Report

Biophysical climate impacts of recent changes in global forest cover

See allHide authors and affiliations

Science  05 Feb 2016:
Vol. 351, Issue 6273, pp. 600-604
DOI: 10.1126/science.aac8083

It's not only the carbon in the trees

Forest loss affects climate not just because of the impacts it has on the carbon cycle, but also because of how it affects the fluxes of energy and water between the land and the atmosphere. Evaluating global impact is complicated because deforestation can produce different results in different climate zones, making it hard to determine large-scale trends rather than more local ones. Alkama and Cescatti conducted a global assessment of the biophysical effects of forest cover change. Forest loss amplifies diurnal temperature variations, increases mean and maximum air temperatures, and causes a significant amount of warming when compared to CO2 emission from land-use change.

Science, this issue p. 600

Abstract

Changes in forest cover affect the local climate by modulating the land-atmosphere fluxes of energy and water. The magnitude of this biophysical effect is still debated in the scientific community and currently ignored in climate treaties. Here we present an observation-driven assessment of the climate impacts of recent forest losses and gains, based on Earth observations of global forest cover and land surface temperatures. Our results show that forest losses amplify the diurnal temperature variation and increase the mean and maximum air temperature, with the largest signal in arid zones, followed by temperate, tropical, and boreal zones. In the decade 2003–2012, variations of forest cover generated a mean biophysical warming on land corresponding to about 18% of the global biogeochemical signal due to CO2 emission from land-use change.

Forests play a relevant role in the climate system by absorbing approximately one-fourth of anthropogenic CO2 emissions (1), storing large carbon pools in tree biomass and forest soils (2), and modulating the land-atmosphere exchange of energy and water vapor (3). Given the important role of forests in the global carbon cycle, climate treaties account for land-based mitigation options such as afforestation, reforestation, and avoided deforestation or forest degradation (4, 5). On the contrary, the climate impacts of biophysical processes, such as the surface exchange of energy and water vapor (6), are still uncertain in sign and magnitude and therefore have not been considered in climate negotiations to date.

Over the past two decades, the biophysical effects of deforestation on climate have been assessed mainly by comparing paired model simulations with contrasting forest cover (712). These analyses have shown that, despite the increase in surface albedo, the net biophysical effects of tropical deforestation may increase surface temperature through the reduction of evapotranspiration (9, 13). On the contrary, boreal deforestation may lead to net climate cooling due to the high snow albedo in cleared areas during winter/spring and to the land-albedo/sea-temperature feedback (11, 12, 14). However, results of these numerical experiments are model-dependent, and the uncertainties in sign, magnitude, and spatial distribution of the predicted effects are very large (1517). Therefore, direct observations of the biophysical climate effects of recent forest losses and gains are required to constrain predictions, reduce the uncertainty of model ensembles, and provide robust recommendations to climate policy.

To date, data-driven assessments based on in situ (1820) or satellite observations (3, 21, 22) have adopted the space-for-time analogy, meaning that spatial differences in surface temperature between areas with contrasting forest cover have been interpreted as the climate signal of hypothetical deforestation/afforestation. The substitution of space for time produces unbiased results only if forests are randomly distributed in the landscape. Conversely, the systematic location of forests in less favorable areas (such as steeper or colder slopes, shallow soils, etc.) may produce spatial gradients in surface climate that should not be attributed to changes in land cover (18). In addition, both model-based and observation-based assessments have focused so far on idealized scenarios of deforestation (10, 11) and on the estimation of climate sensitivities to land-use change (3, 18, 22), but the climate signal generated by the ongoing changes in forest cover has not yet been quantified.

To overcome the limits and uncertainties of past assessments, in this work we focused on areas that underwent recent land cover transitions, with the objective of providing a global, robust, and data-driven assessment of the biophysical climate impacts of observed forest gains and losses. The analysis builds on overlapping satellite retrievals of surface radiometric temperature (23) and of high-resolution variations in forest cover (24). A novel methodology has been developed to disentangle the effect of forest cover change from the global climate signal [details in supplementary materials (SM) text S1.2]. For this purpose, the temperature difference (ΔT) between two years at a given location is expressed as the effect of forest cover change (ΔTfcc) plus the residual signal (ΔTres) due to climate variability (Eq. 1)

ΔT = ΔTfcc + ΔTres → ΔTfcc = ΔT – ΔTres (1)

The temporal variation in air surface temperature (ΔT) is estimated from satellite retrievals of radiometric land surface temperature, evapotranspiration, and albedo, with semi-empirical models calibrated against in situ measurements of air temperature (SM text S1.1 and figs. S1 and S2). For a given location, we derive ΔTfcc from Eq. 1 by estimating ΔTres from adjacent areas with stable forest cover and therefore where ΔTfcc ≃ 0 and ΔT ≃ ΔTres (fig. S4). To estimate ΔTres, areas located within 50 km of the target location were considered, using the inverse distance as a weighting factor (methods in SM text S1.2). At a seasonal time scale, the residual signal ΔTres is typically in the range ±1.5°C, as the temperature signal of forest cover changes (fig. S4). We focused the analysis on the first and last year of the available time series (i.e., 2003 and 2012) in order to maximize the observed land cover change and therefore the spatial extent and robustness of the estimates. In parallel, the interannual variability of the climate signal was investigated by comparing 2003 with each of the other years, and the robustness of the signal was estimated on an ensemble of nine pairs of years (2003–2005 versus 2010–2012; methods in SM text S1.2.2).

Results show that in all climate zones, forest clearing produces a marked increase of mean annual maximum air surface temperatures, slight changes in minimum temperatures, and an overall increase of mean temperatures, except at the northernmost latitudes (Fig. 1, B and C). In fact, the removal of forest cover does not significantly affect the mean air temperature in the boreal zone, whereas it increases the temperature by about 1°C in the temperate and tropical zones and by more than 2°C in the arid zone (Fig. 1C). These signals show limited interannual variability and a decreasing uncertainty at the increase of the time interval and therefore of the area affected by cover change (fig. S9). In the temperate and boreal zones, the warming induced by forest losses declines over time, presumably because of the progressive recovery of vegetation in forest clearings. On the contrary, tropical areas show a stable signal, likely due to the conversion of forests to agriculture (fig. S9).

The methodology used in Fig. 1 to investigate mean annual temperatures has been replicated at a monthly time scale to explore the seasonal temperature sensitivity (Fig. 2 and figs. S5 and S6). These monthly signals show a limited variability between years in the different climate zones and latitudinal bands (figs. S10 and S11). The climate impact of deforestation is modulated by the incoming radiation and, as a consequence, the largest warming occurs during the summer solstice at maximum temperatures, whereas changes in air temperatures during nighttime are negligible. The substantial reduction of evapotranspiration and surface roughness with forest clearing (11, 13) is the most plausible explanation for the substantial local warming under high radiation load. The remarkable daytime warming ultimately leads to an increase in the diurnal variation (the difference between the daily maximum and minimum temperature) of about 1.13 ± 0.1°C, 2.85 ± 0.04°C, 4.4 ± 0.17°C, and 1.95 ± 0.08°C over the boreal, temperate, arid, and tropical climate zones, respectively. In summary, forests show important biophysical mitigation effects on local maximum temperature in all climate zones, by reducing local daytime summer temperatures and substantially decreasing the diurnal and annual temperature variations. The key role of evapotranspiration in the biophysical impacts of forest clearing emerges from the ranking of the climate zones, with the arid areas showing the strongest signal, followed by the temperate, the tropical, and the boreal zones.

Fig. 1 Impacts on surface temperature of changes in forest cover for the different climate zones.

The panels in rows 1 to 4 show the observed climate impacts of changes in forest cover between 2003 and 2012 in the climate zones defined in (A). (B) Observed variations in mean annual air surface temperature due to observed changes in forest cover (seasonal plots are reported in figs. S5 and S6). The sensitivity of the mean (dark red), minimum (black), and maximum (orange) air surface temperatures (C) and land surface temperature (D) to the fraction of the deforested area is shown.

Fig. 2 Seasonal changes in air and land surface temperature due to losses of forest cover.

Expected changes in the monthly maximum, minimum, and mean air (A) and land (B) surface temperature due to the total clearing of a 0.05° grid cell in the different climate zones are shown(mean ± weighted root mean square error, WRMSE).

Climate sensitivity to losses of forest cover was investigated at the regional scale by applying in a 12°-by-12° moving window the same methodology used to investigate the climate zones (Fig. 3 and fig. S7; methods in SM text S1.2.1). This analysis shows that forest losses in tropical areas generate warming across all seasons. Contrasting effects occur at northern latitudes between seasons (winter cooling and summer warming) and continents (warming in North America and cooling in northern Eurasia, Fig. 3, A and B). In accordance with observations performed in North America (18, 20), forest clearing increases the diurnal temperature variation during summer months at all latitudes (Fig. 3, C and D), whereas it has no effects on the diurnal variation during the boreal winter, because of the dominant effect of snow albedo. The changes in diurnal temperature variation (Fig. 3D) are substantially larger than those in mean annual temperature (Fig. 3B). These results highlight the fact that local biophysical processes triggered by forest losses can effectively increase summer temperatures in all world regions, further amplifying the climate trends driven by the increasing greenhouse gas concentrations.

Fig. 3 Regional changes in air surface temperature due to losses in forest cover between 2003 and 2012.

Changes in mean annual air temperature (A) and diurnal variations (C) due to forest losses are shown. The symbol size indicates the magnitude of forest cover losses, and the color specifies the average temperature sensitivity to total deforestation. Points are spaced 4° in both latitude and longitude, and statistics were computed in windows measuring 12° by 12°. (B and D) Zonal averages of the annual, summer, and winter air surface temperature statistics at 4° of latitudinal resolution (the equivalent image for land surface temperature is reported in fig. S7).

The estimated rate of local warming after forest losses is lower than that predicted by observation-driven studies focused on land surface temperature (22). In addition, the low sensitivity of the mean annual temperature observed at northern latitudes is in contrast with model simulations of large-scale deforestation that typically predict a sharp reduction in boreal temperatures (8, 10), possibly amplified by land-ocean interactions (11). Differently from previous assessments (3, 22), our analysis focused on air surface temperature instead of land surface temperature, given the greater relevance of the first parameter in climate science. On this aspect Figs. 1, C and D, and 2 show that the sensitivity of land surface temperature to changes in forest cover is about 50% larger than that of air temperature. This large difference is probably driven by satellite retrievals of land surface temperature that are biased toward clear sky conditions, when the biophysical differences between land covers are maximized. In addition, it is important to consider that our analysis quantifies the local impacts of fine-scale variations in forest cover that are primarily driven by changes in the surface energy budget and related first-order interactions with the boundary layer. Therefore, this assessment cannot capture the signal of large-scale land-atmosphere interactions and regional teleconnections. On the contrary, model experiments of idealized large-scale deforestation also account for second-order effects and feedbacks (such as changes in cloud cover, rainfall, sea surface temperature, etc.) that may amplify and eventually override the local temperature signal of deforestation (25). In particular, significant feedbacks between land surface albedo and sea temperature seem to drive the temperate/boreal cooling in model experiments of global deforestation (11).

Afforestation or reforestation can significantly attenuate the biophysical effect of forest clearing on surface temperature, especially over boreal and temperate zones, where gains in forest cover compensated for more than 60% of forest losses in the decade 2003–2012 (Fig. 4A). On the contrary, over the same period, forest gains offset less than 30% of the losses in the tropics, leading to a significant net deforestation (24, 26) (Fig. 4C). The strong local effect of the changes in forest cover on air surface temperature turns out to be rather minor when averaged over the year at the global scale (0.0062°C, Fig. 4), due to the attenuation effect of the boreal winter and of nighttime temperatures, to the compensatory effect of forest gains, and to the limited extent of forest losses. On average, in the analyzed decade, the global biophysical warming due to changes in forest cover is equal to about 18% (12 to 42%) of the biogeochemical warming due to CO2 emissions from land-use change (methods in SM text S1.3).

Fig. 4 Net impact of deforestation and afforestation on monthly air surface temperatures.

Changes in monthly mean, maximum, and minimum air surface temperature due to forest cover change over the (A) boreal, (B) temperate, and (C) tropical climate zones (the arid zone has experienced minor changes in forest cover) are shown. Yellow and green squares represent the global temperature signal of forest cover losses and gains, and light red and blue bars indicate the warming and cooling due to the net change in forest cover (the equivalent image for land surface temperature is shown in fig. S8).

This analysis reveals that the biophysical effects of changes in forest cover can substantially affect the local climate by altering the average temperature and, even more markedly, the maximum summer temperatures and the diurnal and annual variations (18, 20). In addition to the global mitigation effects of the terrestrial carbon sink (1), the biophysical properties of forests can therefore contribute to the mitigation of climate extremes, in particular by reducing daytime temperatures during summer months (27). These effects are relevant both in the tropics, where deforestation rates are still substantial and forest clearing generates warming throughout the whole year, and in the boreal zone, where forests contribute to the mitigation of rapidly increasing summer temperatures. Overall, the observation-driven global quantification of the biophysical signal of deforestation provided in this study may support accounting for land biophysics in climate negotiations, as well as the definition of novel protocols for the measurement, reporting, and verification of these relevant effects.

Supplementary Materials

www.sciencemag.org/content/351/6273/600/suppl/DC1

Materials and Methods

Supplementary Text

Figs. S1 to S11

Table S1

References (2839)

REFERENCES AND NOTES

  1. ACKNOWLEDGMENTS: The authors gratefully acknowledge G. Duveiller and F. Achard for inspiring discussions and valuable inputs. The authors acknowledge the use of forest cover data from the University of Maryland (http://earthenginepartners.appspot.com/science-2013-global-forest); gridded air surface temperature data TS.3.22 from the Climatic Research Unit, University of East Anglia (http://catalogue.ceda.ac.uk/uuid/3f8944800cc48e1cbc29a5ee12d8542d); and in situ monthly data of air temperature from the Global Historical Climatology Network (GHCN-Monthly, http://www.ncdc.noaa.gov/ghcnm/v3.php) of the National Oceanic and Atmospheric Administration’s National Climatic Data Center. The MODIS products LST MYD11C3 and NDVI MYD13C2 were retrieved through the online data pool of the NASA Land Processes Distributed Active Archive Center (https://lpdaac.usgs.gov). The authors are grateful to C. Martinez for the linguistic revision. The work was supported by the European Commission JRC-IES-H07 ClimEcos (995) and EU-FP7-LUC4C (603542). The authors contributed equally to the conception of the work, design of the analysis, development of the methodology, and interpretation of results. R.A. was responsible for the data processing and artwork, and A.C. finalized the writing of the text.
View Abstract

Navigate This Article