The impact of ecological restoration on ecosystem services change modulated by drought and rising CO2

Ecological restoration projects (ERPs) are an indispensable component of natural climate solutions and have proven to be very important for reversing environmental degradation in vulnerable regions and enhancing ecosystem services. However, the level of enhancement would be inevitably influenced by global drought and rising CO2, which remain less investigated. In this study, we took the Beijing‐Tianjin sand source region (which has experienced long‐term ERPs), China, as an example and combined the process‐based Biome‐BGCMuSo model to set multiple scenarios to address this issue. We found ERP‐induced carbon sequestration (CS), water retention (WR), soil retention (SR), and sandstorm prevention (SP) increased by 22.21%, 2.87%, 2.35%, and 28.77%, respectively. Moreover, the ecosystem services promotion from afforestation was greater than that from grassland planting. Approximately 91.41%, 98.13%, and 64.51% of the increased CS, SR, and SP were contributed by afforestation. However, afforestation also caused the WR to decline. Although rising CO2 amplified ecosystem services contributed by ERPs, it was almost totally offset by drought. The contribution of ERPs to CS, WR, SR, and SP was reduced by 5.74%, 32.62%, 11.74%, and 14.86%, respectively, under combined drought and rising CO2. Our results confirmed the importance of ERPs in strengthening ecosystem services provision. Furthermore, we provide a quantitative way to understand the influence rate of drought and rising CO2 on ERP‐induced ecosystem service dynamics. In addition, the considerable negative climate change impact implied that restoration strategies should be optimized to improve ecosystem resilience to better combat negative climate change impacts.


| INTRODUC TI ON
In recent decades, natural ecosystems have undergone severe degradation due to inappropriate human activities (Mitchard, 2018;Santos et al., 2020). To reverse this trend, ecological restoration projects (ERPs) represented by UN REDD+ (The United Nations Collaborative Programme on Reducing Emissions from Deforestation and Forest Degradation in Developing Countries plus) programs have emerged around the world with the objective of restoring degraded ecosystems (Strassburg et al., 2012). Prior studies have already suggested that these ERPs are efficient in restoring degraded ecosystems and enhancing ecosystem services provision (Erbaugh et al., 2020;Rey Benayas et al., 2009;Tong et al., 2020).
Recent advances have also suggested that the impact of drought and rising CO 2 on ecosystem services is increasingly important (Runting et al., 2017). Unfortunately, it remains unknown how drought and rising CO 2 affect the contribution of ERPs to ecosystem service provision (Constenla-Villoslada et al., 2022;Pereira, 2020;Yang et al., 2022).
Urbanization, agricultural reclamation, and deforestation have caused land use/cover (LUC) changes to occur in almost one-third (32%) of the global land area in the past six decades (Winkler et al., 2021). LUC changes have triggered the area loss and quality decline of natural ecosystems (Mitchard, 2018). This change is particularly evident in developing countries, such as those in East Asia, Southeast Asia, South America, and Africa (Sze et al., 2022).
Ecosystem degradation has caused a substantial decline in ecosystem service provision and has greatly increased the probability of water erosion, floods, sandstorms, and other environmental problems (Borrelli et al., 2017;Ouyang et al., 2016;Rodriguez-Caballero et al., 2022;Tellman et al., 2021). ERPs have been implemented globally to mitigate and prevent these negative impacts on ecosystem services (Mitchard, 2018;Orth Robert et al., 2020;Santos et al., 2020). Afforestation in Europe was shown to significantly increase local water supply services (Meier et al., 2021). It has been reported that global afforestation has the potential to lead to a greater than 25% increase in forested area, including more than 200 Gt of additional carbon at maturity (Bastin et al., 2019). The spatial scale and duration of China's ERPs were the greatest globally (Lu et al., 2018). Satellite photos show that 42% of the greening of the Earth's surface has occurred in China, and this increase was mainly due to the ERPs implemented in recent decades (Chen, Park, et al., 2019). China's ERPs also explained the increase in most ecosystem services (Ouyang et al., 2016). Nevertheless, some studies have noted that high-intensity afforestation in water-scarce areas may cause acute drought and soil moisture declines, which would lead to poor vegetation growth and the decline in ecosystem services Schwärzel et al., 2020;Zhao et al., 2021).
Climate change and rising CO 2 have been found to have great impacts on ecosystem services by affecting ecosystem processes (Morán-Ordóñez et al., 2021;Runting et al., 2017;Walker et al., 2021). Climate change can increase the intensity and frequency of heat waves (Zheng et al., 2021), floods , wildfires (Mollicone et al., 2006), pest diseases , and especially drought (Chiang et al., 2021). These changes would produce a great negative impact on ecosystem services (Shukla et al., 2019;Walther et al., 2002). Among them, droughts in a warming climate have become more common and more extreme, increasing the water stress in vegetation and constraining the vegetation growth rate (Konings et al., 2021). This change would affect ecosystem processes and reduce the provision of ecosystem services (Fu et al., 2013). Morán-Ordóñez et al. (2021) found that future drought would probably cause material services and regulating services to decline by approximately 4.43% and 8.6%, respectively, in Mediterranean forests. The negative impact of drought on ecosystem services provision would be aggravated by the increase in the degree of drought (Han et al., 2019). Moreover, the increased duration of drought would also further amplify the negative impact of drought on ecosystem services (Vaughn et al., 2015). Rising CO 2 can boost photosynthesis to promote vegetation growth and reduce stomatal conductance to improve water use efficiency (Dusenge et al., 2019). Promoted vegetation growth would increase leaf area to absorb more carbon from the air, intercept more rainwater to reduce storm runoff, reduce the erosion of soil by precipitation, and block dust Zhou et al., 2020). Thus, rising CO 2 would be helpful in increasing ecosystem services, including carbon sink, water retention, soil retention, and sandstorm prevention. Devaraju et al. (2016) found that more than 40% of the increased carbon sink was contributed by rising CO 2 in 1850-2005. This rate has further increased to 47% in recent decades (Chen, Ju, et al., 2019). However, rising CO 2 would also increase leaf area to accelerate evapotranspiration and reduce water retention (Ukkola et al., 2016). It has been reported that global water yield has decreased by 13% owing to rising CO 2 (Piao et al., 2007).
Although recent advances have already explored the individual effects of drought, rising CO 2 , and ERPs on specific ecosystem services at the regional and global scales McCauley et al., 2022;Piao et al., 2007), few studies have investigated how the impact of ERPs on ecosystem services change is modulated by drought and rising CO 2 . This deficiency may lead to an incomplete understanding of the influence of the global change environment on ERP-induced ecosystem services change , which would not provide valuable insights for decision-making and ecosystem management optimization.
To quantify the impact of ERPs on regional ecosystem service dynamics and how it was affected by drought and rising CO 2 , we used the ecosystem process-based Biome-BGCMuSo model (BBMS) and took the Beijing-Tianjin Sand Source Region (BTSSR) as a case study area to explore this issue. The BTSSR experiences severe water scarcity and long-term ERPs. Moreover, the growth of vegetation in the BTSSR is sensitive to drought and rising CO 2 (Huang, Lu, et al., 2021). This study aimed to (1) examine the impact of ERPs on ecosystem services change in 2001-2020 and (2) clarify the pattern and extent of the impact of ERPs on ecosystem services, as affected by drought and rising CO 2 .

| Study area description
The BTSSR is located in the arid and semiarid area of northern China and covers approximately 458 × 10 3 km 2 (Figure 1a). This area suffered from serious soil erosion and sandstorms owing to vegetation damage and ecosystem degradation .
The BTSSR is territorially composed of forest (dominated by deciduous broadleaf forest), grassland, cropland, and other types of land uses (Jiang et al., 2018;Li et al., 2014). Meanwhile, the BTSSR is characterized by a plateau, plains, and mountains. The western, northwestern, and northern portions of this area are located on the central Inner Mongolian Plateau, while the southeast portions are plains (Huang, Lu, et al., 2021). To reduce the threat of violent sandstorms to Beijing and the surrounding areas, the Beijing-Tianjin Sand Source Control Project was launched with the goal of reducing dust hazards mainly through afforestation and grassland planting (planting grass species). For details, please refer to Text S1 in Supporting Information. There was a total investment of approximately US$ 8.11 billion during stage I of the BTSSR. The capital investment in stage II of the BTSSR will reach US$13.4 billion (Huang, Lu, et al., 2021).

| Methodology for ecosystem service calculation
Here, the BBMS was used to calculate the ecosystem services.
The BBMS is an ecosystem process-based model developed from the Biome-BGC model (Hidy et al., 2016). It can simulate the storage and fluxes of water, carbon, and nitrogen between ecosystems and the atmosphere (Thornton & Rosenbloom, 2005). The BBMS can provide multiple vegetation, water, and carbon flux parameters that can be used in ecosystem services calculations. Equations to calculate ecosystem services (including carbon sequestration, water retention, soil retention, and sandstorm prevention) are described in Equations (1)-(4). Explicit information to connect the BBMS and ecosystem services calculation is described in Text S3 of Supporting Information.

| Carbon sequestration
Carbon sequestration is defined as the sequestration of CO 2 by terrestrial ecosystems (Ouyang et al., 2020). It was calculated using where NEP is the net ecosystem productivity; 44/12 is the conversion factor from C to CO 2 ; NPP is the net primary productivity; and Rh is the soil heterotrophic respiration. Details to calculate NPP and Rh are described in Text S3 of Supporting Information.

| Water retention
Water balance theory was used to calculate water retention . The calculation of water retention is described in where WR is the water retention (mm); PRE is the precipitation (mm); AET is the actual evapotranspiration (mm); and QF is the storm runoff.
Details to calculate AET and QF are described in Text S3 of Supporting Information.

| Soil retention
The ecosystem service of soil retention refers to the soil retained by ecosystems, which prevents sediments from entering water bodies and causing damage. We measured soil retention as the difference between soil erosion without vegetation cover and soil erosion under the current land cover pattern and soil erosion control practices (Ouyang et al., 2020). The Revised Universal Soil Loss Equation (RUSLE) model was used to calculate the soil retention (Rao et al., 2014). The calculation of soil retention is described in where SR is the soil retention (t hm −2 a −1 ); R is the rainfall erosivity factor (MJ mm hm −2 h −1 a −1 ); K is the erodibility of the soil or the amount of soil lost through erosion per unit area following rainfall of a given intensity (t hm 2 h hm −2 MJ −1 mm −1 ); LS is the topographic factor; and C is the vegetation cover factor. Details to calculate these factors are described in Text S3 of Supporting Information.

| Sandstorm prevention
Sandstorm prevention refers to the sand retained in an ecosystem.
We measure sandstorm prevention as the difference between wind erosion without vegetation cover and wind erosion under the current land cover pattern (Ouyang et al., 2016). Sandstorm prevention was calculated through the RWEQ model and described in Equation (4): where SP is the sandstorm prevention (kg m −2 ); S LP is the potential wind erosion (kg m −2 ); and S L is the actual wind erosion (kg m −2 ). Details to calculate these factors are described in Text S3 of Supporting Information.
To quantify the effect of ERPs on ecosystem services change through the BBMS, we constructed a mechanism by summing the area fraction change of LUCs in each 10 × 10 km 2 grid cell ( Figure S4) (Equation 1) with reference to our prior study .
For more details, please refer to Text S4 in Supporting Information.
where ES i denotes the ecosystem services (including carbon sequestration, water retention, soil retention, and sandstorm prevention) in and grassland in grid cell i, respectively; and ES F,i , ES C,i , and ES G,i denote the ecosystem services per unit area of forest, cropland, and grassland in grid cell i, respectively.

| Data sources
Multiple datasets were used as inputs to drive the calculation of ecosystem services. All datasets are summarized in Table 1. Explicit information for the data sources is described in Text S2 of Supporting Information. As these datasets varied in format and spatial and temporal resolutions, they were integrated to feed the needs of the model input. For the procedures used to integrate the multiple datasets, refer to Huang, Lu, et al. (2021).

| Technical validation
To guarantee the reliability of our simulated results, we compared our estimated leaf area index (LAI) and ecosystem services with other datasets or results from published studies. For LAI, our simulated BBMS-based LAI was consistent with MODIS LAI in trend ( Figure S6), which indicates the reliability of BBMS in catching the vegetation dynamics. For ecosystem services, we compared our simulated carbon sequestration (NEP was used as surrogate), water retention, soil retention (soil erosion was used as surrogate), and sandstorm prevention (sandstorm erosion was used as surrogate) with accessible dataset (Bodesheim et al., 2018;Huntzinger et al., 2013;Jung et al., 2020;Zeng et al., 2020) or values extracted from published articles that implemented in areas with similar climate background and vegetation (Table S6) (Chi et al., 2021;Guo et al., 2015;Jin et al., 2021;Li et al., 2020;Lu et al., 2022;Ouyang et al., 2016;Rao et al., 2015;Xu et al., 2019;Zhang et al., 2018;Zhao et al., 2020). The high consistency between these comparisons implied the confidence of our simulated results ( Figure S7).

| Scenario design
Here, we constructed eight scenarios (Reference (Re), Drought (Dr), Rising CO 2 (CO 2 ), Drought + Rising CO 2 (Dr + CO 2 ), ERPs (ER), ERPs + Drought (ER + Dr), ERPs + Rising CO 2 (ER + CO 2 ), and ERPs + Drought + Rising CO 2 (ER + Dr + CO 2 )), which aimed to quantify the effect of ERPs on ecosystem services change in the condition that with or without the impact of drought and rising CO 2 ( Table 2). Explicit scenario setting rules and the meaning of these scenarios are described in Table 2.
Next, we used the spin-up procedure to make the carbon/water pools and flux reach a steady state to prepare the normal run (Hidy et al., 2016). Then, we ran a set of model simulations in 2001-2020 with the constant or varying climate, atmospheric CO 2 concentration, and LUC, as shown in Table 2. Equations for calculating the ecosystem services based on the results from these scenarios are described in Table S3. 2.3.2 | Scenario comparison to disentangle the impact of drought and rising CO 2 on ERP-induced ecosystem services change First, we calculated the cumulative ecosystem services difference between different scenarios according to drought or rising CO 2 alone, and with joint drought and rising CO 2 . Next, we quantified the effect of drought and rising CO 2 on ERP-induced ecosystem services change based on the rules described in Table 4.  Figure S5. Then, we extracted the area fractions of LUCs involved in different LUC trajectories. Next, we calculated the ecosystem services in different LUC trajectories of these scenarios using the equations described in Tables S4 and S5.
Finally, we quantified the effects of different LUC trajectories on ecosystem services and their interaction with drought and rising CO 2 using the steps described in Text S4 in Supporting Information.

| RE SULTS
Compared with the Reference scenario, ERP-induced carbon sequestration, water retention, soil retention, and sandstorm prevention (3.20 billion t), and 28.77% (8.64 billion t) during 2001-2020, respectively ( Figure 2, Figure S8). Approximately 91.41%, 98.13%, and 64.51% of the increased carbon sequestration, soil retention, and sandstorm prevention were contributed by afforestation (Figure 2). In particular, afforestation caused water retention to decline by 2.39%, while grassland planting increased water retention (+5.26%) and reversed this decline (Figure 2). Increased carbon sequestration and soil retention contributed by ERPs mainly occurred in the southeastern BTSSR, while increased water retention contributed by ERPs mainly occurred in the eastern and western BTSSR (Figure 3). Increased TA B L E 4 Impact of drought and rising CO 2 on ERP-induced ecosystem services change. sandstorm prevention occurred in the southern, southwestern, and southeastern BTSSR (Figure 3). In addition, the impact of ERPs on carbon sequestration and soil retention gradually increased along the precipitation gradients, while the impact of ERPs on water retention and sandstorm prevention gradually increased initially but then decreased along the precipitation gradients ( Figure 4).   Figure 5). The reducing impact from drought on ERP-induced carbon sequestration and soil retention change mainly occurred in the southeastern BTSSR, while that on ERP-induced water retention and sandstorm prevention change mainly occurred in the eastern and western BTSSR ( Figure 6). The negative impacts of drought on ERP-induced carbon sequestration, soil retention, and sandstorm prevention decline were gradually aggravated along the precipitation gradients, while the negative impact of drought on ERP-induced water retention decline was initially aggravated but then reversed along the precipitation gradients (Figure 7).

F I G U R E 3
Rising CO 2 caused the results of ERP on carbon sequestration, water retention, soil retention, and sandstorm prevention to be amplified by 13.49%, 14.53%, 9.87%, and 7.10% in 2001-2020, respectively ( Figure 5). Notably, rising CO 2 promoted grassland planting-induced water retention increases and decreased afforestation-induced water retention losses (Figures 2 and 5). The amplification impact from rising CO 2 on ERP-induced carbon sequestration, water retention, soil retention, and sandstorm prevention change mainly occurred in the southeastern BTSSR ( Figure 6). In addition, the positive impact of rising CO 2 on ERP-induced carbon sequestration, water retention, soil retention, and sandstorm prevention promotion gradually increased along the precipitation gradients ( Figure 7).
The combined drought and rising CO 2 -induced carbon sequestration and soil retention change decline mainly occurred in the southeastern BTSSR, while the induced water retention and sandstorm prevention to decline mainly occurred in the eastern and western BTSSR ( Figure 6). The impact of combined drought and rising CO 2 on ERP-induced carbon sequestration, soil retention, and sandstorm prevention decline was gradually aggravated along the precipitation gradients, while combined drought and rising CO 2 on ERP-induced water retention decline was initially aggravated but then reversed along the precipitation gradients (Figure 7).

| The impacts of ERPs on ecosystem services
Afforestation was the main mission of the ERPs implemented in the BTSSR , which may explain why the impacts of afforestation on carbon sequestration, soil retention, and sandstorm prevention were much higher than those of grassland planting, and the increase in carbon sequestration and soil retention mainly occurred in the southeastern BTSSR. In particular, grassland planting F I G U R E 5 Amplifying/reducing rate of drought (Dr), rising CO 2 (CO 2 ) and combined drought and rising CO 2 (Dr + CO 2 ) on ecological restoration project (ERP)-induced ecosystem services change in 2001-2010, 2011-2020 and 2001-2020.
increased water retention, but afforestation decreased water retention. Extensive afforestation in water-limited areas has been proven to increase evapotranspiration and decrease water storage (Bond et al., 2019;Tian et al., 2021;Veldman et al., 2015). Moreover, water consumption increases with increased forest age under protection Huang, Liu, et al., 2021). Unlike forests, the vegetation structure and species composition in grasslands are much simpler (Pan et al., 2022). The leaf area in grassland was much lower than that in forests, so the water loss through evapotranspiration in grassland was lower than that in forests. This is probably the main cause of the increase in water retention that mainly occurred in the western and eastern BTSSR, but the decrease in water retention mainly occurred in the southeastern BTSSR (Figure 3). Similarly, approximately 95% of the newly planted forests were distributed in areas where precipitation was greater than 400 mm ( Figures S9 and S10). Thus, the per unit area carbon sink, soil retention, and sandstorm prevention improvement contributed by ERPs in areas with higher than 400 mm precipitation was much greater than that in areas with less precipitation (Figure 4a,c,d), while this change in water retention was the opposite (Figure 4b). This truth also told us that the impact of different ERPs actions on different ecosystem services is varied, and this difference varies with differences in climate background. This finding was also proven in prior studies (Griscom Bronson et al., 2017;Huang et al., 2023). In addition, we should note that the per unit area sandstorm prevention improvement contributed by ERPs in areas with >500 mm precipitation was lower than that in areas with 400-500 mm precipitation (Figure 4d). The main cause was probably the gradual decrease in wind speed and surface roughness along the precipitation gradients ( Figure S10), and this negative trend may gradually dominate sandstorm prevention change along the precipitation gradients.

| Impacts of drought on ERP-induced ecosystem services change
The reducing effect of drought on ERP-induced varied ecosystem services change is different, and the variability of the response of the different ecosystem services to drought further increased along the precipitation gradients (Figure 7). This was probably because afforestation was mainly implemented in areas with more precipitation, while grassland planting was mainly implemented in areas with less precipitation (Figure 1b; Figures S9 and S10). Drought influences the impact of ERPs on ecosystem services mainly by constraining F I G U R E 6 Spatial pattern of the difference in ecological restoration project (ERP)-induced cumulative (a1) carbon sequestration, (b1) water retention, (c1) soil retention, (d1) sandstorm prevention change under drought or not; spatial pattern of the difference in ERP-induced cumulative (a2) carbon sequestration, (b2) water retention, (c2) soil retention, (d2) sandstorm prevention change under rising CO 2 or not; spatial pattern of the difference in ERP-induced cumulative (a3) carbon sequestration, (b3) water retention, (c3) soil retention and (d3) sandstorm prevention change under drought and rising CO 2 or not.
Prior studies have already found that the impact of drought on vegetation growth in woody plants is much greater than that in grasslands in water-limited areas Jiao et al., 2021;Pan et al., 2022). Although drought reduced the water retention contributed by ERPs, its impact on afforestation and grassland planting induced water retention change varied (Figures 2 and 5). Drought not only reduced the negative effect of afforestation in decreasing water retention, but also constrained the positive effect of grassland planting in increasing water retention (Figures 2 and 5). This result is mainly due to drought weakening the water loss through evapotranspiration ( Figure S11)  Figure S10).

| Impacts of rising CO 2 on ERP-induced ecosystem services change
In contrast to drought, rising CO 2 amplified the impact of ERPs on ecosystem services change ( Figure 5). Rising CO 2 was proven to reduce stomatal conductance and accelerate photosynthesis in vegetation (Soh et al., 2019;Sun et al., 2016). The reduction in stomatal conductance would decrease evapotranspiration to promote water retention, while the improvement in photosynthesis can increase leaf area to increase carbon uptake and intercept more rainfall and dust to F I G U R E 7 The effect of drought (Dr), rising CO 2 (CO 2 ) and combined drought and rising CO 2 (Dr + CO 2 ) on ecological restoration project (ERP)-induced cumulative ecosystem services change along precipitation gradients: (a) carbon sequestration (CS); (b) water retention (WR); (c) soil retention (SR) and (d) sandstorm prevention (SP).
increase soil retention and sandstorm prevention. Thus, afforestation caused excessive evapotranspiration to be suppressed under rising CO 2 , which, in turn, reduced the water retention decline caused by it (Figures 2 and 5). Likewise, the positive effect of grassland planting in increasing water retention was further amplified under rising CO 2 (Figures 2 and 5). Similar to drought, the impact of ERP-induced ecosystem services change under rising CO 2 also varied along the precipitation gradients (Figure 7). Previous experiments found that rising CO 2 can increase leaf area in woody plants more efficiently than in grassland ecosystems (Pan et al., 2022;Piao et al., 2020). Thus, the amplification rate of rising CO 2 on ERP-induced carbon sequestration and soil retention change continued to increase along the precipitation gradients (Figure 7). Notably, the amplification rate of rising CO 2 in ERP-induced water retention change increased initially but saturated later along the precipitation gradients (Figure 7). Rising CO 2 can increase leaf water use efficiency to reduce evapotranspiration and increase water retention. However, rising CO 2 can also increase leaf area, in turn leading to more water loss through evapotranspiration (Duursma et al., 2019), offsetting the water-saving effect of CO 2 (Fatichi et al., 2016;Fiugre S11b). Prior studies even found that rising CO 2 may increase leaf area to accelerate evapotranspiration, which causes water retention to decline in arid and semiarid regions (Ukkola et al., 2016).

| Implications
Prior studies have already confirmed the effectiveness of ERPs implemented in the BTSSR in increasing carbon sequestration (Huang, Lu, et al., 2021;Liu et al., 2019;Lu et al., 2018). This study further highlights that ERPs were also efficient in increasing multiple other ecosystem services (water retention, soil retention, and sandstorm prevention). Moreover, the fact that increased ecosystem services contributed by ERPs in 2011-2020 were much greater than those in 2001-2010 ( Figure 2) indicated that the implementation of long-term ERPs would result in greater ecosystem services promotion. Prior studies have noted that future climate change tends to increase the frequency of rainstorms and wildfires (Mohamadi & Kavian, 2015;Yu et al., 2022), which would decrease soil retention (Borrelli et al., 2020; and carbon sinks (Anderegg et al., 2020). Thus, the continued implementation of ERPs may help to withstand these negative impacts. However, we should note that afforestation caused the water retention to decline, and the decline rate was further aggravated with the continued implementation of afforestation in water-limited areas ( Figure 2). Prior studies have also shown that planting trees causes soil water to decline, eventually causing trees to wilt or die (Cao et al., 2011;Huang, Liu, et al., 2021). Meanwhile, the implementation of ERPs requires a long-term strategy and huge capital investment (Bond et al., 2019;Huang, Lu, et al., 2021). These newly planted trees are also vulnerable to high severity fires and will become more so as the world warms in drylands (Bond et al., 2019). These results warn us to be cautious in implementing afforestation in water-limited areas.
Although rising CO 2 amplified the contribution of ERPs in increasing ecosystem services, this amplification effect was nearly totally offset by drought ( Figure 5). The fact that the efforts of ERPs in increasing ecosystem services declined under drought and rising CO 2 conditions implied that the negative effect from the changed climate was greater than its positive effect (Figure 7). Likewise, prior studies also concluded that the impact of climate change on most types of services was predominantly negative (Runting et al., 2017). Although drought decreased in 2011-2020, reducing its negative impact on ERP-induced ecosystem services change, prior studies forecasted that drought frequency, duration, and intensity would further increase in the future (Satoh et al., 2022;Yin et al., 2023). Our previous study found that the effort of ERPs and rising CO 2 contributed to carbon sinks totally offset drought-caused carbon loss (Huang, Lu, et al., 2021), this study further demonstrated that the ecosystem services contributed by ERPs were modulated by rising CO 2 and drought. This implies that it is necessary to take the effect of rising CO 2 and drought into consideration when quantifying the contribution of ERPs to ecosystem services change.
Optimizing restoration strategies is helpful to improve ecosystem resilience to decrease the negative impacts from the changed climate environment (Isbell et al., 2015;Smith et al., 2022). Measures including choosing suitable areas for afforestation (Zhang, Gentine, et al., 2022), mixed planting with understory recovery , widening the planting interval, and choosing highly drought-tolerant trees (Brancalion & Chazdon, 2017;Feng et al., 2022;Hua et al., 2022) would better withstand the negative impact of global climate change.
Except for the four ecosystem services we assessed in this study, other overlooked ecosystem services including pollination, temperature regulation, and food production can also be affected by ERPs and modulated by climate change (Noack et al., 2019;Peng et al., 2014;Runting et al., 2017). Focusing on these overlooked ecosystem services in future studies is helpful to strengthen the understanding of the pros and cons of ERPs in the global change environment. were combined, their joint effect reduced ERP-induced carbon sequestration, water retention, soil retention, and sandstorm prevention by 5.74%, 32.62%, 11.74%, and 14.86%, respectively. This study confirmed that ERPs are efficient in enhancing ecosystem services provision. Furthermore, the finding that drought and rising CO 2 modulated the contribution of ERPs to ecosystem services change indicated that the impact of these factors on ERPs is unmissable.

| CON CLUS ION
Thus, the implementation of ERPs should be optimized to reduce or even avoid negative climate change impacts. Program of Hebei Province Grants 21373902D.

CO N FLI C T O F I NTE R E S T S TATE M E NT
The authors declare no competing interests.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are derived from the following resources. Meteorological data were collected from