We combined published data on biophysical effects of deforestation by latitude with our own analysis of CO₂ effects from deforestation by latitude to compare the relative strength of biophysical factors and CO₂ (the dominant biogeochemical factor) affecting global climate. Most modeling experiments available in the literature involve total deforestation at all latitudes, and the ocean feedbacks prove very strong (Davin and de Noblet-Ducoudré, 2010). Here, we consider land-only effects within a given 10° latitudinal band as this scale of impact is more indicative of the effects of regional or more incremental change on global temperature than the combined land/ocean effects. Finer scale, more realistic forest loss scenarios would not trigger massive cooling through albedo effects on the oceans. Area-scaled, land-only biophysical effects from deforestation provide the most realistic comparison with the effects of carbon stored by forests, and released through deforestation, at a given latitude. The biophysical response was derived from the results of Davin and de Noblet-Ducoudré (2010) who simulated total deforestation and decomposed the temperature response, by 10° latitude bands, into the fraction due to albedo, evapotranspiration, roughness and a non-linear response (see Supplementary Table 1).

The biogeochemical response was estimated by accounting for the CO₂ effect of deforestation, using existing biomass data and known equilibrium temperature sensitivity to doubled CO₂. The principal input to our analysis is a 2016 global extension of the 500-m resolution aboveground carbon density (ACD) change (2003–2016) product applied by Walker et al. (2020) to the Amazon basin. It is based on an approach to pantropical ACD change estimation developed by Baccini et al. (2017). The pantropical product combined field measurements with colocated NASA ICESat GLAS spaceborne light detection and ranging (LiDAR) data to calibrate a machine-learning algorithm that produced estimates of ACD using MODIS satellite imagery. This approach was modified for application to the extratropics, principally the temperate and boreal zones but also extratropical South America, Africa and Australia, using 47 allometric equations compiled from 27 unique literature sources for relating field-based measurements of aboveground biomass to airborne LiDAR metrics (Chapman et al., 2020). These equations were used to predict ACD within the footprints of GLAS LiDAR acquisitions in each region with the result being a pseudo-inventory of LiDAR-based estimates of ACD spanning the extratropics. This dataset was then combined with the pantropical dataset first generated by Baccini et al. (2012) to produce a global database of millions of spatially explicit ACD predictions. This database was used to calibrate six ecoregional MODIS-based models for the purposes of generating a global 500-m resolution map of ACD for the year 2016. Additional details on these methods can be found in Chapman et al. (2020).

The total aboveground carbon (GtC) was summed for each 10° latitude band and converted to CO₂ (GtC*44/12 = GtCO₂, Supplementary Information 1). The mass of CO₂ was converted to ppm CO₂ in the atmosphere (2.12 Gt/ppm). The derived CO₂ concentration was reduced by 23% to account for ocean uptake (Global Carbon Project, 2019). We assumed that no uptake occurred on land, as the carbon stock in vegetation was completely removed in our experiment to match what occurred in Davin and de Noblet-Ducoudré (2010). Next, we calculated the global temperature response to the increase in atmospheric CO₂ due to the CO₂ released by completely deforesting each 10° latitudinal band using the equilibrium temperature sensitivity derived from general circulation models. Given the accepted value of 3°C (±1.5°C) for a doubling of atmospheric CO₂ (an increase of 280 ppm) (IPCC, 2013), we determined that temperature sensitivity is equivalent to 0.107°C (±0.054°C) for every 10 ppm increase in atmospheric CO₂ content.

To determine the global temperature response to deforestation of a given band, we calculated the area-weighted values for each biophysical response within each latitude band. The area encompassed by 10° of latitude increases toward the equator. Thus, to determine the contribution of a given band to a global temperature response, scaling by the surface area within the band was essential. We used average temperature responses over the land only to avoid the strong bias associated with ocean feedbacks from global scale implementation of deforestation.

For the global analysis, we also determined the contribution of BVOC to global temperature change for deforestation of each 10° of latitude. Scott et al. (2018) described the warming from deforestation due to BVOC in relation to the amount of cooling due to changes in albedo. For the tropics, the BVOC effect on global temperature was 17% of the albedo effect. For the temperate zone, it was 18% and for the boreal, it was 2% of the albedo effect. We applied these scalars (with an opposite sign) to the albedo figures for each 10° latitude band.

To analyze the effect of deforesting 10° of latitude on the temperature within that latitude zone (‘local’ effect), we did not scale by area within the band. Rather we assessed the average temperature change across the band, locally felt, as reported in the original study. The CO₂ effect was calculated as above and then scaled to reflect the sensitivity of a given latitudinal band to a global forcing. Only the CO₂ emitted by the latitudinal band itself was considered when determining the locally felt effects of CO₂ in a given band. Our experimental design involved global deforestation and all emitted CO₂ would have had an effect in a given band, but the point of the analysis was to isolate the temperature change caused by forests in a given latitude. We determined the latitudinal sensitivity to warming in response to added CO₂ from a re-analysis of global 2 m temperature data (CERA-20C) obtained from the European Centre for Medium-Range Weather https://www.ecmwf.int/en/forecasts/datasets/reanalysis-datasets/cera-20c. We compared average temperatures from 1901 to 1910 and 2001–2010, by latitude on land only (inadequate land only data for 50–60S and 80–90N; for those, we do not report a locally felt CO₂ effect). Then we divided the temperature change for each latitude band by the change in global temperature over the same period. We scaled the effect of CO₂ emitted by a given 10° latitude band by this sensitivity to represent the influence of non-linear responses such as polar amplification (see Supplementary Information 1 and Supplementary Table 2).

Our analysis is the first to compare regional scale biophysical and CO₂ impacts from regional scale deforestation but the literature is replete with data on local biophysical impacts. The results for local biophysical effects (100s of m to 100s of km) agree with our results at the regional scale (below). Figures 2, 3 synthesize local biophysically-driven temperature responses to deforestation, as indicated by forest/no-forest comparisons or forest change over time, from the scientific literature. Satellite and flux tower data indicate that surface temperatures in tropical forests are significantly lower than in cleared areas nearby. On an annual basis, local surface cooling of 0.2–2.4°C has been observed (mean 0.96°C, Figure 2 and Supplementary Information 2). In the temperate zone, satellite studies of land surface temperature (which is more sensitive than the temperature of the air at 2 m) have shown biophysical cooling from forest cover, or biophysical warming from deforestation (0.02–1.0°C, mean of 0.4°C; see Figure 2 and Supplementary Information 2). Both in situ and satellite data generally indicate an average annual cooling of under 1°C from boreal deforestation (Figure 2). Across latitudinal zones, warming from deforestation is generally greater during the day, and during the dry (hot) season (Figure 3).

As expected, the regionally felt effect of regionally (10° band) produced CO₂ is very small compared to any individual biophysical effect or the sum of all non-CO₂ effects (Figure 4). These results indicate that the net impact of all non-CO₂ effects is negligible between 20 and 30N. Beyond 30N the local biophysical response to deforestation is cooling. In the broader literature, this latitude of net zero biophysical effect on local temperature is generally between 30 and 40N (Figure 1).

For most latitudinal bands, the strongest biophysical effect of deforestation is cooling from albedo changes. In the tropics, however, the warming effect of lost roughness is comparable to or greater than the albedo effect (Figure 5A). Adding the warming from lost evapotranspiration, the net biophysical effect from tropical deforestation is global warming, as much as 0.1°C contributed each by latitudes 0°–10°S and 0°–10°N. The net biophysical effect of intact tropical forest, therefore, is global cooling; slightly more cooling if BVOCs are also considered (see Figure 5B). Roughness effects are generally greater than evapotranspiration effects across latitudes providing a strong counterbalance to albedo effects (Davin and de Noblet-Ducoudré, 2010; Burakowski et al., 2018; Winckler et al., 2019b; Figure 5A). Albedo almost balances the combined effect of roughness, evapotranspiration, BVOC and non-linear effects between 20 and 30°N resulting in close to zero net biophysical effect on global temperature (Figure 5B). From 30–40°N and northward, albedo dominates, and the net biophysical effect of deforestation is cooling.

From 30°S to 30°N, the biophysical effect of deforesting a given 10° latitudinal band is about half as great and in the same direction as the CO₂ effect: global warming. Biophysical warming is around 60% as great as warming from released CO₂ in the outer tropics (20°S–10°S and 10°N–20°N) and about 35% as great in the heart of the tropics (10°S–10°N). Biophysical cooling due to deforestation from 30°N to 40°N offsets about 40% of the warming associated with carbon loss from deforestation; from 40°N to 50°N biophysical effects offset 85% of CO₂ effects (Figure 5B). Above 50°N, biophysical global cooling is 3–6 times as great as CO₂ induced global warming. The net impact of deforestation (effects of CO₂, biophysical processes and BVOC combined) is warming at all latitudes up to 50N (Figure 5C). Thus, from 50S to 50N, an area that encompasses approximately 65% of global forests (FAO, 2020), deforestation results in global warming (Figure 5C).

Ignoring biophysical effects on local climate means casting aside a powerful inducement to promote global climate goals and advance forest conservation: local self-interest. The biogeochemical effect of forests tends to dominate the biophysical effect at the global scale because physical effects in one region can cancel out effects in another. Nevertheless, biophysical effects are very important, and can be very large, at the local scale (e.g., Anderson-Teixeira et al., 2012; Bright et al., 2015; Jiao et al., 2017; Figures 2–4). The role of forests in maintaining critical habitat for biodiversity is well known, but new research on extinction confirms the role of forests in maintaining critical climates to support biodiversity. Changes in maximum temperature are driving extinction, not changes in average temperature (Román-Palacios and Wiens, 2020). Deforestation is associated with an increase in the maximum daily temperature throughout the year in the tropics and during the summer in higher latitudes (Lee et al., 2011; Zhang et al., 2014). Of course deforestation also increases average daytime temperatures in boreal, mid-latitude and tropical forests (Figure 3). The biophysical effects of forests also moderate local and regional temperature extremes such that extremely hot days are significantly more common following deforestation even in the mid- and high latitudes (Vogel et al., 2017; Stoy, 2018). Historical deforestation explains ∼1/3 of the present day increase in the intensity of the hottest days of the year at a given location (Lejeune et al., 2018). It has also increased the frequency and intensity of hot dry summers two to four fold (Findell et al., 2017). Local increases in extreme temperatures due to forest loss are of comparable magnitude to changes caused by 0.5°C of global warming (Seneviratne et al., 2018). Forests provide local cooling during the hottest times of the year anywhere on the planet, improving the resilience of cities, agriculture, and conservation areas. Forests are critical for adapting to a warmer world.

Forests also minimize risks due to drought associated with heat extremes. Deep roots, high water use efficiency, and high surface roughness allow trees to continue transpiring during drought conditions and thus to dissipate heat and convey moisture to the atmosphere. In addition to this direct cooling, forest ET can influence cloud formation (Stoy, 2018), enhancing albedo and potentially promoting rainfall. The production of BVOCs and organic aerosols by forests accelerates with increasing temperatures, enhancing direct or indirect (cloud formation) albedo effects. This negative feedback on temperature has been observed to counter anomalous heat events in the mid-latitudes (Paasonen et al., 2013).

Disregarding the biophysical effects of specific forests on global climate means under-selling some forest actions and over-selling others. The response to local forest change is not equivalent for similar sized areas in different latitudes. According to Arora and Montenegro (2011) warming reductions per unit reforested area are three times greater in the tropics than in the boreal and northern temperate zone due to a faster carbon sequestration rate magnified by year-round biophysical cooling. Thus, considering biophysical effects significantly enhances both the local and global climate benefits of land-based mitigation projects in the tropics (see Figures 4, 5).

Climate change is likely to alter the biophysical effect of forests in a variety of ways. Deforestation in a future (warmer) climate could warm the tropical surface 25% more than deforestation in a present-day climate due to stronger decreases in turbulent heat fluxes (Winckler et al., 2017b). In a warmer climate, reduced snow cover in the temperate and boreal regions will lead to a smaller albedo effect and thus less biophysical cooling with high latitude deforestation. In addition to snow cover change, future rainfall regimes will affect the response of climate to changes in forest cover (Pitman et al., 2011) as rainfall limits the supply of moisture available for evaporative cooling. Increases in water use efficiency due to increasing atmospheric CO₂ may reduce evapotranspiration (Keenan et al., 2013), potentially reducing the local cooling effect of forests and altering atmospheric moisture content and dynamics at local to global scales. Future BVOC production may increase due to warming and simultaneously decline due to CO₂ suppression (Lathière et al., 2010; Unger, 2014; Hantson et al., 2017). The physiological and ecological responses of forests to warming, rising atmospheric CO₂ and changing precipitation contribute to uncertainty in the biophysical effect of future forests on climate.

Forest persistence is essential for maintaining the global benefits of carbon removals from the atmosphere and the local and global benefits of the physical processes described above. Changing disturbance regimes may limit forest growth and regrowth in many parts of the world. Dynamic global vegetation models currently show an increasing terrestrial carbon sink in the future. This sink is thought to be due to the effects of fertilization by rising atmospheric CO₂ and N deposition on plant growth as well as the effects of climate change lengthening the growing season in northern temperate and boreal areas (Le Quéré et al., 2018). Free-air carbon dioxide enrichment (FACE) experiments often show increases in biomass accumulation under high CO₂ but results are highly variable due to nutrient limitations and climatic factors (Feng et al., 2015; Paschalis et al., 2017; Terrer et al., 2018). Climate change effects on the frequency and intensity of pest outbreaks are poorly studied, but are likely to be significant, particularly at the margins of host ranges. Warmer springs and winters are already increasing insect-related tree mortality in boreal forests through increased stress on the tree hosts and direct effects on insect populations (Volney and Fleming, 2000; Price et al., 2013).

Climate also affects fire regimes. In the tropics, fire regimes often follow El Niño cycles (van der Werf et al., 2017). As temperatures increase, however, fire and rainfall are decoupled as the flammability of forests increases even in normal rainfall years (Fernandes et al., 2017; Brando et al., 2019). Fire frequency is also increasing in some temperate and boreal forests, with a discernable climate change signal (Abatzoglou and Williams, 2016). Modeling exercises indicate that this trend is expected to continue with increasing damage to forests as temperatures rise and fire intensity increases (De Groot et al., 2013).

In addition to changes induced by warming, continued deforestation could severely stress remaining forests by warming and drying local and regional climates (Lawrence and Vandecar, 2015; Costa et al., 2019; Gatti et al., 2021). In the tropics, a tipping point may occur, potentially resulting in a shift to shorter, more savannah-like vegetation and altering the impact of vast, previously forested areas on global climate (Nobre et al., 2016; Brando et al., 2019). Some of these processes are included in climate models and some are not. The gaps leave considerable uncertainty. Nevertheless, a combination of observations, models, and theory gives us a solid understanding of the biophysical effects of forests on climate at local, regional and global scales. We can use that knowledge to plan forest-based climate mitigation and adaptation.

If instead of focusing on the contrast between biophysical and biochemical impacts of forests and forest loss, we focus on the potential of forests to cool the planet through both pathways, another picture emerges. By our conservative estimate, through the combined effects on CO₂, BVOC, roughness and evapotranspiration, forests up to 50°N provide a net global cooling that is enough to offset warming associated with their low albedo. Given the most realistic pathways of forest change in the future (not complete deforestation of a 10° latitudinal band, or an entire biome), global climate stabilization benefits likely extend beyond 50°N. For the 29% of the global land surface that lies beyond 50°N, forests may warm the planet, but only as inferred from assessing the effects of complete zonal deforestation with all the associated, and powerful, land-ocean feedbacks spawned by largescale forest change in the boreal zone. Forests above 50°N, like forests everywhere, provide essential local climate stabilization benefits by reducing surface temperatures during the warm season as well as periods of extreme heat or drought. Indeed, they also reduce extreme cold.

Creating a fair and effective global arena for market-based solutions to climate change requires attention to all the ways that forests affect climate, including the biophysical effects. Future metrics of forest climate impacts should consider the effects of deforestation beyond CO₂. Only recently have modelers begun to include BVOC. Doing so means that the albedo of intact forests (or the atmosphere above them) is higher due to the creation of SOA and subsequent cloud formation. Modeled deforestation thus results in less of a change in albedo, reducing the biophysical cooling effect. Similarly, accounting for the ozone and methane effects of BVOC reduces the biogeochemical warming from deforestation (Scott et al., 2018). In addition, especially in the tropics, deforestation reduces the strength of the soil CH₄ sink (Dutaur and Verchot, 2007). While a small change relative to the atmospheric pool of CH₄ (the second most important greenhouse gas), the loss of this sink is equivalent to approximately 13% of the current rate of increase in atmospheric CH₄ (Saunois et al., 2016). We already have the data (Figure 5) to begin conceptualizing measures to coarsely scale CO₂ impacts of forest change by latitude. Finer resolution of latitude, background climate (current and future) and forest type would improve any such new, qualifying metric for the climate mitigation value of forests.

The role of forests in addressing climate change extends beyond the traditional concept of CO₂ mitigation which neglects the local climate regulation services they provide. The biophysical effects of forest cover can contribute significantly to solving local adaptation challenges, such as extreme heat and flooding, at any latitude. The carbon benefits of forests at any latitude contribute meaningfully to global climate mitigation. In the tropics, however, where forest carbon stocks and sequestration rates are highest, the biophysical effects of forests amplify the carbon benefits, thus underscoring the critical importance of protecting, expanding, and improving the management of tropical forests. Perhaps it is time to think more broadly about what constitutes global climate mitigation. If climate mitigation means limiting global warming, then clearly the biophysical effects of deforestation must be considered in addition to its effects on atmospheric CO₂. We may further consider whether mitigation is too narrow a scope for considering the climate benefits provided by forests. Climate policy often separates mitigation from adaptation, but the benefits of forests clearly extend into both realms.