9 : 1130–8.10.1002/eco.1701
29. Bastos  A, Ciais  P, Friedlingstein  P  et al.  Direct and seasonal legacy effects of the 2018 heat wave and drought on European ecosystem productivity. Sci Adv  2020; 6 : eaba2724.10.1126/sciadv.aba2724 32577519
30. Jung  M, Reichstein  M, Ciais  P  et al.  Recent decline in the global land evapotranspiration trend due to limited moisture supply. Nature  2010; 467 : 951–4.10.1038/nature09396 20935626
31. Rashid  MA, Andersen  MN, Wollenweber  B  et al.  Acclimation to higher VPD and temperature minimized negative effects on assimilation and grain yield of wheat. Agric For Meteorol  2018; 248 : 119–29.10.1016/j.agrformet.2017.09.018
32. Lin  HQ, Li  Y, Zhao  L. Partitioning of sensible and latent heat fluxes in different vegetation types and their spatiotemporal variations based on 203 FLUXNET sites. J Geophys Res-Atmos  2022; 127 : e2022JD037142.10.1029/2022JD037142
33. Green  JK, Ballantyne  A, Abramoff  R  et al.  Surface temperatures reveal the patterns of vegetation water stress and their environmental drivers across the tropical Americas. Glob Change Biol  2022; 28 : 2940–55.10.1111/gcb.16139
34. Wilson  KB, Baldocchi  DD, Hanson  PJ. Quantifying stomatal and non-stomatal limitations to carbon assimilation resulting from leaf aging and drought in mature deciduous tree species. Tree Physiol  2000; 20 : 787–97.10.1093/treephys/20.12.787 12651499
35. Grossiord  C, Buckley  TN, Cernusak  LA  et al.  Plant responses to rising vapor pressure deficit. New Phytol  2020; 226 : 1550–66.10.1111/nph.16485 32064613
36. Jalakas  P, Takahashi  Y, Waadt  R  et al.  Molecular mechanisms of stomatal closure in response to rising vapour pressure deficit. New Phytol  2021; 232 : 468–75.10.1111/nph.17592 34197630
37. Martin-StPaul  N, Delzon  S, Cochard  H. Plant resistance to drought depends on timely stomatal closure. Ecol Lett  2017; 20 : 1437–47.10.1111/ele.12851 28922708
38. Fu  Z, Ciais  P, Prentice  IC  et al.  Atmospheric dryness reduces photosynthesis along a large range of soil water deficits. Nat Commun  2022; 13 : 989.10.1038/s41467-022-28652-7 35190562
39. Shao  XY, Gao  XD, Zeng  YJ  et al.  Eco-physiological constraints of deep soil desiccation in semiarid tree plantations. Water Resour Res  2023; 59 : e2022WR034246.10.1029/2022WR034246
40. Hasanagić  D, Koleška  I, Kojić  D  et al.  Long term drought effects on tomato leaves: anatomical, gas exchange and antioxidant modifications. Acta Physiol Plant  2020; 42 : 121.10.1007/s11738-020-03114-z
41. Liu  ZB, Zhu  JY, Xia  JY  et al.  Declining resistance of vegetation productivity to droughts across global biomes. Agric For Meteorol  2023; 340 : 109602.10.1016/j.agrformet.2023.109602
42. Pritchard  HD . Asia's shrinking glaciers protect large populations from drought stress. Nature  2019; 569 : 649–54.10.1038/s41586-019-1240-1 31142854
43. Shugar  DH, Burr  A, Haritashya  UK  et al.  Rapid worldwide growth of glacial lakes since 1990. Nat Clim Chang  2020; 10 : 939–45.10.1038/s41558-020-0855-4
44. Pokhrel  Y, Felfelani  F, Satoh  Y  et al.  Global terrestrial water storage and drought severity under climate change. Nat Clim Chang  2021; 11 : 226–33.10.1038/s41558-020-00972-w
45. Guan  YL, Liu  JL, Chen  AF  et al.  Spatial aggregation of global dry and wet patterns based on the standard precipitation index. Earth Future  2022; 10 : e2022EF002720.10.1029/2022EF002720
46. Vicente-Serrano  SM, Beguería  S, López-Moreno  JI. A multiscalar drought index sensitive to global warming: the standardized precipitation evapotranspiration index. J Clim  2010; 23 : 1696–718.10.1175/2009JCLI2909.1
47. Zhang  Y, Keenan  TF, Zhou  S. Exacerbated drought impacts on global ecosystems due to structural overshoot. Nat Ecol Evol  2021; 5 : 1490–8.10.1038/s41559-021-01551-8 34593995
48. Zhang  Y, Li  ZL. Uncertainty analysis of standardized precipitation index due to the effects of probability distributions and parameter errors. Front Earth Sci  2020; 8 : 76.10.3389/feart.2020.00076
49. Zhu  ZC, Bi  J, Pan  YZ  et al.  Global data sets of vegetation leaf area index (LAI)3 g and fraction of photosynthetically active radiation (FPAR)3g derived from Global Inventory Modeling and Mapping Studies (GIMMS) normalized difference vegetation index (NDVI3g) for the period 1981 to 2011. Remote Sens  2013; 5 : 927–48.10.3390/rs5020927
50. Jiao  WZ, Wang  LX, Smith  WK  et al.  Observed increasing water constraint on vegetation growth over the last three decades. Nat Commun  2021; 12 : 3777.10.1038/s41467-021-24016-9 34145253
51. Forzieri  G, Dakos  V, McDowell  NG  et al.  Emerging signals of declining forest resilience under climate change. Nature  2020; 608 : 534–9.10.1038/s41586-022-04959-9
52. Jiao  T, Williams  CA, De Kauwe  MG  et al.  Patterns of post-drought recovery are strongly influenced by drought duration, frequency, post-drought wetness, and bioclimatic setting. Glob Change Biol  2021; 27 : 4630–43.10.1111/gcb.15788
53. Yao  N, Li  Y, Lei  TJ  et al.  Drought evolution, severity and trends in mainland China over 1961–2013. Sci Total Environ  2018; 616–617 : 73–89.10.1016/j.scitotenv.2017.10.327
54. Huang  JP, Yu  HP, Guan  XD  et al.  Accelerated dryland expansion under climate change. Nat Clim Chang  2016; 6 : 166–71.10.1038/nclimate2837
55. He  B, Liu  JJ, Guo  LL  et al.  Recovery of ecosystem carbon and energy fluxes from the 2003 drought in Europe and the 2012 drought in the United States. Geophys Res Lett  2018; 45 : 4879–88.10.1029/2018GL077518
56. Chiang  F, Mazdiyasni  O, AghaKouchak  A. Evidence of anthropogenic impacts on global drought frequency, duration, and intensity. Nat Commun  2021; 12 : 2754.10.1038/s41467-021-22314-w 33980822
57. Henry  C, John  GP, Pan  RH  et al.  A stomatal safety-efficiency trade-off constrains responses to leaf dehydration. Nat Commun  2019; 10 : 3398.10.1038/s41467-019-11006-1 31363097
58. Wang  YX, Li  XY, Liu  NN  et al.  The iTRAQ-based chloroplast proteomic analysis of Triticum aestivum L. leaves subjected to drought stress and 5-aminolevulinic acid alleviation reveals several proteins involved in the protection of photosynthesis. BMC Plant Biol  2020; 20 : 96.10.1186/s12870-020-2297-6 32131734
59. Luo  XZ, Keenan  TF. Global evidence for the acclimation of ecosystem photosynthesis to light. Nat Ecol Evol  2020; 4 : 1351–7.10.1038/s41559-020-1258-7 32747771
60. dos Reis  MG, Ribeiro  A. Conversion factors and general equations applied in agricultural and forest meteorology. AgroM  2019; 27 : 227–58.10.31062/agrom.v27i2.26527
61. Novick  KA, Ficklin  DL, Stoy  PC  et al.  The increasing importance of atmospheric demand for ecosystem water and carbon fluxes. Nat Clim Chang  2016; 6 : 1023–7.10.1038/nclimate3114
62. Pennypacker  S, Baldocchi  D. Seeing the fields and forests: application of surface-layer theory and flux-tower data to calculating vegetation canopy height. Bound-Layer Meteor  2016; 158 : 165–82.10.1007/s10546-015-0090-0
63. Lasslop  G, Reichstein  M, Papale  D  et al.  Separation of net ecosystem exchange into assimilation and respiration using a light response curve approach: critical issues and global evaluation: separation of NEE into GPP and RECO. Glob Change Biol  2010; 16 : 187–208.10.1111/j.1365-2486.2009.02041.x
