Co-regulation of water and energy in the spatial heterogeneity of drought resistance and resilience

Vegetation resistance and resilience to drought are linked to the stability of terrestrial ecosystems under climate change. However, the factors driving the spatial heterogeneity in drought resistance and resilience remain poorly understood. In the study, we utilized multiple satellite-derived vegetation indices to calculate and analyze changes in drought resistance and resilience across various biomes worldwide. Results indicated that drought resistance showed a significant increase with the increase in water availability, but no significant relationship was observed between drought resistance and energy. In contrast, drought resilience exhibited a significant increase with an increase in energy rather than in water. Furthermore, a negative correlation was observed between drought resistance and resilience across different biomes worldwide, indicating a trade-off between resistance and resilience. However, the strength of the negative correlation varied based on water and energy conditions. These findings provide compelling evidence that water and energy co-regulated the spatial heterogeneity in drought resistance and resilience across the globe. The robust linear relationship between drought resistance and resilience and available water and energy demonstrated in our study is critical to accurately predicting and assessing the impact of climate change on vegetation growth and terrestrial carbon cycling in the future.


Introduction
Drought, as an extreme climatic event, can have devastating effects on vegetation growth.Widespread droughts in terrestrial ecosystems can alter the carbon cycle, decrease vegetation productivity, accelerate tree mortality, and cause ecological deterioration (Zhao andRunning 2010, Van der Molen et al 2011).With climate warming, we have observed widespread droughts in terrestrial ecosystems, and the frequency and severity of droughts are expected to rise in the future according to model predictions (Xu et al 2019, Gampe et al 2021).Therefore, understanding how terrestrial ecosystems respond to drought stress is critical for predicting the carbon cycling of terrestrial ecosystems in a warmer future.
To date, resistance and resilience have been commonly employed to measure the effects of drought on terrestrial ecosystems, including concurrent and legacy effects (Liang and Yuan 2021).Drought resistance is defined as the ability of an ecosystem to withstand drought stress (Schwalm et al 2017, Li et al 2019), and resilience reflects its ability to rebound to a normal state after a drought (De Vries et al 2012, Isbell et al 2015).Responses of terrestrial ecosystems to drought are highly variable across different ecoregions (Vicente-Serrano et al 2014).Forests, which harbor most of the long-lived woody species, tend to exhibit greater resistance but less resilience due to the high energy and resource investment for growth after drought.By contrast, annual grasslands may present lower resistance but greater resilience owing to their rapid regrowth and establishment (Hoover et al 2014).Previous research has suggested that variations in vegetation resistance and resilience across different ecoregions may be due to ecological adaptation to local environmental conditions (Knapp andSmith 2001, Huxman et al 2004).Precipitation, as the main source of water availability in the environment, may drive spatial heterogeneity in the resistance and resilience of terrestrial ecosystems (Ma et al 2015, Zhao et al 2015, Wen et al 2018, Zhang et al 2021a, 2021b, Herberich et al 2023).Temperature and solar radiation also influence ecosystem resistance and resilience by promoting photosynthesis (Yang et al 2016, Huang and Xia 2019, Zhang et al 2019).Nevertheless, at the global scale, the relationship between resistance or resilience and local environmental variables remains inconclusive (Stuart-Haentjens et al 2018).Therefore, the underlying factors driving this spatial heterogeneity in drought resistance and resilience across different ecoregions are not yet fully understood.This significantly hinders our prediction of the changes in global terrestrial carbon cycling in future climate scenarios.
In this study, we investigated spatial variations in drought resistance and resilience across various ecoregions at the global scale using multiple remotesensing datasets.To elucidate the drivers of resistance and resilience, we further assessed the effects of water and energy on drought resistance and resilience based on a series of environmental variables, including precipitation, soil moisture, temperature, surface solar radiation, vapor pressure deficit (VPD), and potential evapotranspiration (PET).We hypothesized that drought resistance would increase with water availability, while resilience would increase with energy availability.

Data sources 2.1.1. Satellite Vegetation data
We used two remote-sensing normalized difference vegetation index (NDVI) datasets, including Moderate Resolution Imaging Spectroradiometer (MODIS) NDVI dataset and Global Inventory Monitoring and Modeling System (GIMMS) NDVI dataset, to study the drought resistance and resilience.The NDVI assesses vegetation growth by reflecting the fraction of photosynthetically active radiation absorbed by vegetation canopies.The MODIS NDVI product (MOD13A2), with a spatial resolution of 1 km covering 2000-2020, was obtained from the NASA Land Processes Distributed Active Archive Center (Didan et al 2015).To ensure the robustness of result, the GIMMS NDVI product between 1982 and 2015, with a spatial resolution is 1/12 • , was obtained from the Ecological Forecasting Lab of NASA Ames Research Center.(http://ecocast.arc.nasa.gov/).

Environment datasets
Monthly temperature, precipitation, and PET data with a spatial resolution of 0.5 • between 2000 and 2020 were obtained from the Climate Research Unit (CRU TS version 4.05, https://crudata.uea.ac.uk).VPD is calculated as the difference between saturated vapor pressure (S VAP ) and actual water vapor pressure (VAP).The VAP dataset was provided by the CRU (CRU TS version 4.05, https://crudata.uea.ac.uk).The S VAP was calculated based on mean temperature by using the SVP function in the humidity packages in R (Cai 2018).The monthly surface net solar radiation (SSR) dataset was obtained from the ECMWF Reanalysis 5 (ERA5) covering 2000-2020(Muñoz-Sabater et al 2021).The soil moisture dataset between 2000 and 2020 was obtained from the NASA Global Land Data Assimilation System Version (GLADS-2.1).This soil moisture dataset, with a spatial resolution of 0.25 • , provides soil moisture in four layers: 0-10 cm, 10-40 cm, 40-100 cm, and 100-200 cm.We used the weighted means of soil moisture based on the soil depth of the four layers to estimate soil water conditions in our study.

Drought index
The Palmer drought severity index (PDSI) that considers precipitation and evapotranspiration is a widely-applied meteorological drought index (Burke 2011, Sheffield et al 2012, Yan et al 2016).However, the PDSI is often influenced by different climates and land use properties, and the PDSI overrepresents the frequency of extremely dry or wet spells (Van der Schrier et al 2006).Wells et al (2004) proposed a new formulation called the self-calibrated PDSI (scPDSI).The scPDSI could reduce the excessive frequency of extreme events compared to the original PDSI and has been frequently applied to characterize drought (Cook et al 2014, Trenberth et al 2014, Dorigo et al 2017).Monthly self-calibrating PDSI datasets with a spatial resolution of 0.5 • over 1982 and 2020 were downloaded from the CRU (https://crudata.uea.ac.uk/cru/data/drought/).

Ecoregions classification
Global ecological zone (GEZ) map was commonly used to determine ecoregions in previous studies (Carreiras et al 2017, Alvarenga et al 2020, Feeley et al 2020, Sayre et al 2020).The GEZ ecoregion classification relies on a combination of climate and potential vegetation, specific rules for ecological zones, and highlights the distinction between arid and moist regions, altitudinal belts, and latitudinal regions.The classified ecoregions by GEZ 2010 are provided in the supporting information (figure S1, table S1).

Data analysis 2.2.1. Estimation of drought resistance and resilience
We first identified drought events using the annual scPDSI between1982 and 2020.According to the scP-DSI, we identifed each year as a 'drought' year or a 'normal' year at the pixel level.Many studies have showed that PDSI values below −2 indicate the occurrence of drought events, which reduce photosynthetic productivity (Ditmarová et al 2010, Xiao et al 2010, Rosbach and Andersen 2017).We defined the 'drought' year when the annual mean scPDSI was below −2.Conversely, the annual scPDSI with a score between −2 and 2 was defined as the 'normal' year.We mapped the frequency of drought at the pixel level during 2000-2020 (figure S2).We retained only drought years when both scPDSI and vegetation data have valid records for each grid cell.Pure deserts and tundra in GEZ 2010 were also excluded from the studied area.
Drought resistance is defined as the ability of vegetation to resist drought disturbance, and resilience can be defined as the ability of vegetation productivity to recover after drought (Lloret et al 2011, Dakos et al 2015, Liang et al 2021).According to previous studies (Scheffer et al 2009, Van Ruijven and Berendse 2010), we selected one year period as the recovery time of vegetation from external perturbations in this study.We focused on individual drought events and ignored consequent drought event, which may produce biased results (Li et al 2020), when estimated resistance and resilience between 1982 and 2020.Drought resistance (RT) and resilience (RS) for each drought year and grid cell based on the definitions proposed by Isbell et al (2015) as the following formulas: where Y n is the mean NDVI in 'normal' years, Y e is NDVI in 'drought' years, and Y e + 1 is NDVI in the next year after drought years.We calculated drought resistance and resilience for each pixel during 2000-2020 using MODIS NDVI dataset, and during 1982-2015 using GIMMS NDVI dataset.

The effects of environmental factors on resistance and resilience
To examine the drivers of drought resistance or resilience, we collected six environmental variables, including temperature, precipitation, PET, VPD, SSR, and soil moisture in the study.We calculated mean annual precipitation (MAP, mm•yr −1 ), average annual temperature (MAT, • C), PET (mm d −1 ) and VPD (KPa), SSR (J•m −2 ) and total soil moisture (TSM) between 2000 and 2020.The TSM (kg•m −2 •cm −1 ) was calculated as the weighted average according to the soil depth of each layer.All the spatial data were resampled to 0.5 degree spatial resolution by bilinear interpolation method using the 'raster' package (Hijmans et al 2017) in R (R Core Team 2021).
To reduce the potential bias due to collinearity, principal component analysis (PCA) was first used to reduce the dimensions of the selected six environmental variables (Jackson 1993, Xu et al 2016).We extracted the first (PC1) and second (PC2) axes in the PCA across all selected environmental factors according to a broken-stick stopping rule and loadings as the energy and water components, respectively (table S2).
Pearson correlation analysis was used to demonstrate the relationship between environmental factors and PC1 or PC2.We mapped the spatial distribution of water and energy components as well as drought resistance and resilience across the world.To be consistent of the standardized principal components, we also scaled drought resistance and resilience with Z-scores.The negative and positive RT or RS mean lower or higher resistance or resilience, respectively.To understand changes in drought resistance and resilience in regions under different water and energy conditions, we divided all pixels into regions with low and high water or energy based on the means of these variables across all pixels.Bootstrapping method considers the assumed distribution type and uses parameter estimation to generate resampled data.It can better capture the true distribution of the mean for comparing two groups and avoid false-positive error resulted from large sample size (Gareth et al 2013, Jorgensen andFath 2018).We used bootstrapping method to estimate the mean distribution and calculate 95% confidence intervals (95% CIs) of drought resistance and resilience under different water and energy conditions.The difference between groups was considered statistically significant when their 95% CIs were not overlapped.The aridity index (AI), which is defined as the ratio of precipitation (Pre) to PET, reflects the exchangeability of water and energy.To investigate the relationship between water or energy and resistance or resilience, we also re-classified all grid cells based on AI and compared vegetation resistance and resilience across the aridity gradient.
We used path analysis to investigate the direct and indirect effects of water and energy on drought resistance and resilience.The water-related variables included MAP and TSM, while the energy-related variable included mean annual temperature (MAT), SSR, VPD, and PET.The AI, defined as the ratio of precipitation (Pre) to PET, reflecting the exchangeability of water and energy, was also included in the path analysis.In this study, we employed the partial least squares path modeling (PLS-PM) approach for the path analysis.PLS-PM can deal with complexity and non-normality sample sizes, focusing on prediction and practical significance, accommodating different measurement models, and being robust to model misspecification compared to traditional path analysis (Crocetta et al 2021).The goodnessof-fit index (Gof), which was calculated as the geometric mean of the average community and the average R 2 value, was used to assess the overall prediction performance of the PLS-PM model.The range of Gof is from 0 to 1, and a higher value of Gof means better prediction power for the model (Ravand and Baghaei 2016).In the PLS-PM model, the loading of each index variable was used to estimate the score of the potential variable.The PLS-PM analyses are carried out using plspm packages (Sanchez et al 2013) in R version 4.2.3 (R Core Team 2021).
To explain the spatial difference in resistance and resilience across different ecoregions, we calculated the means of resistance, resilience, energy, and water for each ecoregion.The linear regression model was applied to estimate the relationship between resistance or resilience and water or energy.We then compared the differences in water, energy, resistance and resilience of various ecoregions using oneway analysis of variance (ANOVA).We then used gradient boosting regression tree (GBRT) to estimate the importance of water and energy on vegetation resistance and resilience of various ecoregions.As a highly effective and flexible statistical learning method, GBRT is built by combining multiple regression trees (Ridgeway 2007).We set the n.trees and shrinkage values as 1000 and 0.01, respectively.The GBRT procedure was conducted by using the 'gbm' package (Ridgeway 2007) in R (R Core Team 2021).We further quantified the variances in resistance and resilience explained by water, energy and vegetation type using the 'relaimpo' package (Groemping and Matthias 2018) in R (R Core Team 2021).
Pearson correlation analysis was also used to estimate the relationship between vegetation resistance and resilience to drought in regions with low and high water or energy.The non-parametric bootstrapping (1000 iterations) was used to calculate the CIs of the correlation coefficients.All data analyses were conducted using R version 4.2.3 (R Core Team 2021).

Results
Two independent-dimension variables related to water and energy were determined using the PCA approach by considering six environmental variables: MAT, SSR, VPD, PET, MAP, and TSM (figure 1(a)).PC1 had a strong positive correlation with MAT, SSR, VPD, and PET but showed a weak correlation with MAP.In contrast, PC2 had strong positive correlations with MAP and TSM (table S2 and figure 1(a)).Therefore, PC1 and PC2 were used to represent energy and water, respectively.The water and energy axes explained 21.1% and 63.2% of the total variations, respectively (figure 1(a)).The region near the tropics had high PC1 (energy) values, while the boreal region had low PC1 (energy) values (figure 1(c)).Tropical regions had high water availability, while temperate regions had low water availability (figure 1(d)).
We observed widespread spatial differences in resistance and resilience across the globe from 2000 to 2020 using the MODIS NDVI dataset (figure 2).Lower resistance was found in part of southern South America, Australia, and central Asia (figure 2(a)).Widespread negative drought resilience values were observed in Asia, Europe, and North America (figure 2(b)).When comparing the means of resistance and resilience under different water or energy conditions, mean resistance under high water condition is significantly higher than that under low water condition, as shown by non-overlapping 95% CIs for the two groups (figure 3(a)).Mean resilience under low energy condition is significantly lower than that under high energy condition (figure 3(b)).In contrast, there is no significant difference in mean resistance under low and high energy or mean resilience under low and high water (figures 3(a) and (b)).We obtained similar results using the GIMMS NDVI dataset during 1982-2015 (figure S2).In addition, drought resistance significantly increased with the increase in the AI, whereas drought resilience remained relatively stable in response to the changes in aridity gradient (figure S4).
We also investigated the spatial variations of resistance and resilience under different water and energy conditions across various ecoregions.Results indicated that forests in the tropics, with higher water and energy conditions, showed greater resistance and resilience compared to other ecoregions (figures S5 and 4).Across all selected ecoregions, drought resistance showed a significant increase with increasing water availability (R 2 = 0.82), but no significant relationship was observed between drought resistance and energy (figures 4(b) and (a)).Drought resilience showed a significant increase with the increase in energy (R 2 = 0.58), but no significant correlation was observed between drought resilience and water availability (figures 4(c) and (d)).We further explored the relative importance of water and energy to resistance and resilience for each ecoregion using GBRT.Result indicated that, in most ecoregions, the relative importance of water to resistance was greater than that of energy, while the relative importance of energy to resilience was greater than that of water (figure S6).Overall, water and energy explained around 71.45% and 75.78% of the variance in drought resistance and  resilience, respectively (figure S7).Vegetation type explained around 27.83% and 21.85% of the variance of drought resistance and resilience, respectively (figure S7).
In addition, drought resistance generally showed a negative correlation with drought resilience across the globe (figure 5).This indicated a tradeoff between resistance and resilience.We observed that correlation coefficients between resistance and resilience were significantly lower under low energy condition than those under high water or energy conditions (figure 5).However, the difference in the correlation between resistance and resilience was larger between low-and high-energy conditions compared with between low-and high-water conditions (figure 5).
Using path analysis, we further examined and quantified the direct and indirect effects of water and energy on drought resistance and resilience.Both water and energy have positive direct effect on resistance (figure 6(a)).Both water and energy can directly influence resistance by triggering water stress in environment on vegetation growth.The indirect effect of water on drought resistance was also stronger than that of energy (figures 6(a) and (c)), resulting in a stronger total effect of water on drought resistance than that of energy (figure 6(c)).In contrast,

Discussion
In this study, we examined the effects of water and energy on drought resistance and resilience to understand the drivers of spatial heterogeneity in global vegetation resistance and resilience to drought stress.Our results indicated that the drought resistance of vegetation showed a significant increase with the increase in water availability.In contrast, drought resilience increased with the increase in energy rather than in water (figure 3(a)).These findings suggested that drought resistance and resilience were influenced differently by water and energy (figure 3(b)).Waterand energy-related variables co-regulated the spatial heterogeneity of drought resistance and resilience around the world.Drought resistance described vegetation persistence during the drought (Hoover et al 2014).An earlier study documented that increasing water availability, such as precipitation and soil moisture, effectively improved vegetation resistance to drought stress (Stuart-Haentjens et al 2018).This finding is supported by the observed increase in drought resistance as water increased in our study (figure 3(a)).However, energy has a positive direct effect of energy on resistance through its total effect is very small (figure 6(a)).Carbohydrate storage plays a key role in plant resistance to environmental stress, such as droughts, herbivory, and fires (Piper and Paula 2020).Under water stress, plants tend to close stomata to reduce water loss through transpiration in leaves, thereby decreasing CO 2 uptake and thus reducing photosynthetic carbon assimilation.As such, greater water availability can effectively increase drought resistance by maintaining or improving photosynthetic carbon storage.These are supported by the observed positive direct effects of aridity on resistance in the path analysis and the increased drought resistance with the increase in aridity (figures 6(a) and S4).Alternatively, greater water availability can increase root metabolism associated with the search for water and nutrients, and finally increase vegetation resistance to drought stress (Alvarez et al 2009, Fan et al 2017).In addition, vegetation in regions with more water availability tends to present high water-use efficiency and maintain better photosynthetic capacity (Baldocchi et al 2004, Wolf et al 2013a).Therefore, drought resistance showed a significant increase with the increased water availability.
In contrast to resistance, we observed that drought resilience was associated with the energy condition in the environment instead of the water condition.Drought resilience describes the ability of a vegetation ecosystem to recover to its original state after drought (Scheffer et al 2009).In general, vegetation productivity is reduced by drought stress after the drought.To recover and survive after drought events, plants need enough carbohydrates to support vegetation growth.Indeed, water is critical to plant growth.However, it is anticipated that following drought occurrences, as opposed to during droughts, the water restriction on plant growth will weaken.This is supported by the observed weaker relationship between aridity and resilience compared to that between aridity and resistance (figure 6(d)).With the absence of water limitation, greater energy inputs can effectively increase growth rate by accelerating photosynthetic carbon assimilation in plants.As such, drought resilience showed a significant increase with the increased energy.This is also supported by the positive direct effect of energy on drought resilience in the path analysis (figure 6(b)).
In addition, we found drought resistance and resilience also varied among vegetation types.For example, forests exhibited higher resistance but lower resistance than grasslands.Previous research reported that forests tend to display higher water-use efficiency than grasslands during drought episodes (Baldocchi et al 2004, Wolf et al 2013b).However, grasslands exhibit a notable capacity for rapid regrowth following drought (Ingrisch et al 2018, Stampfli et al 2018).Therefore, different vegetation types often demonstrate distinct physiological adaptations in functional traits that either enhance or diminish resistance and resilience (figure S6, Craine et al 2012, Anderegg and HilleRisLambers 2016).
Furthermore, water and energy condition have also a significant effect on the trade-off between resistance and resilience in vegetation, which showed a significant decrease with the increased water and energy (figure 5).Both resilience and maintaining high resistance can successfully reduce harm, but each characteristic has a different resource cost needed to build and maintain tissue structure, thus one generally comes at the price of the other (Miller et al 2017, Oliveira et al 2021).To reduce growth decline as much as feasible under drought stress, plants must retain a strong resistance to drought.However, to recover from drought as soon as possible, plants need to increase resilience rather than resistance.Due to the disparate effects of water and energy on resistance and resilience, it is anticipated that changes in either of these factors would disrupt the trade-off relationship between resistance and resilience.This may explain why the close relationship between resistance and resilience showed a significant decrease with the variations in water and energy.

Conclusions
Our findings revealed a strong positive correlation between drought resistance and water availability across all ecoregions, rather than energy.In contrast, drought resilience significantly increased with energy availability rather than water.Changes in local water and energy conditions not only affected drought resistance and resilience, but also tradeoff between resistance and resilience.Greater water availability in the environment might improve vegetation resistance through enhancing water efficiency and maintaining higher photosynthetic capacity.After drought events, energy input might increase growth rates and accelerated recovery, increasing vegetation resilience.Changes in water and energy availability also explained spatial heterogeneity in global vegetation resistance and resilience to drought.Our results provide evidence that water and energy co-regulate spatial heterogeneity in drought resistance and resilience globally.The robust linear relationship between drought resistance and resilience and available water and energy demonstrated in our study is crucial for accurately predicting and assessing the impact of climate change on the stability and productivity of terrestrial ecosystems under continued climate warming scenarios.
of Energy, Office of Science, Office of Biological and Environmental Research.Oak Ridge National Laboratory is supported by the Office of Science of the US Department of Energy under Contract No. DE-AC05-00OR22725.

Figure 1 .
Figure 1.Association of climate factors and vegetation with water and energy.Principal component analysis of six environmental factors (a).The correlation relationship among environmental factors, resistance and PC1 (energy) and PC2 (water) (b).(c) and (d) present the spatial distribution of energy (PC1) and water (PC2) during 2000-2020.PC1 and PC2 are the energy and water component by PCA analysis from six environment variables.

Figure 3 .
Figure 3. Mean resistance and resilience under different energy and water conditions during 2000-2020.RT indicates resistance, RS indicates resilience.The two outermost black lines in distribution plot represent the 95% confidence intervals.The difference between groups is considered statistically significant when their 95% CIs are not overlapped.

Figure 4 .
Figure 4. Relationships between the response of vegetation to drought and water and energy among ecoregions.Relationship among mean resistance (±SE, RT), resilience (±SE, RS), energy (PC1) and water (PC2) over ecoregions.PC1 and PC2 are the energy and water component by PCA analysis from six environment variables.The dashed line indicated that the linear regression model is statistically insignificant, and the solid line indicated that the linear regression model is statistically significant.The full names of ecoregion were listed in tableS1.

Figure 5 .
Figure5.Correlation coefficient between resistance and resilience under different low and high energy and water conditions.The two outermost black lines in distribution plot represent the 95% confidence intervals (95% CIs).The difference between groups is considered statistically significant when their 95% CIs are not overlapped.

Figure 6 .
Figure 6.Path analysis results on the direct and indirect effects of environmental factors on vegetation resistance and resilience to drought.(a), (b) present path models on environmental factors on resistance and resilience, respectively.Numbers mean the path coefficients value (PCv), and negative numbers indicate the negative effect.Gof means the goodness of fit in the path model.The direct and indirect effects of environmental factors on resistance ((c), RT) and resilience ((d), RS) of vegetation were shown.
, van Etten J, Cheng J, Mattiuzzi M, Sumner M, Greenberg J A, Lamigueiro O P, Bevan A, Racine E B and Shortridge A 2017 Raster: geographic data analysis and modeling R package version 2.6-7.2017Hoover D L, Knapp A K and Smith M D 2014 Resistance and resilience of a grassland ecosystem to climate extremes Ecology 95 2646-56 Huang K and Xia J 2019 High ecosystem stability of evergreen broadleaf forests under severe droughts Glob.Change Biol. 25 3494-503 Huxman T E, Snyder K A, Tissue D, Leffler A J, Ogle K, Pockman W T, Sandquist D R, Potts D L and Schwinning S 2004 Precipitation pulses and carbon fluxes in semiarid and arid ecosystems Oecologia 141 254-68 Ingrisch J, Karlowsky S, Anadon-Rosell A, Hasibeder R, König A, Augusti A, Gleixner G and Bahn M 2018 Land use alters the drought responses of productivity and CO2 fluxes in mountain grassland Ecosystems 21 689-703 Isbell F, Craven D, Connolly J, Loreau M, Schmid B, Beierkuhnlein C, Bezemer T M, Bonin C, Bruelheide H and De Luca E 2015 Biodiversity increases the resistance of ecosystem productivity to climate extremes Nature 526 574-7 Jackson D A 1993 Stopping rules in principal components analysis: a comparison of heuristical and statistical approaches Ecology 74 2204-14 Jorgensen S E and Fath B 2018 Encyclopedia of Ecology (Elsevier) Knapp A K and Smith M D 2001 Variation among biomes in temporal dynamics of aboveground primary production Science 291 481-4 Li X, Piao S, Wang K, Wang X, Wang T, Ciais P, Chen A, Lian X, Peng S and Peñuelas J 2020 Temporal trade-off between gymnosperm resistance and resilience increases forest sensitivity to extreme drought Nat.Ecol.Evol. 4 1075-83 Li X, Smith R, Choat B and Tissue D T 2019 Drought resistance of cotton (Gossypium hirsutum) is promoted by early stomatal closure and leaf shedding Funct.Plant Biol.47 91-98 Liang M, Cao R, Di K, Han D and Hu Z 2021 Vegetation resistance and resilience to a decade-long dry period in the temperate grasslands in China Ecol.Evol.11 10582-9 University of Arizona: Vegetation Index and Phenology Lab vol 35 pp 2-33 Ditmarová Ĺ, Kurjak D, Palmroth S, Kmet ′ J and Střelcová K 2010 Physiological responses of Norway spruce (Picea abies) seedlings to drought stress Tree Physiol.30 205-13 Dorigo W, Wagner W, Albergel C, Albrecht F, Balsamo G, Brocca L, Chung D, Ertl M, Forkel M and Gruber A 2017 ESA CCI soil moisture for improved Earth system understanding: state-of-the art and future directions Remote Sens. Environ.203 185-215 Fan Y, Miguez-Macho G, Jobbágy E G, Jackson R B and Otero-Casal C 2017 Hydrologic regulation of plant rooting depth Proc.Natl Acad.Sci.114 10572-7 Feeley K J, Bravo-Avila C, Fadrique B, Perez T M and Zuleta D 2020 Climate-driven changes in the composition of New World plant communities Nat.Clim.Change 10 965-70 Gampe D, Zscheischler J, Reichstein M, O'Sullivan M, Smith W K, Sitch S and Buermann W 2021 Increasing impact of warm droughts on northern ecosystem productivity over recent decades Nat.Clim.Change 11 772-9 Gareth J, Daniela W, Trevor H and Robert T 2013 An Introduction to Statistical Learning: With Applications in R (Springer) Groemping U and Matthias L 2018 Package 'relaimpo' Relative importance of regressors in linear models (R package version) Herberich M M, Schadle J E and Tielborger K 2023 Plant community productivity and soil water are not resistant to extreme experimental drought in temperate grasslands but in the understory of temperate forests Sci.Total Environ.