全球变化知识资源中心

Tower, ground-based and satellite observations indicate that tropical deforestation results in warmer, drier conditions at the local scale. Understanding the regional or global impacts of deforestation on climate, and ultimately on agriculture, requires modelling. General circulation models show that completely deforesting the tropics could result in global warming equivalent to that caused by burning of fossil fuels since 1850, with more warming and considerable drying in the tropics. More realistic scenarios of deforestation yield less warming and less drying, suggesting critical thresholds beyond which rainfall is substantially reduced. In regional, mesoscale models that capture topography and vegetation-based discontinuities, small clearings can actually enhance rainfall. At this smaller scale as well, a critical deforestation threshold exists, beyond which rainfall declines. Future agricultural productivity in the tropics is at risk from a deforestation-induced increase in mean temperature and the associated heat extremes and from a decline in mean rainfall or rainfall frequency. Through teleconnections, negative impacts on agriculture could extend well beyond the tropics.

Considerable evidence from both modelling and empirical studies indicates that tropical deforestation on many scales influences local, regional, and even global climate^{1, 2, 3}. Deforestation-driven changes to water availability and climate variability could have strong implications for agricultural production systems and food security in some regions. Here we review the impacts of tropical deforestation on climate in the three major tropical regions at spatial scales from local to global, and we consider the implications for agriculture.

Experiments using general circulation models (GCMs) have greatly influenced our understanding of how the land surface affects climate. GCMs are global, 3D computer models of the climate system that link the atmosphere, oceans, and land surface. They operate on a coarse resolution, often with grids of 1°–5° (latitude and longitude, ~110 km to ~550 km across at the equator). Current models incorporate the hydrologic cycle and an explicit representation of plant canopies and their effect on energy and water fluxes (including radiative and turbulent transfers, and the physical and biological controls on evapotranspiration). The atmosphere and biosphere form a coupled system whereby climate influences vegetation distribution and ecosystem function, which feed back to affect climate⁴. These models have been used to simulate the climatic consequences of global deforestation, pantropical deforestation, and regional-scale deforestation of the Amazon basin, Central Africa, or Southeast Asia.

Global, pantropical, and regional impacts of pantropical deforestation. Complete deforestation of the entire tropics leads to an increase in global mean temperatures, and no change in global mean precipitation. The magnitude of predicted global warming varies from 0.1–0.7 °C (refs 5,6,7). Thus, at the upper end, deforestation of the tropics would effectively double the observed warming since 1850⁸. Mean precipitation remains unchanged, partly owing to the opposing signals of local and regional impacts and because global averaging minimizes these climatic responses^{5, 9, 10, 11}.

Warming is more substantial across the tropics (0.1–1.3 °C), as is drying, (approximately −270 mm yr⁻¹ or up to 10–15% of annual rainfall)^{7, 11}. Amazonia responds more strongly than Southeast Asia or Central Africa. Predicted warming ranges from 0–2 °C (mean of 0.63 °C) over the Amazon basin^{7, 11, 12, 13, 14, 15, 16}. Responses in Southeast Asia and Central Africa are much less severe (–0.2 °C to +0.2 °C and −0.2 °C to +0.4 °C, respectively)^{13, 14, 16, 17}. Pantropical deforestation leads to drying in all tropical regions: −146 to −547 mm yr⁻¹ (mean −337 mm yr⁻¹) for Amazonia, −44 to −251 mm yr⁻¹ (mean −179 mm yr⁻¹) for Southeast Asia, and −63 to −109 mm yr⁻¹ (mean −88 mm yr⁻¹) for Central Africa^{11, 12, 13, 14, 15, 16, 17}.

Regional impacts of regional deforestation. Studies using GCMs agree overwhelmingly that continental-scale deforestation in Amazonia, Africa or Southeast Asia leads to a warmer, drier climate over the deforested area. Many factors, including soil type, vegetation, topography, climatology, and distribution of land and water, determine the sensitivity of regional climate to land-cover change³. This dependency on region-specific characteristics makes it impossible to predict the consequences of deforestation in one region by extrapolating from observations in another. Amazonia and Central Africa both occupy continuous continental landmasses; Amazonian forest covers an area 25% larger. Africa is hotter and drier due to northward diversion of moisture-laden winds from the east and higher elevation¹⁸. More fires generate more aerosols there, potentially affecting rainfall. Southeast Asian forests are the most humid, with highest rainfall¹⁸. They occur on the smallest land area surrounded by vast oceans. This complex land–water distribution makes the ocean response to deforestation very important.

The three tropical areas differ in their sensitivity to deforestation. Simulated complete deforestation of the Amazon alone yields a warmer, drier climate over the deforested area (0.1–3.8 °C, mean 1.9 °C; -140 to -640 mm yr⁻¹ (<10-30%, mean −324 mm yr⁻¹ or ~15% of annual rainfall)^{19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34}. Results are sensitive to model structure. For example, coupling fully interactive oceans to the atmosphere resulted in twice the rainfall reduction observed in non-coupled GCM experiments³⁵. Few GCM studies have explored complete deforestation in Central Africa alone or Southeast Asia alone. In Central Africa, predicted warming and drying are somewhat less than in Amazonia (0.5–2.5 °C, −2 to −3 mm per day during the dry season, approximately −240 mm or 15% of annual rainfall)^{36, 37, 38}. In Southeast Asia, responses are milder still (warming of ~0.23 °C, drying of 132 mm yr⁻¹)³⁹.

Extratropical impacts of regional deforestation. Tropical deforestation can lead to extratropical changes in temperature and rainfall. Changes in large-scale atmospheric circulations (for example, geopotential height, Rossby waves, Hadley circulations, and Ferrel cells) allow the propagation of climatic impacts to geographically remote areas through teleconnections^{6, 9, 40, 41, 42, 43, 44}. Even relatively small land-cover perturbations in the tropics can lead to impacts at higher latitudes^{45, 46}.

Simulating the complete deforestation of Amazonia reduces rainfall in the US Midwest during the agricultural season as well as the US Northwest and portions of the south^{9, 40, 47} (Fig. 1). In another model, it increases rainfall (particularly in winter) over the eastern seaboard of the US and the northeast Atlantic, extending towards western Europe and possibly affecting precipitation in Europe³³. Effects of deforesting the Amazon extend all the way to Asia⁹.

Increasing (circles) and decreasing (triangles) precipitation result from complete deforestation of either Amazonia (red), Africa (yellow), or Southeast Asia (blue). Boxes indicate the area in which tropical forest was removed in each region. Numbers refer to the study from which the data were derived.

Historically, GCM studies have applied unrealistically large perturbations such as one-time complete deforestation of the Amazon basin. Recent modelling has simulated more realistic deforestation scenarios at the regional scale, incorporating different patterns, rates, and scales of deforestation. As expected, climate impacts of slow, prolonged, and incomplete deforestation differ from the impacts of sudden, complete deforestation^{6, 7, 9, 14, 32, 52}.

Partial versus complete deforestation. Medvigy et al.⁵³ simulated both complete and 'business-as-usual' (BAU) deforestation of the Amazon (incremental forest loss with 40% of the Amazon deforested by 2050). Their model operated at a resolution usually limited to regional models: a 25-km grid over South America and the nearby oceans, gradually coarsening to a 200-km grid over the rest of the world. For complete deforestation, this variable resolution model predicted a smaller reduction in precipitation (–62 mm yr⁻¹) than most previous coarse-resolution GCMs. In addition, the northwest Amazon became drier and the southeast Amazon became wetter. BAU deforestation yielded less severe impacts on precipitation than complete deforestation during certain times of the year and in some regions. In both BAU and complete deforestation scenarios, extreme cold events became more common in Columbia and Argentina, far from modelled deforestation. However, widespread climatic impacts occurred outside of the deforested area in the complete deforestation scenario that did not occur in the BAU scenario, including a 400 km northward shift in the Atlantic intertropical convergence zone (ITCZ).

Where previous GCMs of complete pantropical deforestation showed a modest to substantial decrease in rainfall, a gradual decline in forest cover resulted in a modest increase in annual rainfall for Central Africa and Southeast Asia. The increase ranged from 65–105 mm yr⁻¹ in Central Africa (approximately 4–7%) and from −22 to +123 mm yr⁻¹ in Southeast Asia (approximately −1 to +5%)⁵⁴.

The rate of deforestation does not strongly affect climate impacts. Castillo and Gurney⁵⁴ reduced forest across the tropics by 0.5, 1, 2, and 5% per year down to a target of 10% remaining per forest functional type. Temperature and precipitation were sensitive to rate only in Southeast Asia, but the response was not linear. The Amazon was the only region where annual rainfall declined significantly with tree cover, and even there drying did not respond linearly to deforestation rate.

Critical thresholds at large scales. A number of GCM studies that incorporate various levels of deforestation suggest that tropical forest clearing beyond ~30–50% may constitute a critical threshold for Amazonia, beyond which reduced rainfall triggers a significant decline in ecosystem structure and function (Fig. 2)^{4, 21, 55, 56, 57, 58, 59}. In our assessment, this type of threshold may also apply in Africa, where the land-mass is similarly large. However, the threshold may differ in Southeast Asia, where the land–ocean balance is very different.

Global circulation models suggest that for low levels of deforestation, regional rainfall may be similar to or slightly higher than (B) rainfall in undisturbed forest (A). As deforestation expands, rainfall declines (C). A tipping point may occur when deforestation reaches 30 - 50% (D), after which rainfall is substantially reduced. A fully deforested state (E) results in the greatest decline in rainfall. Hatching indicates proportion of undisturbed forest.

Regional scale models have smaller grid sizes than GCMs, but they are often not connected to the entire global system. The smallest grid cells (400 km²) are 30–750 times smaller than cells in the models described above. Regional models are able to simulate convection induced by abrupt spatial changes in vegetation and orographic precipitation^{62, 63}, processes typically not captured by GCMs with ~1°–5° grids^{64, 65}. As in GCMs, the representation of clouds is still limited.

Impacts and critical thresholds at small scales. In contrast to the lower rainfall induced by large clearings modelled at coarse scales, small clearings modelled at regional scales can result in enhanced local rainfall over the deforested area^URL: