Declining coupling between vegetation and drought over the past three decades

1 INTRODUCTION

Droughts can restrict vegetation growth (Jiao, Wang, et al., 2021; Wang et al., 2021) and even cause widespread tree mortality, which undermines ecosystem stability (He et al., 2018) and has a negative impact on land carbon sinks (Reichstein, 2013; Schwalm et al., 2012, 2017). The association between vegetation and drought has long been recognized as a critical component of ecosystem vulnerability (Li et al., 2018). As global warming is expected to increase drought frequency and severity (Dai, 2013), the impact of different types of droughts on vegetation is unparalleled (Gampe et al., 2021; Jiao, Williams, et al., 2021; Xu et al., 2019; Zhang et al., 2021). Therefore, it is essential to understand the current vegetation–drought coupling and its evolutionary pattern to effectively predict future forest risk.

Drought is an indicator with multiscalar features (Vicente-Serrano et al., 2010), and the cumulative and time-lagged effects of drought on vegetation dynamics are widespread (Wen et al., 2019). Over the past decade, there has been an increase in studies examining vegetation sensitivity to droughts (He et al., 2018; Jiao, Wang, et al., 2021; Seddon et al., 2016). Present vegetation–drought coupling studies concentrate on two key topics, regardless of their definitions and spatial scales. In fact, plants can absorb water from past precipitation stored in deeper unsaturated soils or groundwater (Miguez-Macho & Fan, 2021). The response of vegetation to drought is often not due to water deficit during the current month but rather due to water scarcity that lasts months or even years. Therefore, vegetation–drought coupling assessments that do not consider timescales (or fixed drought timescales) will impede our understanding of how vegetation responds to drought to some extent. Vicente-Serrano et al.'s (2013) pioneering work found that vegetation responds to drought at different characteristic timescales across regions and biomes. The length of the drought timescale at which drought has the greatest impact on vegetation is considered a useful metric for quantifying vegetation sensitivity to droughts (Vicente-Serrano et al., 2013). The response time represented by drought timescales itself reflects the resistance and resilience of vegetation to drought (Vicente-Serrano et al., 2013). Longer drought timescales indicate stronger memory effects (Liu et al., 2018; Seddon et al., 2016), which may reduce the impact of more recent water deficits and reduce vegetation sensitivity to drought. In contrast, shorter drought timescales imply that vegetation reacts more quickly in the event of water scarcity below normal conditions, indicating a higher sensitivity to drought (Jiao, Wang, et al., 2021; Vicente-Serrano et al., 2013). There is generally a negative correlation between drought timescales and vegetation sensitivity. Thus, the response of vegetation to drought timescales can indirectly indicate vegetation sensitivity.

Indeed, it is more important to understand how vegetation responds to droughts over time and to uncover the underlying mechanisms behind the increasing or decreasing trend. For future predictions, it is critical to determine whether vegetation in specific regions will become more resistant or more sensitive to droughts. Although many researchers have become interested in determining how vegetation responds to drought over time, most studies show an increasing trend in vegetation sensitivity to droughts (Jiao, Wang, et al., 2021; Li et al., 2022; Wei et al., 2023; Zhang et al., 2022). However, the question of whether vegetation becomes more sensitive to drought remains debated (He et al., 2018; Zeng et al., 2022). Furthermore, all these studies focusing on time evolution trends are conducted on specific or fixed drought timescales (Jiao, Wang, et al., 2021; Li et al., 2022; Wei et al., 2023; Zhang et al., 2022), ignoring the cumulative and time-lagged effects of droughts on vegetation dynamics (Wen et al., 2019). Currently, vegetation–drought relationship studies either consider drought timescales but ignore temporal evolution trends or consider temporal evolution trends but ignore drought timescales. It remains unclear whether vegetation is becoming more sensitive to drought when considering optimal response timescale over the past three decades. Therefore, considering the drought timescales with temporal evolutionary trends would provide a more comprehensive and robust understanding of the vegetation–drought coupling relationship.

Here, we aim investigated the global vegetation–drought relationship and determined whether there is an overall increasing coupling relationship after considering the optimal response timescale. Specifically, we seek to answer two key questions: (1) How has vegetation–drought coupling changed over time after considering the optimal response timescale? (2) What are the factors that determine the increasing or decreasing trend of TS_d and trend of R_d? To achieve these objectives, we calculated the monthly Pearson correlation between the Normalized Difference Vegetation Index (NDVI) and the Standardized Precipitation Evaporation Index (SPEI) across 1–24 months of timescales, to detect vegetation–drought coupling at different timescales. We estimated the optimal drought timescales of vegetation response to drought by selecting timescales that show the maximum correlation coefficient between NDVI and SPEI (Vicente-Serrano et al., 2013). The corresponding correlation coefficient is used as a measure of vegetation–drought coupling (He et al., 2018; Jiao, Wang, et al., 2021). We perform our analysis on 18 moving windows, each with a 17-year time series, to detect changes in this coupling relationship over time. In addition, we use random forest methods to examine 30 possible drivers to explain the observed trends.

2 MATERIALS AND METHODS

3 RESULTS

3.1 Global response of vegetation to drought along time

The timescales at which droughts (SPEI) affect vegetation (GIMMS-NDVI) were performed at each of the 18 moving windows. The coupling between vegetation and drought, as reflected by optimal drought timescales (Figure S1) and maximum Pearson correlation coefficients (Figure S2), changes markedly over time. The mean optimal drought timescales at which droughts affect vegetation (TS_d) in the 18 windows exhibit high spatial variability (Figure S3). Vegetation tends to respond to longer TS_d in high latitudes of the Northern Hemisphere but shorter TS_d in tropical, subtropical areas, and most parts of the Southern Hemisphere (Figure S3a). The mean correlation coefficients between GIMMS-NDVI and SPEI (R_d) in the 18 windows (Figure S3c) exhibit an opposite pattern compared to that of TS_d. Although there is no obvious regularity along a precipitation or temperature gradient for the trend of TS_d (Figure S3b), the high values of R_d are generally distributed in wet and warm areas (Figure S3d) with latitudes below 50° N.

Looking at trends over the past three decades, TS_d shows an overall increasing trend over time, covering 54% of global vegetated areas (Figure 1a). However, there is no obvious regularity along a precipitation or temperature gradient for the trend of TS_d (Figure 1b). An increasing trend of TS_d can be found in almost all geographic zones, but a decreasing trend of TS_d is prominent in wetter zones, such as the northern parts of Eurasian continent, most parts of Brazil, the southern regions of South America, the central and western parts of USA (Figure 1a). Unlike the overall increasing trend of TS_d, the trend of R_d shows an overall declining trend (Figure 1c), accounting for 53% of global vegetated areas. The decreasing trend of R_d tends to be in warm and wet regions (Figure 1d), especially in tropical rainforest regions and most parts of Australia (Figure 1c). Noted that the R_d in this study took the optimal response timescale into consideration. When exploring the relationship between vegetation and droughts using fixed drought timescales, the distributions of R_d slope show local spatial differences among different drought timescales, although the overall spatial layout is roughly similar (Figure S4). The overall vegetation–drought coupling shows an increasing trend in 1-, 12-, 18-, and 24-month timescales of droughts but a decreasing trend in 3- and 6-month timescales of droughts. Thus, our prevailing decrease in vegetation–drought coupling reflected by R_d may be due to the consideration of optimal drought timescales.

Moreover, using GPP_NIRv as proxy for vegetation dynamics, we find the overall increasing trend of TS_d and decreasing trend of R_d (Figure S5). We also repeated the analysis by changing the moving windows to 15 years (Figure S6) and 19 years (Figure S7). We still find the same increasing trend of TS_d and decreasing trend of R_d. The selection of moving windows does not change our main results. The increasing trend of TS_d and decreasing trend of R_d together suggest a weakening relationship between drought and vegetation over time.

3.2 Evolution of drought influence on vegetation between water-deficit and water-surplus regions

There is an obvious regularity along a dryness gradient for both TS_d and R_d (Figure 2a). The mean TS_d tends to show the highest increasing trend in arid regions (Figure 2a, blue line), and this positive trend decreases for regions that are more humid, as indicated by a higher AI. However, the long-term trend of R_d is found to be negative (showing a decreasing trend) in arid regions (Figure 2a, red line) but positive in regions that are more humid. Thus, increasing vegetation–drought coupling tends to be found in humid regions. Counterintuitively, vegetation–drought coupling shows a declining trend in arid regions, although its initial mean correlation is relatively high (Figure S8).

When seeking the trend of TS_d and R_d within the phase space of precipitation trend and static map of AI, we find that vegetation in arid environments tends to increase TS_d, regardless of whether precipitation increases or not (Figure 2b). Vegetation in relative wet environments is prone to increase TS_d when precipitation declines but reduce TS_d when precipitation increases (Figure 2b). This is because plants in arid environment have strong adaptability to droughts and can endure long-term droughts. When local precipitation declines, plants can absorb past precipitation stored in deeper unsaturated soils, rocks, and groundwater (Miguez-Macho & Fan, 2021). While precipitation increases, plants might change their original water use striges in long-term water-deficit conditions and uptake more water from local places, resulting in a shorter response time to drought. Interestingly, vegetation in each bin of arid conditions consistently shows an increasing trend of R_d when precipitation decreases or decreasing trend of R_d when precipitation increases (Figure 2c). This implies that vegetation under similar arid conditions would have the same mechanism for responding to precipitation changes, reflecting the long-term adaptation of vegetation to its arid condition.

We further investigated the trend of TS_d and R_d in water-deficit regions and water-surplus regions. We found a consistent pattern with earlier studies (Jiao, Wang, et al., 2021; Piao et al., 2014) that vegetation water-deficit regions are found in the middle and low latitudes in the Northern Hemisphere and the Southern Hemisphere (Figure S9), while vegetation water-surplus regions are mainly found in high latitudes (above the 50th degree of Northern Latitude) in the Northern Hemisphere and rainforests in the tropics.

Although increasing and decreasing trends of TS_d coexist in water-deficit and water-surplus regions at the pixel scale (Figure 3a), both water-deficit and water-surplus regions show an overall increasing trend of TS_d (Figure 3c,d). However, the trend of R_d is prone to increase in water-surplus regions (Green in Figure 3b,d) but decrease in water-deficit regions (Red in Figure 3b,c). The increasing trend of R_d in water-surplus regions is mainly in the north and eastern parts of Russia, North Europe, and high latitudes of North America (Green in Figure 3b). The decreasing trend of R_d in water-deficit regions (Red in Figure 3b) is likely to occur in most regions of Africa. This result suggests that plants in arid environments have a tendency to adapt to droughts and reduce their drought sensitivity.

3.3 The driving mechanism of drought timescales and vegetation–drought correlation changes

To investigate the underlying factors influencing the TS_d slope and R_d slope, we tested the correlation between TS_d slope (or R_d slope) and 30 climatic, environmental, and plant physiological variables that may correlate with vegetation growth. We find that most drivers have a strong correlation with TS_d slope (Figure S10) and R_d slope (Figure S11). These 30 climatic, environmental, and plant physiological variables show high collinearity. We set a threshold of 0.75 and obtained the 10 variables with low correlations (Figure S12).

We find that the CO₂ trend is the most predictive driving factor for both TS_d slope and R_d slope (Figure 4) from the RF models. However, the trend of TS_d and the trend of R_d are determined by different climatic, environmental, and plant physiological variables. TS_d slope is sensitive to external environmental and climatic factors (Figure 4a). The top three predictive driving factors for the prediction of slope of TS_d only include environmental and climatic variables: CO₂ trend, temperature trend, and precipitation trend (Figure 4a). While the R_d slope tends to be mainly determined by the environmental condition and inherent physiological attributes (Figure 4b). The CO₂ trend, the foliar nitrogen concentration per unit dry mass (Nm), and WTD below land surface are the three most predictive driving factors (Figure 4b).

We further investigated the partial dependence plot of the 10 variables on trend of TS_d and trend of R_d (Figure 4). Higher CO₂ and Nm tend to increase both the trend of TS_d and the trend of R_d. Higher CO₂ and Nm can enhance vegetation growth and need to absorb more water. On the one hand, the enhanced adaptability of plants to water shortages increases root water uptake from deeper soil, even groundwater, resulting in longer TS_d. On the other hand, increasing vegetation water demand promotes coupling between vegetation and drought (R_d).

The dominant drivers are different between water-surplus regions (Figure S13) and water-deficit regions (Figure S14). For predicting the trend of TS_d, the three most important factors are CO₂ trend, mean temperature (TEM) slope, and RAD slope in water-surplus regions, but RAD slope, WTD, and TEM slope in water-deficit regions. Among the same drivers, TEM slope and Rad slope show opposite dependency between water-surplus regions and water-deficit regions (Figures S13a and S14a). When predicting the trend of R_d, the three most important factors are Nm, CO₂ trend, and WTD in water-surplus regions, but CO₂ trend, Nm, and precipitation (PRE) slope in water-deficit regions (Figures S13b and S14b). The different relative importance and dependencies between water-surplus regions and water-deficit regions indicate distinctive response mechanisms of vegetation to drought.

3.4 Prevailing decoupling between drought timescales and correlation coefficients

Both drought TS_d and R_d can reflect the coupling between vegetation and drought. As earlier studies suggested, the TS_d generally shows a negative relationship with vegetation–drought sensitivity (Vicente-Serrano et al., 2013), while R_d has a positive connection with vegetation–drought sensitivity (He et al., 2018). Thus, there should be a negative coupling relationship between TS_d and R_d. In space, we tested this relationship between TS_d and R_d with 5 × 5 moving windows at each 18 successive 17-year bins (Figure S15). We find that the coupling relationship between TS_d and R_d changes over time, with the proportion of coupling areas (negative correlation) decreasing from 53% in the first 17 years (1982–1998) to 50% in the last 17 years (1999–2015) (Figure S15). We further tested this relationship between TS_d and R_d both in temporal dimensions (Figure 5a) and the changing trend of spatial correlation (Figure 5c). The strong coupling between TS_d and R_d at the pixel only accounts for 45% of global vegetated areas, suggesting a relative weak coupling relationship (Figure 5a). In addition, the relationship between TS_d and R_d based on moving 17-year windows shows overall increasing trends (Figure 5c,e). The negative relationship between TS_d and R_d in space can be found in earlier stages of the past three decades, but this negative relationship tends to approach zero (Figure 5e) and even turns positive over time.

When inspecting the evolution of the coupling relationship in climate phase space, there is no obvious regularity along a precipitation or temperature gradient for the temporal correlation between TS_d and R_d (Figure 5b). However, the changing trend of spatial correlation tends to be strengthened in cold regions (temperatures below 10°C). The prevailing negative relationship of temporal correlation and increasing trends of spatial correlation together suggest a weakening coupling relationship between TS_d and R_d. We can conclude that the strong negative relationship between TS_d and R_d coupling, previously reported only spatially, will not occur in the temporal dimension.

4 DISCUSSION

We extended previous studies by investigating how vegetation–drought coupling changed over time over the past three decades (1982–2015) while considering optimal response timescale. We found an overall increasing trend in the timescales of vegetation responding to drought (Figure 1). This overall increasing trend of TS_d suggests that vegetation tends to be affected by long-term droughts. This finding indirectly indicates a weakening vegetation–drought coupling trend over time, which indicates that plants have a tendency to adapt to droughts and might adjust their water use strategies by using past precipitation stored in deeper unsaturated soils, rocks, and groundwater to overcome seasonal and irregular droughts (Miguez-Macho & Fan, 2021). With the optimal response timescale considered, we find that the increasing vegetation–drought coupling is mainly observed in water-surplus regions, while vegetation in water-deficit regions is prone to show a decreasing trend of vegetation–drought coupling. These findings challenge earlier conclusions that vegetation sensitivity to droughts has increased over the past decades (Jiao, Wang, et al., 2021; Li et al., 2022; Wei et al., 2023; Zhang et al., 2022). However, these studies either ignored the variable timescales of droughts or focused solely on the plant response to precipitation and soil moisture rather than drought. The varying definitions, data sources, methods, study periods, and regions might lead to different trends in the coupling of vegetation and droughts (Table S2). The biggest reason for the declining trend may be the consideration of drought timescales, which is consistent with He et al.'s (2018) conclusion of weakening sensitivity of vegetation to droughts. Although they took timescales into account, we question He et al.'s (2018) conclusion for two reasons. On the one hand, they solely investigated long-term droughts (12–24 months) but ignored short- and mid-term droughts (less than 12 months). On the other hand, they directly equated correlation coefficients with vegetation sensitivity to droughts. Our comparison between TS_d and R_d indicated that correlation coefficients do not represent vegetation sensitivity over time (Figure 5). Thus, we can only conclude a declining coupling between vegetation and drought, rather than a decreasing sensitivity.

Plants in water-deficit zones tend to reduce their sensitivity to drought over time. Plants in arid environments have three important strategies for actively maintaining physiological water balance: (1) increasing root water uptake from soil; (2) reducing transpiration by closing stomata; and (3) adjusting osmotic processes within tissues (Gupta et al., 2020). We find that rising CO₂ is the most important driver affecting the change in vegetation–drought coupling (Figure 4). On the one hand, elevated CO₂ leads to a reduction in stomatal conductance, an increase in water-use efficiency (Keenan et al., 2013), and a decrease in the amount of water needed by plants (Farrior Caroline et al., 2015). On the other hand, plants tend to acclimate to water stress to some extent over time (Vicente-Serrano et al., 2013). Plants in arid environments could maximize their hydrological benefits through proper hydraulic traits (Anderegg et al., 2018), root zone storage capacity, and tree cover (Singh et al., 2020). For instance, a meta-analysis of the global distribution of xylem vulnerability showed that trees growing in arid conditions around the world are better at withstanding xylem embolism (Choat et al., 2012). Studies have also shown that in arid conditions, plants have relatively deep rooting systems allowing them to absorb groundwater from deeper soil or rocks (McDowell, 2008; Miguez-Macho & Fan, 2021). Felton et al. (2021) found that the temporal sensitivity of vegetation in drylands is better predicted by biotic factors related to vegetation structure than by mean annual precipitation, which further verifies plant adaptation and acclimation after frequent and severe droughts. Plants will strengthen the growth of their root systems, and their main roots will extend downstream to deeper depths in search of water sources, which can be verified by the increasing response time to drought.

Although wet regions are not normally considered to be at drought risk, vegetation in wet environments is showing increasing vegetation–drought coupling. CO₂ fertilization is the main driver of global greening (Piao et al., 2020). Rising CO₂-induced leaf area increase, in turn, leads to an increase in water demand and water loss (Duursma et al., 2019), offsetting the CO₂ water saving effect due to reduced stomatal aperture (Fatichi et al., 2016). Another possible explanation may be that plants in humid regions lack adaptation mechanisms to drought but may encounter severe droughts due to rising VPD and temperature (Yuan et al., 2019). Indeed, drought-induced declines in forest growth are occurring not only in arid regions but also in wet forests (Meir & Woodward, 2010; Phillips, 2009). Choat et al. (2012) also showed a global convergence in the vulnerability of forests to drought, with all forest biomes equally vulnerable to hydraulic failure regardless of their current rainfall environment. Attention should also be paid to wet regions, which are not normally considered at drought risk due to abundant water availability.

Besides, the prevailing decoupling between TS_d and R_d suggests that the response time should not be replaced by the correlation-derived sensitivity indicator. As earlier studies have shown, shorter timescales mean more rapid vegetation reactions as soon as water deficits fall below normal conditions, representing a higher sensitivity of vegetation to drought (Vicente-Serrano et al., 2013). The sensitivity of vegetation to drought can also be measured by the vegetation–drought coupling (correlation coefficients; De Keersmaecker, 2015; Seddon et al., 2016), with a higher correlation representing a higher sensitivity. If this assumption is true, the TS_d and R_d should exhibit a negative coupling relationship. However, our results confirm that the negative coupling relationship becomes positive over time. We find that the change in TS_d along time is mainly determined by environmental and climatic drivers, while the change in R_d is more controlled by environmental and plant physiological traits. This suggests that optimal drought timescales can reflect the information that vegetation–drought correlation fails to provide. Furthermore, the correlation-derived indicator would fail to represent vegetation sensitivity. For instance, when the response of vegetation to drought is non-linear and near the key arid threshold (turning points), a minor increase in drought intensity may lead to a huge loss of vegetation production (Berdugo et al., 2020). In this case, the actual high vegetation–drought sensitivity might be marked by a low correlation from a statistical perspective. It should be noted that the reduction in R_d in water-deficit regions can be attributed to the cumulative effects of chronic water scarcity and the high sensitivity of plants to droughts. This is because trees are likely to exhibit a minimal response to water stress in subsequent months or years after they have been affected by water deficit-induced die-off. Considering the possibility that R_d is low but the vegetation–drought sensitivity is high, other indicators representing vegetation–drought coupling are needed. Although our optimal drought timescales can only indirectly represent vegetation's sensitivity to droughts, the response time we propose may serve as an effective supplement to current vegetation–drought coupling assessments.

Our work focused on dynamic changes in vegetation response to drought rather than static spatial patterns. However, many predictive drivers are static, which may not reflect the influence of changing drivers on changing vegetation–drought coupling. Although the importance of predicting factors was briefly discussed, the specific mechanism has not been verified. The differential drought timescale is also linked to drought-induced changes in the hydrological cycle (Geruo, 2017). Thus, we need to further understand how vegetation under drought stress responds to changes in hydrological components at different soil depths and with varying temporal signatures. Nevertheless, our results reveal the contradictory trend in vegetation sensitivity to droughts between water-surplus regions and water-deficit regions and stress the importance of paying more attention to wet regions. The next challenge is to reveal specific plant water use strategies (Brodribb et al., 2020) and understand the deep soil water uptake process (Werner et al., 2021) as response time increases over time. It is also critical to apply other robust methods to quantify vegetation's sensitivity to droughts, such as regression slope methods (Rao et al., 2022) and tipping point theory (Au et al., 2023).

5 CONCLUSIONS

Our optimal drought timescales quantified the length of time over which past drought conditions significantly affect current vegetation growth. We find a remarkable increase in the timescale of vegetation response to droughts globally over the last three decades from 1982 to 2015, which suggests that plants tend to adapt to droughts and might adjust their water use strategies. With the optimal drought timescales considered, we find that the increasing vegetation–drought coupling is mainly observed in water-surplus regions, while the vegetation in water-deficit regions is prone to show a declining coupling relationship, although its initial correlation is relatively high. These two opposite trends indicate that more attention should be paid to wet regions. Furthermore, the prevailing decoupling between TS_d and R_d over time suggests the importance of measuring vegetation–drought relationships in different ways. Overall, temporal linkages between vegetation–drought improve our understanding of historically observed vegetation–drought coupling and can contribute to our understanding of upcoming risk in the future.

AUTHOR CONTRIBUTIONS

Delong Li: Conceptualization; data curation; formal analysis; funding acquisition; investigation; methodology; software; visualization; writing – original draft; writing – review and editing. Li An: Data curation; software; writing – review and editing. Shuai Zhong: Formal analysis; software; writing – review and editing. Lei Shen: Conceptualization; funding acquisition; investigation; supervision; writing – review and editing. Shuyao Wu: Data curation; investigation; writing – review and editing.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflict of interest.

DATA AVAILABILITY STATEMENT

The input data, output data, and main codes underpinning this paper are openly available from Zenodo at https://doi.org/10.5281/zenodo.10428840. The data we used are all publicly available. The Global GIMMS NDVI3g v1 dataset (1981–2015, 8 km) are available from the National Tibetan Plateau Data Center at https://data.tpdc.ac.cn/zh-hans/data/9775f2b4-7370-4e5e-a537-3482c9a83d88. The high-resolution (10 km) global surface solar radiation are also available from the National Tibetan Plateau Data Center at https://data.tpdc.ac.cn/en/data/be562de3-6367-402f-956d-59f7c21ad294/. The gridded SPEI data (SPEIbase v2.6) are available from https://spei.csic.es/spei_database. The Climatic Research Unit (CRU), version TS 4.03, is available from https://crudata.uea.ac.uk/cru/data/hrg/. The 0.05° GPP_NIRv from 1982 to 2018 are available from https://doi.org/10.1016/j.scitotenv.2020.142569. GLEAM soil moisture reanalysis is from https://www.gleam.eu/. The Gridded 1° atmospheric carbon dioxide concentrations from 1850 to 2013 can be accessed via the Zenodo at https://doi.org/10.5281/zenodo.5021361. Global Human Modification map is obtained from http://s3.amazonaws.com/DevByDesign-Web/Apps/gHM/index.html. The global above biomass carbon density in the year 2010 is available on Figshare at https://doi.org/10.6084/m9.figshare.c.4561940. The Arid Index is obtained from https://cgiarcsi.community/data/global-aridity-and-pet-database/. Yearly TC cover (1982–2016) is obtained from https://lpdaac.usgs.gov/products/vcf5kyrv001/. The 1 km canopy height is available from https://webmap.ornl.gov/ogc/dataset.jsp?dg_id=10023. The maximum rooting depth is from https://wci.earth2observe.eu/thredds/catalog/usc/root-depth/catalog.html. The tree density is from https://elischolar.library.yale.edu/yale_fes_data/1/. The proportion of water uptake of plant from current month is available from http://thredds-gfnl.usc.es/thredds/catalog/DATA_TRANSPSOURCES/catalog.html. The SLA, Nm, and Pm are available from https://github.com/abhirupdatta/global_maps_of_plant_traits. Water Table Depth is available from https://aquaknow.jrc.ec.europa.eu/content/global-patterns-groundwater-table-depth-wtd. Elevation is available from http://viewfinderpanoramas.org/dem3.html. Saturated hydraulic conductivity and Soil porosity are available from https://doi.org/10.1594/PANGAEA.870605. Root zone storage capacity is accessible on GitHub at https://github.com/chandrakant6492/Drought-coping-strategy. The Wood Density is obtained from the Global Wood Density Database on Dryad https://doi.org/10.5061/dryad.234/1. The Isohydricity is derived from GitHub at https://github.com/agkonings/isohydricity.