The role of high-biodiversity regions in preserving Nature’s Contributions to People

Increasing human pressures are driving a global loss of biodiversity and of Nature’s Contributions to People (NCP)—the contributions of living nature to people’s quality of life. Understanding the spatial relationship between biodiversity and NCP is essential for securing Earth’s life support systems. Here we estimate the importance of high-biodiversity regions in maintaining the provision of three NCP under four scenarios of climate change. We focus on critical regulatory NCP which are currently facing decline: regulation of air quality, climate and freshwater quantity. We estimate the current and future value of NCP using a suite of environmental indicators and evaluate whether risk from environmental change is higher or lower within high-biodiversity regions compared with control regions. We find higher levels of NCP within high-biodiversity regions both in the present and the future for all indicators, which highlights the spatial congruence between biodiversity and NCP. Moreover, air quality and climate regulation indicators show rapidly increasing levels within high-biodiversity regions, especially under higher-emission scenarios. Our results point to a substantial contribution of high-biodiversity areas to the provision of NCP. Protecting areas of high biodiversity value will synergistically contribute to the preservation of many of nature’s contributions humanity depends on. Understanding the role of biodiversity in maintaining the provision of Nature’s Contributions to People is critical to sustainability. This study finds a substantial contribution of high-biodiversity areas to the regulation of air quality, climate and freshwater quantity, with important implications for conservation efforts.


Article
https://doi.org/10.1038/s41893-023-01179-5quality (NCP3), climate (NCP4) and freshwater quantity (NCP6).We chose seven different indicators to represent global trends in each potential NCP (Supplementary Table 1) and used one representative indicator for each NCP to be reported in the main text, highlighting any discrepancy with other indicators (reported in Supplementary Information) when present.Specifically, we used the leaf area index (LAI) to represent the regulation of air quality (NCP3), as the density and structural complexity of leaves determine air pollutants interception 27 ; the density of total carbon biomass in vegetation (cVeg) as an indicator of climate regulation (NCP4); and water availability (WA) as an indicator of freshwater quantity (NCP6).

Global trend of NCP proxies
Overall, we found an increasing global trend of both air quality and climate regulation (NCP 3 and 4, respectively) between the present and the future, while water quantity regulation (NCP6) declines slightly.We found that the global mean LAI increases from a present value of 0.85 (s.d.= 0.4, among different climate models) to a future value of 0.95 (s.d.= 0.44) m 2 m −2 or 0.99 (s.d.= 0.46) m 2 m −2 , depending on the scenario (lowest under SSP1-2.6,highest under SSP5-8.5).The global mean cVeg increases from 1.85 (s.d.= 0.91) kg m −2 to 2.16 (s.d.= 0.96) or 2.17 (s.d.= 0.94) kg m −2 under scenarios SSP1-2.6 and SSP5-8.5, respectively.The mean net primary productivity (NPP), an alternative climate regulation indicator, shows similar trends in future scenarios.In contrast, mean global WA is projected to decline from the current −5.6 mm per month (s.d.= 6.9) to −6.04 (s.d.= 7.2) mm per month under scenarios SSP1-2.6, and even more (−6.14mm per month, s.d.= 7.1) under scenario SSP5-8.5.This is due to actual evapotranspiration (AET) surpassing precipitation; in fact, AET and potential evapotranspiration (PET) increase under each scenario.Accordingly, the aridity index (AI), an alternative indicator of freshwater regulation, reports a similar declining global trend.
These average global trends conceal high spatial variability in distribution of many of these NCP.While LAI (NCP3) increases in many regions under all scenarios, extreme declines are found in southeast Brazil, and (for SSP5-8.5) in the Gulf of Guinea and part of Tanzania (Fig. 1a and Supplementary Fig. 2a).Similarly, cVeg (NCP4) shows positive trends in almost every region, except for Brazil and Central Africa, especially under the most pessimistic scenario SSP5-8.5 (Fig. 1b and Supplementary Fig. 2b).In contrast, NPP increases in every region under each scenario (Supplementary Fig. 3a,b).WA (NCP6) shows even higher spatial heterogeneity, with large declines especially in the Amazon (an area of high biodiversity value) and large increases in much of Central and Southern Africa, and especially parts of China and Southeast Asia (Fig. 1c and Supplementary Fig. 2c).Other indicators of water regulation show similarly heterogeneous trends, except for PET which increases in every region excluding Central Africa and South India under scenario SSP5-8.5 (see Supplementary Fig. 3).

Effect of high-biodiversity regions on NCP trends
To verify whether NCP change was significantly different inside versus outside of high-biodiversity regions, we first downscaled the NCP indicators with bilinear interpolation to 10 km resolutions and then run a propensity score matching analysis to identify control areas with environmental characteristics similar to the biodiversity regions, within the same country.After the matching, the covariate imbalance for countries comprising the top 10% of high-biodiversity regions was reduced successfully from a mean C-statistic value of 0.98 (0.95 if considering the top 30% priority regions) before the matching to 0.76 (0.79 if considering the top 30% priority regions) after the matching procedure.This reduction in imbalance was also individually achieved in 43 out of 60 countries when looking at the top 10% regions (58 out of 85 countries if looking at the top 30%).The C-statistic has the advantage of automatically evaluating the balance regulation, carbon sequestration, agricultural pest control and human health regulation [12][13][14] .Through the modification of such regulating services, biodiversity indirectly supports and affects the long-term viability of material contributions such as the production of food, fibres, shelter and medicines [15][16][17] .Therefore, the decline of current regulating NCP might forshadow a decline in material NCP.Here we focus on the regulating NCP that have shown a decrease in the past 50 yr and might determine a decline in several other NCP (including important material and non-material NCP for which we lack a long time-series of data) that may depend on them directly or indirectly 9 .
While global biodiversity conservation can help preserve many NCP 3 , rapid rates of climate change risk jeopardizing such interventions 6,18 .In fact, the effect of climate change in determining future biodiversity loss might surpass that of land-use change 11,19,20 .At the same time, climate change might reduce the availability of NCP, especially in areas such as South Asia and Africa, under the more aggressive development scenarios 21 .But can biodiversity conservation help maintain NCP under climate change?
Measuring the current and future role of important biodiversity regions in preserving the provisioning of regulating NCP provides insight into the potential for environmental conservation to deliver both biodiversity and NCP benefits under climate change.Recent studies have used an integrated approach to spatial conservation planning to include both biodiversity and NCP objectives, although these are limited to only few NCP indicators or focused on specific regions 3,22,23 .Crucially, none of these studies have directly investigated whether regions of high biodiversity value can play a global role in preserving NCP, and whether such role will be altered by global change.In addition, past efforts to represent NCP globally have used models that are driven by categorical changes in land use/land cover and are not sensitive to changes in biodiversity (although biodiversity may of course respond to many of the same land-use and climate drivers as the NCP).In contrast, the NCP represented here are based on Earth Systems models that capture more of the heterogeneity within different ecosystems that can determine both biodiversity and NCP.Reducing risks to NCP provision is fundamental to achieving global conservation targets, and ensures continuing delivery of ecosystem services 3,24 .
Here we measure the importance of high-biodiversity regions for the maintenance of different regulating NCP at a global scale and evaluate whether such contributions will be altered under future scenarios of climate change.
We define 'high-biodiversity regions' as those emerging from a comprehensive analysis of 63 published global maps 25 , which were combined on the basis of their underlying methodology and input data.Our goal is to estimate the avoided risk to humanity, in terms of NCP loss, that might result from conserving important biodiversity regions under global change.The IPBES Conceptual Framework makes a distinction between 'potential' and 'realized' NCP, where potential NCP are defined as the possibility for an ecosystem to provide an NCP independently of its demand, while the realized contribution is the actual NCP experienced or delivered, which is determined by the combination of the potential NCP and the human demand for the resulting benefits 10,11 .Since trends in potential and realized contribution usually follow the same path for regulating NCP 11 , and since future changes in population, consumption and technology (determining realized NCP) are highly uncertain, we limit our analysis to potential NCP estimates.
We analysed different scenarios of global environmental change from the sixth Coupled Model Intercomparison Project (CMIP6) on the basis of the five socio-economic narratives (Shared Socio-economic Pathways (SSPs)) 26 that describe the possible combination of challenges for societal mitigation and adaptation to climate change.We considered four climate scenarios, from a low-emission scenario compatible with the Paris' Goal of 2 °C (scenario SSP1-2.6) to a high-emission scenario that will largely fail that goal (SSP5-8.5).We evaluated the risk to the maintenance of three regulating NCP: the regulation of air Article https://doi.org/10.1038/s41893-023-01179-5 on all covariates simultaneously, but since LAI, cVeg and WA have a strong relationship with forest cover, we also checked for the balance of this particular variable by measuring the standardized mean difference (SMD), and the mean and maximum empirical cumulative density functions (eCDFs).All the metrics showed a high reduction in forest cover imbalance after the matching (from 1.1 to 0.08 for SMD, and from 0.32 to 0.06 and from 0.53 to 0.21 for the mean and maximum eCDFs, respectively), demonstrating the ability of the matching procedure to discount the confounding effect of forest cover from our results (Supplementary Tables 5 and 6).
Most of the increases in air quality regulation, as represented by LAI, occurs within high-biodiversity regions (Supplementary Fig. 1 and Fig. 2).Similarly, the increases in climate regulation, as represented by carbon biomass in vegetation, under each scenario are found within high-biodiversity regions.The mean NPP, also an indicator of climate regulation, shows a similar trend (see Supplementary Figs. 5 and 6).
Declines in mean WA are predicted in all future scenarios and are stronger within high-biodiversity regions compared with control areas mainly due to the high risk of water loss in the Amazon, which represents the largest high-biodiversity-value region.Nonetheless, absolute WA remains higher within biodiversity regions than outside under each scenario (Fig. 2 and Supplementary Fig. 4).Correspondingly, the mean AET and PET, also indicators of freshwater quantity, are always higher inside the high-biodiversity regions than outside under each scenario.However, while AET increases more in the future within the biodiversity regions, PET increases are larger in control regions (see Supplementary Figs. 5 and 6).We found that high-biodiversity regions are less arid than control regions, both in the present and the future; however, aridity will decrease relatively more within the former (Supplementary Figs. 5 and 6).
The trend for the 10 countries with the largest extent of highbiodiversity regions generally follows the global trend across the different scenarios, with greater mean changes for LAI and cVeg within the high-biodiversity regions compared with the control regions and the reverse being expected for WA (Fig. 2 and Supplementary Figs. 7 and 8).However, there are some important differences in certain countries (Supplementary Figs.7-11).The greatest differences between priority and control regions are in Colombia, Indonesia and Brazil.Australia shows a smaller difference between biodiversity and control regions and Cameroon shows a strong decrease in the median value of cVeg in particular under high-emission scenarios SSP3-7.0 and SPP5-8.5, with a slight increase in WA under the same scenarios.These differences in NCP indicators between priority regions and control regions are always statistically significant at the country scale, with very few exceptions (Supplementary Table 7).Our results are qualitatively confirmed when looking at the top 30% high-biodiversity regions (rather than the top 10%), even if in this case, differences between high-biodiversity versus control regions are less stark.This larger selection of high biodiversity areas includes several regions outside the tropics, such as high-biodiversity regions in North America, central Australia, parts of the Mediterranean, and parts of Siberia and China (Supplementary Figs. 1, 11 and 12).10.

Article
https://doi.org/10.1038/s41893-023-01179-5 While the spatial congruence between high biodiversity value and high NCP values can be altered by land-use change, we found that climate was always a stronger correlate of NCP change than land-use change (Supplementary Tables 8 and 9), once again confirming the rising importance of climate in regulating these relationships.

Discussion
Biodiversity loss could considerably undermine the provision of regulating NCP 12,15,28 and coordinated global responses that simultaneously mitigate global change impacts on biodiversity and NCP loss are increasingly important 3,23,28,29 .While our work did not aim to establish causality between biodiversity value and NCP value, we did find high spatial congruence between high biodiversity value and high NCP provision both in the present and in the future.
According to CMIP6 estimates, levels of air quality and climate regulation will increase in the future, and such increase will be higher within biodiversity regions (especially under higher-emission scenarios).WA will conversely decrease in the future, especially within high-value biodiversity regions.Areas of high biodiversity value in South and central America will especially face decreasing values of mean WA (indicator for NCP6), except for the southeast coast of Brazil, while areas of central Africa, Indonesia and Malaysia will be characterized by increasing WA values.High-biodiversity regions play an increasingly important role in ensuring high LAI under increasing levels of climate change (as the difference in LAI is higher under scenarios of higher emission), meaning that the conservation of these areas will be even more important if carbon emission levels will surpass the Paris targets.Our results also demonstrate that high-biodiversity regions will overlap with critical carbon sinks even if global warming surpasses 2 °C above pre-industrial levels, as cVeg increases more under the two more distant scenarios, SSP1-2.6 and SSP5-8.5, compared with the others.

The complexity of Earth System models
The discrepancies between the cVeg and LAI trends, with more consistent differences in cVeg between biodiversity versus control areas seen across scenarios, is probably due to a mix of factors.Changes in cVeg result from changes in NPP but also from the carbon residence time in living vegetation 30,31 .Residence time responds to CO 2 levels, which vary across scenarios, and vegetation biomass is influenced by natural disturbances, such as natural fires, which are also simulated interactively in CMIP6 for each scenario.In contrast, LAI results from the carbon balance of the leaves 32 , which is projected to increase 'linearly' under each scenario (see Supplementary Fig. 13).This is different from cVeg, which does not increase much under intermediate scenarios SSP2-4.5 and SSP3-7.0;this is probably related to the influence of land-use change on cVeg.In fact, while all scenarios project some increase in cropland extent, scenario SSP3-7.0 also projects a substantial increase in grazing land 33 .Thus, the land use of this particular scenario is dominated by a type of vegetation with higher leaf carbon content but lower amount of carbon stored belowground and in woody parts aboveground, leading to lower total carbon in vegetation (cVeg).
The global loss of WA under each scenario, which is even more pronounced in the high-biodiversity regions, is due to the combination of reduced precipitation and increased AET.This is especially true in the Amazon (which covers large extents of high-biodiversity regions), meaning that one of the currently most important areas for biodiversity will not be able to preserve its water regulation ability in the future 34,35 .However, it is also important to recognize the contribution the evapotranspiration from the Amazon (and other extensive tropical forested regions such as the Congo and Indonesia) makes in generating precipitation elsewhere.This vegetation-mediated moisture recycling accounts for a substantial portion of the precipitation in downwind systems 36 , and increased evapotranspiration in these sending ecosystems could actually increase the resilience of the receiving systems, and the NCP they generate.

The regulating role of important biodiversity areas
Understanding the relationship between NCP and biodiversity, and their potential conservation synergies, is essential for sustaining human well-being and securing Earth's life support systems 21,37 .Nevertheless, even as the number of studies focused on biodiversityecosystem-functioning relationship has increased in recent years 15,38 , many remaining uncertainties hinder clear conclusions 12,28,39 .Indeed, the relationship between biodiversity and NCP provision can be very hard to define due to the complexity of processes and interactions present in ecosystems that are seldom fully described and have remained inadequately investigated 24,40 .However, different authors have used indirect analysis 3,22,41 aimed at linking biodiversity and NCP either through a relevant indicator or a proxy derived from spatial layers such as land-use/land-cover classifications, vegetation maps or remote sensing indices.Indirect analyses are assumed to be informative about the quality of the habitat/ecosystem, its contributions to people and its ability to support the species.
The fact that important biodiversity areas are strongly associated with the provision and increase in biomass-related NCP might be related to the fact that both processes are ultimately driven by the same underlying environmental drivers.For example, it has been found that the correlation between species richness and carbon content is higher when both variables can be independently predicted by climate, soil and topography 37 .This is not always the case, as many open ecosystems (grasslands, savannahs) are important for biodiversity conservation but have limited carbon content.This means that in principle, our results might be driven by global forest cover rather than biodiversity value.However, we have reduced this risk because we have explicitly accounted for forest cover (and other confounding factors) while selecting control areas to be compared to our biodiversity value regions, and we successfully reduced the global imbalance in those variables through statistical matching.Moreover, our selection of high-biodiversity regions includes large areas with limited forest cover, such as the North American west coast, Patagonia, parts of Spain, South Africa, India and Australia.Even in these areas, we showed high NCP values in high-biodiversity regions.
The pattern observed with WA, where areas of high biodiversity value face higher risk of increased evapotranspiration than control areas, is linked to a potential overestimation of LAI due to CO 2 fertilization effects and their impact on the hydrologic cycle.
To reproduce the response and feedback of global changes in Earth's ecosystems, some CMIP6 models simulated land-cover structure (such as EC-Earth3-Veg and MPI-ESM1-2-LR used in our analysis), while others used prescribed natural vegetation distribution based on a forcing dataset of land-use and land-cover change (for example, LUH2).In general, CMIP6 models represent vegetation structure primarily in terms LAI, tree height and vegetation distribution by fractional coverage of plant functional types.Therefore, biases in these three variables could lead to inaccurate simulation of biogeochemical processes.However, the CO 2 fertilization effect is considered the main driver of the projected increment in global LAI, which is partly offset by the negative effects of global warming, although all CMIP6 models are known to overestimate global mean LAI 42,43 .Such overestimation will lead to overestimation of carbon input to the terrestrial ecosystem (gross primary production) 44 , transpiration and canopy evaporation 44 .The AET increase induced by vegetation greening through an extended area of leaves performing transpiration will then reduce soil moisture and run-off, which can intensify droughts at the catchment scale 45 .Thus, notwithstanding the continuous improvement of CMIP6 models, there is still a need to ameliorate the parameterization of dynamic vegetation within Earth System models, while future research on global circulation models (GCMs) should be focused on better defining and critically evaluating the terrestrial carbon cycle model using updated observations data to validate the models 32,46 .

Understanding realized versus potential NCP levels
Analysing how the increases or decreases in potential NCP coincide with current and future human population centres is an important next step for ultimately mapping realized NCP, the actual flow of NCP that humanity receives, which could better inform policies that drive national mitigation and adaptation actions.This is especially important given the heterogeneity found in the levels of potential NCP provision between countries even within the same subregion.How future changes in population might interact with and intensify these ecological changes is a key consideration for future policy.For example, strong differences in NCP values are seen between high-biodiversity regions and control areas in Indonesia, while the difference is less evident in Malaysia.Nevertheless, population density is projected to increase strongly under all future scenarios in areas outside high-biodiversity regions in both countries, making the differences in realized NCP even larger.Other countries are projected to suffer a reduction in NCP provision with concurrent population increases: a reduction of cVeg is seen in Cameroon, Gabon, Zambia and South Africa under various scenarios as population grows in all these countries.In contrast, despite overall declines in water quantity regulation globally, a few countries are characterized by an increase in WA, such as Zimbabwe, Venezuela, Suriname and Sri Lanka.In particular in these countries, both precipitation and AET will decrease, but AET will decrease more substantially, leading to a positive change in WA; in Sri Lanka, a higher increase in precipitation compared with the increase in evapotranspiration is projected.Sri Lanka is also projected to move towards a major population increase compared with other countries characterized by a WA increase, where smaller increases in population are projected.On the other hand, there are also countries, such as China, where a reduction in human population will offset (to some extent) the increase in potential NCP.Thus, fully accounting for realized NCP requires an examination of not only population but also demand (or need) for NCP, which may vary among countries or different social groups on the basis of their vulnerability; this is an important area for further work.
While both land-use change and climate change play a role in determining the change in NCP levels predicted under alternative scenarios, we found a dominant role of climate over land use.This means that area-based conservation interventions must be coupled with bold climate mitigation policies, or risk being ineffective at preserving the crucial NCP provision role played by several important biodiversity areas.However, we found a risk of reduced WA in these areas even under the most optimistic scenario considered in our analysis, which is compliant with the 2 °C Paris target.This suggests that adaptation to climate change will assume increasing importance even under sustainability scenarios where local communities and nations will need to safeguard water resources, increase water-use efficiency, and change practices and behaviours where necessary to continue to thrive under changing precipitation patterns.Indeed, trade-offs will need to be weighed in the role of nature to provide mitigation versus adaptation benefits for coping with climate change 47 .
Spatial guidance is needed to identify areas of potential co-benefits between conserving biodiversity and NCP, to promote the implementation of global climate and biodiversity commitments at local levels.General conclusions have been difficult to draw from past work in which 'biodiversity' has generally been based on subsets of global biodiversity, typically mammals and birds 48 , often to the exclusion of reptiles, invertebrates and plant species, as well as other dimensions of biodiversity 3,23,37,49 .Our work synthesizes high-biodiversity regions resulting from the combination of published biodiversity conservation maps focusing on different taxa, as well as phylogenetic and functional diversity 25 , thus representing many different elements of biodiversity.

Need for a coupled approach to reach different global targets
Our results show spatial congruence between biodiversity value and NCP value.Regardless of whether this spatial congruence is due to correlation (via underlying environmental mechanisms) or causation, the areas we identified are important from a conservation policy perspective, allowing us to identify the relative contribution of high-biodiversity regions to NCP provision.Our results show the existence of substantial synergies between the achievement of goals set under different conventions (for example, Convention on Biological Diversity and the United Nations Framework Convention on Climate Change) and different Sustainable Development Goals (SDGs).Conserving areas of high biodiversity value would protect life on land (SDG15) while delivering a high contribution to good health and well-being (SDG3), availability of clean water (SDG6) and mitigation of climate change (SDG 13).Thus, it is now fundamental to improve the mapping of other (less studied) NCP to discover whether other synergistic patterns (similar to those we describe here) emerge 50 .Under accelerating climate change and under high risk of global geopolitical instability from recent humanitarian catastrophes (such as COVID-19, the war in Ukraine and many regional-scale extreme weather events), it is imperative to quickly consolidate an integrated human-nature paradigm shift incorporating NCP into the assessment of SDGs while guiding investments and implementing nature-based solutions for climate change adaptation and mitigation policies 51 .Here we show that spatial options for win-win strategies that achieve human and nature benefits are available and rather substantial, and should be pursued before being eroded by human-induced environmental change.

Methods
Regulating NCP are not easily measured since different abiotic factors are intertwined in the generation of several regulating NCP.Nevertheless, the biophysical processes behind NCP provision can be measured by evaluating different indicators.Thus, we chose different indicators to represent global trends in potential NCP when available (Supplementary Table 1).We selected indicators starting with those reported in chapter 2.3 of the IPBES global assessment report on biodiversity and ecosystem services 11 , and expanded the selection on the basis of data availability across multiple climate models and multiple scenarios, following literature sources reported in the IPBES global assessment report chapter 2.3 27,52,53 .The data selected to evaluate the trend of different NCP were retrieved from the CMIP6 and WorldClim datasets 54,55 .We measured indicators for three NCP representing the regulation of air quality, climate and water quantity.
We measured air quality regulation (NCP3) using the LAI.LAI is defined as one side of the green leaf area per unit ground area in broad-leaf canopies and as one half of the total needle surface area per unit ground area in coniferous canopies 27 .LAI is used as an indicator of air quality regulation because vegetated areas with higher LAI values have a higher surface roughness, structural complexity and density of leaves, which all contribute to the interception of air pollutants and favour their further deposition and absorption 27 .Thus, a higher LAI value is associated with higher provision of NCP3.
Similarly, we used spatial estimates of the density of the total carbon (aboveground and belowground) in vegetation (cVeg) as an index of climate regulation.The regulation of climate (NCP4) depends on ecosystems through either sequestration or release of greenhouse gases such CO 2 .Consequently, changes in carbon stored in vegetation can potentially mitigate warming caused by increasing concentrations of CO 2 in the atmosphere 56 .
Climate change also influences ecological processes that contribute to atmospheric carbon balance, such as plant respiration and decomposition of organic matter 30 , so we added NPP to our list of indicators for NCP4.NPP is one of the main components of carbon balance and indicates the rate at which energy is stored as biomass, measuring the degree of accumulation of atmospheric CO 2 into terrestrial ecosystems.NPP is equal to the difference between the carbon assimilated during photosynthesis and that released during plant respiration, hence it is an important indicator of terrestrial carbon 57 .

Article
https://doi.org/10.1038/s41893-023-01179-5 Human and terrestrial wildlife survival relies on freshwater supply and the regulation of this contribution (NCP6) is an essential ecosystem function.The water cycle is an extremely complex system based on different processes including precipitation, evapotranspiration and run-off.Changes in the spatio-temporal distribution of precipitation and evapotranspiration alter the availability of water 58 , and anthropogenic climate change affects water balance by altering these variables 59 .Furthermore, aridity (the lack of moisture) depends on the same variables and many studies predicted a global terrestrial drying in the future 60,61 .Thus, to try to include all the different aspects characterizing the complexity of the water cycle, we chose to include in our analyses different indicators to evaluate freshwater quantity regulation, namely WA, AET, PET and AI.
We reported WA as a simple function of precipitation (P) and AET 58 .WA is then expressed as the total available water that can be in the form of run-off, soil moisture and groundwater recharge in terms of mm per month 62 , as shown in equation ( 1): Data on AET and precipitation were retrieved from CMIP6 54 .We also recalculated the WA and PET index using precipitation and temperature data from WorldClim 55 as a sensitivity test of the influence of dataset resolution, since data from WorldClim are available as downscaled data at a resolution of 30 s (~1 km).
AET is defined as the actual amount of water removed from a surface area through a combined process of both evaporation from soil and plant surfaces and transpiration through plant canopies.AET is expected to increase in the future due to the warming climate and this would probably result in more frequent and intense extreme events 63 .PET can be considered as another indicator of water regulation because it is a measure of the atmospheric demand for evaporation and is independent of the supply of water itself.A higher PET value represents more arid, evaporative conditions 64 ; thus, an increase in PET would indicate a reduction in the provision of water quantity regulation (NCP6).Specifically, PET is defined as the amount of water that would potentially be removed from a vegetated surface through the processes of evaporation or transpiration when there is no water limitation, that is, with no forcing other than atmospheric demand 65 .Many different equations have been adopted for PET estimation 66 , but we chose the Hargreaves model 67 for our study.The Hargreaves method perform almost as well as the FAO Penman-Monteith method but requires less parameterization 53 since it is based only on solar radiation and temperature parameters 67 following equation (2): where T mean is the mean monthly temperature, TD is the mean monthly temperature range and RA is the radiation on top of the atmosphere.PET can also be used to calculate a variety of aridity, drought and soil moisture indices, as the AI.AI is a quantitative indicator for the background climatological dryness or wetness of the land surface at given climate conditions 61 , and is defined as the ratio of the mean annual precipitation to the PET following equation (3): Using AI, the climate of a region can be categorized into one of five different classes: hyper-arid (AI < 0.05), arid (0.05 ≤ AI < 0.20), semi-arid (0.20 ≤ AI < 0.50), dry sub-humid (0.50 ≤ AI < 0.65) and humid (AI ≥ 0.65).
All the AET and PET products were converted from their respective units (kg m −2 s −1 and W m −2 ) to millimetres per month (mm per month).

Data manipulation and calculation of trends
We collected data for the selected proxies from the CMIP6 dataset available from the archive of the Earth System grid federation online system (https://esgf-node.llnl.gov/search/cmip6/)using the R package 'epwshiftr' 68 .The CMIP6 dataset gathers results of many different GCMs that run different experiments to simulate climates of the past, present and future.Specifically, we downloaded the data for the 'historical' and the future (SSP1-2.6,SSP2-4.5,SSP3-7.0 and SSP5-8.5)experiments.The SSP experiments represent alternative socio-economic development scenarios, from lowest (SSP1-2.6) to highest (SSP5-8.5)levels of development intensity and hence climate change.
For each NCP indicator, we used the GCMs that had data for the historical experiment and all four future experiments (Supplementary Table 2).Since the spatial resolution of the different models differs, the NCP indicator data were resampled to a 10 km grid using bilinear interpolation and re-projected into Mollweide equal-area projection.Then, the different indicator values from individual GCMs were averaged.Finally, we estimated the difference in each NCP indicator's projections between the future time window (2041-2070) and the baseline period (1985-2014) under the different development scenarios (SSP1-2.6,SSP2-4.5,SSP3-7.0 and SSP5-8.5).To compare the change in different NCP, all the indicator values were rescaled between 0 and 1 using the function 'rescaleImage' from the R package RStoolbox 69 , which performs linear shifts of value ranges (in our case the minimum and maximum values of the raster stack of each indicator comprising the baseline period and all the four scenarios) to match new minimum and maximum values.

Comparing NCP values inside and outside high-biodiversity regions
We tested whether high-biodiversity regions showed different NCP trends compared with other areas with comparable environmental characteristics.Several maps of biodiversity conservation priorities have been developed in recent decades, each with partially different characteristics, and built on different approaches of spatial prioritization and different data.We represented high-biodiversity regions using the map of 'biodiversity consensus' developed in ref. 25, which synthesizes global conservation priorities and ranks regions on the basis of their level of inclusion in independently generated biodiversity priority maps.The map of biodiversity consensus is a global map with continuous value, but we chose to report results only for the top 10% of high-biodiversity regions, with the highest level of biodiversity importance covering 10.6% of the land area (Supplementary Fig. 1); otherwise, we would have had difficulties with propensity score matching at the country level because there are few countries, especially in the tropics, for which almost their entire surface could be considered as a biodiversity priority region.As a sensitivity test, we ran the analysis for the top 30% of high-biodiversity regions, which instead cover 31.6% of the land mass, excluding countries for which it was impossible to sample control, not priority, areas (Supplementary Table 4).
To assess the role of high-biodiversity regions in preventing declines in NCP provision, we compared scenarios of NCP inside and outside these regions.To determine the effectiveness of high-biodiversity regions in retaining higher levels of NCP, it is necessary to account for the characteristics of these regions.The top-ranked high-biodiversity regions are disproportionately located in the tropics where there are usually high levels of taxonomic/phylogenetic diversity, and in wilderness areas distant from human infrastructures and urban areas, with low agricultural value 19,70 .Thus, to control for such selection bias and to account for other confounding factors in biodiversity conservation priority region location, we used the propensity score matching technique 71,72 to identify areas with socio-environmental conditions similar to those of priority biodiversity regions 73 .We selected 10 bioclimatic and 3 topographic variables available from WorldClim 55 , land-cover variables reclassified from the European Space Agency (ESA) Article https://doi.org/10.1038/s41893-023-01179-5and a global map of accessibility describing the travel time to cities 74 , all logit transformed and standardized (that is, mean = 0, s.d.= 1) (see Supplementary Table 3).Specifically, the propensity score, which is defined as the probability of receiving treatment given the observed covariates 75 , allows matching of individuals in the control and treatment conditions with the same likelihood of receiving treatment 76 .We used a nearest-neighbour method with a ratio of 1:1 without replacement, such that each grid cell inside a biodiversity priority region is matched to a different non-priority grid cell with similar characteristics and within the same country.We made an exception for countries where there are not enough non-priority areas (that is, control units) for the matching, in which case we allowed replacement in the sampling.To control for low-quality matches, we used a 0.25 s.d.caliper when possible (see Supplementary Table 5), that is, we selected only the matches where the distance in propensity scores between treatment and controls is <0.25 s.d. of the estimated propensity scores of the sample 71,76 .
The highest resolution available for CMIP6 data is 100 km, while the resolution of the biodiversity priority map is 10 km; to avoid selecting priority and non-priority grid cells from within the same 100 km cell, we applied a filter and selected only biodiversity priority cells falling within 100 km resolution grids with at least 50% consensus coverage and applied the same procedure to the selection of non-priority cells.We also ran a sensitivity analysis to control for the effect of the filtering process using a softer threshold, selecting 10 km priority cells only if >25% of the corresponding 100 km cell is covered by a priority region.
To assess the quality of the matching procedure, we checked for covariate balance before and after the matching within the original dataset and in the matched sample.For this purpose, we used the C-statistic, which has the advantage of automatically evaluating balance on all covariates simultaneously.This metric is given by the area under the receiver operating characteristic curve (or C-statistic from logistic regression) from a propensity score model estimated in the matched sample 77 .The C-statistic ranges from 0.5 to 1.0, with the minimum indicating that the propensity score model has no ability to discriminate between treated and untreated units after matching, which means perfect balance of covariates, having very similar values between the two groups 78 .In addition to the cumulative C-statistics, we also performed a more specific test for the variable 'forest cover', which is highly correlated to LAI and cVeg, and risks confounding our results on NCP values inside versus outside high-biodiversity regions.We used the SMD and the eCDFs of each covariate between groups.SMDs close to zero indicate good balance 79 , with recommended values of SMDs being below 0.1.The eCDF statistics (that is, mean and maximum eCDF) correspond to the difference in the overall distributions of the covariates between the treatment groups.The values of both statistics range from 0 to 1, with values closer to zero indicating better balance.There are no specific recommendations for the values these statistics should take, although notably high values may indicate imbalance on higher moments of the covariates.After the statistical matching procedure was completed, we were able to identify treatment (that is, high biodiversity) and control areas with similar characteristics, and we evaluated their difference in NCP levels using the two-sided unpaired Wilcoxon signed-rank test.

Fig. 1 |Fig. 2 |
Fig. 1 | Projections of change in the value of NCP at 10 km resolution between the baseline period 1985-2014 and the future 2041-2070 under scenario SSP5-8.5.The absolute change in levels of three NCP indicators.a, NCP3, represented by leaf area index (LAI, m 2 m −2 ).b, NCP4, represented by carbon in vegetation (cVeg, kg m −2 ).c, NCP6, represented by water availability (WA, mm per month).Masked opaque areas indicate regions outside the top 10% biodiversity conservation priorities.See Supplementary Fig. 2 for scenario SSP1-2.6 and Supplementary Fig. 3 for other indicators.Regions of high biodiversity value are indicated in Supplementary Fig.1.