Recent climate change strengthens the local cooling of European forests

Recent climate change strengthens the local cooling of European forests | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Physical Sciences - Article Recent climate change strengthens the local cooling of European forests Zhao-Liang Li, Yitao Li, Jun Ge, Hua Wu, Ronglin Tang, Yuanliang Cheng, and 11 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5281378/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 25 Sep, 2025 Read the published version in Nature Communications → Version 1 posted You are reading this latest preprint version Abstract Forests exhibit cooling or warming effects compared to adjacent openlands through biophysical processes. The local temperature effects are predicted by earth system models to evolve in response to climate change. However, these temporal dynamic patterns remain unconstrained by observations and have not been detected in historical records. Here, we provide satellite evidence of emergent negative trends in local land surface temperature (LST) effects of European forests from 2003–2023. The daytime cooling effects have significantly intensified in both winter (-0.17 K/decade) and summer (-0.22 K/decade). The enhanced winter cooling is attributed to the reduced shortwave radiative forcing in forests due to decreasing snow cover. In the summertime, the vegetation physiological response to increased atmospheric vapor pressure deficit boosts evaporative cooling in forests. The negative trends in LST effects of European forests are roughly supported by four state-of-the-art earth system models. However, considerable biases and intermodel spread in the representation of underlying biophysical processes. Given the continued climate change, we emphasize the need to consider their impacts on biophysical effect dynamics when comprehensive forest-related climate mitigation policies are formed. Earth and environmental sciences/Climate sciences/Climate change/Climate-change impacts/Environmental health Earth and environmental sciences/Ecology/Forest ecology Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Introduction Forests provide numerous benefits to humans and the planet, such as producing food and energy, reducing soil erosion, and increasing water availability in downwind areas 1 , 2 . Through conservation, proper management, and restoration practices, forest ecosystems play a crucial role in addressing global climate warming 3 , 4 . Current evidence supporting forest-related climate mitigation is primarily based on the biochemical process of carbon sequestration 5 – 8 . However, forests can impact the climate system through other processes that are not comprehensively considered in current mitigation or adaptation strategies. For instance, forests release biogenic volatile organic compounds, which affect local or non-local radiative forcings 9 . Compared with other ecosystems, forests also present unique biophysical characteristics (e.g., albedo, aerodynamic roughness and evaporation efficiency), which can reshape surface energy and water balance processes 10 . The biophysical process of forests has garnered particular attention in recent years, as it can significantly intensify or offset the global climate mitigation effects of carbon sequestration 11 , 12 . Using numerical models or muti-scale observations, previous studies have thoroughly investigated the impacts of re/afforestation, deforestation, and forest degradation on local land surface temperature (LST) and their biophysical mechanisms 10 , 13 – 15 . There is broad consensus that forests can have two direct effects with opposing signs compared to non-forest vegetation: (a) land surface cooling due to higher evapotranspiration (ET) rates 16 and aerodynamic roughness, and (b) land surface warming due to lower albedo 17 . The sign and magnitude of the temperature signal depend on the counterbalancing of the albedo-related radiative process and the turbulence-related non-radiative process 18 . The relative dominance of these two processes is fundamentally determined by the background climate 19 , leading to varying temperature effects of forests with different latitudes 20 , seasons 21 or elevations 22 . Forests are usually warmer than surrounding openlands in cold environments because bright snow covers low vegetation but is masked by dark forest canopies 23 . Conversely, forests exhibit cooling effects under warm and wet environments because their surface roughness and transpiration rates are larger than those of other vegetation types 24 . Additionally, arid conditions may limit the water availability and subsequent evaporative cooling effects of forests, making the radiative process dominant 25 . Given that the background climate profoundly impacts the biophysical processes of forests 19 , it is reasonable to infer that, in the context of global change, the local temperature effect of forests may change correspondingly. In recent years, studies have revealed the role of rising atmospheric CO 2 , changing aerosols or internal climate variability in the temporal dynamics of such biophysical temperature effects 26 – 29 . However, these findings are based on model simulations, which are subject to inaccurate physical parameterization schemes and representations of biophysical processes in forests 30 , 31 . Besides, the prior simulations only show significant changes in the biophysical temperature effects under high-emission or other ideal scenarios 26 – 29 . On the other hand, observational studies that assess the biophysical effects of forests are generally performed in a relatively static manner 15 , 18 , 21 . These studies focused on spatial, seasonal or diurnal heterogeneity in the mean state of the biophysical effects over a specific year or study period but ignored the potential long-term variations. Overall, direct observational evidence of temporal dynamics in local temperature effects of forests is still lacking in historical records. Evaluating these dynamic patterns and their driving mechanisms is useful for constraining model results and forming better global mitigation or regional adaptation policies. This study aims to fill the knowledge gap regarding the temporal variations (especially trends) in the local land surface temperature (LST) effects of forests and the underlying biophysical mechanisms. We focus on Europe for the following reasons: (1) Both radiative and non-radiative processes play important roles in the local LST effects of European forests, facilitating the exploration of different biophysical mechanisms underlying the long-term variation of the forests’ LST effects. (2) Thanks to the dense distribution of weather stations, high-quality gridded meteorological data in Europe can help to explore potential links between changes in biophysical processes and the background climate. (3) Europe is a hotspot of climate change and forestation projects (3 billion trees to be planted in the European Union by 2030 according to the European Green Deal 32 ), making our assessment more relevant and instructive. Our study aims to answer the following three questions: (1) What are the temporal patterns of the observed local LST effects of European forests in recent two decades? (2) What climate factors drive these temporal patterns? (3) What is the biophysical mechanism behind this climatic control? To address these questions, we first use observations from the Moderate Resolution Imaging Spectroradiometer (MODIS) and auxiliary satellite data to estimate the local LST effects of European forests. We then assess the temporal dynamics of the local LST effects and explore their relationships with different biophysical processes and potential climate drivers using statistical methods. To examine whether models can capture the observed temporal patterns, we evaluate the simulated trend of local LST effects of European forests in four state-of-the-art earth system models against satellite observations. Our results could inform assessments of climate mitigation or adaptation effects through future forestation in Europe. Results and Discussion 3.1 Intensified local daytime cooling effects of forests We first investigate the multi-year (2003–2023) winter and summer LST effects of European forests (ΔLST w and ΔLST s ) (Supplementary Figs. 1 to 2). Compared with nearby openlands, forests have a slight warming effect in winter (0.14 ± 0.61 K, spatial mean ± standard deviation) due to the counterbalancing of the nighttime warming signal (0.39 ± 0.55 K) and the daytime cooling signal (-0.20 ± 1.05 K). In contrast, forests exhibit intense land surface cooling in summer (-1.14 ± 0.81 K), as the strong daytime negative LST effect (-3.16 ± 1.69 K) is partially offset by the nighttime positive effect (0.34 ± 0.52 K). Our results confirm previous conclusions of the contrasting LST effects of mid-latitude forests at seasonal and diurnal scales driven by different biophysical processes 20 , 21 , 33 . We then focus on the temporal variations in ΔLST w and ΔLST s over the recent two decades. Results show the declining of winter warming effects at the daily scale, with a trend of -0.06 K/decade ( p = 0.10) (Fig. 1 a). This decrease in the daily mean ΔLST w is dominated by a negative trend of the daytime ΔLST w (-0.17 K/decade, p = 0.04), while the nighttime ΔLST w demonstrates only interannual variability with a negligible temporal trend (0.02 K/decade, p = 0.65) (Fig. 1 a). Spatially, 75% of grids show negative daytime ΔLST w trends, with 12% of grids, which are mainly concentrated in southeastern Europe, showing significant trends ( p < 0.05) (Fig. 1 b). During the summer nighttime, the warming effect of forests shows an increasing trend of 0.05 K/decade ( p = 0.06), but the magnitude is only about one-fifth of the daytime ΔLSTs negative trend (-0.22 K/decade, p = 0.04) (Fig. 1 c). As a result, the daily mean cooling effects of forests have been enhanced in the recent two decades, with a trend of -0.08 K/decade ( p = 0.07) (Fig. 1 c). Spatially, about 83% of grids show negative daytime ΔLST s trends, and 18% of grids (mainly distributed in central Europe) show significant negative trends ( p < 0.05) (Fig. 1 d). These results show that in both summer and winter, the daytime cooling effects of forests were significantly enhanced, leading to negative trends of daily mean ΔLST s and ΔLST w . Further analyses mainly focus on the biophysical mechanisms behind the negative trends in daytime temperature effects. 3.2 Impact of snow cover on daytime ΔLST w dynamics The local LST effects of mid-latitude forests in winter are generally dominated by the radiative process 2 , 10 , 18 . Here, we hypothesize that the variation of the forest’s albedo effect drives the observed daytime ΔLST w dynamics (Fig. 1 a). We first examine the observed albedo difference between forests and nearby openlands (Δα w ) and confirm that forests are darker than openlands, as evidenced by the negative Δα w (Supplementary Fig. 3), but Δα w show a positive trend (0.031 unitless/decade, p = 0.02) (Fig. 2 a). Spatially, Δα w trends are positive in most areas, and about 32% of grids (concentrated in central and eastern Europe) are statistically significant ( p < 0.05) (Fig. 2 b). These results indicate that the forest darkening effect has become weaker and the additional solar radiation absorbed by forests (compared to that absorbed by openlands) has decreased over the last two decades. The aforementioned positive Δα w trend can be traced back to the stronger albedo decrease in openlands (-0.047 unitless /decade, p = 0.02) than in forests (-0.015 unitless /decade, p = 0.04) (Fig. 2 a). These surface darkening trends can be attributed to the reduced snow cover (SC w ) under the global warming trend 34 – 36 (Supplementary Fig. 4). Here, we confirm that the surface darkening is more susceptible to snow cover decrease in openlands than forests, as evident by the larger spatial regression slope value between α w trends and SC w trends (0.59 vs 0.21, Fig. 2 c). This is because snow tends to be masked by tree canopies in forest ecosystems, resulting in a lower sensitivity of forest albedo to snowpack. The divergent responses of albedo to SC w are also supported by the temporal regression results, which suggest that openland albedo is about 3 times more sensitive than forest albedo to SC w (Supplementary Fig. 5). Notably, the used SC w is a function of snow mass (represented by the snow water equivalent) 37 , which is more governed by the macroclimate and thus shows no difference between forests and nearby openlands. The positive Δα w trend induced by snow dynamics further exhibits strong temporal control on ΔLST w . Specifically, in those grids with more pronounced SC w reductions, the Δα w trends is larger and the negative daytime ΔLST w trends are also stronger (Fig. 2 c and d). Meanwhile, the temporal relationships between SC w and ΔLST w suggest that in those years with lower snow contents, the daytime cooling effect of forests is stronger (slope = 1.47 ± 0.59), resulting in a weaker daily warming effect (slope = 0.73 ± 0.29) (Fig. 2 e). However, the impact of SC w on the nighttime ΔLST w is insignificant (slope = 0.19 ± 0.28) (Fig. 2 e). In addition to Δα w , the shortwave radiative forcing of forests is related to the downward solar radiation in winter (DSR w ). However, we show a poor correlation between DSR w and ΔLST w (Supplementary Fig. 6). Overall, our statistical evidence supports the temporal control of SC w on daytime ΔLST w dynamics through weakening the albedo difference between forests and openlands. 3.3 Dominant role of vapor pressure deficit in daytime ΔLST s dynamics In contrast to the net warming effect observed in winter, European forests exhibit strong daytime and daily mean cooling effects during the summer (Supplementary Fig. 2), driven by non-radiative processes 18 . Over the past two decades, Europe has experienced significant increases in summer air temperature (AT s ) and atmospheric vapor pressure deficit (VPD s ). While there has also been an increase in summer downward solar radiation (DSR s ) and decreases in precipitation (P s ) and wind speeds (WS s ), the trend values for these variables are not significant ( p > 0.1) (Supplementary Fig. 4). We hypothesize that climate change has distinct effects on turbulent fluxes of forests and openlands, thus contributing to the temporal dynamics of ΔLST s . Temporal correlation analyses of interannual variabilities reveal a strong relationship between the daytime ΔLST s and four potential climatic drivers at the grid or regional scales: VPD s (mean r = -0.68), DSR s (mean r = -0.52), P s (mean r = 0.55), and AT s (mean r = -0.55) (Fig. 3 a, Supplementary Fig. 7). After accounting for covariates, the partial correlations between the daytime ΔLST s and DSR s , P s , and AT s are greatly reduced and sometimes even reverse in sign (Fig. 3 a). VPD s , however, maintain a predominant negative correlation with the daytime ΔLST s , with a spatial mean partial correlation coefficient of -0.43 (Fig. 3 a). About 91.3% of grids show a negative partial correlation between the VPDs and daytime ΔLST s , with 44.8% being statistically significant ( p < 0.05) (Fig. 3 b). On the basis of the maximum values of the absolute correlation coefficients, VPD s is identified as the most important driver of the daytime ΔLST s in most grids (Supplementary Fig. 8). We then built a random forest (RF) model using elevation and multi-year meteorological data to predict the daytime ΔLST s at the grid scale (total of 19,656 samples). We excluded AT s from the model inputs due to its high correlation with VPD s . The model shows good accuracy on the test dataset (R 2 = 0.75, RMSE = 0.79 K, Supplementary Fig. 9), supporting our further analysis. The Gini importance and mean absolute Shapley Additive Explanations (SHAP) values both suggest that VPD s is the most important driver, followed by elevation, DSR s , P s , and WS s (Fig. 3 c). The marginal contributions quantified by the SHAP value further reveal different impacts of climate variables (Fig. 3 d, Supplementary Fig. 10). VPD s exhibits the most pronounced linear negative effect, while DSR s and P s show complex non-linear effects. Specifically, under low-radiation conditions (DSR s < 200 W/m 2 ), the influence of DSR s is negligible; under moderate-radiation conditions (200 W/m 2 < DSR s < 250 W/m 2 ), DSR s shows a negative effect; under high-radiation conditions, the impact of DSR s is highly uncertain. For P s , we find a positive effect only when P s < 300 mm. On the basis of the RF model, we estimate the contributions of climate variables to the long-term trend of daytime ΔLST s . We first compared the reconstructed trends with all forcings and the observed trends, showing good consistency at the grid scale (R 2 = 0.64, RMSE = 0.11 K/decade, Supplementary Fig. 9). For the regional mean, the simulated trend is -0.26 K/decade ( p < 0.05), which is close to the observed trend of -0.22 K/decade ( p < 0.05) (Fig. 1 c). Model experiments suggest that increasing VPD s contributes the most to the decreasing trend in daytime ΔLST s (-0.17 K/decade, p = 0.06), while P s (-0.02 K/decade, p = 0.32), DSR s (-0.02 K/decade, p = 0.34), and WS s (0.02 K/decade, p = 0.65) show insignificant contributions (Fig. 3 e). To understand the biophysical mechanisms behind the influence of climate changes, especially VPD s changes, on the temporal dynamics of daytime ΔLST s , we further applied structural equation modeling (SEM) analysis at the local scale (Fig. 4 a). We adopt differences in the leaf area index (LAI) and latent heat (LE) between forests and nearby openlands (ΔLAI s and ΔLE s ), two key factors related to the biophysical effects of forests, as intermediaries. Notably, we set ΔLAI s and ΔLE s as intermediaries rather than potential drivers because their variation can also be essentially attributed to distinctive responses of different vegetation types to climate changes. Results show that climate changes show different effects on forest and non-forest canopy structures: higher VPD s (0.42 ± 0.18, path value ± standard error), AT s (0.29 ± 0.15), or DSR s (0.14 ± 0.12) enhance ΔLAI s , while P s has a negative effect on ΔLAI s (-0.13 ± 0.10). The resultant ΔLAI s further demonstrate a strong positive impact on ΔLE s (0.74 ± 0.05). Moreover, climate changes can directly affect ΔLE s , with positive effects resulting from VPD s (0.42 ± 0.10) and WS s (0.07 ± 0.05) changes and negative effects resulting from AT s (-0.24 ± 0.09), DSR s (-0.16 ± 0.08), and P s (-0.14 ± 0.06) changes. ΔLE s show a predominantly negative effect on daytime ΔLST s (-0.73 ± 0.10) compared to the small direct effect of ΔLAI s (0.06 ± 0.10), suggesting that it is evaporative cooling, rather than roughness- or albedo-related processes (related to the vegetation canopy structure), dominates the temporal dynamics of ΔLST s . Each driver could affect ΔLST s via three pathways through ΔLAI s and ΔLE s , and we calculate the individual and summed pathway effects on the daytime ΔLST s for all the climate variables (Fig. 4 b). Results show that pathways only through ΔLAI s are insignificant for all drivers (yellow bar). The summed negative effect of VPD s is the strongest in magnitude (-0.51 ± 0.14). The two pathway effects of VPD s through ΔLE s are comparable (via ΔLAIs: -0.23 ± 0.10; not via ΔLAIs: -0.30 ± 0.08). For P s , the two path effects through ΔLE s are positive, with a summed path effect of 0.17 ± 0.07. For DSR s and AT s , the two path effects have different signs, resulting in slight total effects. Overall, our partial correlation analysis, RF analysis, and SEM analysis all indicate the predominant role of VPDs in the dynamics of daytime ΔLST s , which is achieved through two pathways affecting the evaporative cooling effect of the forest. 3.4 Evaluation of ΔLST s and ΔLST w trends in four CMIP6 models To examine whether the observed trends of the biophysical LST effects of European forests can be reproduced by earth system models (ESMs), we use the output from the simulation of the historical period in the Coupled Model Intercomparison Project Phase 6 (CMIP6) archive. The simulated effect of forests is isolated by a similar method for observational data, using differences between the forest and opeland tiles (forest minus openland) within each model grid cell (see Methods) in four state-of-art ESMs that provide subgrid information (CESM2, CNRM-ESM2-1, GFDL-ESM4 and UKESM1-0-LL). The multi-model mean suggests widespread declining trends of daily mean ΔLST w (-0.068 K/decade, p < 0.01) and ΔLST s (-0.054 K/decade, p < 0.01) over Europe during 1985–2014 (Fig. 5 a, d). These trends largely align well with the observed ones (Fig. 1 a, c), and the discrepancy in magnitudes can be partly attributed to differences in the period and spatial coverage of the observation and simulations (Supplementary Figs. 11 and 12). We should emphasize that the current models perform better in simulating negative trends in ΔLST w than in ΔLST s , as evidenced by the large intermodel differences in ΔLST s trends simulated by four models. Specifically, the ΔLST w trend values are similar across the four models (Fig. 5 a). While for ΔLST s , CESM2 simulates a significant declining trend of -0.133 K/decade, which dominates the muti-model mean signal. However, CNRM-ESM2-1, GFDL-ESM4 and UKESM1-0-LL simulate much weaker trends of -0.042, -0.018, and − 0.027 K/decade, respectively (Fig. 5 d). In addition, differences are found in the multi-year mean ΔLST w or ΔLST s from the four models (Fig. 5 a, d), which can be attributed to the biases in surface energy partitioning and albedo responses to land cover changes 38 . We further evaluate the model representation of biophysical processes behind the negative trends in ΔLST w . Observational results indicate that the decreasing trend in the openland albedo is 3.13 times greater than that in the forest albedo, whereas all four models underestimate this ratio, with the muti-model mean of 1.94 (Fig. 5 b). Moreover, we show that only CESM2 (-2.41 ± 0.70) and UKESM1-0-LL (-2.93 ± 2.22) can approximately reproduce the observed slope (2.42 ± 1.01) between Δα w and ΔLST w (Fig. 5 c). CNRM-ESM2-1 (-6.29 ± 1.18), GFDL-ESM4 (-12.37 ± 3.05) and the multi-model mean (5.57 ± 1.39) significantly overestimate this slope (Fig. 5 c). These may lead to bias in estimations of ΔLST w variation in response to climate changes, especially the snow cover decline. For the summer results, we conduct partial correlation analysis between ΔLST s and climate variables (Fig. 3 a). The wind speed is excluded here considering the model data availability. Among all four models, only CNRM-ESM2-1 captures the dominant role of VPD s on ΔLST s dynamics and reproduces the impacts of other climate variables (Fig. 5 e). Overall, although negative trends in ΔLST w and ΔLST s can be captured by the models, there are still discrepancies between the observations and simulations as well as intermodel differences in the mechanisms underlying the biophysical effect of forests. Our results are informative for improving current models for better simulations of local climate effects of European forests and their temporal dynamics in a warming world. 3.5 Predicted local LST effect of European forests under four SSP scenarios In the recent two decades, Europe has experienced a remarkable warming trend, accompanied by stronger atmospheric drought in summer and less snow cover (Supplementary Fig. 4). Since the warming trend is unlikely to be reversed until carbon neutrality, we anticipate that the negative trends in ΔLST s and ΔLST s will persist. Here, we project the daily mean ΔLST s and ΔLST s under four SSPs (SSP126, SSP245, SSP370, and SSP585) by combining the simulated future climate changes, as well as the established relationships between climate drivers and satellite ΔLST w and ΔLST s in historical records (see Methods). We then compare the predicted trends of ΔLST w and ΔLST s with the future trends of winter LST (LST w ) and summer LST (LST s ) within two periods (2020–2060 and 2060–2100). We calculate the ratio of ΔLST w to LST w and ΔLST s to LST s to assess the future relative magnitude of the biophysical mitigation of European forests. Over 2020–2060, the negative trends in ΔLST w and ΔLST s persist, generally with stronger trends in warmer SSPs (Fig. 6). In winter, the decreasing trends in ΔLST w can offset -2.9 ± 1.9%, -5.3 ± 2.3% and -4.5 ± 2.6% of the LST warming trends under the SSP245, SSP370 and SSP585 scenarios, respectively (Fig. 6c, e and g). For the SSP126 scenario, however, the negative trend in ΔLST w and its mitigation effect are insignificant (Fig. 6a). In summer, the ratios of ΔLST s trends to warming are more pronounced, with values of -8.4 ± 3.4%, -9.1 ± 2.7%, -8.8 ± 2.4% and -9.9 ± 2.0% under the SSP126, SSP245, SSP370 and SSP585 scenarios, respectively (Fig. 6b, d, f, g). Over 2060–2100, trends in ΔLST w and ΔLST s are found to be insignificant in the SSP126 scenario, as the climate change slows down due to reduced emissions. For the remaining three scenarios, the negative trends in ΔLST w are weakened compared with those from 2020–2060, and the ratios are significant only under the SSP370 and SSP585 scenarios (-2.2 ± 1.4% and -2.2 ± 0.6%, respectively) (Fig. 6c, e and g). This could be attributed to lower SC w sensitivity to temperature under warmer conditions (Supplementary Fig. 13), leading to weaker ΔLST w trends in the warming world. For ΔLST s , the magnitudes during 2060-2100 are close to those during 2020-2060, with ratios of -8.8 ± 6.4%, -11.1 ± 3.0% and -11.7 ± 1.6% under the SSP2-4-5, SSP370 and SSP585 scenarios, respectively (Fig. 6d, f, g). Overall, the roles of negative biophysical LST trends in LST mitigation are projected to be stronger in summer than in winter. Notably, the revealed mitigation effects arise from varying biophysical feedbacks to climate change in existing stable forests. The biophysical cooling of forest ecosystems may be stronger (especially in summer with negative ΔLST s ) if more land areas are forested in the future. 3.6 Discussion Previous observational studies only focused on the spatial, diurnal or seasonal heterogeneity of biophysical LST sensitivity to forest cover change or the mechanisms behind these patterns 10 , 15 , 21 , 39 . These studies are performed in a relatively static manner, using the results of a single year or the mean of a period, ignoring potential annual temporal variability. Here, our satellite evidence shows the emergent trend in the biophysical temperature effects of European forests over the last two decades (Fig. 1 ). Our results suggest that previous observational assessments may underestimate the daytime cooling effect of forests in the changing world. At least in the European region, the temporal dynamic of forest biophysical effects should be taken into consideration. Meanwhile, prior simulations mostly reported responses of the biophysical effects of forests to climate changes in high-emissions or ideal scenarios 26 , 27 , 40 . However, we have detected robust trends in the biophysical effects of forests in historical records, despite the limited climate change during the study period. This implies that the sensitivity of the biophysical effect to climate change may be much stronger than previously thought, which merits the attention of researchers in the fields of climate change and earth system modeling. Since we estimate the LST effects based on fixed and stable forest and openland pixels with minimum disturbance, these negative trends can be solely attributed to background climate changes in Europe. We link the temporal change of the winter LST effect to the radiative process, represented by Δα w , and further attribute the negative ΔLST w trend to the decreasing SC w in Europe. Specifically, the albedo of openland is about three times more sensitive to SC w than that of forests. Thus, decreasing SC w in the recent two decades has weakened Δα w (Fig. 2 c), and, consequently the radiative forcing of European forests. As a result, the relative contribution from non-radiative processes (i.e., higher heat exchange efficiency in the rougher forested land surface) may become more dominant, leading to stronger daytime cooling and attenuated daily warming effects of forests in European winter (Fig. 1 a, b). This result suggests that the transitional latitudes of current forest cooling and warming effects will migrate northwards in a warming world 10 . The obtained relationships between SC w and biophysical factors of forests are useful for making reasonable predictions or constraining model simulations of future forest-snow interactions. In addition, existing studies have suggested that high latitudes and altitudes are considered unsuitable for forest restoration, as the positive radiative forcing driven by the albedo effect surpasses the negative radiative forcing driven by the carbon effect 4 , 41 . Our satellite evidence implies that the timing of tree growth and future climate change (especially the snow amount) should be fully considered when prioritizing the geolocation of forestation practices. In the context of global warming, the snow cover in winter or spring is projected to maintain a decreasing trend, which may significantly reduce the biophysical radiative forcings of forests and make high-latitude and high-altitude regions becoming suitable for forestation. In summer, the biophysical mechanisms behind enhanced forest cooling are more complicated than those in winter. Previous studies have shown that the summer cooling effect of forests is dominated by non-radiative processes 18 . The evaporative cooling effect (represented by ΔLE s here) plays the most important role 21 , which is influenced by both vegetation structure differences in forests and openlands (ΔLAI s ) and meteorological conditions 42 . Using a range of statistical methods, we find that increasing VPD s is the predominant factor contributing to both the interannual variation and the negative trend in ΔLST s (Fig. 3 ). Path analysis shows that VPD s affects ΔLST s dynamics through two main pathways (Fig. 4 ). First, a positive effect of VPD s on ΔLAI s implies that, as the atmosphere becomes drier, the vegetation structure difference between forests and openlands becomes more evident. The higher ΔLAI s could further stimulate ΔLE s and boost the cooling efficiency of forests. The impact of VPD s on vegetation has been well-documented: increasing vapor pressure deficit (VPD) could trigger stomatal closure, inhibit photosynthesis, and increase vegetation mortality 43 , 44 . Our results suggest a stronger negative effect of VPD on non-forest vegetation than on forests, which is consistent with previous findings of heterogeneity in the vegetation response to VPD 45 , 46 . The potential reason is that forest ecosystems with high species richness and deep roots are more stable and resilient to drought 47 , 48 . Second, even under constant vegetation conditions, higher VPD s directly amplifies ΔLE s and enhances the evaporative cooling effect of forests. ET in Europe is generally driven by the atmospheric demand rather than the water supply. Thus, ET tends to increase in response to higher VPD 49 . Conversely, plants could also reduce water loss through stomatal closure with increasing VPD 50 . Our results support previous conclusions concerning the diverse effects of VPD on ET in different vegetation types 50 , 51 and suggest that the atmospheric water demand-driven positive effect overwhelms the stomatal conductance-driven negative effect. Overall, the revealed impact of VPD s on ΔLST s essentially reflects the diverse vegetation physiological responses to climate changes. Through these two pathways, the rising VPD s in the recent two decades positively affected the evaporative cooling effect of European forests and drove the observed negative trend in the daytime ΔLST s . In addition to LST, the biophysical effects of European forests also affect the hydrological process. In winter, forest cover can delay or accelerate the melting of snowpack compared to the adjacent openlands, depending on the background climate and canopy density 52 . However, the revealed increasing daytime cooling effect favors the retention of winter snow in forests and further increases streamflow from snowmelt in the following season. As a result, forests may buffer the impact of global warming on fresh-water availability 53 , plant phenology 54 and food production 55 through altering the snow melting process in Europe. During summer, forests maintain high ET, which further promotes precipitation in downwind areas and accelerates the water cycle process 56 . The enhanced cooling effect is achieved at the expense of water loss through transpiration, which may increase the pressure on terrestrial water availability and pose threats to ecosystem productivity 57 and freshwater resources 58 . The vegetation-climate feedback can promote precipitation through local moisture recycling and thus have a positive effect on water availability, which can offset the direct negative effect of ET on water availability 59 . Considering the drying or wetting trend in the future, the impact of potential forestation in Europe should be discussed according to the specific background climatic conditions due to the trade-off between greater cooling effects and less water availability 60 . There are several potential caveats or issues when interpreting our results. First, the adopted “space-for-time” method provides priori estimates of local temperature effects 61 , which ignore the atmospheric feedback of potential afforestation (e.g., cloud formation 62 and atmospheric circulation 63 ), as well as the asymmetric patterns in the biophysical effects of forest gain and loss 64 , 65 . Second, the projected stronger forest local cooling effect in the future implies that planting trees in proper areas remains a promising local solution against the risk of warming (Supplementary Fig. 13), especially in highly populated regions. Nonetheless, tree restoration is not a panacea for climate change. Although our results highlight the stronger negative biophysical feedback to the warming trend in forests, this negative feedback is impossible to fully offset the warming trend driven by rising atmospheric CO 2 (Supplementary Text 1 and 2). Therefore, reducing greenhouse gas emissions and developing clean energy remain fundamental to limiting climate change. Third, while we reveal the influences of changes in climate variables on the biophysical effects of forests, the role of increasing atmospheric CO 2 is not considered in this study. The CO 2 fertilization effect can enhance vegetation leaf area 66 , boost transpiration 67 and cool the land surface 68 . Meanwhile, the CO 2 physiological forcing also stimulates stomatal closure 69 , reduces transpiration 70 and increases local temperature 71 . These two opposing mechanisms have diverse effects on forest and non-forest vegetation and affect forest cooling efficiency. The stomatal closure-driven negative effect may overwhelm the LAI-driven positive effect 72 . Thus, ignoring the physiological response of vegetation may result in the overestimation of the forest cooling effect (especially in summer) under high atmospheric CO 2 concentration conditions. Finally, the evaluated LST is the thermal radiative temperature of the land (vegetation foliage, canopy, and soil), which is biologically relevant and has advantages in describing energy, water, and carbon fluxes 73 . However, in climate change assessments, the screen height air temperature, rather than the LST, is a more widely used metric in terrestrial regions 74 . Converting LST effects to corresponding air temperature effects is of great importance but remains challenging. Nonetheless, our LST-based results are useful for model improvement and informing the increasing climate benefits of European forests. Overall, our findings provide solid satellite evidence that the local LST effects of European forests have varied with climate changes and have shown significant negative trends in the last two decades. This temporal variation pattern may be even more evident when different forest age dynamics are considered 75 . Our results challenge the previous view that biophysical effects are marginally time-dependent in the recent two decades and advocate that temporal dynamics should be considered in biophysical effect assessments. Moreover, the revealed dominant role of snow cover and atmospheric moisture deficit in controlling biophysical sensitivity emphasizes the importance of accurate parameterization of related processes in land surface models or vegetation dynamics models (e.g., snow masking effects and climatic constraints on vegetation leaf area), which are directly relevant to the projected future climate effects of forests and forming forest-based climate policies. Materials and methods 4.1 Satellite data Satellite observations are adopted to estimate the potential biophysical effects of forestation. The satellite data used in this study are divided into two categories: MODIS products and other auxiliary land surface data. All satellite data are processed and outputted with a spatial resolution of 1 km (0.0083°). LST data are obtained from the MYD11A1 dataset, which is retrieved using the split-window algorithm 73 . The MYD11 LST provides daily instantaneous observations at approximately 1:30 PM (daytime) and 1:30 AM (nighttime), corresponding to the daily maximum and minimum temperatures. Observations with LST errors greater than 2 K are filtered out according to the quality control flag. All high-quality observations are used to aggregate the summer (June, July, and August) and winter mean LST (December, January, and February) for both daytime and nighttime. We generate summer and winter LST from 2003 to 2023. It is worth noting that winter LST for a given year is averaged from the January and February LST of that year and the December LST of the previous year. For example, the 2003 winter LST is the mean value from December 2002 to February 2003. This aggregation principle is applied to all seasonal variables. We calculate the daily mean LST of summer and winter through the weighted combination of daytime and nighttime LST for both seasons 76 . Specifically, the weight for daytime and nighttime LSTs are 0.5637 and 0.4244, respectively, and the constant term is 2.75 K. LE data are derived from the MOD16A2GF (MOD16A2 Gap-filled) dataset, with an original resolution of 8-day and 500 m. We select all observations retrieved by the main algorithm (Penman-Monteith model) and convert the LE unit to W/m². Then, all available records are spatially and temporally aggregated to 1 km summer means from 2003 to 2023. Leaf area index (LAI) data are obtained from the MCD15A3H dataset with a 4-day and 500 m resolution. All LAI values retrieved by the main Look-up-Table (LUT) method are aggregated to 1 km summer means from 2003 to 2023. The albedo data are from the MCD43A3 dataset, which contains daily 500 m black-sky and white-sky albedo for the shortwave band. We average the black-sky and white-sky albedo and then aggregate them to 1 km winter means from 2003 to 2023. Yearly land use/land cover data from 2003 to 2022 are derived from the MCD12Q1 dataset, with an original spatial resolution of 250 m. The land cover results under the International Geosphere-Biosphere Program (IGBP) classification scheme are used. The 250 m land cover maps are spatially aggregated to a 1 km scale based on the majority type. Other auxiliary land surface data include forest loss data from the Global Forest Change (GFC) dataset 77 , impervious surface data from the Global Artificial Impervious Area (GAIA) dataset 78 , and digital elevation model (DEM) from the GMTED2010 dataset. The GFC dataset provides annual forest loss masks from 2000 to 2023 with 30 m spatial resolution based on time series analysis. All layers from 2003 to 2023 are composited, and a mask of forest loss during the study period is generated. The composited map is then spatially aggregated to 1 km resolution to match the MODIS data. The aggregated values range between 0 to 1, indicating the proportion of 30 m pixels where forest loss has occurred. GAIA data document the impervious surface mask from 1985 to 2018 with a spatial resolution of 30 m. Similar to the GFC datasets, the data from the last year (2018) are aggregated to a 1 km resolution impervious surface percentage map for further analysis. The DEM data of GMTED2010 has a spatial resolution of approximately 7.5 arc-seconds. We aggregate the original DEM to a 1 km spatial resolution for further analysis. 4.2 Reanalysis data Meteorological variables are obtained from the monthly ERA-5 Land reanalysis dataset 79 with a spatial resolution of 0.1°. ERA5-Land datasets have been widely used for analysing the interaction between climate and terrestrial ecosystems. Here, several variables, including 2 m air temperature (AT), 2 m dewpoint temperature (DT), precipitation, downward shortwave radiation, wind speed and snow cover, are downloaded. We aggregate these climatic variables into winter and summer means with a spatial resolution of 0.5°. AT and DT were used to calculate the atmospheric vapor pressure deficit (VPD) using the following equations 43 : $$\:VPD=e\left(AT\right)-e\left(DT\right)$$ 1 $$\:e\left(T\right)=6.112\cdot\:{f}_{w}\cdot\:exp\left(\frac{17.67\cdot\:T}{T+243.5}\right)$$ 2 $$\:{f}_{w}=1.0007+3.46\times\:1{0}^{-6}\cdot\:AP$$ 3 $$\:AP=1013.15{\left(\frac{T+273.16}{T+273.16+0.0065\cdot\:Z}\right)}^{5.625}$$ 4 where AT and DT indicate the 2 m air temperature and dewpoint temperature (Celsius), respectively; \(\:e\left(T\right)\) indicates the saturation water vapor (hPa) at a given temperature T (Celsius); and AP denotes the air pressure (hPa); Z indicates elevation (m). 4.3 Observed biophysical effects of forests on LST This paper explores the temporal dynamics of forests’ biophysical processes under a changing climate. To ensure a sufficient sample size for robust results, we employ a space-for-time substitution strategy to estimate the a priori potential effect, rather than using the space-and-time approach for the a posteriori actual effect 61 , 80 . Specifically, we assume that all 1 km pixels within a 0.25° grid (comprising 30×30 pixels) share the same background climate 21 , and the potential effect of forests on the LST (ΔLST) can be estimated by the difference between the mean LSTs of forested areas ( \(\:\stackrel{-}{{LST}_{f}}\) ) and non-forest openlands ( \(\:\stackrel{-}{{LST}_{o}}\) ) within the grid: $$\:\varDelta\:LST=\stackrel{-}{{LST}_{f}}-\stackrel{-}{{LST}_{o}}$$ 5 According to the MCD12Q1 land cover product, forest pixels should belong to evergreen needleleaf forests (ENFs), deciduous broadleaf forests (DBFs), evergreen broadleaf forests (EBFs) and mixed forests (MFs). Openland pixels include grasslands (GRA) and croplands (CRO). To minimize the potential impact of direct human activities, such as anthropogenic heat emissions and land use/land cover changes, on the ΔLST time series, we apply additional filtering criteria to the potential samples: (a) the land cover type remains unchanged from 2003 to 2022; (b) the forest loss area should be less than 5% during 2003–2023 based on GFC data; and (c) the impervious surface percentage should be less than 5% according to GAIA data. In addition, we calculate ΔLST only if both the forest and openland sample sizes exceed 5 pixels. We discard ΔLST calculations for grids where the mean elevation exceeds 500 m to avoid potential bias induced by mountainous terrain. The resultant mapping of the biophysical LST effect is spatially aggregated from the 0.25° scale to the 0.5° scale to match the meteorological data and mitigate the impact of potential outliers. We calculate ΔLST for both winter and summer from 2003 to 2023. The criteria ensure that the forest and non-forest pixels remain consistent and stable during the study period, allowing us to attribute the temporal dynamics of ΔLST primarily to background climate change. By substituting the target variable, we also calculate the winter albedo, summer LE and summer LAI difference between forest and non-forest areas. 4.4 Simulated biophysical effects of forests on LST To obtain the simulated biophysical effects of forests on LST over the historical period, we use the output from the historical simulation of the CMIP6 81 . The historical simulation covers the period of 1850–2014 and is forced by externally imposed and time-varying natural (e.g., solar variability and volcanic aerosols) and anthropogenic (e.g., greenhouse gases, aerosols, and land use/land cover changes) forcings. The last 30-year (i.e., 1985–2014) simulation output is used to calculate the biophysical effects of forests and their trends. Here, we choose the last 30 years because it is closest to our observational period (i.e., 2003–2023). Moreover, the 30-year time series can be considered long enough to minimize the influence of internal climate variabilities on the long-term trend. A few models that participate in the historical simulation report the subgrid information following the data reporting protocol of the Land Use Model Intercomparison Project (LUMIP) 82 . Specifically, these models provide surface diagnostic variables outputs on four subgrid tiles, including primary and secondary land (psl, including trees, grasslands, barrens and vegetated wetlands), cropland (crp), pastureland (pst) and urban land, in each model grid cell. To create forest and openland comparisons within a grid cell for further analysis, we only retain the grid cells with the ratio of the tree cover fraction to the psl fraction exceeding 10%; this threshold is consistent with that used by the Food and Agricultural Organization to define forest cover 83 . For these selected grid cells, therefore, the psl tile can be considered as a forest tile. The crp and pst tiles are combined into an openland tile to keep consistency with the satellite observations. Since the forest and openland tiles in a grid cell are forced by the same atmospheric inputs (e.g., downward radiation, air temperature and precipitation), the background climate is the same between the two tiles within a grid cell. As such, differences in surface diagnostic variables (e.g., LST) between the two tiles are solely attributed to differences in land cover 38 , 84 . The local biophysical effects of forests on LST ( \(\:\varDelta\:LST\) ) in a grid cell can be expressed as follows 38 , 84 : $$\:\varDelta\:LST={LST}_{psl}-{LST}_{crp/pst}$$ 6 where \(\:{LST}_{psl}\) denotes the forest LST value (psl tile), and \(\:{LST}_{crp/pst}\) indicates the openland LST value (crp and pst combination). Specifically, if the outputs on both the crp and pst tiles are available, their arithmetic mean value is used; otherwise, the available one is used. LST can be replaced by other surface diagnostic variables (e.g., albedo) to quantify the biophysical effects of forests on corresponding variables. It should be emphasized that the forest effects identified by this method should be interpreted as the result of a complete conversion from openlands to forests in a grid cell. In other words, the identified forest effects are insensitive to the area fraction of the forest or openland tile in a grid cell. Therefore, while historical land use and land cover changes are prevalent in the simulations, they do not influence the identified forest effects. Four models (CESM2, CNRM-ESM2-1, GFDL-ESM4 and UKESM1-0-LL) are used for the subgrid analysis, as they provide the required output for the historical simulation. CESM2 and GFDL-ESM4 each have only one member that provides the subgrid output, and the available one (r10i1p1f1 for CESM2 and r1i1p1f1 for GFDL-ESM4) is used. CNRM-ESM2-1 and UKESM1-0-LL each have more than one member, and the first one (r1i1p1f2 for CNRM-ESM2-1 and r2i1p1f2 for UKESM1-0-LL) is used for the sake of keeping consistency with other models as well as saving computational cost. 4.5 Predicted local LST effects of forest in future We also predict the biophysical effects of forests on LST base on the statistical relationships between climate variables and observation derived ΔLST w or ΔLST s , as well as the simulated climate changes in future. Specifically, the relationships are established using the temporal samples of ΔLST w or ΔLST s (21 samples from 2003–2023) and corresponding reanalysis climate data. For winter, we build a simple linear regression model to link ΔLST w and SC w (statsmodels package in Python). For summer, we build a ridge regression model for ΔLST s and four climate variables, including AT s , VPD s , DSR s and P s (sklearn package in Python). A ridge regression model is used to obtain robust estimations when the input variables are highly correlated. The established regression models are then combined with projected climate data to obtain future ΔLST w or ΔLST s . The projected climate data are from CMIP6 monthly simulations of four SSP scenarios (2015–2100), including SSP126, SSP245, SSP370 and SSP585. SSP126 indicates a low-emission path in which the global society prioritizes sustainable development; SSP245 suggests a moderate scenario in which social, economic, and technological trends follow historical patterns; SSP370 describes a regional rivalry pathway in which global competition leads to slower technological progress and continued reliance on fossil fuels. SSP585 is the most extreme scenario, with minimal efforts to reduce greenhouse gas emissions. We use the simulations from the r1i1p1f1 series to reduce the differences among different models. The climate variables used for predicting local LST effects of forest in future includes winter snow area fraction, summer air temperature, specific humidity, downward shortwave radiation and precipitation from ten models (BCC-CSM2-MR, CAS-ESM2-0, CESM2-WACCM, CMCC-CM2-SR5, CMCC-ESM2, EC-Earth3-Veg, MPI-ESM1-2-HR, MRI-ESM2-0, NorESM2-MM and TaiESM1). All model outputs are resampled to the 1° spatial resolution. Summer VPD data are calculated using saturation water vapor at a given air temperature and actual water vapor (e a ) estimated by specific humidity (q): The CMIP6 historical (1950–2014) simulations are adopted to bias calibration of future climate in four SSPs scenarios, using ERA-5 Land reanalysis as the benchmark. This step is designed to ensure the projected climate variables share the same baseline with the reanalysis data. We further calculate the predicted ΔLST w or ΔLST s trends in four SSPs scenarios and compare them with the projected skin temperature warming trends in the corresponding scenarios. The ratios suggest the relative magnitude to which the enhanced cooling effects of forests account for the land surface warming driven by climate change. 4.6 Statistical analyses We use the Theil-Sen estimator and Mann-Kendall test to analyse temporal trends 85 (scipy and pymannkendall packages in Python). The Theil-Sen estimator is a non-parametric method for robust estimation of the slope by using the median value of a range of possible slopes. The Mann-Kendall test, which is also nonparametric, statistically assesses whether a temporal trend is monotonically upward or downward. Correlation analyses are employed to examine the potential influence of given drivers on the target variables (numpy package in Python). These analyses are based on either spatial or temporal samples. We perform spatial correlation using trend values to verify whether changes in the independent variables could trigger changes in the dependent variables. Temporal correlation is conducted at both pixel and regional scales. For pixel-scale correlation, we calculate the spatial mean and standard deviation of correlation coefficients to represent regional results. For regional-scale correlation, we first calculate the mean value for the study area and then determine the correlation coefficient. Partial temporal correlation analysis is used to explore the relative impacts of multiple climate drivers on ΔLST s at the pixel scale (pingouin package in Python). The partial correlation coefficient describes the relationship between two variables after removing the effects of other potential drivers. We apply a non-linear machine learning method, Random Forest (RF), to evaluate the importance of multiple potential climate drivers for predicting ΔLST s (sklearn package in Python). Specifically, we use all spatial samples over 21 years to build the dataset. The dataset is divided into training (80% of the samples) and test sets (20% of the samples), with model parameters confirmed based on validation results from the test set (number of trees = 50; maximum depth of the tree = 20). All the samples are then used to fit the best model. The relative importance of each variable is indicated by the Gini coefficient. Moreover, we use the SHAP method to quantify the individual contribution of each variable in each sample (shap package in Python). SHAP is an approach to explain machine learning models based on cooperative game theory, where SHAP values quantify the marginal contributions of predictors. We calculate the mean absolute SHAP values for each input variable as importance metrics and explore the non-linear impact of climate variables on ΔLST s by plotting their SHAP values against input values. Based on the RF model, we set several scenarios to separate the relative contributions of climate variables to the ΔLST s trend. Specifically, we perform four experimental simulations: (S1) varying VPD s only; (S2) varying VPD s and P s while fixing DSR s and WS s ; (S3) varying VPD s , P s , and DSR s while fixing WS s ; and (S4) varying VPD s , P s , DSR s , and WS s . The Theil-Sen slope and Mann-Kendall P-value of simulations S1, S2 − S1, S3 − S2, and S4 − S3 are used to evaluate the individual contributions of the changing climate to the ΔLST s trend. We use SEM to reveal the biophysical mechanism underlying the ΔLST s dynamics (semopy package in Python). We assume that climate variables influence the temporal dynamics of ΔLST s through three pathways: (P1) climate change affects ΔLST s via ΔLAI s ; (P2) climate change affects ΔLAI s , which further alters ΔLE s and then ΔLST s ; and (P3) climate change affects ΔLST s via ΔLE s . For P1 and P2, the impact of climate variables on ΔLAI s arises from the diverse sensitivities of different vegetation types to climate changes. Specifically, biophysical parameters (such as albedo and surface roughness) vary differently in forested and non-forested areas under a changing climate, which further affects ΔLST s (P1). Additionally, climate changes promote or inhibit forest and non-forest vegetation growth differently, further influencing the evapotranspiration effect of forests and ΔLST s (P2). Furthermore, climate variables can directly modulate ΔLE s , even when the biophysical and physiological conditions of vegetation remain constant. This is due to inherent differences between different vegetation types 48 , 51 , 86 , resulting in diverse evapotranspiration responses to climate change and, consequently, affecting ΔLST s (P3). In practice, SEM is performed at the pixel level. We use all the data within a 3×3 spatial moving window for each 0.5° grid to ensure a sufficient sample size for establishing a robust model. SEM is conducted in grids where the sample size is larger than 150. The data are normalized using the Z-score method before modeling. Following previous study 87 , the goodness-of-fit index (GFI) is employed to evaluate the applicability and effectiveness of each model. We filter out paths with P values lower than 0.05 and corresponding model GFIs greater than 0.9 and calculate the mean path values along with the standard error. The contributions of pathways P1, P2, and P3 are measured by their path effects, which are calculated by multiplying the contained path values. Uncertainties are estimated based on error propagation (error_propagation package in Python). Declarations Data Availability All 1 km MODIS products, including MYD11, MOD16A2 Gap-filled, MCD15A3H, MCD15A3H, MCD43A3, MCD12Q1 can be downloaded from https://ladsweb.modaps.eosdis.nasa.gov/search/. The GFC data is available on the Google Earth Engine at https://developers.google.com/earth-engine/datasets/catalog/UMD_hansen_global_forest_change_2023_v1_11. The GAIA products are available at https://data-starcloud.pcl.ac.cn/resource/13. The GMTED2010 data are from https://developers.google.com/earth-engine/datasets/catalog/USGS_GMTED2010_FULL. ERA5 reanalysis data is accessible at https://cds.climate.copernicus.eu/datasets/reanalysis-era5-land-monthly-means?tab=overview. CMIP6 simulations are available at https://aims2.llnl.gov/search/cmip6/. Code Availability The Python codes used to generate all the results are available at https://zenodo.org/records/13944220 . References Seddon, N., Turner, B., Berry, P., Chausson, A. & Girardin, C. A. J. Grounding nature-based climate solutions in sound biodiversity science. Nat. Clim. Chang. 9 , 84–87 (2019). Bonan, G. B. Forests and climate change: Forcings, feedbacks, and the climate benefits of forests. Science (80-. ). 320 , 1444–1449 (2008). Novick, K. A. et al. We need a solid scientific basis for nature-based climate solutions in the United States. Proc. Natl. Acad. Sci. 121 , 1–4 (2024). Griscom, B. W. et al. Natural climate solutions. Proc. Natl. Acad. Sci. U. S. A. 114 , 11645–11650 (2017). Walker, W. S. et al. The global potential for increased storage of carbon on land. Proc. Natl. Acad. Sci. U. S. A. 119 , 1–12 (2022). Lewis, S. L., Wheeler, C. E., Mitchard, E. T. A. & Koch, A. Restoring natural forests is the best way to remove atmospheric carbon. Nature 568 , 25–28 (2019). Roebroek, C. T. J., Duveiller, G., Seneviratne, S. I., Davin, E. L. & Cescatti, A. Releasing global forests from human management: How much more carbon could be stored? Science (80-. ). 380 , 749–753 (2023). Mo, L. et al. Integrated global assessment of the natural forest carbon potential. Nature 624 , 92–101 (2023). Weber, J. et al. Chemistry-albedo feedbacks offset up to a third of forestation’s CO 2 removal benefits. Science (80-. ). 383 , 860–864 (2024). Lawrence, D., Coe, M., Walker, W., Verchot, L. & Vandecar, K. The Unseen Effects of Deforestation: Biophysical Effects on Climate. Front. For. Glob. Chang. 5 , 1–13 (2022). Windisch, M. G., Davin, E. L. & Seneviratne, S. I. Prioritizing forestation based on biogeochemical and local biogeophysical impacts. Nat. Clim. Chang. 11 , 867–871 (2021). Zhu, L. et al. Comparable biophysical and biogeochemical feedbacks on warming from tropical moist forest degradation. Nat. Geosci. 16 , 244–249 (2023). Lawrence, D. & Vandecar, K. Effects of tropical deforestation on climate and agriculture. Nat. Clim. Chang. 5 , 27–36 (2015). Alkama, R. & Cescatti, A. Biophysical climate impacts of recent changes in global forest cover. Science (80-. ). 351 , 600–604 (2016). Duveiller, G., Hooker, J. & Cescatti, A. The mark of vegetation change on Earth’s surface energy balance. Nat. Commun. 9 , 64–75 (2018). Mildrexler, D. J., Zhao, M. & Running, S. W. A global comparison between station air temperatures and MODIS land surface temperatures reveals the cooling role of forests. J. Geophys. Res. Biogeosciences 116 , 1–15 (2011). R.A.Betts. Offset of the potential carbon sink from boreal forestation by decreases in surface albedo. Nature 408 , 187–190 (2000). Bright, R. M. et al. Local temperature response to land cover and management change driven by non-radiative processes. Nat. Clim. Chang. 7 , 296–302 (2017). Pitman, A. J. et al. Importance of background climate in determining impact of land-cover change on regional climate. Nat. Clim. Chang. 1 , 472–475 (2011). Lee, X. et al. Observed increase in local cooling effect of deforestation at higher latitudes. Nature 479 , 384–387 (2011). Li, Y. et al. Local cooling and warming effects of forests based on satellite observations. Nat. Commun. 6 , 1–10 (2015). Zeng, Z. et al. Deforestation-induced warming over tropical mountain regions regulated by elevation. Nat. Geosci. 14 , 23–29 (2021). Essery, R. Large-scale simulations of snow albedo masking by forests. Geophys. Res. Lett. 40 , 5521–5525 (2013). Prevedello, J. A., Winck, G. R., Weber, M. M., Nichols, E. & Sinervo, B. Impacts of forestation and deforestation on local temperature across the globe. PLoS One 14 , 1–18 (2019). Rohatyn, S., Rotenberg, E., Tatarinov, F., Carmel, Y. & Yakir, D. Large variations in afforestation-related climate cooling and warming effects across short distances. Commun. Earth Environ. 4 , 1–10 (2023). Alkama, R. et al. Vegetation-based climate mitigation in a warmer and greener World. Nat. Commun. 13 , 1–10 (2022). Ge, J. et al. Local surface cooling from afforestation amplified by lower aerosol pollution. Nat. Geosci. (2023) doi:10.1038/s41561-023-01251-x. Lee, T. H. & Lo, M. H. The role of El Niño in modulating the effects of deforestation in the Maritime Continent. Environ. Res. Lett. 16 , (2021). Singh, A. et al. Land-Use Feedback under Global Warming—A Transition from Radiative to Hydrological Feedback Regime. J. Clim. 37 , 3847–3866 (2024). Pitman, A. J. et al. Uncertainties in climate responses to past land cover change: First results from the LUCID intercomparison study. Geophys. Res. Lett. 36 , 1–6 (2009). Forzieri, G. et al. Evaluating the Interplay Between Biophysical Processes and Leaf Area Changes in Land Surface Models. J. Adv. Model. Earth Syst. 10 , 1102–1126 (2018). Fetting, C. ‘The European Green Deal’, ESDN Report. Eur. Comm. 53 , 24 (2020). Schultz, N. M., Lawrence, P. J. & Lee, X. Global satellite data highlights the diurnal asymmetry of the surface temperature response to deforestation. J. Geophys. Res. Biogeosciences 122 , 903–917 (2017). Manninen, T. et al. Monitoring changes in forestry and seasonal snow using surface albedo during 1982-2016 as an indicator. Biogeosciences 16 , 223–240 (2019). Atlaskina, K., Berninger, F. & de Leeuw, G. Satellite observations of changes in snow-covered land surface albedo during spring in the Northern Hemisphere. Cryosph. 9 , 1879–1893 (2015). Li, Q., Ma, M., Wu, X. & Yang, H. Snow Cover and Vegetation-Induced Decrease in Global Albedo From 2002 to 2016. J. Geophys. Res. Atmos. 124–138 (2017) doi:10.1002/2017JD027010. Kouki, K., Luojus, K. & Riihelä, A. Evaluation of snow cover properties in ERA5 and ERA5-Land with several satellite-based datasets in the Northern Hemisphere in spring 1982-2018. Cryosphere 17 , 5007–5026 (2023). Luo, X. et al. An Evaluation of CMIP6 Models in Representing the Biophysical Effects of Deforestation With Satellite‐Based Observations. J. Geophys. Res. Atmos. 128 , 1–20 (2023). Li, Y. et al. Biophysical impacts of earth greening can substantially mitigate regional land surface temperature warming. Nat. Commun. 14 , 121 (2023). Kumar, S. et al. Land use/cover change impacts in CMIP5 climate simulations: A new methodology and 21st century challenges. J. Geophys. Res. Atmos. 118 , 6337–6353 (2013). Hasler, N. et al. Accounting for albedo change to identify climate-positive tree cover restoration. Nat. Commun. 15 , 2275 (2024). Forzieri, G. et al. Increased control of vegetation on global terrestrial energy fluxes. Nat. Clim. Chang. 10 , 356–362 (2020). Yuan, W. et al. Increased atmospheric vapor pressure deficit reduces global vegetation growth. Sci. Adv. 5 , 1–13 (2019). Zhong, Z. et al. Disentangling the effects of vapor pressure deficit on northern terrestrial vegetation productivity. Sci. Adv. 9 , (2023). Zhong, Z. et al. Higher Sensitivity of Planted Forests’ Productivity Than Natural Forests to Droughts in China. J. Geophys. Res. Biogeosciences 126 , 1–13 (2021). Liu, M. et al. Overridingly increasing vegetation sensitivity to vapor pressure deficit over the recent two decades in China. Ecol. Indic. 161 , 111977 (2024). Isbell, F. et al. Biodiversity increases the resistance of ecosystem productivity to climate extremes. Nature 526 , 574–577 (2015). Tumber-Dávila, S. J., Schenk, H. J., Du, E. & Jackson, R. B. Plant sizes and shapes above and belowground and their interactions with climate. New Phytol. 235 , 1032–1056 (2022). Un, N. Natural evaporation from open water, bare soil and grass. Proc. R. Soc. London. Ser. A. Math. Phys. Sci. 193 , 120–145 (1948). Grossiord, C. et al. Plant responses to rising vapor pressure deficit. New Phytol. 226 , 1550–1566 (2020). Massmann, A., Gentine, P. & Lin, C. When Does Vapor Pressure Deficit Drive or Reduce Evapotranspiration? J. Adv. Model. Earth Syst. 11 , 3305–3320 (2019). Dickerson-Lange, S. E., Vano, J. A., Gersonde, R. & Lundquist, J. D. Ranking Forest Effects on Snow Storage: A Decision Tool for Forest Management. Water Resour. Res. 57 , (2021). Adam, J. C., Hamlet, A. F. & Lettenmaier, D. P. Implications of global climate change for snowmelt hydrology in the twenty‐first century. Hydrol. Process. 23 , 962–972 (2009). Bjorkman, A. D., Elmendorf, S. C., Beamish, A. L., Vellend, M. & Henry, G. H. R. Contrasting effects of warming and increased snowfall on Arctic tundra plant phenology over the past two decades. Glob. Chang. Biol. 21 , 4651–4661 (2015). Qin, Y. et al. Agricultural risks from changing snowmelt. Nat. Clim. Chang. 10 , 459–465 (2020). Cui, J. et al. Global water availability boosted by vegetation-driven changes in atmospheric moisture transport. Nat. Geosci. 15 , 982–988 (2022). Li, Y. et al. Deforestation-induced climate change reduces carbon storage in remaining tropical forests. Nat. Commun. 13 , (2022). Ma, S., Zhou, S., Yu, B. & Song, J. Deforestation-induced runoff changes dominated by forest-climate feedbacks. Sci. Adv. 10 , (2024). Hoek van Dijke, A. J. et al. Shifts in regional water availability due to global tree restoration. Nat. Geosci. 15 , 363–368 (2022). Zan, B. et al. Spatiotemporal inequality in land water availability amplified by global tree restoration. Nat. Water (2024) doi:10.1038/s44221-024-00296-5. Wang, H., Yue, C. & Luyssaert, S. Reconciling different approaches to quantifying land surface temperature impacts of afforestation using satellite observations. Biogeosciences 20 , 75–92 (2023). Xu, R. et al. Contrasting impacts of forests on cloud cover based on satellite observations. Nat. Commun. 13 , 670 (2022). Li, Y., Piao, S., Chen, A., Ciais, P. & Li, L. Z. X. Local and teleconnected temperature effects of afforestation and vegetation greening in China. Natl. Sci. Rev. 7 , 897–912 (2020). Zhang, Y. et al. Asymmetric impacts of forest gain and loss on tropical land surface temperature. Nat. Geosci. 13 , 823–831 (2024). Su, Y. et al. Asymmetric influence of forest cover gain and loss on land surface temperature. Nat. Clim. Chang. 13 , 823–831 (2023). Zhu, Z. et al. Greening of the Earth and its drivers. Nat. Clim. Chang. 6 , 791–795 (2016). Yang, Y. et al. Evapotranspiration on a greening Earth. Nature Reviews Earth and Environment vol. 4 626–641 (2023). Zeng, Z. et al. Climate mitigation from vegetation biophysical feedbacks during the past three decades. Nat. Clim. Chang. 7 , 432–436 (2017). FIELD, C. B., JACKSON, R. B. & MOONEY, H. A. Stomatal responses to increased CO2: implications from the plant to the global scale. Plant. Cell Environ. 18 , 1214–1225 (1995). Betts, R. A. et al. Projected increase in continental runoff due to plant responses to increasing carbon dioxide. Nature 448 , 1037–1041 (2007). He, M. et al. Amplified warming from physiological responses to carbon dioxide reduces the potential of vegetation for climate change mitigation. Commun. Earth Environ. 3 , 1–10 (2022). Skinner, C. B., Poulsen, C. J. & Mankin, J. S. Amplification of heat extremes by plant CO2 physiological forcing. Nat. Commun. 9 , 1–11 (2018). Li, Z. et al. Satellite Remote Sensing of Global Land Surface Temperature: Definition, Methods, Products, and Applications. Rev. Geophys. 61 , 1–77 (2023). Novick, K. A. & Katul, G. G. The Duality of Reforestation Impacts on Surface and Air Temperature. J. Geophys. Res. Biogeosciences 125 , 1–15 (2020). Zhang, Z., Zhang, F., Wang, L., Lin, A. & Zhao, L. Biophysical climate impact of forests with different age classes in mid- and high-latitude North America. For. Ecol. Manage. 494 , 119327 (2021). Xing, Z. et al. Estimation of daily mean land surface temperature at global scale using pairs of daytime and nighttime MODIS instantaneous observations. ISPRS J. Photogramm. Remote Sens. 178 , 51–67 (2021). Hansen, M. C. et al. High-Resolution Global Maps of 21st-Century Forest Cover Change. Science (80-. ). 342 , 850–853 (2013). Gong, P. et al. Annual maps of global artificial impervious area (GAIA) between 1985 and 2018. Remote Sens. Environ. 236 , 111510 (2020). Muñoz-Sabater, J. et al. ERA5-Land: A state-of-the-art global reanalysis dataset for land applications. Earth Syst. Sci. Data 13 , 4349–4383 (2021). Liu, X. et al. Local temperature responses to actual land cover changes present significant latitudinal variability and asymmetry. Sci. Bull. 68 , 2849–2861 (2023). Eyring, V. et al. Overview of the Coupled Model Intercomparison Project Phase 6 (CMIP6) experimental design and organization. Geosci. Model Dev. 9 , 1937–1958 (2016). Lawrence, D. M. et al. The Land Use Model Intercomparison Project (LUMIP) contribution to CMIP6: Rationale and experimental design. Geosci. Model Dev. 9 , 2973–2998 (2016). Keenan, R. J. et al. Dynamics of global forest area: Results from the FAO Global Forest Resources Assessment 2015. Forest Ecology and Management vol. 352 9–20 (2015). Luo, X. et al. Local and Nonlocal Biophysical Effects of Historical Land Use and Land Cover Changes in CMIP6 Models and the Intermodel Uncertainty. Earth’s Futur. 12 , (2024). Sen, P. K. Estimates of the Regression Coefficient Based on Kendall’s Tau. J. Am. Stat. Assoc. 63 , 1379–1389 (1968). Huang, M. et al. Air temperature optima of vegetation productivity across global biomes. Nat. Ecol. Evol. 3 , 772–779 (2019). Additional Declarations There is NO Competing Interest. Supplementary Files 1SupplementaryfinalLYT.docx Supplementary information Cite Share Download PDF Status: Published Journal Publication published 25 Sep, 2025 Read the published version in Nature Communications → Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-5281378","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Physical Sciences - Article","associatedPublications":[],"authors":[{"id":370381856,"identity":"6096f4cd-3f02-443d-9fde-4f598cffeb3d","order_by":0,"name":"Zhao-Liang Li","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAs0lEQVRIie3OoQoCQRDG8W9ZmCuLVy8I+woHFgXFV9l0zSSYBeGSdpuvcsfA2swGkbtivnjRPcQgGMZm2D8MzIQfDBCL/WEjDVV1gHmdcwEhDdTHgdBwFhIShs17k5EkaXhR3sb2dGg7uLvoMcer8mFyn0wyuLWIVIGwyYmQqd5JiNryLBBbku7hRESDVSDwRJmQEOr9JTzmC5o6CUnTc9v0G17andfXTkI++xnEYrFY7HtPW5UvRQwfCPoAAAAASUVORK5CYII=","orcid":"","institution":"ICube Laboratory","correspondingAuthor":true,"prefix":"","firstName":"Zhao-Liang","middleName":"","lastName":"Li","suffix":""},{"id":370381857,"identity":"7cf8bb93-caad-4b40-879c-35afa88c1266","order_by":1,"name":"Yitao Li","email":"","orcid":"https://orcid.org/0000-0001-7496-2956","institution":"Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences","correspondingAuthor":false,"prefix":"","firstName":"Yitao","middleName":"","lastName":"Li","suffix":""},{"id":370381858,"identity":"98a3d5b3-8be3-4a1c-905f-915bb37d0cb8","order_by":2,"name":"Jun Ge","email":"","orcid":"https://orcid.org/0000-0002-8876-1650","institution":"Nanjing University","correspondingAuthor":false,"prefix":"","firstName":"Jun","middleName":"","lastName":"Ge","suffix":""},{"id":370381859,"identity":"96a93fc7-b6ba-4e70-b25f-eb0c946230be","order_by":3,"name":"Hua Wu","email":"","orcid":"","institution":"University of Electronic Science and Technology of China","correspondingAuthor":false,"prefix":"","firstName":"Hua","middleName":"","lastName":"Wu","suffix":""},{"id":370381860,"identity":"90a2822b-443b-4348-b8e9-942fd5c60766","order_by":4,"name":"Ronglin Tang","email":"","orcid":"","institution":"Institute of Geographic Sciences and Natural Resources Research","correspondingAuthor":false,"prefix":"","firstName":"Ronglin","middleName":"","lastName":"Tang","suffix":""},{"id":370381861,"identity":"fc0cc310-d635-4824-b94f-d0e6f646f642","order_by":5,"name":"Yuanliang Cheng","email":"","orcid":"","institution":"Institute of Geographic Sciences and Natural Resources Research","correspondingAuthor":false,"prefix":"","firstName":"Yuanliang","middleName":"","lastName":"Cheng","suffix":""},{"id":370381862,"identity":"2820175d-2080-41e4-8111-4c4e481c9b70","order_by":6,"name":"Xiangyang Liu","email":"","orcid":"","institution":"State Key Laboratory of Efficient Utilization of Arid and Semi-arid Arable Land in Northern China/Institute of Agricultural Resources and Regional Planning","correspondingAuthor":false,"prefix":"","firstName":"Xiangyang","middleName":"","lastName":"Liu","suffix":""},{"id":370381863,"identity":"5bdcb7fd-8c73-422a-b37d-c76b4cd8cfdb","order_by":7,"name":"Yi Wang","email":"","orcid":"","institution":"Institute of Geographic Sciences and Natural Resources Research","correspondingAuthor":false,"prefix":"","firstName":"Yi","middleName":"","lastName":"Wang","suffix":""},{"id":370381864,"identity":"5b9c520c-e502-482b-a3a0-cd049a2371a4","order_by":8,"name":"Menglin Si","email":"","orcid":"","institution":"Institute of Geographic Sciences and Natural Resources Research","correspondingAuthor":false,"prefix":"","firstName":"Menglin","middleName":"","lastName":"Si","suffix":""},{"id":370381865,"identity":"c3ebea28-3857-4978-8fb1-637143b6c96a","order_by":9,"name":"Wei Zhao","email":"","orcid":"https://orcid.org/0000-0002-4839-6791","institution":"Institute of Mountain Hazards and Environment","correspondingAuthor":false,"prefix":"","firstName":"Wei","middleName":"","lastName":"Zhao","suffix":""},{"id":370381866,"identity":"9afb024b-8259-4dd6-a68e-3666d90a05f1","order_by":10,"name":"Caixia Gao","email":"","orcid":"","institution":"Aerospace Information Research Institute, Chinese Academy of Sciences","correspondingAuthor":false,"prefix":"","firstName":"Caixia","middleName":"","lastName":"Gao","suffix":""},{"id":370381867,"identity":"f76c9019-df79-4613-9543-e894fbcf5175","order_by":11,"name":"Si-Bo Duan","email":"","orcid":"","institution":"Institute of Agricultural Resources and Regional Planning, Chinese Academy of Agricultural Sciences","correspondingAuthor":false,"prefix":"","firstName":"Si-Bo","middleName":"","lastName":"Duan","suffix":""},{"id":370381868,"identity":"ebb2231d-7688-4d20-ad7b-0bb3f3fa8164","order_by":12,"name":"Qian Shi","email":"","orcid":"","institution":"Sun Yat-sen University","correspondingAuthor":false,"prefix":"","firstName":"Qian","middleName":"","lastName":"Shi","suffix":""},{"id":370381869,"identity":"f3e5b7e7-cc1a-4957-902d-d5ea6bef0c1f","order_by":13,"name":"Pei Leng","email":"","orcid":"","institution":"State Key Laboratory of Efficient Utilization of Arid and Semi-arid Arable Land in Northern China/Institute of Agricultural Resources and Regional Planning","correspondingAuthor":false,"prefix":"","firstName":"Pei","middleName":"","lastName":"Leng","suffix":""},{"id":370381870,"identity":"3a5b5214-6843-4faa-b133-35ff39140479","order_by":14,"name":"Enyu Zhao","email":"","orcid":"","institution":"Information Science and Technology College, Dalian Maritime University; Dalian","correspondingAuthor":false,"prefix":"","firstName":"Enyu","middleName":"","lastName":"Zhao","suffix":""},{"id":370381871,"identity":"9399c6f6-c232-43d7-ad68-56064bc54c3b","order_by":15,"name":"Jing Li","email":"","orcid":"","institution":"Institute of Agricultural Resources and Regional Planning, Chinese Academy of Agricultural Sciences","correspondingAuthor":false,"prefix":"","firstName":"Jing","middleName":"","lastName":"Li","suffix":""},{"id":370381872,"identity":"cab0f519-15c4-467f-aae4-f774e34cd7bb","order_by":16,"name":"Chenghu Zhou","email":"","orcid":"https://orcid.org/0000-0003-3331-2302","institution":"Nanjing University","correspondingAuthor":false,"prefix":"","firstName":"Chenghu","middleName":"","lastName":"Zhou","suffix":""}],"badges":[],"createdAt":"2024-10-17 09:11:09","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-5281378/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5281378/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1038/s41467-025-63556-2","type":"published","date":"2025-09-25T04:00:00+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":76779066,"identity":"e9927a26-4baf-46b9-8528-a0e4649b5a67","added_by":"auto","created_at":"2025-02-20 15:56:21","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":2113250,"visible":true,"origin":"","legend":"\u003cp\u003eTemporal dynamics in winter and summer land surface temperature effects of European forests (ΔLST\u003csub\u003ew\u003c/sub\u003e and ΔLST\u003csub\u003es\u003c/sub\u003e) from 2003 to 2023. (a) Interannual variations of the daily mean, daytime and nighttime ΔLST\u003csub\u003ew\u003c/sub\u003e in Europe. The dotted lines indicate the temporal trends. (b) The spatial pattern of the ΔLST\u003csub\u003ew\u003c/sub\u003e trend. The subplot indicates the percentage of grids with different significance levels and signs, in which asterisks indicate that the correlation is significant (\u003cem\u003ep\u003c/em\u003e\u0026lt;0.05). The dots indicate the grids with significant trends (\u003cem\u003ep\u003c/em\u003e\u0026lt;0.05). (c to d) Similar to (a to b) but for ΔLST\u003csub\u003es\u003c/sub\u003e.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-5281378/v1/8cd9e016dcf25cfb134c079e.png"},{"id":76779068,"identity":"1e38c2a9-e8da-402c-9217-f4205834bf02","added_by":"auto","created_at":"2025-02-20 15:56:21","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":1881108,"visible":true,"origin":"","legend":"\u003cp\u003eTemporal control of snow cover (SC\u003csub\u003ew\u003c/sub\u003e) on albedo effect (Δα\u003csub\u003ew\u003c/sub\u003e) and land surface temperature effect (ΔLST\u003csub\u003ew\u003c/sub\u003e) of European forest in winter. (a) Interannual variation of Δα\u003csub\u003ew\u003c/sub\u003e and the individual winter albedo (α\u003csub\u003ew\u003c/sub\u003e) for forests and openland. (b) The spatial pattern of the Δα\u003csub\u003ew\u003c/sub\u003e trend. The subplot indicates the percentage of grids with different significance levels and signs, in which asterisks indicate that the correlation is significant (\u003cem\u003ep\u003c/em\u003e\u0026lt;0.05). The dots indicate the grids with significant trends (p\u0026lt;0.05). (c) Spatial relationships between SC\u003csub\u003ew\u003c/sub\u003e trends and α\u003csub\u003ew\u003c/sub\u003e trends in forests and openlands, and the resultant spatial relationships between SC\u003csub\u003ew\u003c/sub\u003e trends and Δα\u003csub\u003ew\u003c/sub\u003e trends. The error bars show half of the standard deviation of the ΔLST\u003csub\u003ew\u003c/sub\u003e trend within the bin of the SC\u003csub\u003ew\u003c/sub\u003e trend (1 %/decade). (d) Spatial relationships between Δα\u003csub\u003ew\u003c/sub\u003e trends and daytime ΔLST\u003csub\u003ew\u003c/sub\u003e trends. The error bars show half of the standard deviation of the ΔLST\u003csub\u003ew\u003c/sub\u003e trend within the bin of the Δα\u003csub\u003ew\u003c/sub\u003e trend (0.005 unitless/decade). (e) Temporal relationships between the regional mean SC\u003csub\u003ew\u003c/sub\u003e and the daily mean, daytime and nighttime ΔLST\u003csub\u003ew\u003c/sub\u003e. The shaded areas indicate the 95% confidence intervals for the regression estimates. In (c–e), the solid lines indicate the regression line.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-5281378/v1/1a25b7a27dff14dd3c28cf26.png"},{"id":76779069,"identity":"53607e36-4007-49e3-b7cf-18b9e09278fa","added_by":"auto","created_at":"2025-02-20 15:56:21","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":1460318,"visible":true,"origin":"","legend":"\u003cp\u003eDominant role vapor pressure deficit (VPD) in the temporal dynamics in the summer daytime cooling effect (ΔLST\u003csub\u003es\u003c/sub\u003e) of forests. (a) Regional mean correlation and partial correlation coefficients between climate variables and the daytime ΔLST\u003csub\u003es\u003c/sub\u003e. The error bars indicate the standard deviations. (b) Spatial pattern of the partial correlation coefficient between the summer vapor pressure deficit (VPD\u003csub\u003es\u003c/sub\u003e) and daytime ΔLST\u003csub\u003es\u003c/sub\u003e. The subplot indicates the percentage of grids with different significance levels and signs, in which asterisks indicate that the correlation is significant (\u003cem\u003ep\u003c/em\u003e\u0026lt;0.05). (c) The importance of climate variables is indicated by Shapley Additive Explanations (SHAP) values and Gini coefficients. (d) Plots of SHAP values based on the established random forest model. (e) Observed and modelled ΔLST\u003csub\u003es\u003c/sub\u003e trends, and the isolated contributions from climate variables.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-5281378/v1/a02fdaf043a1277a60662342.png"},{"id":76779404,"identity":"2241f438-ed1c-4e59-adf5-2eacc10dc5d4","added_by":"auto","created_at":"2025-02-20 16:04:21","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":2751575,"visible":true,"origin":"","legend":"\u003cp\u003eMechanisms underlying the effects of climate variables on the daytime LST effects of European forests (ΔLST\u003csub\u003es\u003c/sub\u003e) in summer. (a) Structure equation models describing the biophysical relationships between climate variables and ΔLST\u003csub\u003es\u003c/sub\u003e. The numbers denote the path value (mean ± standard error). The arrow color indicates the sign of the path (red for positive and blue for negative), and the thickness indicates the magnitude of the corresponding path value. (b) Bar plot of the pathway effects of climate variables affecting ΔLST\u003csub\u003es\u003c/sub\u003e. The error bars indicate the uncertainty based on error propagation.\u0026nbsp;\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-5281378/v1/c94ff07b3dd8d87cd7c4cd11.png"},{"id":76779070,"identity":"6f4b4d0b-45f0-4fb0-9bff-60e97fd1bc10","added_by":"auto","created_at":"2025-02-20 15:56:21","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":1715542,"visible":true,"origin":"","legend":"\u003cp\u003eSimulated temporal dynamics in wintertime and summertime land surface temperature effects of European forests (ΔLST\u003csub\u003ew\u003c/sub\u003e and ΔLST\u003csub\u003es\u003c/sub\u003e) from 1985 to 2014. (a) Interannual variations of the simulated daily mean ΔLST\u003csub\u003ew\u003c/sub\u003e in Europe. The dotted lines indicate the temporal trends. (b) Observed and simulated winter albedo α\u003csub\u003ew\u003c/sub\u003e trends in forests and openlands. (c) Observed and simulated regression slopes between the albedo effect (Δα\u003csub\u003ew\u003c/sub\u003e) and ΔLST\u003csub\u003ew\u003c/sub\u003e. (d) Similar to (a), but for ΔLST\u003csub\u003es\u003c/sub\u003e. (e) Simulated regional mean correlation and partial correlation coefficients between climate variables and daily mean ΔLST\u003csub\u003es\u003c/sub\u003e. The error bars indicate the spatial standard deviations.\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-5281378/v1/bfaa18342f2e7f30cb622e8a.png"},{"id":76779067,"identity":"b75b44c8-8ece-48c9-8460-f259cd7274a7","added_by":"auto","created_at":"2025-02-20 15:56:21","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":2207328,"visible":true,"origin":"","legend":"\u003cp\u003eProjected daily mean land surface temperature effect of European forests over 2020–2100 (ΔLST\u003csub\u003ew\u003c/sub\u003e and ΔLST\u003csub\u003es\u003c/sub\u003e). (a) Projected daily mean ΔLST\u003csub\u003ew\u003c/sub\u003e and winter land surface temperature (LST\u003csub\u003ew\u003c/sub\u003e) changes in the SSP126 scenario. (b) Projected daily mean ΔLST\u003csub\u003es\u003c/sub\u003e and the summer land surface temperature (LST\u003csub\u003es\u003c/sub\u003e) changes in the SSP126 scenario. (c to d), (e to f) and (g to h) are similar with (a to b), but for the SSP245, SSP370, and SSP585 scenarios, respectively. The shaded area indicates the 95% confidence interval of the mean. The dotted lines indicate the temporal trends. The figures near the dotted lines indicate the temporal trend values. Levels of significance: ***, \u003cem\u003ep\u003c/em\u003e \u0026lt; 0.001; **, \u003cem\u003ep\u003c/em\u003e \u0026lt; 0.01; *, \u003cem\u003ep\u003c/em\u003e \u0026lt; 0.05; NS, \u003cem\u003ep\u003c/em\u003e \u0026gt; 0.05. The percentages indicate the ratio of the ΔLST\u003csub\u003ew\u003c/sub\u003e trend to the LST\u003csub\u003ew\u003c/sub\u003e trend (or the ΔLST\u003csub\u003es\u003c/sub\u003e trend to the LST\u003csub\u003es\u003c/sub\u003e trend).\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-5281378/v1/dbd564bde79aa2cb16304e6a.png"},{"id":92235088,"identity":"b21486ef-3bb7-42df-ae54-b8a783674958","added_by":"auto","created_at":"2025-09-26 07:09:25","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":15780386,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5281378/v1/f63e188d-c6e2-4dac-bf21-908a9516b554.pdf"},{"id":76779072,"identity":"e03f8767-e362-4f8d-b3bd-6359cd7c06af","added_by":"auto","created_at":"2025-02-20 15:56:22","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":4026722,"visible":true,"origin":"","legend":"Supplementary information","description":"","filename":"1SupplementaryfinalLYT.docx","url":"https://assets-eu.researchsquare.com/files/rs-5281378/v1/b95b43be5d047f5d831898f2.docx"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"Recent climate change strengthens the local cooling of European forests","fulltext":[{"header":"Introduction","content":"\u003cp\u003eForests provide numerous benefits to humans and the planet, such as producing food and energy, reducing soil erosion, and increasing water availability in downwind areas\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e,\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e. Through conservation, proper management, and restoration practices, forest ecosystems play a crucial role in addressing global climate warming\u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e,\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e. Current evidence supporting forest-related climate mitigation is primarily based on the biochemical process of carbon sequestration\u003csup\u003e\u003cspan additionalcitationids=\"CR6 CR7\" citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e. However, forests can impact the climate system through other processes that are not comprehensively considered in current mitigation or adaptation strategies. For instance, forests release biogenic volatile organic compounds, which affect local or non-local radiative forcings\u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u003c/sup\u003e. Compared with other ecosystems, forests also present unique biophysical characteristics (e.g., albedo, aerodynamic roughness and evaporation efficiency), which can reshape surface energy and water balance processes\u003csup\u003e\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e. The biophysical process of forests has garnered particular attention in recent years, as it can significantly intensify or offset the global climate mitigation effects of carbon sequestration\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e,\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eUsing numerical models or muti-scale observations, previous studies have thoroughly investigated the impacts of re/afforestation, deforestation, and forest degradation on local land surface temperature (LST) and their biophysical mechanisms\u003csup\u003e\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e,\u003cspan additionalcitationids=\"CR14\" citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e. There is broad consensus that forests can have two direct effects with opposing signs compared to non-forest vegetation: (a) land surface cooling due to higher evapotranspiration (ET) rates\u003csup\u003e\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u003c/sup\u003e and aerodynamic roughness, and (b) land surface warming due to lower albedo\u003csup\u003e\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u003c/sup\u003e. The sign and magnitude of the temperature signal depend on the counterbalancing of the albedo-related radiative process and the turbulence-related non-radiative process\u003csup\u003e\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u003c/sup\u003e. The relative dominance of these two processes is fundamentally determined by the background climate\u003csup\u003e\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u003c/sup\u003e, leading to varying temperature effects of forests with different latitudes\u003csup\u003e\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u003c/sup\u003e, seasons\u003csup\u003e\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e or elevations\u003csup\u003e\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u003c/sup\u003e. Forests are usually warmer than surrounding openlands in cold environments because bright snow covers low vegetation but is masked by dark forest canopies\u003csup\u003e\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e\u003c/sup\u003e. Conversely, forests exhibit cooling effects under warm and wet environments because their surface roughness and transpiration rates are larger than those of other vegetation types\u003csup\u003e\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u003c/sup\u003e. Additionally, arid conditions may limit the water availability and subsequent evaporative cooling effects of forests, making the radiative process dominant\u003csup\u003e\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eGiven that the background climate profoundly impacts the biophysical processes of forests\u003csup\u003e\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u003c/sup\u003e, it is reasonable to infer that, in the context of global change, the local temperature effect of forests may change correspondingly. In recent years, studies have revealed the role of rising atmospheric CO\u003csub\u003e2\u003c/sub\u003e, changing aerosols or internal climate variability in the temporal dynamics of such biophysical temperature effects\u003csup\u003e\u003cspan additionalcitationids=\"CR27 CR28\" citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e\u003c/sup\u003e. However, these findings are based on model simulations, which are subject to inaccurate physical parameterization schemes and representations of biophysical processes in forests\u003csup\u003e\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e,\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e\u003c/sup\u003e. Besides, the prior simulations only show significant changes in the biophysical temperature effects under high-emission or other ideal scenarios\u003csup\u003e\u003cspan additionalcitationids=\"CR27 CR28\" citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e\u003c/sup\u003e. On the other hand, observational studies that assess the biophysical effects of forests are generally performed in a relatively static manner\u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e,\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e,\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e. These studies focused on spatial, seasonal or diurnal heterogeneity in the mean state of the biophysical effects over a specific year or study period but ignored the potential long-term variations. Overall, direct observational evidence of temporal dynamics in local temperature effects of forests is still lacking in historical records. Evaluating these dynamic patterns and their driving mechanisms is useful for constraining model results and forming better global mitigation or regional adaptation policies.\u003c/p\u003e \u003cp\u003eThis study aims to fill the knowledge gap regarding the temporal variations (especially trends) in the local land surface temperature (LST) effects of forests and the underlying biophysical mechanisms. We focus on Europe for the following reasons: (1) Both radiative and non-radiative processes play important roles in the local LST effects of European forests, facilitating the exploration of different biophysical mechanisms underlying the long-term variation of the forests\u0026rsquo; LST effects. (2) Thanks to the dense distribution of weather stations, high-quality gridded meteorological data in Europe can help to explore potential links between changes in biophysical processes and the background climate. (3) Europe is a hotspot of climate change and forestation projects (3\u0026nbsp;billion trees to be planted in the European Union by 2030 according to the European Green Deal\u003csup\u003e\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e\u003c/sup\u003e), making our assessment more relevant and instructive. Our study aims to answer the following three questions: (1) What are the temporal patterns of the observed local LST effects of European forests in recent two decades? (2) What climate factors drive these temporal patterns? (3) What is the biophysical mechanism behind this climatic control? To address these questions, we first use observations from the Moderate Resolution Imaging Spectroradiometer (MODIS) and auxiliary satellite data to estimate the local LST effects of European forests. We then assess the temporal dynamics of the local LST effects and explore their relationships with different biophysical processes and potential climate drivers using statistical methods. To examine whether models can capture the observed temporal patterns, we evaluate the simulated trend of local LST effects of European forests in four state-of-the-art earth system models against satellite observations. Our results could inform assessments of climate mitigation or adaptation effects through future forestation in Europe.\u003c/p\u003e"},{"header":"Results and Discussion","content":"\u003cdiv id=\"Sec3\"\u003e\n \u003ch2\u003e3.1 Intensified local daytime cooling effects of forests\u003c/h2\u003e\n \u003cp\u003eWe first investigate the multi-year (2003\u0026ndash;2023) winter and summer LST effects of European forests (\u0026Delta;LST\u003csub\u003ew\u003c/sub\u003e and \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e) (Supplementary Figs.\u0026nbsp;1 to 2). Compared with nearby openlands, forests have a slight warming effect in winter (0.14\u0026thinsp;\u0026plusmn;\u0026thinsp;0.61 K, spatial mean\u0026thinsp;\u0026plusmn;\u0026thinsp;standard deviation) due to the counterbalancing of the nighttime warming signal (0.39\u0026thinsp;\u0026plusmn;\u0026thinsp;0.55 K) and the daytime cooling signal (-0.20\u0026thinsp;\u0026plusmn;\u0026thinsp;1.05 K). In contrast, forests exhibit intense land surface cooling in summer (-1.14\u0026thinsp;\u0026plusmn;\u0026thinsp;0.81 K), as the strong daytime negative LST effect (-3.16\u0026thinsp;\u0026plusmn;\u0026thinsp;1.69 K) is partially offset by the nighttime positive effect (0.34\u0026thinsp;\u0026plusmn;\u0026thinsp;0.52 K). Our results confirm previous conclusions of the contrasting LST effects of mid-latitude forests at seasonal and diurnal scales driven by different biophysical processes\u003csup\u003e\u003cspan\u003e20\u003c/span\u003e,\u003cspan\u003e21\u003c/span\u003e,\u003cspan\u003e33\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\n \u003cp\u003eWe then focus on the temporal variations in \u0026Delta;LST\u003csub\u003ew\u003c/sub\u003e and \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e over the recent two decades. Results show the declining of winter warming effects at the daily scale, with a trend of -0.06 K/decade (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.10) (Fig. \u003cspan\u003e1\u003c/span\u003ea). This decrease in the daily mean \u0026Delta;LST\u003csub\u003ew\u003c/sub\u003e is dominated by a negative trend of the daytime \u0026Delta;LST\u003csub\u003ew\u003c/sub\u003e (-0.17 K/decade, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.04), while the nighttime \u0026Delta;LST\u003csub\u003ew\u003c/sub\u003e demonstrates only interannual variability with a negligible temporal trend (0.02 K/decade, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.65) (Fig. \u003cspan\u003e1\u003c/span\u003ea). Spatially, 75% of grids show negative daytime \u0026Delta;LST\u003csub\u003ew\u003c/sub\u003e trends, with 12% of grids, which are mainly concentrated in southeastern Europe, showing significant trends (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05) (Fig. \u003cspan\u003e1\u003c/span\u003eb).\u003c/p\u003e\n \u003cp\u003eDuring the summer nighttime, the warming effect of forests shows an increasing trend of 0.05 K/decade (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.06), but the magnitude is only about one-fifth of the daytime \u0026Delta;LSTs negative trend (-0.22 K/decade, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.04) (Fig. \u003cspan\u003e1\u003c/span\u003ec). As a result, the daily mean cooling effects of forests have been enhanced in the recent two decades, with a trend of -0.08 K/decade (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.07) (Fig. \u003cspan\u003e1\u003c/span\u003ec). Spatially, about 83% of grids show negative daytime \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e trends, and 18% of grids (mainly distributed in central Europe) show significant negative trends (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05) (Fig. \u003cspan\u003e1\u003c/span\u003ed). These results show that in both summer and winter, the daytime cooling effects of forests were significantly enhanced, leading to negative trends of daily mean \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e and \u0026Delta;LST\u003csub\u003ew\u003c/sub\u003e. Further analyses mainly focus on the biophysical mechanisms behind the negative trends in daytime temperature effects.\u003c/p\u003e\n\u003c/div\u003e\n\u003ch3\u003e3.2 Impact of snow cover on daytime \u0026Delta;LST\u003csub\u003ew\u003c/sub\u003e dynamics\u003c/h3\u003e\n\u003cp\u003eThe local LST effects of mid-latitude forests in winter are generally dominated by the radiative process\u003csup\u003e\u003cspan\u003e2\u003c/span\u003e,\u003cspan\u003e10\u003c/span\u003e,\u003cspan\u003e18\u003c/span\u003e\u003c/sup\u003e. Here, we hypothesize that the variation of the forest\u0026rsquo;s albedo effect drives the observed daytime \u0026Delta;LST\u003csub\u003ew\u003c/sub\u003e dynamics (Fig. \u003cspan\u003e1\u003c/span\u003ea). We first examine the observed albedo difference between forests and nearby openlands (\u0026Delta;\u0026alpha;\u003csub\u003ew\u003c/sub\u003e) and confirm that forests are darker than openlands, as evidenced by the negative \u0026Delta;\u0026alpha;\u003csub\u003ew\u003c/sub\u003e (Supplementary Fig. 3), but \u0026Delta;\u0026alpha;\u003csub\u003ew\u003c/sub\u003e show a positive trend (0.031 unitless/decade, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.02) (Fig. \u003cspan\u003e2\u003c/span\u003ea). Spatially, \u0026Delta;\u0026alpha;\u003csub\u003ew\u003c/sub\u003e trends are positive in most areas, and about 32% of grids (concentrated in central and eastern Europe) are statistically significant (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05) (Fig. \u003cspan\u003e2\u003c/span\u003eb). These results indicate that the forest darkening effect has become weaker and the additional solar radiation absorbed by forests (compared to that absorbed by openlands) has decreased over the last two decades.\u003c/p\u003e\n\u003cp\u003eThe aforementioned positive \u0026Delta;\u0026alpha;\u003csub\u003ew\u003c/sub\u003e trend can be traced back to the stronger albedo decrease in openlands (-0.047 unitless /decade, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.02) than in forests (-0.015 unitless /decade, p\u0026thinsp;=\u0026thinsp;0.04) (Fig. \u003cspan\u003e2\u003c/span\u003ea). These surface darkening trends can be attributed to the reduced snow cover (SC\u003csub\u003ew\u003c/sub\u003e) under the global warming trend\u003csup\u003e\u003cspan\u003e34\u003c/span\u003e\u0026ndash;\u003cspan\u003e36\u003c/span\u003e\u003c/sup\u003e (Supplementary Fig. 4). Here, we confirm that the surface darkening is more susceptible to snow cover decrease in openlands than forests, as evident by the larger spatial regression slope value between \u0026alpha;\u003csub\u003ew\u003c/sub\u003e trends and SC\u003csub\u003ew\u003c/sub\u003e trends (0.59 vs 0.21, Fig. \u003cspan\u003e2\u003c/span\u003ec). This is because snow tends to be masked by tree canopies in forest ecosystems, resulting in a lower sensitivity of forest albedo to snowpack. The divergent responses of albedo to SC\u003csub\u003ew\u003c/sub\u003e are also supported by the temporal regression results, which suggest that openland albedo is about 3 times more sensitive than forest albedo to SC\u003csub\u003ew\u003c/sub\u003e (Supplementary Fig. 5). Notably, the used SC\u003csub\u003ew\u003c/sub\u003e is a function of snow mass (represented by the snow water equivalent)\u003csup\u003e\u003cspan\u003e37\u003c/span\u003e\u003c/sup\u003e, which is more governed by the macroclimate and thus shows no difference between forests and nearby openlands.\u003c/p\u003e\n\u003cp\u003eThe positive \u0026Delta;\u0026alpha;\u003csub\u003ew\u003c/sub\u003e trend induced by snow dynamics further exhibits strong temporal control on \u0026Delta;LST\u003csub\u003ew\u003c/sub\u003e. Specifically, in those grids with more pronounced SC\u003csub\u003ew\u003c/sub\u003e reductions, the \u0026Delta;\u0026alpha;\u003csub\u003ew\u003c/sub\u003e trends is larger and the negative daytime \u0026Delta;LST\u003csub\u003ew\u003c/sub\u003e trends are also stronger (Fig. \u003cspan\u003e2\u003c/span\u003ec and d). Meanwhile, the temporal relationships between SC\u003csub\u003ew\u003c/sub\u003e and \u0026Delta;LST\u003csub\u003ew\u003c/sub\u003e suggest that in those years with lower snow contents, the daytime cooling effect of forests is stronger (slope\u0026thinsp;=\u0026thinsp;1.47\u0026thinsp;\u0026plusmn;\u0026thinsp;0.59), resulting in a weaker daily warming effect (slope\u0026thinsp;=\u0026thinsp;0.73\u0026thinsp;\u0026plusmn;\u0026thinsp;0.29) (Fig. \u003cspan\u003e2\u003c/span\u003ee). However, the impact of SC\u003csub\u003ew\u003c/sub\u003e on the nighttime \u0026Delta;LST\u003csub\u003ew\u003c/sub\u003e is insignificant (slope\u0026thinsp;=\u0026thinsp;0.19\u0026thinsp;\u0026plusmn;\u0026thinsp;0.28) (Fig. \u003cspan\u003e2\u003c/span\u003ee). In addition to \u0026Delta;\u0026alpha;\u003csub\u003ew\u003c/sub\u003e, the shortwave radiative forcing of forests is related to the downward solar radiation in winter (DSR\u003csub\u003ew\u003c/sub\u003e). However, we show a poor correlation between DSR\u003csub\u003ew\u003c/sub\u003e and \u0026Delta;LST\u003csub\u003ew\u003c/sub\u003e (Supplementary Fig. 6). Overall, our statistical evidence supports the temporal control of SC\u003csub\u003ew\u003c/sub\u003e on daytime \u0026Delta;LST\u003csub\u003ew\u003c/sub\u003e dynamics through weakening the albedo difference between forests and openlands.\u003c/p\u003e\n\u003ch3\u003e3.3 Dominant role of vapor pressure deficit in daytime \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e dynamics\u003c/h3\u003e\n\u003cp\u003eIn contrast to the net warming effect observed in winter, European forests exhibit strong daytime and daily mean cooling effects during the summer (Supplementary Fig.\u0026nbsp;2), driven by non-radiative processes\u003csup\u003e\u003cspan\u003e18\u003c/span\u003e\u003c/sup\u003e. Over the past two decades, Europe has experienced significant increases in summer air temperature (AT\u003csub\u003es\u003c/sub\u003e) and atmospheric vapor pressure deficit (VPD\u003csub\u003es\u003c/sub\u003e). While there has also been an increase in summer downward solar radiation (DSR\u003csub\u003es\u003c/sub\u003e) and decreases in precipitation (P\u003csub\u003es\u003c/sub\u003e) and wind speeds (WS\u003csub\u003es\u003c/sub\u003e), the trend values for these variables are not significant (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026gt;\u0026thinsp;0.1) (Supplementary Fig.\u0026nbsp;4). We hypothesize that climate change has distinct effects on turbulent fluxes of forests and openlands, thus contributing to the temporal dynamics of \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e.\u003c/p\u003e\n\u003cp\u003eTemporal correlation analyses of interannual variabilities reveal a strong relationship between the daytime \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e and four potential climatic drivers at the grid or regional scales: VPD\u003csub\u003es\u003c/sub\u003e (mean r = -0.68), DSR\u003csub\u003es\u003c/sub\u003e (mean r = -0.52), P\u003csub\u003es\u003c/sub\u003e (mean r\u0026thinsp;=\u0026thinsp;0.55), and AT\u003csub\u003es\u003c/sub\u003e (mean r = -0.55) (Fig. \u003cspan\u003e3\u003c/span\u003ea, Supplementary Fig.\u0026nbsp;7). After accounting for covariates, the partial correlations between the daytime \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e and DSR\u003csub\u003es\u003c/sub\u003e, P\u003csub\u003es\u003c/sub\u003e, and AT\u003csub\u003es\u003c/sub\u003e are greatly reduced and sometimes even reverse in sign (Fig. \u003cspan\u003e3\u003c/span\u003ea). VPD\u003csub\u003es\u003c/sub\u003e, however, maintain a predominant negative correlation with the daytime \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e, with a spatial mean partial correlation coefficient of -0.43 (Fig. \u003cspan\u003e3\u003c/span\u003ea). About 91.3% of grids show a negative partial correlation between the VPDs and daytime \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e, with 44.8% being statistically significant (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05) (Fig. \u003cspan\u003e3\u003c/span\u003eb). On the basis of the maximum values of the absolute correlation coefficients, VPD\u003csub\u003es\u003c/sub\u003e is identified as the most important driver of the daytime \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e in most grids (Supplementary Fig. 8).\u003c/p\u003e\n\u003cp\u003eWe then built a random forest (RF) model using elevation and multi-year meteorological data to predict the daytime \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e at the grid scale (total of 19,656 samples). We excluded AT\u003csub\u003es\u003c/sub\u003e from the model inputs due to its high correlation with VPD\u003csub\u003es\u003c/sub\u003e. The model shows good accuracy on the test dataset (R\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;0.75, RMSE\u0026thinsp;=\u0026thinsp;0.79 K, Supplementary Fig.\u0026nbsp;9), supporting our further analysis. The Gini importance and mean absolute Shapley Additive Explanations (SHAP) values both suggest that VPD\u003csub\u003es\u003c/sub\u003e is the most important driver, followed by elevation, DSR\u003csub\u003es\u003c/sub\u003e, P\u003csub\u003es\u003c/sub\u003e, and WS\u003csub\u003es\u003c/sub\u003e (Fig. \u003cspan\u003e3\u003c/span\u003ec). The marginal contributions quantified by the SHAP value further reveal different impacts of climate variables (Fig. \u003cspan\u003e3\u003c/span\u003ed, Supplementary Fig.\u0026nbsp;10). VPD\u003csub\u003es\u003c/sub\u003e exhibits the most pronounced linear negative effect, while DSR\u003csub\u003es\u003c/sub\u003e and P\u003csub\u003es\u003c/sub\u003e show complex non-linear effects. Specifically, under low-radiation conditions (DSR\u003csub\u003es\u003c/sub\u003e \u0026lt; 200 W/m\u003csup\u003e2\u003c/sup\u003e), the influence of DSR\u003csub\u003es\u003c/sub\u003e is negligible; under moderate-radiation conditions (200 W/m\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;\u0026lt;\u0026thinsp;DSR\u003csub\u003es\u003c/sub\u003e \u0026lt; 250 W/m\u003csup\u003e2\u003c/sup\u003e), DSR\u003csub\u003es\u003c/sub\u003e shows a negative effect; under high-radiation conditions, the impact of DSR\u003csub\u003es\u003c/sub\u003e is highly uncertain. For P\u003csub\u003es\u003c/sub\u003e, we find a positive effect only when P\u003csub\u003es\u003c/sub\u003e \u0026lt; 300 mm.\u003c/p\u003e\n\u003cp\u003eOn the basis of the RF model, we estimate the contributions of climate variables to the long-term trend of daytime \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e. We first compared the reconstructed trends with all forcings and the observed trends, showing good consistency at the grid scale (R\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;0.64, RMSE\u0026thinsp;=\u0026thinsp;0.11 K/decade, Supplementary Fig.\u0026nbsp;9). For the regional mean, the simulated trend is -0.26 K/decade (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05), which is close to the observed trend of -0.22 K/decade (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05) (Fig. \u003cspan\u003e1\u003c/span\u003ec). Model experiments suggest that increasing VPD\u003csub\u003es\u003c/sub\u003e contributes the most to the decreasing trend in daytime \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e (-0.17 K/decade, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.06), while P\u003csub\u003es\u003c/sub\u003e (-0.02 K/decade, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.32), DSR\u003csub\u003es\u003c/sub\u003e (-0.02 K/decade, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.34), and WS\u003csub\u003es\u003c/sub\u003e (0.02 K/decade, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.65) show insignificant contributions (Fig. \u003cspan\u003e3\u003c/span\u003ee).\u003c/p\u003e\n\u003cp\u003eTo understand the biophysical mechanisms behind the influence of climate changes, especially VPD\u003csub\u003es\u003c/sub\u003e changes, on the temporal dynamics of daytime \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e, we further applied structural equation modeling (SEM) analysis at the local scale (Fig. \u003cspan\u003e4\u003c/span\u003ea). We adopt differences in the leaf area index (LAI) and latent heat (LE) between forests and nearby openlands (\u0026Delta;LAI\u003csub\u003es\u003c/sub\u003e and \u0026Delta;LE\u003csub\u003es\u003c/sub\u003e), two key factors related to the biophysical effects of forests, as intermediaries. Notably, we set \u0026Delta;LAI\u003csub\u003es\u003c/sub\u003e and \u0026Delta;LE\u003csub\u003es\u003c/sub\u003e as intermediaries rather than potential drivers because their variation can also be essentially attributed to distinctive responses of different vegetation types to climate changes. Results show that climate changes show different effects on forest and non-forest canopy structures: higher VPD\u003csub\u003es\u003c/sub\u003e (0.42\u0026thinsp;\u0026plusmn;\u0026thinsp;0.18, path value\u0026thinsp;\u0026plusmn;\u0026thinsp;standard error), AT\u003csub\u003es\u003c/sub\u003e (0.29\u0026thinsp;\u0026plusmn;\u0026thinsp;0.15), or DSR\u003csub\u003es\u003c/sub\u003e (0.14\u0026thinsp;\u0026plusmn;\u0026thinsp;0.12) enhance \u0026Delta;LAI\u003csub\u003es\u003c/sub\u003e, while P\u003csub\u003es\u003c/sub\u003e has a negative effect on \u0026Delta;LAI\u003csub\u003es\u003c/sub\u003e (-0.13\u0026thinsp;\u0026plusmn;\u0026thinsp;0.10). The resultant \u0026Delta;LAI\u003csub\u003es\u003c/sub\u003e further demonstrate a strong positive impact on \u0026Delta;LE\u003csub\u003es\u003c/sub\u003e (0.74\u0026thinsp;\u0026plusmn;\u0026thinsp;0.05). Moreover, climate changes can directly affect \u0026Delta;LE\u003csub\u003es\u003c/sub\u003e, with positive effects resulting from VPD\u003csub\u003es\u003c/sub\u003e (0.42\u0026thinsp;\u0026plusmn;\u0026thinsp;0.10) and WS\u003csub\u003es\u003c/sub\u003e (0.07\u0026thinsp;\u0026plusmn;\u0026thinsp;0.05) changes and negative effects resulting from AT\u003csub\u003es\u003c/sub\u003e (-0.24\u0026thinsp;\u0026plusmn;\u0026thinsp;0.09), DSR\u003csub\u003es\u003c/sub\u003e (-0.16\u0026thinsp;\u0026plusmn;\u0026thinsp;0.08), and P\u003csub\u003es\u003c/sub\u003e (-0.14\u0026thinsp;\u0026plusmn;\u0026thinsp;0.06) changes. \u0026Delta;LE\u003csub\u003es\u003c/sub\u003e show a predominantly negative effect on daytime \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e (-0.73\u0026thinsp;\u0026plusmn;\u0026thinsp;0.10) compared to the small direct effect of \u0026Delta;LAI\u003csub\u003es\u003c/sub\u003e (0.06\u0026thinsp;\u0026plusmn;\u0026thinsp;0.10), suggesting that it is evaporative cooling, rather than roughness- or albedo-related processes (related to the vegetation canopy structure), dominates the temporal dynamics of \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e.\u003c/p\u003e\n\u003cp\u003eEach driver could affect \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e via three pathways through \u0026Delta;LAI\u003csub\u003es\u003c/sub\u003e and \u0026Delta;LE\u003csub\u003es\u003c/sub\u003e, and we calculate the individual and summed pathway effects on the daytime \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e for all the climate variables (Fig. \u003cspan\u003e4\u003c/span\u003eb). Results show that pathways only through \u0026Delta;LAI\u003csub\u003es\u003c/sub\u003e are insignificant for all drivers (yellow bar). The summed negative effect of VPD\u003csub\u003es\u003c/sub\u003e is the strongest in magnitude (-0.51\u0026thinsp;\u0026plusmn;\u0026thinsp;0.14). The two pathway effects of VPD\u003csub\u003es\u003c/sub\u003e through \u0026Delta;LE\u003csub\u003es\u003c/sub\u003e are comparable (via \u0026Delta;LAIs: -0.23\u0026thinsp;\u0026plusmn;\u0026thinsp;0.10; not via \u0026Delta;LAIs: -0.30\u0026thinsp;\u0026plusmn;\u0026thinsp;0.08). For P\u003csub\u003es\u003c/sub\u003e, the two path effects through \u0026Delta;LE\u003csub\u003es\u003c/sub\u003e are positive, with a summed path effect of 0.17\u0026thinsp;\u0026plusmn;\u0026thinsp;0.07. For DSR\u003csub\u003es\u003c/sub\u003e and AT\u003csub\u003es\u003c/sub\u003e, the two path effects have different signs, resulting in slight total effects. Overall, our partial correlation analysis, RF analysis, and SEM analysis all indicate the predominant role of VPDs in the dynamics of daytime \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e, which is achieved through two pathways affecting the evaporative cooling effect of the forest.\u003c/p\u003e\n\u003ch3\u003e\u003cstrong\u003e3.4 Evaluation of \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e and \u0026Delta;LST\u003csub\u003ew\u003c/sub\u003e trends in four CMIP6 models\u003c/strong\u003e\u003c/h3\u003e\n\u003cp\u003eTo examine whether the observed trends of the biophysical LST effects of European forests can be reproduced by earth system models (ESMs), we use the output from the simulation of the historical period in the Coupled Model Intercomparison Project Phase 6 (CMIP6) archive. The simulated effect of forests is isolated by a similar method for observational data, using differences between the forest and opeland tiles (forest minus openland) within each model grid cell (see Methods) in four state-of-art ESMs that provide subgrid information (CESM2, CNRM-ESM2-1, GFDL-ESM4 and UKESM1-0-LL).\u003c/p\u003e\n\u003cp\u003eThe multi-model mean suggests widespread declining trends of daily mean \u0026Delta;LST\u003csub\u003ew\u003c/sub\u003e (-0.068 K/decade, p\u0026thinsp;\u0026lt;\u0026thinsp;0.01) and \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e (-0.054 K/decade, p\u0026thinsp;\u0026lt;\u0026thinsp;0.01) over Europe during 1985\u0026ndash;2014 (Fig. \u003cspan\u003e5\u003c/span\u003ea, d). These trends largely align well with the observed ones (Fig. \u003cspan\u003e1\u003c/span\u003ea, c), and the discrepancy in magnitudes can be partly attributed to differences in the period and spatial coverage of the observation and simulations (Supplementary Figs.\u0026nbsp;11 and 12). We should emphasize that the current models perform better in simulating negative trends in \u0026Delta;LST\u003csub\u003ew\u003c/sub\u003e than in \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e, as evidenced by the large intermodel differences in \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e trends simulated by four models. Specifically, the \u0026Delta;LST\u003csub\u003ew\u003c/sub\u003e trend values are similar across the four models (Fig. \u003cspan\u003e5\u003c/span\u003ea). While for \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e, CESM2 simulates a significant declining trend of -0.133 K/decade, which dominates the muti-model mean signal. However, CNRM-ESM2-1, GFDL-ESM4 and UKESM1-0-LL simulate much weaker trends of -0.042, -0.018, and \u0026minus;\u0026thinsp;0.027 K/decade, respectively (Fig. \u003cspan\u003e5\u003c/span\u003ed). In addition, differences are found in the multi-year mean \u0026Delta;LST\u003csub\u003ew\u003c/sub\u003e or \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e from the four models (Fig. \u003cspan\u003e5\u003c/span\u003ea, d), which can be attributed to the biases in surface energy partitioning and albedo responses to land cover changes\u003csup\u003e\u003cspan\u003e38\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eWe further evaluate the model representation of biophysical processes behind the negative trends in \u0026Delta;LST\u003csub\u003ew\u003c/sub\u003e. Observational results indicate that the decreasing trend in the openland albedo is 3.13 times greater than that in the forest albedo, whereas all four models underestimate this ratio, with the muti-model mean of 1.94 (Fig. \u003cspan\u003e5\u003c/span\u003eb). Moreover, we show that only CESM2 (-2.41\u0026thinsp;\u0026plusmn;\u0026thinsp;0.70) and UKESM1-0-LL (-2.93\u0026thinsp;\u0026plusmn;\u0026thinsp;2.22) can approximately reproduce the observed slope (2.42\u0026thinsp;\u0026plusmn;\u0026thinsp;1.01) between \u0026Delta;\u0026alpha;\u003csub\u003ew\u003c/sub\u003e and \u0026Delta;LST\u003csub\u003ew\u003c/sub\u003e (Fig. \u003cspan\u003e5\u003c/span\u003ec). CNRM-ESM2-1 (-6.29\u0026thinsp;\u0026plusmn;\u0026thinsp;1.18), GFDL-ESM4 (-12.37\u0026thinsp;\u0026plusmn;\u0026thinsp;3.05) and the multi-model mean (5.57\u0026thinsp;\u0026plusmn;\u0026thinsp;1.39) significantly overestimate this slope (Fig. \u003cspan\u003e5\u003c/span\u003ec). These may lead to bias in estimations of \u0026Delta;LST\u003csub\u003ew\u003c/sub\u003e variation in response to climate changes, especially the snow cover decline. For the summer results, we conduct partial correlation analysis between \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e and climate variables (Fig. \u003cspan\u003e3\u003c/span\u003ea). The wind speed is excluded here considering the model data availability. Among all four models, only CNRM-ESM2-1 captures the dominant role of VPD\u003csub\u003es\u003c/sub\u003e on \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e dynamics and reproduces the impacts of other climate variables (Fig. \u003cspan\u003e5\u003c/span\u003ee). Overall, although negative trends in \u0026Delta;LST\u003csub\u003ew\u003c/sub\u003e and \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e can be captured by the models, there are still discrepancies between the observations and simulations as well as intermodel differences in the mechanisms underlying the biophysical effect of forests. Our results are informative for improving current models for better simulations of local climate effects of European forests and their temporal dynamics in a warming world.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e3.5 Predicted local LST effect of European forests under four SSP scenarios\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIn the recent two decades, Europe has experienced a remarkable warming trend, accompanied by stronger atmospheric drought in summer and less snow cover (Supplementary Fig. 4). Since the warming trend is unlikely to be reversed until carbon neutrality, we anticipate that the negative trends in \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e and \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e will persist. Here, we project the daily mean \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e and \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e under four SSPs (SSP126, SSP245, SSP370, and SSP585) by combining the simulated future climate changes, as well as the established relationships between climate drivers and satellite \u0026Delta;LST\u003csub\u003ew\u003c/sub\u003e and \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e in historical records (see Methods). We then compare the predicted trends of \u0026Delta;LST\u003csub\u003ew\u003c/sub\u003e and \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e with the future trends of winter LST (LST\u003csub\u003ew\u003c/sub\u003e) and summer LST (LST\u003csub\u003es\u003c/sub\u003e) within two periods (2020\u0026ndash;2060 and 2060\u0026ndash;2100). We calculate the ratio of\u0026nbsp;\u0026Delta;LST\u003csub\u003ew\u003c/sub\u003e to LST\u003csub\u003ew\u003c/sub\u003e and \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e to LST\u003csub\u003es\u003c/sub\u003e to assess the future relative magnitude of the biophysical mitigation of European forests.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eOver 2020\u0026ndash;2060, the negative trends in \u0026Delta;LST\u003csub\u003ew\u003c/sub\u003e and \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e persist, generally with stronger trends in warmer SSPs (Fig. 6). In winter, the decreasing trends in \u0026Delta;LST\u003csub\u003ew\u003c/sub\u003e can offset -2.9 \u0026plusmn; 1.9%, -5.3 \u0026plusmn; 2.3% and -4.5 \u0026plusmn; 2.6% of the LST warming trends under the SSP245, SSP370 and SSP585 scenarios, respectively (Fig. 6c, e and g). For the SSP126 scenario, however, the negative trend in \u0026Delta;LST\u003csub\u003ew\u003c/sub\u003e and its mitigation effect are insignificant (Fig. 6a). In summer, the ratios of \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e trends to warming are more pronounced, with values of -8.4 \u0026plusmn; 3.4%, -9.1 \u0026plusmn; 2.7%, -8.8 \u0026plusmn; 2.4% and -9.9 \u0026plusmn; 2.0% under the SSP126, SSP245, SSP370 and SSP585 scenarios, respectively (Fig. 6b, d, f, g).\u003c/p\u003e\n\u003cp\u003eOver 2060\u0026ndash;2100, trends in \u0026Delta;LST\u003csub\u003ew\u003c/sub\u003e and \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e are found to be insignificant in the SSP126 scenario, as the climate change slows down due to reduced emissions. For the remaining three scenarios, the negative trends in \u0026Delta;LST\u003csub\u003ew\u003c/sub\u003e are weakened compared with those from 2020\u0026ndash;2060, and the ratios are significant only under the SSP370 and SSP585 scenarios (-2.2 \u0026plusmn; 1.4% and -2.2 \u0026plusmn; 0.6%, respectively) (Fig. 6c, e and g). This could be attributed to lower SC\u003csub\u003ew\u003c/sub\u003e sensitivity to temperature under warmer conditions (Supplementary Fig. 13), leading to weaker \u0026Delta;LST\u003csub\u003ew\u003c/sub\u003e trends in the warming world. For \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e, the magnitudes during\u0026nbsp;2060-2100\u0026nbsp;are close to those during 2020-2060, with\u0026nbsp;ratios of -8.8 \u0026plusmn; 6.4%, -11.1 \u0026plusmn; 3.0% and -11.7 \u0026plusmn; 1.6% under the SSP2-4-5, SSP370 and SSP585 scenarios, respectively (Fig. 6d, f, g). Overall, the roles of negative biophysical LST trends in LST mitigation are projected to be stronger in summer than in winter. Notably, the revealed mitigation effects arise from varying biophysical feedbacks to climate change in existing stable forests. The biophysical cooling of forest ecosystems may be stronger (especially in summer with negative \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e) if more land areas are forested in the future.\u003c/p\u003e\n\u003ch3\u003e3.6 Discussion\u003c/h3\u003e\n\u003cp\u003ePrevious observational studies only focused on the spatial, diurnal or seasonal heterogeneity of biophysical LST sensitivity to forest cover change or the mechanisms behind these patterns\u003csup\u003e\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e,\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e,\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e,\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e\u003c/sup\u003e. These studies are performed in a relatively static manner, using the results of a single year or the mean of a period, ignoring potential annual temporal variability. Here, our satellite evidence shows the emergent trend in the biophysical temperature effects of European forests over the last two decades (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). Our results suggest that previous observational assessments may underestimate the daytime cooling effect of forests in the changing world. At least in the European region, the temporal dynamic of forest biophysical effects should be taken into consideration. Meanwhile, prior simulations mostly reported responses of the biophysical effects of forests to climate changes in high-emissions or ideal scenarios\u003csup\u003e\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e,\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e,\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e\u003c/sup\u003e. However, we have detected robust trends in the biophysical effects of forests in historical records, despite the limited climate change during the study period. This implies that the sensitivity of the biophysical effect to climate change may be much stronger than previously thought, which merits the attention of researchers in the fields of climate change and earth system modeling.\u003c/p\u003e \u003cp\u003eSince we estimate the LST effects based on fixed and stable forest and openland pixels with minimum disturbance, these negative trends can be solely attributed to background climate changes in Europe. We link the temporal change of the winter LST effect to the radiative process, represented by Δα\u003csub\u003ew\u003c/sub\u003e, and further attribute the negative ΔLST\u003csub\u003ew\u003c/sub\u003e trend to the decreasing SC\u003csub\u003ew\u003c/sub\u003e in Europe. Specifically, the albedo of openland is about three times more sensitive to SC\u003csub\u003ew\u003c/sub\u003e than that of forests. Thus, decreasing SC\u003csub\u003ew\u003c/sub\u003e in the recent two decades has weakened Δα\u003csub\u003ew\u003c/sub\u003e (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ec), and, consequently the radiative forcing of European forests. As a result, the relative contribution from non-radiative processes (i.e., higher heat exchange efficiency in the rougher forested land surface) may become more dominant, leading to stronger daytime cooling and attenuated daily warming effects of forests in European winter (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ea, b). This result suggests that the transitional latitudes of current forest cooling and warming effects will migrate northwards in a warming world\u003csup\u003e\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e. The obtained relationships between SC\u003csub\u003ew\u003c/sub\u003e and biophysical factors of forests are useful for making reasonable predictions or constraining model simulations of future forest-snow interactions. In addition, existing studies have suggested that high latitudes and altitudes are considered unsuitable for forest restoration, as the positive radiative forcing driven by the albedo effect surpasses the negative radiative forcing driven by the carbon effect\u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e,\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e\u003c/sup\u003e. Our satellite evidence implies that the timing of tree growth and future climate change (especially the snow amount) should be fully considered when prioritizing the geolocation of forestation practices. In the context of global warming, the snow cover in winter or spring is projected to maintain a decreasing trend, which may significantly reduce the biophysical radiative forcings of forests and make high-latitude and high-altitude regions becoming suitable for forestation.\u003c/p\u003e \u003cp\u003eIn summer, the biophysical mechanisms behind enhanced forest cooling are more complicated than those in winter. Previous studies have shown that the summer cooling effect of forests is dominated by non-radiative processes\u003csup\u003e\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u003c/sup\u003e. The evaporative cooling effect (represented by ΔLE\u003csub\u003es\u003c/sub\u003e here) plays the most important role\u003csup\u003e\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e, which is influenced by both vegetation structure differences in forests and openlands (ΔLAI\u003csub\u003es\u003c/sub\u003e) and meteorological conditions\u003csup\u003e\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e\u003c/sup\u003e. Using a range of statistical methods, we find that increasing VPD\u003csub\u003es\u003c/sub\u003e is the predominant factor contributing to both the interannual variation and the negative trend in ΔLST\u003csub\u003es\u003c/sub\u003e (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). Path analysis shows that VPD\u003csub\u003es\u003c/sub\u003e affects ΔLST\u003csub\u003es\u003c/sub\u003e dynamics through two main pathways (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e). First, a positive effect of VPD\u003csub\u003es\u003c/sub\u003e on ΔLAI\u003csub\u003es\u003c/sub\u003e implies that, as the atmosphere becomes drier, the vegetation structure difference between forests and openlands becomes more evident. The higher ΔLAI\u003csub\u003es\u003c/sub\u003e could further stimulate ΔLE\u003csub\u003es\u003c/sub\u003e and boost the cooling efficiency of forests. The impact of VPD\u003csub\u003es\u003c/sub\u003e on vegetation has been well-documented: increasing vapor pressure deficit (VPD) could trigger stomatal closure, inhibit photosynthesis, and increase vegetation mortality\u003csup\u003e\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e,\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e\u003c/sup\u003e. Our results suggest a stronger negative effect of VPD on non-forest vegetation than on forests, which is consistent with previous findings of heterogeneity in the vegetation response to VPD\u003csup\u003e\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e,\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e\u003c/sup\u003e. The potential reason is that forest ecosystems with high species richness and deep roots are more stable and resilient to drought\u003csup\u003e\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e,\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e\u003c/sup\u003e. Second, even under constant vegetation conditions, higher VPD\u003csub\u003es\u003c/sub\u003e directly amplifies ΔLE\u003csub\u003es\u003c/sub\u003e and enhances the evaporative cooling effect of forests. ET in Europe is generally driven by the atmospheric demand rather than the water supply. Thus, ET tends to increase in response to higher VPD\u003csup\u003e\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e\u003c/sup\u003e. Conversely, plants could also reduce water loss through stomatal closure with increasing VPD\u003csup\u003e\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e\u003c/sup\u003e. Our results support previous conclusions concerning the diverse effects of VPD on ET in different vegetation types\u003csup\u003e\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e,\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e\u003c/sup\u003e and suggest that the atmospheric water demand-driven positive effect overwhelms the stomatal conductance-driven negative effect. Overall, the revealed impact of VPD\u003csub\u003es\u003c/sub\u003e on ΔLST\u003csub\u003es\u003c/sub\u003e essentially reflects the diverse vegetation physiological responses to climate changes. Through these two pathways, the rising VPD\u003csub\u003es\u003c/sub\u003e in the recent two decades positively affected the evaporative cooling effect of European forests and drove the observed negative trend in the daytime ΔLST\u003csub\u003es\u003c/sub\u003e.\u003c/p\u003e \u003cp\u003eIn addition to LST, the biophysical effects of European forests also affect the hydrological process. In winter, forest cover can delay or accelerate the melting of snowpack compared to the adjacent openlands, depending on the background climate and canopy density\u003csup\u003e\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e\u003c/sup\u003e. However, the revealed increasing daytime cooling effect favors the retention of winter snow in forests and further increases streamflow from snowmelt in the following season. As a result, forests may buffer the impact of global warming on fresh-water availability\u003csup\u003e\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e\u003c/sup\u003e, plant phenology\u003csup\u003e\u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e54\u003c/span\u003e\u003c/sup\u003e and food production\u003csup\u003e\u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e55\u003c/span\u003e\u003c/sup\u003e through altering the snow melting process in Europe. During summer, forests maintain high ET, which further promotes precipitation in downwind areas and accelerates the water cycle process\u003csup\u003e\u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e56\u003c/span\u003e\u003c/sup\u003e. The enhanced cooling effect is achieved at the expense of water loss through transpiration, which may increase the pressure on terrestrial water availability and pose threats to ecosystem productivity\u003csup\u003e\u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e57\u003c/span\u003e\u003c/sup\u003e and freshwater resources\u003csup\u003e\u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e58\u003c/span\u003e\u003c/sup\u003e. The vegetation-climate feedback can promote precipitation through local moisture recycling and thus have a positive effect on water availability, which can offset the direct negative effect of ET on water availability\u003csup\u003e\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e59\u003c/span\u003e\u003c/sup\u003e. Considering the drying or wetting trend in the future, the impact of potential forestation in Europe should be discussed according to the specific background climatic conditions due to the trade-off between greater cooling effects and less water availability\u003csup\u003e\u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e60\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eThere are several potential caveats or issues when interpreting our results. First, the adopted “space-for-time” method provides priori estimates of local temperature effects\u003csup\u003e\u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e61\u003c/span\u003e\u003c/sup\u003e, which ignore the atmospheric feedback of potential afforestation (e.g., cloud formation\u003csup\u003e\u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e62\u003c/span\u003e\u003c/sup\u003e and atmospheric circulation\u003csup\u003e\u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e63\u003c/span\u003e\u003c/sup\u003e), as well as the asymmetric patterns in the biophysical effects of forest gain and loss\u003csup\u003e\u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e64\u003c/span\u003e,\u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e65\u003c/span\u003e\u003c/sup\u003e. Second, the projected stronger forest local cooling effect in the future implies that planting trees in proper areas remains a promising local solution against the risk of warming (Supplementary Fig.\u0026nbsp;13), especially in highly populated regions. Nonetheless, tree restoration is not a panacea for climate change. Although our results highlight the stronger negative biophysical feedback to the warming trend in forests, this negative feedback is impossible to fully offset the warming trend driven by rising atmospheric CO\u003csub\u003e2\u003c/sub\u003e (Supplementary Text 1 and 2). Therefore, reducing greenhouse gas emissions and developing clean energy remain fundamental to limiting climate change. Third, while we reveal the influences of changes in climate variables on the biophysical effects of forests, the role of increasing atmospheric CO\u003csub\u003e2\u003c/sub\u003e is not considered in this study. The CO\u003csub\u003e2\u003c/sub\u003e fertilization effect can enhance vegetation leaf area\u003csup\u003e\u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e66\u003c/span\u003e\u003c/sup\u003e, boost transpiration\u003csup\u003e\u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e67\u003c/span\u003e\u003c/sup\u003e and cool the land surface\u003csup\u003e\u003cspan citationid=\"CR68\" class=\"CitationRef\"\u003e68\u003c/span\u003e\u003c/sup\u003e. Meanwhile, the CO\u003csub\u003e2\u003c/sub\u003e physiological forcing also stimulates stomatal closure\u003csup\u003e\u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e69\u003c/span\u003e\u003c/sup\u003e, reduces transpiration\u003csup\u003e\u003cspan citationid=\"CR70\" class=\"CitationRef\"\u003e70\u003c/span\u003e\u003c/sup\u003e and increases local temperature\u003csup\u003e\u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e71\u003c/span\u003e\u003c/sup\u003e. These two opposing mechanisms have diverse effects on forest and non-forest vegetation and affect forest cooling efficiency. The stomatal closure-driven negative effect may overwhelm the LAI-driven positive effect\u003csup\u003e\u003cspan citationid=\"CR72\" class=\"CitationRef\"\u003e72\u003c/span\u003e\u003c/sup\u003e. Thus, ignoring the physiological response of vegetation may result in the overestimation of the forest cooling effect (especially in summer) under high atmospheric CO\u003csub\u003e2\u003c/sub\u003e concentration conditions. Finally, the evaluated LST is the thermal radiative temperature of the land (vegetation foliage, canopy, and soil), which is biologically relevant and has advantages in describing energy, water, and carbon fluxes\u003csup\u003e\u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e73\u003c/span\u003e\u003c/sup\u003e. However, in climate change assessments, the screen height air temperature, rather than the LST, is a more widely used metric in terrestrial regions\u003csup\u003e\u003cspan citationid=\"CR74\" class=\"CitationRef\"\u003e74\u003c/span\u003e\u003c/sup\u003e. Converting LST effects to corresponding air temperature effects is of great importance but remains challenging. Nonetheless, our LST-based results are useful for model improvement and informing the increasing climate benefits of European forests.\u003c/p\u003e \u003cp\u003eOverall, our findings provide solid satellite evidence that the local LST effects of European forests have varied with climate changes and have shown significant negative trends in the last two decades. This temporal variation pattern may be even more evident when different forest age dynamics are considered\u003csup\u003e\u003cspan citationid=\"CR75\" class=\"CitationRef\"\u003e75\u003c/span\u003e\u003c/sup\u003e. Our results challenge the previous view that biophysical effects are marginally time-dependent in the recent two decades and advocate that temporal dynamics should be considered in biophysical effect assessments. Moreover, the revealed dominant role of snow cover and atmospheric moisture deficit in controlling biophysical sensitivity emphasizes the importance of accurate parameterization of related processes in land surface models or vegetation dynamics models (e.g., snow masking effects and climatic constraints on vegetation leaf area), which are directly relevant to the projected future climate effects of forests and forming forest-based climate policies.\u003c/p\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003cdiv id=\"Sec9\" class=\"Section3\"\u003e \u003c/div\u003e \u003c/div\u003e"},{"header":"Materials and methods","content":"\u003ch2\u003e4.1 Satellite data\u003c/h2\u003e\n\u003cp\u003eSatellite observations are adopted to estimate the potential biophysical effects of forestation. The satellite data used in this study are divided into two categories: MODIS products and other auxiliary land surface data. All satellite data are processed and outputted with a spatial resolution of 1 km (0.0083\u0026deg;).\u003c/p\u003e\n\u003cp\u003eLST data are obtained from the MYD11A1 dataset, which is retrieved using the split-window algorithm\u003csup\u003e\u003cspan\u003e73\u003c/span\u003e\u003c/sup\u003e. The MYD11 LST provides daily instantaneous observations at approximately 1:30 PM (daytime) and 1:30 AM (nighttime), corresponding to the daily maximum and minimum temperatures. Observations with LST errors greater than 2 K are filtered out according to the quality control flag. All high-quality observations are used to aggregate the summer (June, July, and August) and winter mean LST (December, January, and February) for both daytime and nighttime. We generate summer and winter LST from 2003 to 2023. It is worth noting that winter LST for a given year is averaged from the January and February LST of that year and the December LST of the previous year. For example, the 2003 winter LST is the mean value from December 2002 to February 2003. This aggregation principle is applied to all seasonal variables. We calculate the daily mean LST of summer and winter through the weighted combination of daytime and nighttime LST for both seasons\u003csup\u003e\u003cspan\u003e76\u003c/span\u003e\u003c/sup\u003e. Specifically, the weight for daytime and nighttime LSTs are 0.5637 and 0.4244, respectively, and the constant term is 2.75 K.\u003c/p\u003e\n\u003cp\u003eLE data are derived from the MOD16A2GF (MOD16A2 Gap-filled) dataset, with an original resolution of 8-day and 500 m. We select all observations retrieved by the main algorithm (Penman-Monteith model) and convert the LE unit to W/m\u0026sup2;. Then, all available records are spatially and temporally aggregated to 1 km summer means from 2003 to 2023. Leaf area index (LAI) data are obtained from the MCD15A3H dataset with a 4-day and 500 m resolution. All LAI values retrieved by the main Look-up-Table (LUT) method are aggregated to 1 km summer means from 2003 to 2023. The albedo data are from the MCD43A3 dataset, which contains daily 500 m black-sky and white-sky albedo for the shortwave band. We average the black-sky and white-sky albedo and then aggregate them to 1 km winter means from 2003 to 2023. Yearly land use/land cover data from 2003 to 2022 are derived from the MCD12Q1 dataset, with an original spatial resolution of 250 m. The land cover results under the International Geosphere-Biosphere Program (IGBP) classification scheme are used. The 250 m land cover maps are spatially aggregated to a 1 km scale based on the majority type.\u003c/p\u003e\n\u003cp\u003eOther auxiliary land surface data include forest loss data from the Global Forest Change (GFC) dataset\u003csup\u003e\u003cspan\u003e77\u003c/span\u003e\u003c/sup\u003e, impervious surface data from the Global Artificial Impervious Area (GAIA) dataset\u003csup\u003e\u003cspan\u003e78\u003c/span\u003e\u003c/sup\u003e, and digital elevation model (DEM) from the GMTED2010 dataset. The GFC dataset provides annual forest loss masks from 2000 to 2023 with 30 m spatial resolution based on time series analysis. All layers from 2003 to 2023 are composited, and a mask of forest loss during the study period is generated. The composited map is then spatially aggregated to 1 km resolution to match the MODIS data. The aggregated values range between 0 to 1, indicating the proportion of 30 m pixels where forest loss has occurred. GAIA data document the impervious surface mask from 1985 to 2018 with a spatial resolution of 30 m. Similar to the GFC datasets, the data from the last year (2018) are aggregated to a 1 km resolution impervious surface percentage map for further analysis. The DEM data of GMTED2010 has a spatial resolution of approximately 7.5 arc-seconds. We aggregate the original DEM to a 1 km spatial resolution for further analysis.\u003c/p\u003e\n\u003ch3\u003e4.2 Reanalysis data\u003c/h3\u003e\n\u003cp\u003eMeteorological variables are obtained from the monthly ERA-5 Land reanalysis dataset\u003csup\u003e\u003cspan\u003e79\u003c/span\u003e\u003c/sup\u003e with a spatial resolution of 0.1\u0026deg;. ERA5-Land datasets have been widely used for analysing the interaction between climate and terrestrial ecosystems. Here, several variables, including 2 m air temperature (AT), 2 m dewpoint temperature (DT), precipitation, downward shortwave radiation, wind speed and snow cover, are downloaded. We aggregate these climatic variables into winter and summer means with a spatial resolution of 0.5\u0026deg;. AT and DT were used to calculate the atmospheric vapor pressure deficit (VPD) using the following equations\u003csup\u003e\u003cspan\u003e43\u003c/span\u003e\u003c/sup\u003e:\u003c/p\u003e\n\u003cdiv id=\"Equ1\"\u003e\n \u003cdiv id=\"FileID_Equ1\" name=\"EquationSource\"\u003e$$\\:VPD=e\\left(AT\\right)-e\\left(DT\\right)$$\u003c/div\u003e\n \u003cdiv\u003e1\u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Equ2\"\u003e\n \u003cdiv id=\"FileID_Equ2\" name=\"EquationSource\"\u003e$$\\:e\\left(T\\right)=6.112\\cdot\\:{f}_{w}\\cdot\\:exp\\left(\\frac{17.67\\cdot\\:T}{T+243.5}\\right)$$\u003c/div\u003e\n \u003cdiv\u003e2\u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Equ3\"\u003e\n \u003cdiv id=\"FileID_Equ3\" name=\"EquationSource\"\u003e$$\\:{f}_{w}=1.0007+3.46\\times\\:1{0}^{-6}\\cdot\\:AP$$\u003c/div\u003e\n \u003cdiv\u003e3\u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Equ4\"\u003e\n \u003cdiv id=\"FileID_Equ4\" name=\"EquationSource\"\u003e$$\\:AP=1013.15{\\left(\\frac{T+273.16}{T+273.16+0.0065\\cdot\\:Z}\\right)}^{5.625}$$\u003c/div\u003e\n \u003cdiv\u003e4\u003c/div\u003e\n\u003c/div\u003e\n\u003cp\u003ewhere AT and DT indicate the 2 m air temperature and dewpoint temperature (Celsius), respectively; \u003cspan\u003e\u003cspan\u003e\\(\\:e\\left(T\\right)\\)\u003c/span\u003e\u003c/span\u003e indicates the saturation water vapor (hPa) at a given temperature T (Celsius); and AP denotes the air pressure (hPa); Z indicates elevation (m).\u003c/p\u003e\n\u003cdiv id=\"Sec11\"\u003e\n \u003ch2\u003e4.3 Observed biophysical effects of forests on LST\u003c/h2\u003e\n \u003cp\u003eThis paper explores the temporal dynamics of forests\u0026rsquo; biophysical processes under a changing climate. To ensure a sufficient sample size for robust results, we employ a space-for-time substitution strategy to estimate the a priori potential effect, rather than using the space-and-time approach for the a posteriori actual effect\u003csup\u003e\u003cspan\u003e61\u003c/span\u003e,\u003cspan\u003e80\u003c/span\u003e\u003c/sup\u003e. Specifically, we assume that all 1 km pixels within a 0.25\u0026deg; grid (comprising 30\u0026times;30 pixels) share the same background climate\u003csup\u003e\u003cspan\u003e21\u003c/span\u003e\u003c/sup\u003e, and the potential effect of forests on the LST (\u0026Delta;LST) can be estimated by the difference between the mean LSTs of forested areas (\u003cspan\u003e\u003cspan\u003e\\(\\:\\stackrel{-}{{LST}_{f}}\\)\u003c/span\u003e\u003c/span\u003e) and non-forest openlands (\u003cspan\u003e\u003cspan\u003e\\(\\:\\stackrel{-}{{LST}_{o}}\\)\u003c/span\u003e\u003c/span\u003e) within the grid:\u003c/p\u003e\n \u003cdiv id=\"Equ5\"\u003e\n \u003cdiv id=\"FileID_Equ5\" name=\"EquationSource\"\u003e$$\\:\\varDelta\\:LST=\\stackrel{-}{{LST}_{f}}-\\stackrel{-}{{LST}_{o}}$$\u003c/div\u003e\n \u003cdiv\u003e5\u003c/div\u003e\n \u003c/div\u003e\n \u003cp\u003eAccording to the MCD12Q1 land cover product, forest pixels should belong to evergreen needleleaf forests (ENFs), deciduous broadleaf forests (DBFs), evergreen broadleaf forests (EBFs) and mixed forests (MFs). Openland pixels include grasslands (GRA) and croplands (CRO). To minimize the potential impact of direct human activities, such as anthropogenic heat emissions and land use/land cover changes, on the \u0026Delta;LST time series, we apply additional filtering criteria to the potential samples: (a) the land cover type remains unchanged from 2003 to 2022; (b) the forest loss area should be less than 5% during 2003\u0026ndash;2023 based on GFC data; and (c) the impervious surface percentage should be less than 5% according to GAIA data. In addition, we calculate \u0026Delta;LST only if both the forest and openland sample sizes exceed 5 pixels. We discard \u0026Delta;LST calculations for grids where the mean elevation exceeds 500 m to avoid potential bias induced by mountainous terrain. The resultant mapping of the biophysical LST effect is spatially aggregated from the 0.25\u0026deg; scale to the 0.5\u0026deg; scale to match the meteorological data and mitigate the impact of potential outliers.\u003c/p\u003e\n \u003cp\u003eWe calculate \u0026Delta;LST for both winter and summer from 2003 to 2023. The criteria ensure that the forest and non-forest pixels remain consistent and stable during the study period, allowing us to attribute the temporal dynamics of \u0026Delta;LST primarily to background climate change. By substituting the target variable, we also calculate the winter albedo, summer LE and summer LAI difference between forest and non-forest areas.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec12\"\u003e\n \u003ch2\u003e4.4 Simulated biophysical effects of forests on LST\u003c/h2\u003e\n \u003cp\u003eTo obtain the simulated biophysical effects of forests on LST over the historical period, we use the output from the historical simulation of the CMIP6\u003csup\u003e81\u003c/sup\u003e. The historical simulation covers the period of 1850\u0026ndash;2014 and is forced by externally imposed and time-varying natural (e.g., solar variability and volcanic aerosols) and anthropogenic (e.g., greenhouse gases, aerosols, and land use/land cover changes) forcings. The last 30-year (i.e., 1985\u0026ndash;2014) simulation output is used to calculate the biophysical effects of forests and their trends. Here, we choose the last 30 years because it is closest to our observational period (i.e., 2003\u0026ndash;2023). Moreover, the 30-year time series can be considered long enough to minimize the influence of internal climate variabilities on the long-term trend.\u003c/p\u003e\n \u003cp\u003eA few models that participate in the historical simulation report the subgrid information following the data reporting protocol of the Land Use Model Intercomparison Project (LUMIP)\u003csup\u003e\u003cspan\u003e82\u003c/span\u003e\u003c/sup\u003e. Specifically, these models provide surface diagnostic variables outputs on four subgrid tiles, including primary and secondary land (psl, including trees, grasslands, barrens and vegetated wetlands), cropland (crp), pastureland (pst) and urban land, in each model grid cell. To create forest and openland comparisons within a grid cell for further analysis, we only retain the grid cells with the ratio of the tree cover fraction to the psl fraction exceeding 10%; this threshold is consistent with that used by the Food and Agricultural Organization to define forest cover\u003csup\u003e\u003cspan\u003e83\u003c/span\u003e\u003c/sup\u003e. For these selected grid cells, therefore, the psl tile can be considered as a forest tile. The crp and pst tiles are combined into an openland tile to keep consistency with the satellite observations. Since the forest and openland tiles in a grid cell are forced by the same atmospheric inputs (e.g., downward radiation, air temperature and precipitation), the background climate is the same between the two tiles within a grid cell. As such, differences in surface diagnostic variables (e.g., LST) between the two tiles are solely attributed to differences in land cover\u003csup\u003e\u003cspan\u003e38\u003c/span\u003e,\u003cspan\u003e84\u003c/span\u003e\u003c/sup\u003e. The local biophysical effects of forests on LST (\u003cspan\u003e\u003cspan\u003e\\(\\:\\varDelta\\:LST\\)\u003c/span\u003e\u003c/span\u003e) in a grid cell can be expressed as follows\u003csup\u003e\u003cspan\u003e38\u003c/span\u003e,\u003cspan\u003e84\u003c/span\u003e\u003c/sup\u003e:\u003c/p\u003e\n \u003cdiv id=\"Equ6\"\u003e\n \u003cdiv id=\"FileID_Equ6\" name=\"EquationSource\"\u003e$$\\:\\varDelta\\:LST={LST}_{psl}-{LST}_{crp/pst}$$\u003c/div\u003e\n \u003cdiv\u003e6\u003c/div\u003e\n \u003c/div\u003e\n \u003cp\u003ewhere \u003cspan\u003e\u003cspan\u003e\\(\\:{LST}_{psl}\\)\u003c/span\u003e\u003c/span\u003e denotes the forest LST value (psl tile), and \u003cspan\u003e\u003cspan\u003e\\(\\:{LST}_{crp/pst}\\)\u003c/span\u003e\u003c/span\u003e indicates the openland LST value (crp and pst combination). Specifically, if the outputs on both the crp and pst tiles are available, their arithmetic mean value is used; otherwise, the available one is used. LST can be replaced by other surface diagnostic variables (e.g., albedo) to quantify the biophysical effects of forests on corresponding variables. It should be emphasized that the forest effects identified by this method should be interpreted as the result of a complete conversion from openlands to forests in a grid cell. In other words, the identified forest effects are insensitive to the area fraction of the forest or openland tile in a grid cell. Therefore, while historical land use and land cover changes are prevalent in the simulations, they do not influence the identified forest effects.\u003c/p\u003e\n \u003cp\u003eFour models (CESM2, CNRM-ESM2-1, GFDL-ESM4 and UKESM1-0-LL) are used for the subgrid analysis, as they provide the required output for the historical simulation. CESM2 and GFDL-ESM4 each have only one member that provides the subgrid output, and the available one (r10i1p1f1 for CESM2 and r1i1p1f1 for GFDL-ESM4) is used. CNRM-ESM2-1 and UKESM1-0-LL each have more than one member, and the first one (r1i1p1f2 for CNRM-ESM2-1 and r2i1p1f2 for UKESM1-0-LL) is used for the sake of keeping consistency with other models as well as saving computational cost.\u003c/p\u003e\n \u003ch2\u003e4.5 Predicted local LST effects of forest in future\u003c/h2\u003e\n \u003cp\u003eWe also predict the biophysical effects of forests on LST base on the statistical relationships between climate variables and observation derived \u0026Delta;LST\u003csub\u003ew\u003c/sub\u003e or \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e, as well as the simulated climate changes in future. Specifically, the relationships are established using the temporal samples of \u0026Delta;LST\u003csub\u003ew\u003c/sub\u003e or \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e (21 samples from 2003\u0026ndash;2023) and corresponding reanalysis climate data. For winter, we build a simple linear regression model to link \u0026Delta;LST\u003csub\u003ew\u003c/sub\u003e and SC\u003csub\u003ew\u003c/sub\u003e (statsmodels package in Python). For summer, we build a ridge regression model for \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e and four climate variables, including AT\u003csub\u003es\u003c/sub\u003e, VPD\u003csub\u003es\u003c/sub\u003e, DSR\u003csub\u003es\u003c/sub\u003e and P\u003csub\u003es\u003c/sub\u003e (sklearn package in Python). A ridge regression model is used to obtain robust estimations when the input variables are highly correlated.\u0026nbsp;\u003c/p\u003e\n \u003cp\u003eThe established regression models are then combined with projected climate data to obtain future \u0026Delta;LST\u003csub\u003ew\u003c/sub\u003e or \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e. The projected climate data are from CMIP6 monthly simulations of four SSP scenarios (2015\u0026ndash;2100), including SSP126, SSP245, SSP370 and SSP585. SSP126 indicates a low-emission path in which the global society prioritizes sustainable development; SSP245 suggests a moderate scenario in which social, economic, and technological trends follow historical patterns; SSP370 describes a regional rivalry pathway in which global competition leads to slower technological progress and continued reliance on fossil fuels. SSP585 is the most extreme scenario, with minimal efforts to reduce greenhouse gas emissions. We use the simulations from the r1i1p1f1 series to reduce the differences among different models.\u003c/p\u003e\n \u003cp\u003eThe climate variables used for predicting local LST effects of forest in future includes winter snow area fraction, summer air temperature, specific humidity, downward shortwave radiation and precipitation from ten models (BCC-CSM2-MR, CAS-ESM2-0, CESM2-WACCM, CMCC-CM2-SR5, CMCC-ESM2, EC-Earth3-Veg, MPI-ESM1-2-HR, MRI-ESM2-0, NorESM2-MM and TaiESM1). All model outputs are resampled to the 1\u0026deg; spatial resolution. Summer VPD data are calculated using saturation water vapor at a given air temperature and actual water vapor (e\u003csub\u003ea\u003c/sub\u003e) estimated by specific humidity (q):\u003c/p\u003e\n \u003cp\u003e\u003cimg src=\"data:image/png;base64,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\"\u003e\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec13\"\u003e\n \u003cp\u003eThe CMIP6 historical (1950\u0026ndash;2014) simulations are adopted to bias calibration of future climate in four SSPs scenarios, using ERA-5 Land reanalysis as the benchmark. This step is designed to ensure the projected climate variables share the same baseline with the reanalysis data.\u003c/p\u003e\n \u003cp\u003eWe further calculate the predicted \u0026Delta;LST\u003csub\u003ew\u003c/sub\u003e or \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e trends in four SSPs scenarios and compare them with the projected skin temperature warming trends in the corresponding scenarios. The ratios suggest the relative magnitude to which the enhanced cooling effects of forests account for the land surface warming driven by climate change.\u003c/p\u003e\n \u003ch2\u003e4.6 Statistical analyses\u003c/h2\u003e\n \u003cp\u003eWe use the Theil-Sen estimator and Mann-Kendall test to analyse temporal trends\u003csup\u003e\u003cspan\u003e85\u003c/span\u003e\u003c/sup\u003e (scipy and pymannkendall packages in Python). The Theil-Sen estimator is a non-parametric method for robust estimation of the slope by using the median value of a range of possible slopes. The Mann-Kendall test, which is also nonparametric, statistically assesses whether a temporal trend is monotonically upward or downward.\u003c/p\u003e\n \u003cp\u003eCorrelation analyses are employed to examine the potential influence of given drivers on the target variables (numpy package in Python). These analyses are based on either spatial or temporal samples. We perform spatial correlation using trend values to verify whether changes in the independent variables could trigger changes in the dependent variables. Temporal correlation is conducted at both pixel and regional scales. For pixel-scale correlation, we calculate the spatial mean and standard deviation of correlation coefficients to represent regional results. For regional-scale correlation, we first calculate the mean value for the study area and then determine the correlation coefficient. Partial temporal correlation analysis is used to explore the relative impacts of multiple climate drivers on \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e at the pixel scale (pingouin package in Python). The partial correlation coefficient describes the relationship between two variables after removing the effects of other potential drivers.\u003c/p\u003e\n \u003cp\u003eWe apply a non-linear machine learning method, Random Forest (RF), to evaluate the importance of multiple potential climate drivers for predicting \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e (sklearn package in Python). Specifically, we use all spatial samples over 21 years to build the dataset. The dataset is divided into training (80% of the samples) and test sets (20% of the samples), with model parameters confirmed based on validation results from the test set (number of trees\u0026thinsp;=\u0026thinsp;50; maximum depth of the tree\u0026thinsp;=\u0026thinsp;20). All the samples are then used to fit the best model. The relative importance of each variable is indicated by the Gini coefficient. Moreover, we use the SHAP method to quantify the individual contribution of each variable in each sample (shap package in Python). SHAP is an approach to explain machine learning models based on cooperative game theory, where SHAP values quantify the marginal contributions of predictors. We calculate the mean absolute SHAP values for each input variable as importance metrics and explore the non-linear impact of climate variables on \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e by plotting their SHAP values against input values.\u003c/p\u003e\n \u003cp\u003eBased on the RF model, we set several scenarios to separate the relative contributions of climate variables to the \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e trend. Specifically, we perform four experimental simulations: (S1) varying VPD\u003csub\u003es\u003c/sub\u003e only; (S2) varying VPD\u003csub\u003es\u003c/sub\u003e and P\u003csub\u003es\u003c/sub\u003e while fixing DSR\u003csub\u003es\u003c/sub\u003e and WS\u003csub\u003es\u003c/sub\u003e; (S3) varying VPD\u003csub\u003es\u003c/sub\u003e, P\u003csub\u003es\u003c/sub\u003e, and DSR\u003csub\u003es\u003c/sub\u003e while fixing WS\u003csub\u003es\u003c/sub\u003e; and (S4) varying VPD\u003csub\u003es\u003c/sub\u003e, P\u003csub\u003es\u003c/sub\u003e, DSR\u003csub\u003es\u003c/sub\u003e, and WS\u003csub\u003es\u003c/sub\u003e. The Theil-Sen slope and Mann-Kendall P-value of simulations S1, S2\u0026thinsp;\u0026minus;\u0026thinsp;S1, S3\u0026thinsp;\u0026minus;\u0026thinsp;S2, and S4\u0026thinsp;\u0026minus;\u0026thinsp;S3 are used to evaluate the individual contributions of the changing climate to the \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e trend.\u003c/p\u003e\n \u003cp\u003eWe use SEM to reveal the biophysical mechanism underlying the \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e dynamics (semopy package in Python). We assume that climate variables influence the temporal dynamics of \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e through three pathways: (P1) climate change affects \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e via \u0026Delta;LAI\u003csub\u003es\u003c/sub\u003e; (P2) climate change affects \u0026Delta;LAI\u003csub\u003es\u003c/sub\u003e, which further alters \u0026Delta;LE\u003csub\u003es\u003c/sub\u003e and then \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e; and (P3) climate change affects \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e via \u0026Delta;LE\u003csub\u003es\u003c/sub\u003e. For P1 and P2, the impact of climate variables on \u0026Delta;LAI\u003csub\u003es\u003c/sub\u003e arises from the diverse sensitivities of different vegetation types to climate changes. Specifically, biophysical parameters (such as albedo and surface roughness) vary differently in forested and non-forested areas under a changing climate, which further affects \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e (P1). Additionally, climate changes promote or inhibit forest and non-forest vegetation growth differently, further influencing the evapotranspiration effect of forests and \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e (P2). Furthermore, climate variables can directly modulate \u0026Delta;LE\u003csub\u003es\u003c/sub\u003e, even when the biophysical and physiological conditions of vegetation remain constant. This is due to inherent differences between different vegetation types\u003csup\u003e\u003cspan\u003e48\u003c/span\u003e,\u003cspan\u003e51\u003c/span\u003e,\u003cspan\u003e86\u003c/span\u003e\u003c/sup\u003e, resulting in diverse evapotranspiration responses to climate change and, consequently, affecting \u0026Delta;LST\u003csub\u003es\u003c/sub\u003e (P3).\u003c/p\u003e\n \u003cp\u003eIn practice, SEM is performed at the pixel level. We use all the data within a 3\u0026times;3 spatial moving window for each 0.5\u0026deg; grid to ensure a sufficient sample size for establishing a robust model. SEM is conducted in grids where the sample size is larger than 150. The data are normalized using the Z-score method before modeling. Following previous study\u003csup\u003e\u003cspan\u003e87\u003c/span\u003e\u003c/sup\u003e, the goodness-of-fit index (GFI) is employed to evaluate the applicability and effectiveness of each model. We filter out paths with P values lower than 0.05 and corresponding model GFIs greater than 0.9 and calculate the mean path values along with the standard error. The contributions of pathways P1, P2, and P3 are measured by their path effects, which are calculated by multiplying the contained path values. Uncertainties are estimated based on error propagation (error_propagation package in Python).\u003c/p\u003e\n\u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eData Availability\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll 1 km MODIS products, including MYD11, MOD16A2 Gap-filled, MCD15A3H, MCD15A3H, MCD43A3, MCD12Q1 can be downloaded from https://ladsweb.modaps.eosdis.nasa.gov/search/. The GFC data is available on the Google Earth Engine at https://developers.google.com/earth-engine/datasets/catalog/UMD_hansen_global_forest_change_2023_v1_11. The GAIA products are available at https://data-starcloud.pcl.ac.cn/resource/13. The GMTED2010 data are from https://developers.google.com/earth-engine/datasets/catalog/USGS_GMTED2010_FULL. ERA5 reanalysis data is accessible at https://cds.climate.copernicus.eu/datasets/reanalysis-era5-land-monthly-means?tab=overview. CMIP6 simulations are available at https://aims2.llnl.gov/search/cmip6/.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCode Availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe Python codes used to generate all the results are available at \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://zenodo.org/records/13944220\u003c/span\u003e\u003c/span\u003e.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eSeddon, N., Turner, B., Berry, P., Chausson, A. \u0026amp; Girardin, C. A. J. Grounding nature-based climate solutions in sound biodiversity science. \u003cem\u003eNat. Clim. Chang.\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, 84\u0026ndash;87 (2019).\u003c/li\u003e\n\u003cli\u003eBonan, G. B. Forests and climate change: Forcings, feedbacks, and the climate benefits of forests. \u003cem\u003eScience (80-. ).\u003c/em\u003e \u003cstrong\u003e320\u003c/strong\u003e, 1444\u0026ndash;1449 (2008).\u003c/li\u003e\n\u003cli\u003eNovick, K. A. \u003cem\u003eet al.\u003c/em\u003e We need a solid scientific basis for nature-based climate solutions in the United States. \u003cem\u003eProc. Natl. Acad. Sci.\u003c/em\u003e \u003cstrong\u003e121\u003c/strong\u003e, 1\u0026ndash;4 (2024).\u003c/li\u003e\n\u003cli\u003eGriscom, B. W. \u003cem\u003eet al.\u003c/em\u003e Natural climate solutions. \u003cem\u003eProc. Natl. Acad. Sci. U. S. A.\u003c/em\u003e \u003cstrong\u003e114\u003c/strong\u003e, 11645\u0026ndash;11650 (2017).\u003c/li\u003e\n\u003cli\u003eWalker, W. S. \u003cem\u003eet al.\u003c/em\u003e The global potential for increased storage of carbon on land. \u003cem\u003eProc. Natl. Acad. Sci. U. S. A.\u003c/em\u003e \u003cstrong\u003e119\u003c/strong\u003e, 1\u0026ndash;12 (2022).\u003c/li\u003e\n\u003cli\u003eLewis, S. L., Wheeler, C. E., Mitchard, E. T. A. \u0026amp; Koch, A. Restoring natural forests is the best way to remove atmospheric carbon. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e568\u003c/strong\u003e, 25\u0026ndash;28 (2019).\u003c/li\u003e\n\u003cli\u003eRoebroek, C. T. J., Duveiller, G., Seneviratne, S. I., Davin, E. L. \u0026amp; Cescatti, A. Releasing global forests from human management: How much more carbon could be stored? \u003cem\u003eScience (80-. ).\u003c/em\u003e \u003cstrong\u003e380\u003c/strong\u003e, 749\u0026ndash;753 (2023).\u003c/li\u003e\n\u003cli\u003eMo, L. \u003cem\u003eet al.\u003c/em\u003e Integrated global assessment of the natural forest carbon potential. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e624\u003c/strong\u003e, 92\u0026ndash;101 (2023).\u003c/li\u003e\n\u003cli\u003eWeber, J. \u003cem\u003eet al.\u003c/em\u003e Chemistry-albedo feedbacks offset up to a third of forestation\u0026rsquo;s CO 2 removal benefits. \u003cem\u003eScience (80-. ).\u003c/em\u003e \u003cstrong\u003e383\u003c/strong\u003e, 860\u0026ndash;864 (2024).\u003c/li\u003e\n\u003cli\u003eLawrence, D., Coe, M., Walker, W., Verchot, L. \u0026amp; Vandecar, K. The Unseen Effects of Deforestation: Biophysical Effects on Climate. \u003cem\u003eFront. For. Glob. Chang.\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e, 1\u0026ndash;13 (2022).\u003c/li\u003e\n\u003cli\u003eWindisch, M. G., Davin, E. L. \u0026amp; Seneviratne, S. I. Prioritizing forestation based on biogeochemical and local biogeophysical impacts. \u003cem\u003eNat. Clim. Chang.\u003c/em\u003e \u003cstrong\u003e11\u003c/strong\u003e, 867\u0026ndash;871 (2021).\u003c/li\u003e\n\u003cli\u003eZhu, L. \u003cem\u003eet al.\u003c/em\u003e Comparable biophysical and biogeochemical feedbacks on warming from tropical moist forest degradation. \u003cem\u003eNat. Geosci.\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, 244\u0026ndash;249 (2023).\u003c/li\u003e\n\u003cli\u003eLawrence, D. \u0026amp; Vandecar, K. Effects of tropical deforestation on climate and agriculture. \u003cem\u003eNat. Clim. Chang.\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e, 27\u0026ndash;36 (2015).\u003c/li\u003e\n\u003cli\u003eAlkama, R. \u0026amp; Cescatti, A. Biophysical climate impacts of recent changes in global forest cover. \u003cem\u003eScience (80-. ).\u003c/em\u003e \u003cstrong\u003e351\u003c/strong\u003e, 600\u0026ndash;604 (2016).\u003c/li\u003e\n\u003cli\u003eDuveiller, G., Hooker, J. \u0026amp; Cescatti, A. The mark of vegetation change on Earth\u0026rsquo;s surface energy balance. \u003cem\u003eNat. Commun.\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, 64\u0026ndash;75 (2018).\u003c/li\u003e\n\u003cli\u003eMildrexler, D. J., Zhao, M. \u0026amp; Running, S. W. A global comparison between station air temperatures and MODIS land surface temperatures reveals the cooling role of forests. \u003cem\u003eJ. Geophys. Res. Biogeosciences\u003c/em\u003e \u003cstrong\u003e116\u003c/strong\u003e, 1\u0026ndash;15 (2011).\u003c/li\u003e\n\u003cli\u003eR.A.Betts. Offset of the potential carbon sink from boreal forestation by decreases in surface albedo. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e408\u003c/strong\u003e, 187\u0026ndash;190 (2000).\u003c/li\u003e\n\u003cli\u003eBright, R. M. \u003cem\u003eet al.\u003c/em\u003e Local temperature response to land cover and management change driven by non-radiative processes. \u003cem\u003eNat. Clim. Chang.\u003c/em\u003e \u003cstrong\u003e7\u003c/strong\u003e, 296\u0026ndash;302 (2017).\u003c/li\u003e\n\u003cli\u003ePitman, A. J. \u003cem\u003eet al.\u003c/em\u003e Importance of background climate in determining impact of land-cover change on regional climate. \u003cem\u003eNat. Clim. Chang.\u003c/em\u003e \u003cstrong\u003e1\u003c/strong\u003e, 472\u0026ndash;475 (2011).\u003c/li\u003e\n\u003cli\u003eLee, X. \u003cem\u003eet al.\u003c/em\u003e Observed increase in local cooling effect of deforestation at higher latitudes. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e479\u003c/strong\u003e, 384\u0026ndash;387 (2011).\u003c/li\u003e\n\u003cli\u003eLi, Y. \u003cem\u003eet al.\u003c/em\u003e Local cooling and warming effects of forests based on satellite observations. \u003cem\u003eNat. Commun.\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, 1\u0026ndash;10 (2015).\u003c/li\u003e\n\u003cli\u003eZeng, Z. \u003cem\u003eet al.\u003c/em\u003e Deforestation-induced warming over tropical mountain regions regulated by elevation. \u003cem\u003eNat. Geosci.\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 23\u0026ndash;29 (2021).\u003c/li\u003e\n\u003cli\u003eEssery, R. Large-scale simulations of snow albedo masking by forests. \u003cem\u003eGeophys. Res. Lett.\u003c/em\u003e \u003cstrong\u003e40\u003c/strong\u003e, 5521\u0026ndash;5525 (2013).\u003c/li\u003e\n\u003cli\u003ePrevedello, J. A., Winck, G. R., Weber, M. M., Nichols, E. \u0026amp; Sinervo, B. Impacts of forestation and deforestation on local temperature across the globe. \u003cem\u003ePLoS One\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 1\u0026ndash;18 (2019).\u003c/li\u003e\n\u003cli\u003eRohatyn, S., Rotenberg, E., Tatarinov, F., Carmel, Y. \u0026amp; Yakir, D. Large variations in afforestation-related climate cooling and warming effects across short distances. \u003cem\u003eCommun. Earth Environ.\u003c/em\u003e \u003cstrong\u003e4\u003c/strong\u003e, 1\u0026ndash;10 (2023).\u003c/li\u003e\n\u003cli\u003eAlkama, R. \u003cem\u003eet al.\u003c/em\u003e Vegetation-based climate mitigation in a warmer and greener World. \u003cem\u003eNat. Commun.\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 1\u0026ndash;10 (2022).\u003c/li\u003e\n\u003cli\u003eGe, J. \u003cem\u003eet al.\u003c/em\u003e Local surface cooling from afforestation amplified by lower aerosol pollution. \u003cem\u003eNat. Geosci.\u003c/em\u003e (2023) doi:10.1038/s41561-023-01251-x.\u003c/li\u003e\n\u003cli\u003eLee, T. H. \u0026amp; Lo, M. H. The role of El Ni\u0026ntilde;o in modulating the effects of deforestation in the Maritime Continent. \u003cem\u003eEnviron. Res. Lett.\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, (2021).\u003c/li\u003e\n\u003cli\u003eSingh, A. \u003cem\u003eet al.\u003c/em\u003e Land-Use Feedback under Global Warming\u0026mdash;A Transition from Radiative to Hydrological Feedback Regime. \u003cem\u003eJ. Clim.\u003c/em\u003e \u003cstrong\u003e37\u003c/strong\u003e, 3847\u0026ndash;3866 (2024).\u003c/li\u003e\n\u003cli\u003ePitman, A. J. \u003cem\u003eet al.\u003c/em\u003e Uncertainties in climate responses to past land cover change: First results from the LUCID intercomparison study. \u003cem\u003eGeophys. Res. Lett.\u003c/em\u003e \u003cstrong\u003e36\u003c/strong\u003e, 1\u0026ndash;6 (2009).\u003c/li\u003e\n\u003cli\u003eForzieri, G. \u003cem\u003eet al.\u003c/em\u003e Evaluating the Interplay Between Biophysical Processes and Leaf Area Changes in Land Surface Models. \u003cem\u003eJ. Adv. Model. Earth Syst.\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 1102\u0026ndash;1126 (2018).\u003c/li\u003e\n\u003cli\u003eFetting, C. \u0026lsquo;The European Green Deal\u0026rsquo;, ESDN Report. \u003cem\u003eEur. Comm.\u003c/em\u003e \u003cstrong\u003e53\u003c/strong\u003e, 24 (2020).\u003c/li\u003e\n\u003cli\u003eSchultz, N. M., Lawrence, P. J. \u0026amp; Lee, X. Global satellite data highlights the diurnal asymmetry of the surface temperature response to deforestation. \u003cem\u003eJ. Geophys. Res. Biogeosciences\u003c/em\u003e \u003cstrong\u003e122\u003c/strong\u003e, 903\u0026ndash;917 (2017).\u003c/li\u003e\n\u003cli\u003eManninen, T. \u003cem\u003eet al.\u003c/em\u003e Monitoring changes in forestry and seasonal snow using surface albedo during 1982-2016 as an indicator. \u003cem\u003eBiogeosciences\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, 223\u0026ndash;240 (2019).\u003c/li\u003e\n\u003cli\u003eAtlaskina, K., Berninger, F. \u0026amp; de Leeuw, G. Satellite observations of changes in snow-covered land surface albedo during spring in the Northern Hemisphere. \u003cem\u003eCryosph.\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, 1879\u0026ndash;1893 (2015).\u003c/li\u003e\n\u003cli\u003eLi, Q., Ma, M., Wu, X. \u0026amp; Yang, H. Snow Cover and Vegetation-Induced Decrease in Global Albedo From 2002 to 2016. \u003cem\u003eJ. Geophys. Res. Atmos.\u003c/em\u003e 124\u0026ndash;138 (2017) doi:10.1002/2017JD027010.\u003c/li\u003e\n\u003cli\u003eKouki, K., Luojus, K. \u0026amp; Riihel\u0026auml;, A. Evaluation of snow cover properties in ERA5 and ERA5-Land with several satellite-based datasets in the Northern Hemisphere in spring 1982-2018. \u003cem\u003eCryosphere\u003c/em\u003e \u003cstrong\u003e17\u003c/strong\u003e, 5007\u0026ndash;5026 (2023).\u003c/li\u003e\n\u003cli\u003eLuo, X. \u003cem\u003eet al.\u003c/em\u003e An Evaluation of CMIP6 Models in Representing the Biophysical Effects of Deforestation With Satellite‐Based Observations. \u003cem\u003eJ. Geophys. Res. Atmos.\u003c/em\u003e \u003cstrong\u003e128\u003c/strong\u003e, 1\u0026ndash;20 (2023).\u003c/li\u003e\n\u003cli\u003eLi, Y. \u003cem\u003eet al.\u003c/em\u003e Biophysical impacts of earth greening can substantially mitigate regional land surface temperature warming. \u003cem\u003eNat. Commun.\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 121 (2023).\u003c/li\u003e\n\u003cli\u003eKumar, S. \u003cem\u003eet al.\u003c/em\u003e Land use/cover change impacts in CMIP5 climate simulations: A new methodology and 21st century challenges. \u003cem\u003eJ. Geophys. Res. Atmos.\u003c/em\u003e \u003cstrong\u003e118\u003c/strong\u003e, 6337\u0026ndash;6353 (2013).\u003c/li\u003e\n\u003cli\u003eHasler, N. \u003cem\u003eet al.\u003c/em\u003e Accounting for albedo change to identify climate-positive tree cover restoration. \u003cem\u003eNat. Commun.\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, 2275 (2024).\u003c/li\u003e\n\u003cli\u003eForzieri, G. \u003cem\u003eet al.\u003c/em\u003e Increased control of vegetation on global terrestrial energy fluxes. \u003cem\u003eNat. Clim. Chang.\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 356\u0026ndash;362 (2020).\u003c/li\u003e\n\u003cli\u003eYuan, W. \u003cem\u003eet al.\u003c/em\u003e Increased atmospheric vapor pressure deficit reduces global vegetation growth. \u003cem\u003eSci. Adv.\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e, 1\u0026ndash;13 (2019).\u003c/li\u003e\n\u003cli\u003eZhong, Z. \u003cem\u003eet al.\u003c/em\u003e Disentangling the effects of vapor pressure deficit on northern terrestrial vegetation productivity. \u003cem\u003eSci. Adv.\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, (2023).\u003c/li\u003e\n\u003cli\u003eZhong, Z. \u003cem\u003eet al.\u003c/em\u003e Higher Sensitivity of Planted Forests\u0026rsquo; Productivity Than Natural Forests to Droughts in China. \u003cem\u003eJ. Geophys. Res. Biogeosciences\u003c/em\u003e \u003cstrong\u003e126\u003c/strong\u003e, 1\u0026ndash;13 (2021).\u003c/li\u003e\n\u003cli\u003eLiu, M. \u003cem\u003eet al.\u003c/em\u003e Overridingly increasing vegetation sensitivity to vapor pressure deficit over the recent two decades in China. \u003cem\u003eEcol. Indic.\u003c/em\u003e \u003cstrong\u003e161\u003c/strong\u003e, 111977 (2024).\u003c/li\u003e\n\u003cli\u003eIsbell, F. \u003cem\u003eet al.\u003c/em\u003e Biodiversity increases the resistance of ecosystem productivity to climate extremes. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e526\u003c/strong\u003e, 574\u0026ndash;577 (2015).\u003c/li\u003e\n\u003cli\u003eTumber-D\u0026aacute;vila, S. J., Schenk, H. J., Du, E. \u0026amp; Jackson, R. B. Plant sizes and shapes above and belowground and their interactions with climate. \u003cem\u003eNew Phytol.\u003c/em\u003e \u003cstrong\u003e235\u003c/strong\u003e, 1032\u0026ndash;1056 (2022).\u003c/li\u003e\n\u003cli\u003eUn, N. Natural evaporation from open water, bare soil and grass. \u003cem\u003eProc. R. Soc. London. Ser. A. Math. Phys. Sci.\u003c/em\u003e \u003cstrong\u003e193\u003c/strong\u003e, 120\u0026ndash;145 (1948).\u003c/li\u003e\n\u003cli\u003eGrossiord, C. \u003cem\u003eet al.\u003c/em\u003e Plant responses to rising vapor pressure deficit. \u003cem\u003eNew Phytol.\u003c/em\u003e \u003cstrong\u003e226\u003c/strong\u003e, 1550\u0026ndash;1566 (2020).\u003c/li\u003e\n\u003cli\u003eMassmann, A., Gentine, P. \u0026amp; Lin, C. When Does Vapor Pressure Deficit Drive or Reduce Evapotranspiration? \u003cem\u003eJ. Adv. Model. Earth Syst.\u003c/em\u003e \u003cstrong\u003e11\u003c/strong\u003e, 3305\u0026ndash;3320 (2019).\u003c/li\u003e\n\u003cli\u003eDickerson-Lange, S. E., Vano, J. A., Gersonde, R. \u0026amp; Lundquist, J. D. Ranking Forest Effects on Snow Storage: A Decision Tool for Forest Management. \u003cem\u003eWater Resour. Res.\u003c/em\u003e \u003cstrong\u003e57\u003c/strong\u003e, (2021).\u003c/li\u003e\n\u003cli\u003eAdam, J. C., Hamlet, A. F. \u0026amp; Lettenmaier, D. P. Implications of global climate change for snowmelt hydrology in the twenty‐first century. \u003cem\u003eHydrol. Process.\u003c/em\u003e \u003cstrong\u003e23\u003c/strong\u003e, 962\u0026ndash;972 (2009).\u003c/li\u003e\n\u003cli\u003eBjorkman, A. D., Elmendorf, S. C., Beamish, A. L., Vellend, M. \u0026amp; Henry, G. H. R. Contrasting effects of warming and increased snowfall on Arctic tundra plant phenology over the past two decades. \u003cem\u003eGlob. Chang. Biol.\u003c/em\u003e \u003cstrong\u003e21\u003c/strong\u003e, 4651\u0026ndash;4661 (2015).\u003c/li\u003e\n\u003cli\u003eQin, Y. \u003cem\u003eet al.\u003c/em\u003e Agricultural risks from changing snowmelt. \u003cem\u003eNat. Clim. Chang.\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 459\u0026ndash;465 (2020).\u003c/li\u003e\n\u003cli\u003eCui, J. \u003cem\u003eet al.\u003c/em\u003e Global water availability boosted by vegetation-driven changes in atmospheric moisture transport. \u003cem\u003eNat. Geosci.\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, 982\u0026ndash;988 (2022).\u003c/li\u003e\n\u003cli\u003eLi, Y. \u003cem\u003eet al.\u003c/em\u003e Deforestation-induced climate change reduces carbon storage in remaining tropical forests. \u003cem\u003eNat. Commun.\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, (2022).\u003c/li\u003e\n\u003cli\u003eMa, S., Zhou, S., Yu, B. \u0026amp; Song, J. Deforestation-induced runoff changes dominated by forest-climate feedbacks. \u003cem\u003eSci. Adv.\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, (2024).\u003c/li\u003e\n\u003cli\u003eHoek van Dijke, A. J. \u003cem\u003eet al.\u003c/em\u003e Shifts in regional water availability due to global tree restoration. \u003cem\u003eNat. Geosci.\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, 363\u0026ndash;368 (2022).\u003c/li\u003e\n\u003cli\u003eZan, B. \u003cem\u003eet al.\u003c/em\u003e Spatiotemporal inequality in land water availability amplified by global tree restoration. \u003cem\u003eNat. Water\u003c/em\u003e (2024) doi:10.1038/s44221-024-00296-5.\u003c/li\u003e\n\u003cli\u003eWang, H., Yue, C. \u0026amp; Luyssaert, S. Reconciling different approaches to quantifying land surface temperature impacts of afforestation using satellite observations. \u003cem\u003eBiogeosciences\u003c/em\u003e \u003cstrong\u003e20\u003c/strong\u003e, 75\u0026ndash;92 (2023).\u003c/li\u003e\n\u003cli\u003eXu, R. \u003cem\u003eet al.\u003c/em\u003e Contrasting impacts of forests on cloud cover based on satellite observations. \u003cem\u003eNat. Commun.\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 670 (2022).\u003c/li\u003e\n\u003cli\u003eLi, Y., Piao, S., Chen, A., Ciais, P. \u0026amp; Li, L. Z. X. Local and teleconnected temperature effects of afforestation and vegetation greening in China. \u003cem\u003eNatl. Sci. Rev.\u003c/em\u003e \u003cstrong\u003e7\u003c/strong\u003e, 897\u0026ndash;912 (2020).\u003c/li\u003e\n\u003cli\u003eZhang, Y. \u003cem\u003eet al.\u003c/em\u003e Asymmetric impacts of forest gain and loss on tropical land surface temperature. \u003cem\u003eNat. Geosci.\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 823\u0026ndash;831 (2024).\u003c/li\u003e\n\u003cli\u003eSu, Y. \u003cem\u003eet al.\u003c/em\u003e Asymmetric influence of forest cover gain and loss on land surface temperature. \u003cem\u003eNat. Clim. Chang.\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 823\u0026ndash;831 (2023).\u003c/li\u003e\n\u003cli\u003eZhu, Z. \u003cem\u003eet al.\u003c/em\u003e Greening of the Earth and its drivers. \u003cem\u003eNat. Clim. Chang.\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, 791\u0026ndash;795 (2016).\u003c/li\u003e\n\u003cli\u003eYang, Y. \u003cem\u003eet al.\u003c/em\u003e Evapotranspiration on a greening Earth. \u003cem\u003eNature Reviews Earth and Environment\u003c/em\u003e vol. 4 626\u0026ndash;641 (2023).\u003c/li\u003e\n\u003cli\u003eZeng, Z. \u003cem\u003eet al.\u003c/em\u003e Climate mitigation from vegetation biophysical feedbacks during the past three decades. \u003cem\u003eNat. Clim. Chang.\u003c/em\u003e \u003cstrong\u003e7\u003c/strong\u003e, 432\u0026ndash;436 (2017).\u003c/li\u003e\n\u003cli\u003eFIELD, C. B., JACKSON, R. B. \u0026amp; MOONEY, H. A. Stomatal responses to increased CO2: implications from the plant to the global scale. \u003cem\u003ePlant. Cell Environ.\u003c/em\u003e \u003cstrong\u003e18\u003c/strong\u003e, 1214\u0026ndash;1225 (1995).\u003c/li\u003e\n\u003cli\u003eBetts, R. A. \u003cem\u003eet al.\u003c/em\u003e Projected increase in continental runoff due to plant responses to increasing carbon dioxide. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e448\u003c/strong\u003e, 1037\u0026ndash;1041 (2007).\u003c/li\u003e\n\u003cli\u003eHe, M. \u003cem\u003eet al.\u003c/em\u003e Amplified warming from physiological responses to carbon dioxide reduces the potential of vegetation for climate change mitigation. \u003cem\u003eCommun. Earth Environ.\u003c/em\u003e \u003cstrong\u003e3\u003c/strong\u003e, 1\u0026ndash;10 (2022).\u003c/li\u003e\n\u003cli\u003eSkinner, C. B., Poulsen, C. J. \u0026amp; Mankin, J. S. Amplification of heat extremes by plant CO2 physiological forcing. \u003cem\u003eNat. Commun.\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, 1\u0026ndash;11 (2018).\u003c/li\u003e\n\u003cli\u003eLi, Z. \u003cem\u003eet al.\u003c/em\u003e Satellite Remote Sensing of Global Land Surface Temperature: Definition, Methods, Products, and Applications. \u003cem\u003eRev. Geophys.\u003c/em\u003e \u003cstrong\u003e61\u003c/strong\u003e, 1\u0026ndash;77 (2023).\u003c/li\u003e\n\u003cli\u003eNovick, K. A. \u0026amp; Katul, G. G. The Duality of Reforestation Impacts on Surface and Air Temperature. \u003cem\u003eJ. Geophys. Res. Biogeosciences\u003c/em\u003e \u003cstrong\u003e125\u003c/strong\u003e, 1\u0026ndash;15 (2020).\u003c/li\u003e\n\u003cli\u003eZhang, Z., Zhang, F., Wang, L., Lin, A. \u0026amp; Zhao, L. Biophysical climate impact of forests with different age classes in mid- and high-latitude North America. \u003cem\u003eFor. Ecol. Manage.\u003c/em\u003e \u003cstrong\u003e494\u003c/strong\u003e, 119327 (2021).\u003c/li\u003e\n\u003cli\u003eXing, Z. \u003cem\u003eet al.\u003c/em\u003e Estimation of daily mean land surface temperature at global scale using pairs of daytime and nighttime MODIS instantaneous observations. \u003cem\u003eISPRS J. Photogramm. Remote Sens.\u003c/em\u003e \u003cstrong\u003e178\u003c/strong\u003e, 51\u0026ndash;67 (2021).\u003c/li\u003e\n\u003cli\u003eHansen, M. C. \u003cem\u003eet al.\u003c/em\u003e High-Resolution Global Maps of 21st-Century Forest Cover Change. \u003cem\u003eScience (80-. ).\u003c/em\u003e \u003cstrong\u003e342\u003c/strong\u003e, 850\u0026ndash;853 (2013).\u003c/li\u003e\n\u003cli\u003eGong, P. \u003cem\u003eet al.\u003c/em\u003e Annual maps of global artificial impervious area (GAIA) between 1985 and 2018. \u003cem\u003eRemote Sens. Environ.\u003c/em\u003e \u003cstrong\u003e236\u003c/strong\u003e, 111510 (2020).\u003c/li\u003e\n\u003cli\u003eMu\u0026ntilde;oz-Sabater, J. \u003cem\u003eet al.\u003c/em\u003e ERA5-Land: A state-of-the-art global reanalysis dataset for land applications. \u003cem\u003eEarth Syst. Sci. Data\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 4349\u0026ndash;4383 (2021).\u003c/li\u003e\n\u003cli\u003eLiu, X. \u003cem\u003eet al.\u003c/em\u003e Local temperature responses to actual land cover changes present significant latitudinal variability and asymmetry. \u003cem\u003eSci. Bull.\u003c/em\u003e \u003cstrong\u003e68\u003c/strong\u003e, 2849\u0026ndash;2861 (2023).\u003c/li\u003e\n\u003cli\u003eEyring, V. \u003cem\u003eet al.\u003c/em\u003e Overview of the Coupled Model Intercomparison Project Phase 6 (CMIP6) experimental design and organization. \u003cem\u003eGeosci. Model Dev.\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, 1937\u0026ndash;1958 (2016).\u003c/li\u003e\n\u003cli\u003eLawrence, D. M. \u003cem\u003eet al.\u003c/em\u003e The Land Use Model Intercomparison Project (LUMIP) contribution to CMIP6: Rationale and experimental design. \u003cem\u003eGeosci. Model Dev.\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, 2973\u0026ndash;2998 (2016).\u003c/li\u003e\n\u003cli\u003eKeenan, R. J. \u003cem\u003eet al.\u003c/em\u003e Dynamics of global forest area: Results from the FAO Global Forest Resources Assessment 2015. \u003cem\u003eForest Ecology and Management\u003c/em\u003e vol. 352 9\u0026ndash;20 (2015).\u003c/li\u003e\n\u003cli\u003eLuo, X. \u003cem\u003eet al.\u003c/em\u003e Local and Nonlocal Biophysical Effects of Historical Land Use and Land Cover Changes in CMIP6 Models and the Intermodel Uncertainty. \u003cem\u003eEarth\u0026rsquo;s Futur.\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, (2024).\u003c/li\u003e\n\u003cli\u003eSen, P. K. Estimates of the Regression Coefficient Based on Kendall\u0026rsquo;s Tau. \u003cem\u003eJ. Am. Stat. Assoc.\u003c/em\u003e \u003cstrong\u003e63\u003c/strong\u003e, 1379\u0026ndash;1389 (1968).\u003c/li\u003e\n\u003cli\u003eHuang, M. \u003cem\u003eet al.\u003c/em\u003e Air temperature optima of vegetation productivity across global biomes. \u003cem\u003eNat. Ecol. Evol.\u003c/em\u003e \u003cstrong\u003e3\u003c/strong\u003e, 772\u0026ndash;779 (2019).\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"nature-portfolio","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"","title":"Nature Portfolio","twitterHandle":"","acdcEnabled":false,"dfaEnabled":false,"editorialSystem":"ejp","reportingPortfolio":"","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-5281378/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5281378/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eForests exhibit cooling or warming effects compared to adjacent openlands through biophysical processes. The local temperature effects are predicted by earth system models to evolve in response to climate change. However, these temporal dynamic patterns remain unconstrained by observations and have not been detected in historical records. Here, we provide satellite evidence of emergent negative trends in local land surface temperature (LST) effects of European forests from 2003\u0026ndash;2023. The daytime cooling effects have significantly intensified in both winter (-0.17 K/decade) and summer (-0.22 K/decade). The enhanced winter cooling is attributed to the reduced shortwave radiative forcing in forests due to decreasing snow cover. In the summertime, the vegetation physiological response to increased atmospheric vapor pressure deficit boosts evaporative cooling in forests. The negative trends in LST effects of European forests are roughly supported by four state-of-the-art earth system models. However, considerable biases and intermodel spread in the representation of underlying biophysical processes. Given the continued climate change, we emphasize the need to consider their impacts on biophysical effect dynamics when comprehensive forest-related climate mitigation policies are formed.\u003c/p\u003e","manuscriptTitle":"Recent climate change strengthens the local cooling of European forests","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-02-20 15:56:16","doi":"10.21203/rs.3.rs-5281378/v1","editorialEvents":[],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"nature-communications","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"NCOMMS","sideBox":"Learn more about [Nature Communications](http://www.nature.com/ncomms/)","snPcode":"","submissionUrl":"https://mts-ncomms.nature.com/","title":"Nature Communications","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Nature Communications","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"dd12919e-f751-4e1f-8cb0-936f0034e78b","owner":[],"postedDate":"February 20th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":39422326,"name":"Earth and environmental sciences/Climate sciences/Climate change/Climate-change impacts/Environmental health"},{"id":39422327,"name":"Earth and environmental sciences/Ecology/Forest ecology"}],"tags":[],"updatedAt":"2025-09-26T07:09:07+00:00","versionOfRecord":{"articleIdentity":"rs-5281378","link":"https://doi.org/10.1038/s41467-025-63556-2","journal":{"identity":"nature-communications","isVorOnly":false,"title":"Nature Communications"},"publishedOn":"2025-09-25 04:00:00","publishedOnDateReadable":"September 25th, 2025"},"versionCreatedAt":"2025-02-20 15:56:16","video":"","vorDoi":"10.1038/s41467-025-63556-2","vorDoiUrl":"https://doi.org/10.1038/s41467-025-63556-2","workflowStages":[]},"version":"v1","identity":"rs-5281378","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5281378","identity":"rs-5281378","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}