Accounting for albedo change to identify climate-positive tree cover restoration

Restoring tree cover changes albedo, which is the fraction of sunlight reflected from the Earth’s surface. In most locations, these changes in albedo offset or even negate the carbon removal benefits with the latter leading to global warming. Previous efforts to quantify the global climate mitigation benefit of restoring tree cover have not accounted robustly for albedo given a lack of spatially explicit data. Here we produce maps that show that carbon-only estimates may be up to 81% too high. While dryland and boreal settings have especially severe albedo offsets, it is possible to find places that provide net-positive climate mitigation benefits in all biomes. We further find that on-the-ground projects are concentrated in these more climate-positive locations, but that the majority still face at least a 20% albedo offset. Thus, strategically deploying restoration of tree cover for maximum climate benefit requires accounting for albedo change and we provide the tools to do so.


DNF
Lands dominated by woody vegetation with a percent cover >60% and height exceeding 2 m.Consists of seasonal needleleaf tree communities with an annual cycle of leaf-on and leaf-off periods.

DBF
Lands dominated by woody vegetation with a percent cover >60% and height exceeding 2 m.Consists of broadleaf tree communities with an annual cycle of leaf-on and leaf-off periods.

Mixed forests (5)
MF Lands dominated by trees with a percent cover >60% and height exceeding 2 m.Consists of tree communities with interspersed mixtures or mosaics of the other four forest types.None of the forest types exceeds 60% of landscape.

Woody savannas (8)
WSA Lands with herbaceous and other understory systems, and with forest canopy cover between 30% and 60%.The forest cover height exceeds 2 m.

Savannas (9)
SAV Lands with herbaceous and other understory systems, and with forest canopy cover between 10% and 30%.The forest cover height exceeds 2 m.Open shrublands (7) OSH Lands with woody vegetation less than 2 m tall and with shrub canopy cover between 10% and 60%.The shrub foliage can be either evergreen or deciduous.

Grasslands (10)
GRA Lands with herbaceous types of cover.Tree and shrub cover is less than 10%.

Croplands (12)
CRO Lands covered with temporary crops followed by harvest and a bare soil period (e.g., single and multiple cropping systems).Note that perennial woody crops will be classified as the appropriate forest or shrub land cover type.
Table S2: Area, carbon, and net climate impacts (NCI) across the total biome and places identified as an opportunity to restore tree cover: the Griscom opportunity map 3 , the Bastin opportunity map 4 , the Walker opportunity map 5 , and unfiltered by an opportunity map (TotalBiome).To be comprehensive, we show all biomes for which we have data, but in the main results we focus on forest and woody savanna biomes that are suitable for upland trees and that represent larger areas.We also show the percentage of the area within each biome that experiences less than 50% albedo offset (AO), as well as overall Pg CO2e and Mg CO2e ha -1 for those areas, and the percentage of the area that is net climate-negative (i.e., >100% albedo offset).Uncertainties in square brackets reflect the maximum and minimum values across the six radiative kernels.S3: Proportion of maximum carbon storage offset by albedo in project-pixels.We show pixel counts and percentage of project pixels that fall into different albedo offset bins.The column on the right indicates only project-pixels that overlapped with at least one of the published opportunity maps 3-5 .When we report statistics in the main text, we assume that locations where our model does not predict an albedo offset (i.e., NA) would have a >100% albedo offset since these are generally in places like deserts where carbon accumulation is very low and the open lands are highly reflective.

Figure S3. Proportion of project pixels with different net climate impacts (Mg CO2e ha -1 ).
We also show the cumulative proportion of pixels (red line, right y-axis).This figure does not include NA values.

Figure S4. Distribution of on-the-ground projects.
From past, on-going, and planned projects that are part of the Grain for Green Program 12 (large gray circles) or uploaded onto Restor 13 (black dots).Project locations are shown on top of the net-climate impact map (Fig. 1a).

Figure S5. Areas experiencing a greater than 50% albedo offset (AO) based on the range observed across six radiative kernels.
We identified the areas with a greater than 50% albedo offset using the minimum radiative forcing observed across all radiative kernels, the median value, and the maximum value.Some locations always have less than a 50% albedo offset (teal) or a greater than 50% offset (darkest brown), regardless of the kernel used.However, a few locations will transition to greater than or less than 50% offset depending on the kernel used.Brown indicates that pixels have greater than 50% offset in both median and maximum radiative forcing maps, while lightest brown indicate pixel that only shown substantial offset when considering maximum value.We provide the spatial data (see Data Availability).and c) percent difference between both net climate impact maps. 5Places with the greatest difference are those where the Walker map may either over-predict biomass (e.g., northern Russia) or where the current biomass may be lower than potential (e.g., China).The scale bar immediately to the right of the maps indicates the 5%, 10%, 25%, 50%, 75%, 90%, and 95% land-area percentiles (top to bottom).6).Within each case we also vary how soon albedo reaches its maximum before carbon.Ocean and land releases of CO2 in response to carbon removals from tree cover restoration are also included in the simulations as an additional term, and this is included in the net cumulative CO2e.We find that the magnitude of the net climate effect changes through time but the sign stays consistent in all but one of the twelve scenarios, with the one exception showing only a slight climate-positive effect before giving way to a climate-negative effect in the long-term.
Where the albedo offset is large (cases 1 and 4), the net climate impact is climate-negative (net cumulative CO2e flux is positive indicating an effect equivalent to a release of CO2 to the atmosphere) regardless of assumptions around time horizon.When the albedo response precedes the carbon response by a lot (50% earlier), we see an earlier warming response (net CO2e emission).Even at longer time frames we see an enduring warming effect because the carbon that was removed from the atmosphere is compensated by ocean and land releases while the large albedo warming is sustained.When the carbon accumulation is faster, so too is the albedo approach to its maximum, and thus we find an earlier net climate warming response.
Where the albedo offset is small (cases 3 and 6), the net climate impact is consistently negative and is insensitive to the timing of albedo relative to carbon.In this situation, atmospheric carbon decreases to a minimum and then sees a rebound as ocean and land outgassing respond over time.We see a maximum climate cooling effect at around the time that the carbon removals saturate, followed by a slow decline as the ocean and land response kick in over time.When the carbon accumulation is faster, in this case we naturally find an earlier net climate cooling effect.
Where the albedo offset is intermediate (cases 2 and 4) we find the most complex temporal dynamics.The short-term effect (20 to 50 years) is a net warming when the albedo effect leads the carbon effect by a lot (50% earlier).However, the medium-term effect can be neutral, a modest warming, or even a slight cooling, as the carbon removals reach their maximum rate, outpacing compensation by ocean and land releases.The long-term effect (150 -200 years) is consistently a net warming response of the climate system as the albedo effect persists but the carbon removals become compensated by the ocean and land release of CO2.From light to dark, areas are assigned either the present land cover or the land cover that is dominant within the respective ecoregion within a 0.01° grid (≤0.01°), within the 0.025° or 0.05° grids (≤0.05°), within the 0.1° grid (≤0.01°), within the 0.25° or 0.5° grids (≤0.5°), within the 1° grid (≤1°), within the 2.5° or 5° grids (≤5°), within the 10° grid (≤10°), the dominant land cover over the entire ecoregion, the dominant land cover within their respective biome, climate-zone and region, or finally the dominant land cover for their respective biome world-wide (see Methods).We provide the data for these histograms (see Data Availability).

Figure S2 .
Figure S2.Global maps of albedo change and maximum potential carbon storage (in MgCO2e ha --1 ) We use the same color scale as Figure1a.(a) Potential albedo change from the most likely open land to woody savanna/forest cover transition, irrespective of actual opportunity or suitability for restoration of tree cover.We show and provide the median value across six radiative kernels 6-11 , but also provide maximum and minimum values (see Data Availability).(b) Maximum carbon storage in above and belowground biomass from Walker et al5 .In both maps the scale bar to the left of the color ramp indicates 5%, 10%, 25%, 50%, 75%, 90%, and 95% land-area percentiles (top to bottom) in Mg CO2e ha -1 .For comparison to other studies, we also label the color ramp in units of Mg Ce ha -1 (italicized text on left side).

Figure S6 .
Figure S6.Net climate impact map with ESA-CCI.a) Net climate impact calculated with a version of Walker et al. 5 where the high values are truncated at the 85% value observed in a current biomass map (ESA-CCI) 14 ("ESA-truncated Walker", see Data Availability), b) absolute difference between ESA-truncated Walker and original Walker et al.net climate impact maps,and c) percent difference between both net climate impact maps.5 Places with the greatest difference are those where the Walker map may either over-predict biomass (e.g., northern Russia) or where the current biomass may be lower than potential (e.g., China).The scale bar immediately to the right of the maps indicates the 5%, 10%, 25%, 50%, 75%, 90%, and 95% land-area percentiles (top to bottom).

Figure S7 .
Figure S7.Different scenarios of albedo and carbon change through time.The combined, cumulative CO2e flux for each is shown in the upper panel and the net climate effect of their combination is shown in the lower panel.Results are divided into six cases that vary the pace of carbon accumulation (fast = cases 1, 2, 3; slow = cases 4, 5, 6), and vary the albedo offset (100% = cases 1, 4; 50% = cases 2, 5; 10% = cases 3, 6).Within each case we also vary how soon albedo reaches its maximum before carbon.Ocean and land releases of CO2 in response to carbon removals from tree cover restoration are also included in the simulations as an additional term, and this is included in the net cumulative CO2e.We find that the magnitude of the net climate effect changes through time but the sign stays consistent in all but one of the twelve scenarios, with the one exception showing only a slight climate-positive effect before giving way to a climate-negative effect in the long-term.

Figure S8 .
Figure S8.Histogram of the spatial proximity used to establish the most likely (a) open land or (b) forest class, expressed as percent of each biome's area.From light to dark, areas are assigned either the present land cover or the land cover that is dominant within the respective ecoregion within a 0.01° grid (≤0.01°), within the 0.025° or 0.05° grids (≤0.05°), within the 0.1° grid (≤0.01°), within the 0.25° or 0.5° grids (≤0.5°), within the 1° grid (≤1°), within the 2.5° or 5° grids (≤5°), within the 10° grid (≤10°), the dominant land cover over the entire ecoregion, the dominant land cover within their respective biome, climate-zone and region, or finally the dominant land cover for their respective biome world-wide (see Methods).We provide the data for these histograms (see Data Availability).

Figure S9 .
Figure S9.Scenarios of forest carbon accumulation over time.Code for recreating these figures is available (see Code Availability).