Introduction

Anthropogenic land use and land cover change (LULCC) is widely recognized as one of the most direct human-induced alterations to the natural environment1. Prominent LULCC has been observed in more than 50% of ice-free land regions worldwide during the past three centuries2. Anticipated future LULCC will also lead to noticeable changes due to the increasing demands of a growing population for food, fiber, and other essential resources2. This widespread LULCC has attracted increased attention from the scientific community and the general public due to its remarkable climate impacts both in the historical period and in the future3,4,5,6,7.

LULCC modifies near-surface geophysical and geochemical properties, thereby altering the exchange of water, energy, and momentum exchange between the land and atmosphere. Consequently, it results in shifts in local convection and large-scale atmospheric circulation, leading to changes in surface temperature, particularly in regions experiencing intense LULCC, as well as neighboring and remote areas affected by these changes8,9,10. The surface temperature serves as a critical variable for measuring and monitoring the Earth’s surface climate11. It is influenced by land-atmosphere feedback mechanisms and can effectively represent perturbations in the surface energy balance intuitively and sensitively12,13,14.

Previous studies have indicated that surface albedo is the dominant factor in modulating mid- to high-latitude surface temperature. This is because taller, darker, and denser forests can absorb more solar radiation than adjacent open lands, particularly during winter when there is seasonal snow cover15,16. In tropical regions, however, surface energy loss primarily occurs through evapotranspiration, making latent heat flux the crucial factor in surface temperature responses to LULCC17. In other regions, different factors may also come into play as well. For example, turbulent energy dissipation, radiation absorption, and sensible heat flux can be more influential in water-limited dryland environments18. Previous studies have mostly focused on the long-term mean surface temperature responses to LULCC rather than when surface temperature responses are different from the long-term mean. Furthermore, less attention has been given to the relative importance of radiative (shortwave and longwave radiation) and non-radiative (latent and sensible heat flux) factors when surface temperature responses are stronger than the long-term mean.

The long-term mean can effectively filter out unwanted signals; however, it may inadvertently obscure valuable information if the upper and lower tails are averaged together. The primary objective of this study is to enhance the understanding of why in some months, surface temperature responds much more strongly to LULCC than to the long-term mean responses and what factors contribute to those extreme responses. Extreme surface temperature responses to LULCC can aggregate high-impact extreme events, especially when combined with other internal and/or external forcing factors. Furthermore, quantifying the contributions of the individual factors to the extreme surface temperature responses can improve the mechanistic understanding of both radiative and non-radiative effects associated with LULCC.

Here, we provide insights from a state-of-the-art multi-model ensemble from Coupled Model Intercomparison Project Phase 6 (CMIP6)-Land Use Model Intercomparison Project (LUMIP) to investigate changes in surface temperature induced by LULCC over two time periods: the historical period from 1995 to 2014 and the future period from 2080 to 2099. During the historical period, the climatic effects of LULCC are defined as differences between model simulation based on historical land use and simulation with land use fixed at 1850. During the future period, the LULCC impacts are differences between SSP1-2.6 (afforestation scenario) and SSP3-7.0 (deforestation scenario) (see Methods). We present results obtained through the surface energy budget to provide a comprehensive understanding of the magnitude of surface temperature change, the relative importance of radiative and non-radiative contributions, and potential mechanisms under both average surface temperature responses and extreme surface temperature responses13,19,20. Our findings build upon previous knowledge gained from CMIP5-LUCID (Land-Use and Climate, IDentification of robust impacts) and LUCAS (Land Use and Climate Across Scales) intercomparison projects regarding LULCC impacts on surface temperature21,22.

Results

Magnitudes of changes in land surface temperature and contributing factors

The widespread forest and natural non-forest vegetation in North America, South America, Central Asia, and the coastal regions of Australia have declined during the historical period (Fig. 1a, b). Future deforestation is projected to be more prevalent in tropical Africa (Fig. 1c, d). In both the Northern Hemisphere at mid-high latitudes and in the tropics, the dominant type of LULCC is to transform forest and natural non-forest vegetation into grazing land and cropland (Fig. 1e, f). This land use/cover transformation has resulted in a decrease of 0.2–0.6 °C in the average surface temperature across a wide land area of mid-high latitudes in the Northern Hemisphere. In contrast, there are also tropical regions that experienced some increase (~0.2–0.4 °C) in surface temperature during the historical period (Fig. 2a), and in the future (Supplementary Fig. 1a). The latitudes separating the negative surface temperature from the positive surface temperature responses due to LULCC (i.e. zero effect line) are generally located between 20°N and 30°N23. As anthropogenic LULCC activities in tropical regions (mid-high latitudes of the Northern Hemisphere), primarily through deforestation since the Industrial Revolution, tend to cause warming (cooling) effects on surface temperature16,24, climate responses in the tropical (20°N–20°S) and mid-high latitudes of the Northern Hemisphere (30°N–90°N) should be analyzed separately.

Fig. 1: Spatial distribution and time series of land use and land cover change (LULCC).
figure 1

a The spatial distribution of the combination of forest and non-forest change caused by LULCC during the historical period (1995–2014) is expressed as a percentage. b The same as (a), but for the combination of grazing land and cropland. c The same as (a), but during the future period (2080–2099). d The same as (c), but for the combination of grazingland and cropland. e Time series of changes of the five land use categories (forest, non-forest, grazing land, cropland, and urban land) caused by LULCC in the mid-high latitudes in the Northern Hemisphere from 1850 to 2099, expressed as a percentage. f The same as (e), but in the tropics. The shade represents the historical period and the future period, respectively.

Full size image
Fig. 2: Spatial distribution of surface temperature change and related contributing factors based on surface budget equation over the historical period.
figure 2

a The spatial distribution of the CMIP6-LUMIP multi-model ensemble mean surface temperature change caused by LULCC. b The annual averaged spatial distribution of contributions from downward longwave radiation (LWd) to surface temperature change. c The spatial distribution of the first contributing factor to surface temperature change. d–f The same as (a)–(c), but for the June–July–August (JJA) average. g–i The same as (a)–(c), but for the December–January–February (DJF) average. The dot denotes that the sign of the change in the CMIP6-LUMIP multi-model ensemble mean is consistent with the sign of change in at least 5 out of the 7 individual models.

Full size image

When averaged across the historical period, the magnitudes of five individual contributions (due to changes in surface albedo, downward shortwave radiation, downward longwave radiation, latent heat flux, and sensible heat flux) were all within 1 °C (Fig. 2b and Supplementary Fig. 2). Downward longwave radiation accounted for 0.2–0.4 °C in the widespread Northern Hemisphere mid- to high-latitude regions (Fig. 2b), as supported by more than 2/3 of the models (Fig. 2c). In most of the mid- to high-latitude regions, downward longwave radiation emerged as the primary contributing factor instead of surface albedo. The remaining areas in mid-high latitudes of the Northern Hemisphere (mainly North America) were influenced by surface albedo and downward shortwave radiation. The latent heat flux was found to be the dominant contributing factor in most tropical regions with an increase of up to 0.4 °C (Fig. 2c). This advantage of latent heat flux can be attributed to the higher evaporation rates in forests than in grazing land or cropland following deforestation. The spatial pattern and magnitude of both factors remained similar during the future period as compared to those during the historical period; however, there was a greater increase in the intensity of latent heat flux within tropical deforestation regions in the future, which also expanded northward. Correspondingly, regions with downward longwave radiation as the dominant contributor contracted toward Siberia (Supplementary Fig. 1b, c, and Supplementary Fig. 3).

Considering the seasonal response of surface temperature change to LULCC during the historical period, the changes in June–July–August (JJA) in the Northern Hemisphere mid- to high-latitude regions were slight and heterogeneous (Fig. 2d). The contribution of downward longwave radiation was within \(\pm\)0.2 °C, and the first factor became the latent heat and downward shortwave radiation in the mid-high latitudes of the Northern Hemisphere (Fig. 2e, f, and Supplementary Fig. 4). In contrast, widespread consistent decreasing regions with a maximum of more than 1 °C of cooling in the surface temperature can be seen in December–January–February (DJF). The spatial pattern of the observed surface temperature change in DJF was similar to that of the mean state, but the magnitude reached up to five times that of the average change (Fig. 2g). Downward longwave radiation dominated the mid-high latitudes of the Northern Hemisphere, and its dominant regions (the first factor) were larger than those of the mean state (Fig. 2h, i). There were fewer other contributing factors than downward longwave radiation in the mid- to high-latitude regions of the Northern Hemisphere (Supplementary Fig. 5). The differences among JJA, DJF, and the mean spatial patterns of surface temperature changes during the future period are projected to be much smaller than those over the historical period. However, the primary contributing factor regions belong to downward longwave radiation are projected to be similar to those of the historical period (Supplementary Fig. 1d–i, Supplementary Fig. 6, and Supplementary Fig. 7).

The largest surface temperature decrease response to LULCC in the boreal cold season is of great interest to the public because of its potentially high impacts (Fig. 3a and Supplementary Fig. 8a). A contribution ranging from 100% to 150% of the downward longwave radiation can be seen in the mid- to high-latitude regions of the Northern Hemisphere (Fig. 3b and Supplementary Fig. 8b). Downward longwave radiation was the first contributing factor and was much greater than the other contributing factors when the average surface temperature change was in the bottom 5% (Bot5%, see Methods) (Fig. 3c and Supplementary Fig. 8c). The spatial pattern of the surface temperature changes and the contributions of downward longwave radiation are similar to those in DJF, but the magnitude of these changes has increased by more than three times in Bot5% (Fig. 3d, e, Supplementary Fig. 8d, e). The effects of other contributing factors are much lower than those of downward longwave radiation in the mid-high latitudes of the Northern Hemisphere (Supplementary Fig. 9 and Supplementary Fig. 10). The dominant regions of downward longwave radiation have expanded southward and encompassed almost the entire Northern Hemisphere over both historical and future periods. There was a remarkable reduction in the area dominated by latent heat flux contributions. The remaining tropical and near tropical areas are governed by downward shortwave radiation (Fig. 3f and Supplementary Fig. 8f).

Fig. 3: Spatial distribution of the minimum surface temperature change and related contributing factors based on surface energy budget equation over the historical period.
figure 3

a The month when surface temperature change reaches its minimum. b The percentage of LWd contribution to the minimum of surface temperature change. c The contributions of five factors (LWd, downward shortwave radiation (SWd), surface albedo (Albedo), latent heat flux (LH), and sensible heat flux (SH)) to surface temperature change weighted averaged in the global land as a function of percentiles are colored lightgreen, skyblue, darkgreen, orange, and red. The surface temperature change is colored gray. All data are categorized by surface temperature change percentiles. d The spatial distribution of surface temperature change caused by LULCC when surface temperature change is bottom 5% average (Bot5%). e The spatial distribution of contributions from LWd to the Bot5% surface temperature change. f The spatial distribution of the first contributing factor to the Bot5% surface temperature change. The dot or asterisk denotes that the sign of the change in the CMIP6-LUMIP multi-model ensemble mean is consistent with the sign of change in at least 5 out of the 7 individual models.

Full size image

The temperature change ranges of the surface temperature exhibited shifts from negative (minimum of −1.5 °C) to positive values (maximum of 1 °C) in the regional average for the mid-high latitudes of the Northern Hemisphere during the historical period. The effects of LULCC are closely related to the background climate (e.g., atmospheric circulation and seasonal snow cover), with positive and negative offsets due to changes in contributing factors at different spatial and temporal scales. The asymmetry between LULCC cooling and warming effects led to final cooling in the mid-high latitudes of the Northern Hemisphere during the historical period. Downward longwave radiation can explain most of the surface temperature change in Bot5% and further extends to Q20 (Fig. 4a). In the tropics, the range of surface temperature change was limited to approximately \(\pm\)1 °C, which is lower than that simulated in the mid- to high-latitude regions of the Northern Hemisphere. The asymmetry of LULCC cooling and warming effects also existed but with final warming effects. Downward longwave radiation also dominated the surface temperature response to LULCC in the lower and upper tails (Q99 and the largest surface temperature increase (maximum)). However, it may be destabilized because only a few months were included (Fig. 4b). In the future, the magnitude of change in the mid-high latitudes of the Northern Hemisphere is projected to be lower than that of the historical period. The advantage of downward longwave radiation in Bot5% is also smaller (Supplementary Fig. 11a). In the tropics, downward longwave radiation will extend its dominant percentiles, which suggests that downward longwave radiation should receive much attention in the future (Supplementary Fig. 11b).

Fig. 4: Contributions of five related factors to surface temperature change along with rank of surface temperature change over the historical period.
figure 4

a The contributions of five factors (LWd, SWd, Albedo, LH, and SH) to surface temperature change weighted averaged in mid-high latitudes of the Northern Hemisphere as a function of percentiles are colored lightgreen, skyblue, darkgreen, orange, and red. The surface temperature change is colored gray. b The same as (a), but weighted averaged in tropics. All data are categorized by surface temperature change percentiles. The asterisk denotes that the sign of the change in the CMIP6-LUMIP multi-model ensemble mean is consistent with the sign of change in at least 5 out of the 7 individual models.

Full size image

Possible mechanism underlying changes in downward longwave radiation

LULCC can affect surface temperature changes in the mid-high latitudes of the Northern Hemisphere through the effects of surface temperature itself, atmospheric temperatures, water vapor, and clouds. LULCC starts its initial influence on surface albedo, then alters the surface energy budget reflected in net radiation, surface temperature, and surface air pressure25. Longwave radiation emitted from the surface is proportional to the fourth power of the surface temperature according to the Stefan-Boltzmann law, which can ultimately affect the downward longwave radiation. Atmospheric temperatures would be corresponding change and can alter the downward longwave radiation by the atmosphere. Meanwhile, these surface/near surface changes in energy and pressure are transmitted to the upper air and affect atmospheric circulation by altering the geopotential height, which in turn changes the amount of water vapor transported from the ocean and lower latitudes to the mid-high latitudes. As an important greenhouse gas, water vapor absorbs and traps a portion of the longwave radiation released from the Earth’s surface, while heating the surface through downward longwave radiation26. Changes in water vapor content and cloud fraction directly affect the amount of downward longwave radiation, which is reflected in changes in the surface temperature27,28.

Averaged over the historical period, we show the LULCC slightly increased the surface albedo in some LULCC-intensive regions in mid-high latitudes of the North Hemisphere (Fig. 5a). There would be a corresponding decrease of net radiation in these mid-high latitude regions (Fig. 5c). The surface temperature would also decrease (0.2–0.6 °C) due to the net radiation reduction in the mid-high latitudes (Fig. 2a). The decrease of net radiation and surface temperature would increase the density of the near-surface atmosphere, thereby the surface air pressure would increase (Fig. 5e). The above feedback which start from land-surface/near-surface would amplify when surface temperature change is Bot5% (Fig. 3d, Fig. 5b, d, f). The near-surface air temperature changes were in line with those of surface temperature in most regions, indicating the non-negligible role of atmospheric temperatures in regulating downward longwave radiation (Fig. 6). LULCC-induced surface thermal and pressure anomalies can change the geopotential height field and adjust atmospheric circulation. The mid-high latitudes in the Northern Hemisphere were controlled by westerly airflow when averaged in DJF across the historical period (Fig. 7a). The LULCC results in an anomaly with a “north increase and south decrease” in the 500 hPa geopotential height (Fig. 7b). This anomaly at the 500 hPa geopotential height in Bot5% was similar to the result in DJF, but of greater magnitude (approximately 4 times greater) (Fig. 7c). The spatial pattern of geopotential height at 850 hPa was similar to that at 500 hPa (Fig. 7d). The water vapor source in the inner regions of the mid-high latitudes in the Northern Hemisphere was the ocean, and the water vapor content was lower than that in the surrounding areas (Fig. 7e). These geopotential height anomalies brought stronger easterly winds to the northwest, weakening the background westerly airflow and inducing low values of water vapor in that region (Fig. 7f). The contribution of downward longwave radiation was greater in Bot5% than in DJF due to the corresponding greater geopotential height anomaly (Fig. 7g). The cloud cover can also contribute to the decrease in downward longwave radiation, as it decreased largely in Eastern Europe and northern Siberia (Fig. 7h).

Fig. 5: Spatial distribution of changes in surface albedo, net radiation, and surface air pressure over the historical period.
figure 5

a The annual averaged surface albedo (Albedo) change. b The Albedo change when the surface temperature change is Bot5%. c The same as (a), but for net radiation change. d The same as (b), but for net radiation change. e The same as (a), but for surface air pressure (PS) change. f The same as (b), but for PS change.

Full size image
Fig. 6: Spatial distribution of the differences between the changes in surface temperature and 2-m surface air temperature (T2m) over the historical period.
figure 6

a Averaged over the historical period. b Bot5%.

Full size image
Fig. 7: Geopotential height, wind vector, water vapor, and total cloud fraction over the historical period.
figure 7

a The spatial distribution of 500 hPa geopotential height (units: gpm) and 500 hPa wind vector (units: m/s) obtained from the “Hist-noLu” experiment averaged in DJF. b The spatial distribution of changes in the 500 hPa geopotential height and 500 hPa wind vector caused by LULCC averaged in DJF. c The spatial distribution of changes in the 500 hPa geopotential height and 500 hPa wind vector caused by LULCC when the surface temperature change is Bot5%. d The same as (c), but for 850 hPa. e The spatial distribution of 850 hPa wind vector (units: m/s) obtained from the “Hist-noLu” experiment averaged in DJF. Color filling represents changes in the water vapor path (atmosphere mass content of water vapor, units: mm). f The spatial distribution of changes in the 850 hPa wind vector and water vapor path caused by LULCC averaged in DJF. g The spatial distribution of changes in the 850 hPa wind vector and water vapor path caused by LULCC when the surface temperature change is Bot5%. h The spatial distribution of changes in the total cloud fraction caused by LULCC when the surface temperature change is Bot5%. The dot denotes that the sign of the change in the CMIP6-LUMIP multi-model ensemble mean is consistent with the sign of change in at least 5 out of the 7 individual models.

Full size image

In the future period, the surface/near surface changes in energy and pressure averaged over the whole period and in the lower tails (Bot5%) are projected to be similar to those over the historical period, but with smaller magnitude (Supplementary Fig. 12). This smaller magnitude might be due to the LULCC is projected to be weaker at mid-high latitudes of the Northern Hemisphere in the future period than in the historical period, resulting in corresponding weaker local effects. Therefore, the surface temperature change in mid-high latitudes will largely rely on LULCC non-local effects (e.g., LULCC teleconnection effects from the tropics) in the future period. These similar surface energy and pressure changes result in the corresponding similar small differences between the surface temperature and near-surface air temperature in the mid-high latitudes (Supplementary Fig. 13). There will also be similar spatial patterns of the geopotential height during the future period compared with those over the historical period, but with lower magnitudes of change (Supplementary Fig. 14a–d). This lower change magnitude in geopotential height will cause a corresponding lower decrease in wind and water vapor (Supplementary Fig. 14e–g). The cloud cover change in the mid-high latitudes of the Northern Hemisphere is not projected to be obvious in most intensive areas of surface temperature decrease except Central Asia, suggesting that water vapor may be a major factor influencing the contribution of downward longwave radiation in mid-high latitude regions with marked changes (Supplementary Fig. 14h).

Discussion

Previous studies have primarily focused on investigating prominent radiative and non-radiative factors at the mean state using in-situ and/or satellite observations29,30. The outputs from model simulations are typically derived from idealized-world experiments31,32,33. In contrast, our study contributes to the existing knowledge by utilizing realistic model simulations to examine the climatic impacts of LULCC. Our findings suggest that downward longwave radiation is the primary contributing factor in the mid-high latitudes of the Northern Hemisphere over both historical and future periods, particularly under lower/upper tails of surface temperature changes. Surface albedo only dominates certain regions with intense LULCC in North America, rather than being the first factor at mid-high latitudes. Note that the reduced downward longwave radiation would not be the initial reason for the surface temperature decrease. The initial disturbance due to LULCC would firstly affect surface albedo, inducing changes in energy budget and surface temperature. The contribution of downward longwave radiation can intensify this initial surface temperature change and become the dominating factor through complex ocean–land–atmosphere interactions.

In the mid- to high-latitude regions of the Northern Hemisphere, surface albedo has been primarily identified as playing an important role in previous studies. However, it is important to note that these studies may underestimate the impact of downward longwave radiation on surface temperature due to the water vapor content may not be accurately measured34. Our findings confirm that under lower/upper tails of surface temperature changes, water vapor change exhibits greater magnitudes and consequently contributes more greatly to changes in downward longwave radiation. Cloud cover is associated with the atmospheric moisture content, and can exert influence on surface temperature change by regulating radiation35. We show that the cloud cover would decrease corresponding to water vapor reduction in some intensive areas of decreasing surface temperature in mid-high latitudes of the Northern Hemisphere. There should be a synergistic effect of water vapor and cloud cover changes modulating downward longwave radiation. Since the cloud cover change in adjacent areas is not as remarkable as water vapor change, it can be inferred that water vapor change may be more important for surface temperature change relative to cloud cover. In addition to water vapor and clouds, downward longwave radiation change is also affected by the atmospheric and surface temperatures36. The downward longwave radiation of the atmosphere is closely related to the near-surface air temperature. Surface temperature change indirectly affects the downward longwave radiation by influencing the upward longwave radiation. The lower the surface temperature, the less upward longwave radiation, resulting in less absorption of water vapor and other greenhouse gases in the atmosphere and, ultimately, less corresponding emission downward. Moreover, the decrease in atmospheric temperature would correspond to a decrease in the atmospheric water holding capacity. The final downward longwave radiation change would be the combined effects of water vapor, clouds, and atmospheric and surface temperatures.

The dominant factor driving climate change in most of the tropics has traditionally been attributed to latent heat flux, as indicated by previous studies37. The downward shortwave radiation is the first contributing factor to surface temperature change in other regions of the tropics. This dominance of downward shortwave radiation in those regions are also largely affected by changes in water vapor and cloud cover. Ongoing deforestation and expansion of cropland in this region have consistently reduced evapotranspiration rates and contributed to the overall warming of the climate system. This warming effect induced by an increase in Bowen ratio and downward shortwave radiation surpasses the cooling influences exerted by changes in surface albedo and downward longwave radiation under average conditions. Our findings extend these observations to extreme surface temperature change conditions and reveal that downward longwave radiation has emerged as a primary driver in certain tropical areas.

Our findings indicate that LULCC can influence climate in distant areas through atmospheric circulation and teleconnections9,10. We show agreement between most individual models and the multi-model ensemble results in this study. Teleconnections are not obvious in a few models, possibly due to different sensitivity of model responses to land-use change and lack of refinement of subgrids LULCC38,39,40. This study specifically focused on the biogeophysical impacts of realistic anthropogenic LULCC on surface temperature change without differentiating between individual land use types41. We agree that analyzing specific land use/cover conversions in detail can help isolate climate impacts associated with LULCC23. For instance, investigating adjoining areas of forests and open land based on observations is important42. Although these individual effects are largely hypothetical, they warrant further investigation. Combining both approaches can effectively enhance the accuracy and persuasiveness of research in the field of the impacts of LULCC on climate.

Future LULCC is increasingly acknowledged as a potentially effective strategy for mitigating climate change in human society43. The Paris Agreement of the United Nations Framework Convention on Climate Change explicitly promotes afforestation and sustainable forest management to reap mitigation and adaptation benefits in the face of a warming future. However, not all land use/land management activities contribute to cooling the climate system or mitigating losses associated with climate disasters44. Therefore, accurate assessment and prediction of potential climate and environmental impacts are crucial before implementing large-scale anthropogenic LULCC activities15. This study highlights the dominant role of downward longwave radiation under both historical and future conditions, particularly during extreme events. We anticipate that our findings will enhance the understanding of interactions between land use/cover, radiation, and atmospheric circulation while providing a scientific basis for targeted measures to adaptively modify the land surface environment.

Methods

Datasets and experimental design

The latest iteration of CMIP, CMIP6, upholds the defining characteristics of its predecessors and serves as a valuable tool for global climate simulation45. In addition to the fundamental experiments conducted in all CMIPs, CMIP6 introduces specialized experiments known as MIPs that cater to the interests of the scientific community. LUMIP is one such CMIP6-Endorsed MIP initiated by the scientific community to quantify past and future climate impacts, interactions, and feedback resulting from LULCC, while also evaluating the potential of land management for mitigating climate change effects45,46.

Detailed information about the seven models employed and the two-pair experiments can be found in Supplementary Table 1. All models that included the required outputs were used to enhance the credibility of our results. The “Hist” and “SSP126” experiments correspond to the means of the CMIP6 historical runs and SSP1-2.6 scenario, respectively. The experiment named “Hist-noLu” is identical to “Hist”, except that land use/cover is held constant at the 1850 level. Similarly, the experiment named “SSP126-SSP370Lu” refers to the same scenario as “SSP126”, but land use/cover is consistent with SSP3-7.0. Consequently, we refer to the “SSP126” experiment as the afforestation scenario and the “SSP126-SSP370Lu” experiment as the deforestation scenario. In this study, LULCC climate impacts were calculated by comparing differences between the “Hist” and “Hist-noLu” experiments during the historical period and between the “SSP126-SSP370Lu” and “SSP126” experiments during the future period.

The biogeophysical impact of LULCC can be distinguished by comparing the results obtained from two pairs of sensitivity experiments. It should be noted that the biogeochemical effect of LULCC is excluded because carbon dioxide concentrations are annually set to actual values47. In the variant label, letters represent the realization (r), initialization (i) schemes, physics (p), and forcing (f) schemes, respectively48. To prevent overweighting a single model, multi-model ensemble means are calculated by averaging intra-model results for individual models with more than one running member, followed by averaging across ensemble means49. The first-order conservative remapping scheme was employed to standardize the spatial resolution of the models to 1.25o\(\times\)1.875o latitude-longitude for improved calculation and comparison purposes50. For more detailed information regarding the LUMIP experiments, please refer to Lawrence et al. 46.

The land-use harmonization 2 (LUH2) data, utilized for the Earth system models participating in the CMIP6 program, were employed as the fundamental land use/cover dataset2. This choice ensures consistency with the land use dataset used by CMIP6 model simulations. The LUH2 data possess several advantages, including reliable historical period reconstruction data, high spatial resolution (0.25° × 0.25°), and long time series spanning from 850 to 2100 years. To simplify the classification, the original twelve land use types were consolidated into five categories: forest, non-forest, grazing land, cropland, and urban land46. The new forest and non-forest types encompass both primary and secondary forest as well as non-forest areas. The grazing land type combines natural pasture and management pasture51. The cropland type is an aggregation of five crop-related functional types52. while the urban land type remains unchanged. Detailed information regarding land use data characteristics and processing can be found in Hurtt et al.2.

The probability distribution of surface temperature change

The calculation processes of the rank of surface temperature change (minimum, Q1, Q5, Q10,…, Q90, Q95, Q99, maximum) caused by LULCC were as follows53. First, the monthly multi-model ensemble mean surface temperature change data can be obtained via the subtraction between the “Hist” and “Hist-noLu” experiments over the historical period or between the “SSP126” and “SSP126-SSP370Lu” experiments over the future period. The surface temperature change data included 12 (months per year) multiplied by 20 (years), which is equal to 240 months of data values. These surface temperature change data were spatially averaged in the specific analyzed regions using the latitude-weighted method. Then, these surface temperature change data were sorted in increasing order from negative to positive values (e.g., approximately −1.5 °C to 1 °C in the mid-high latitudes of the Northern Hemisphere over the historical period, Fig. 4a). The Bot5% of the surface temperature change represents the lower tail month caused by LULCC, which is the average of 240 × 5% = 12 values at the beginning of the sorted surface temperature change data. Most of these data are from the boreal cold season.

Surface energy budget equation

The dominant energy redistribution based on the surface energy budget equation is the primary mechanism through which LULCC biogeophysical processes exert their climatic effects13,54. The surface energy budget equation decomposes land‒atmosphere interactions into surface albedo, downward shortwave radiation, downward longwave radiation, latent heat flux, and sensible heat flux to account for a relatively long period of surface energy balance. The surface energy budget equation begins with the land surface energy balance, expressed as

$${{{\rm{SW}}}}_{u}-{{{\rm{SW}}}}_{d}+{{{\rm{LW}}}}_{u}-{{{\rm{LW}}}}_{d}+{{{\rm{LH+SH}}}}+G\approx 0$$
(1)

Where \({{{\rm{SW}}}}_{u}\) and \({{{\rm{SW}}}}_{d}\) are the upward shortwave radiation and downward shortwave radiation, respectively; \({{{\rm{LW}}}}_{u}\) and \({{{\rm{LW}}}}_{d}\) are the upward longwave radiation and downward longwave radiation, respectively; \({{{\rm{LH}}}}\) represents the latent heat flux, \({{{\rm{SH}}}}\) represents the sensible heat flux; and \(G\) is the ground heat flux which can be ignored when considering long-term periods13,16. All of the above terms are in units of \({{{\rm{W\cdot {m}}}^{-2}}}\).

According to the Stefan–Boltzmann law, \({T}_{s}\) represents the surface temperature is proportional to the fourth power of upward longwave radiation, expressed as

$${{{\rm{LW}}}}_{u}=\varepsilon \sigma {T}_{s}^{4}$$
(2)

\(\varepsilon\) and \(\sigma\) are the surface emissivity (approximately 0.97, dimensionless) and the Stefan–Boltzmann constant (approximately \(5.67\times 1{0}^{-8}\,{{{\rm{W\cdot {m}}}^{-2}\cdot {K}^{-4}}}\)), respectively. Combining Eqs. (1) and (2), the formula can be rewritten as

$$\varepsilon \sigma {T}_{s}^{4}={{{\rm{SW}}}}_{d}+{{{\rm{LW}}}}_{d}-{{{\rm{SW}}}}_{u}-{{{\rm{LH}}}}-{{{\rm{SH}}}}$$
(3)

Commonly, we use the surface albedo (\(\alpha\)) to replace \({{{\rm{SW}}}}_{u}\). Hence, the final formula form is shown as (4).

$$\varepsilon \sigma {T}_{s}^{4}=(1-\alpha ){{{\rm{SW}}}}_{d}+{{{\rm{LW}}}}_{d}-{{{\rm{LH}}}}-{{{\rm{SH}}}}$$
(4)

The change between the sensitivity and control experiments (denoted as \(\delta\)) is calculated by taking the derivative of the equation. We employed the first-order Taylor series expansion for solving \(\delta {T}_{s}\)12.

$$\delta {T}_{s}= \frac{1}{4\varepsilon \sigma {T}_{s,{{{\rm{CTL}}}}}^{3}}\left[-{{{\rm{SW}}}}_{d,{{{\rm{CTL}}}}}\delta \alpha +(1-{\alpha }_{{{\rm{CTL}}}})\delta {{{\rm{SW}}}}_{d}+\delta {{{\rm{LW}}}}_{d}-\delta {{{\rm{LH}}}}-\delta {{{\rm{SH}}}}\right]\\ {{\rm{I}}} \qquad\qquad\qquad{{\rm{II}}} \qquad\qquad\quad{{\rm{III}}}\qquad\;\; {{\rm{IV}}}\qquad {{\rm{V}}}\quad$$
(5)

Formula (5) is employed as a metric to quantify the contributions of five meteorological factors (I–V) to the change in surface temperature. It should be noted that a previous study has already demonstrated the adequacy of utilizing the first-order Taylor series expansion, although disregarding higher-order terms inevitably introduces errors55. Furthermore, the surface energy budget equation method implicitly assumes independence among the contributions to changes in surface temperature from different factors56. The results confirm that cross-order terms containing non-linear effects are negligible within \(\pm\)0.1 °C under mean state surface temperature change conditions. Despite some residual values under extreme surface temperature change conditions at mid-high latitudes in the Northern Hemisphere, our findings remain valid (Supplementary Fig. 15 and Supplementary Fig. 16).