Deforestation causes different subregional effects on the Amazon bioclimatic equilibrium
Abstract
[1] Deforestation on Amazonia and central Brazil Cerrado could change regional climate, possibly shifting forest equilibrium into a bioclimatic envelope typical of savannas. Although impacts of climate change induced by deforestation are likely to vary subregionally, the potential geographic variation of these effects and the thresholds of rainforest and Cerrado removal that will affect Amazonian bioclimatic equilibrium remain unknown. We evaluate the effects of deforestation scenarios of increasing severity on the bioclimatic equilibrium of Amazon subregions. Results indicate that subregional precipitation responds in three distinct ways to progressive deforestation: a near-constant rate of reduction, a rapid drop for low deforestation levels, and a decrease after intermediate deforestation levels. Additionally, while inner forest regions remain inside rainforest bioclimatic envelope, outer forest regions may cross forest-savanna bioclimatic threshold even at low deforestation levels. We argue that at least 90% of Amazonia and 40% of Cerrado should be sustained to avoid subregional bioclimatic savannization.
Key Points
- The difference in the climate response of distinct Amazon regions
- Cerrado deforestation is decisive to anticipate a bioclimatic transition
- At least 90% of Amazonia and 60% of Cerrado should remain preserved
1 Introduction
[2] The Amazon rainforest has been undergoing significant changes on its physical environment due to anthropogenic activity. Substitution of forest with cattle pasture is historically the main driver of environmental change [Leite et al., 2012], with potentially serious impacts on the region's climatic equilibrium [Lenton et al., 2008; Nobre and Borma, 2009; Malhado et al., 2010]. Global and regional climate models have typically indicated that the entire Amazon may cross a tipping point if deforestation exceeds 40% [Sampaio et al., 2007; Nobre and Borma, 2009; Davidson et al., 2012] or 50% [Costa et al., 2007]. However, these studies have neglected the role that deforestation in the Cerrado—a savanna-like vegetation of central Brazil—might play in regulating regional climate. Specifically, land use change in the Cerrado may drive increased water deficits, especially in the forest-savanna frontier [Costa and Pires, 2010]. Another frequently neglected issue is that the climate of the entire forest region might not respond homogenously to deforestation [Avissar et al., 2002]. Regional variations of this type could drive some Amazonian forests over ecological tipping points before deforestation reaches the lower critical thresholds predicted by models. The ability to predict and pinpoint such subregional bioclimatic responses to deforestation is therefore essential for the development of effective forest conservation policies.
2 Methods
[3] Drawing on Malhi et al.'s [2009] foundational research, it is possible to estimate to what extent direct human pressure, such as deforestation, impacts the equilibrium state of the Amazon rainforest. These authors constructed a bioclimatic diagram that combined regional precipitation data and vegetation type. They were thus able to identify broad climatic thresholds (tipping points) for rainforest viability based on annual mean precipitation (AP) and the most negative value of the Maximum Climatological Water Deficit (MCWD) [Malhi et al., 2009]. The climatic boundaries were converted into the following basic typology: (i) evergreen rainforest: AP > 1500 mm or MCWD < 300 mm and (ii) savanna: MCWD > 300 mm and rainfall ≤ 1500 mm. As noted by the authors, this simplified land use classification does not distinguish evergreen from semideciduous forests. The bioclimatic boundaries can be schematically represented in a diagram, as in Figure 4.
[4] Here we adopt Malhi et al.'s [2009] typology to evaluate the bioclimatic equilibrium of different Amazon subregions after different levels of deforestation. We use a well-tested climate-biosphere model [Senna et al., 2009; Costa and Pires, 2010] (see also supporting information—sections S1, S2, and S3) to generate climate scenarios (40 years of numerical simulations for an ensemble with five members defined the average climate of each scenario) derived from progressive Amazon and Cerrado deforestation scenarios (see also supporting information—section S1). The initial range of Amazon deforestation scenarios (10–30%) is based on Soares-Filho et al.'s [2006] scenarios that were compatible with the 10% increments in our study. The 20% Amazon deforestations scenario is based on their GOV2028 scenario, and similarly the 30% scenario is based on their GOV2050 scenario. The scenarios beyond 30% were developed considering two main conditions for Amazon deforestation: the presence of roads and protected areas. While the presence of roads is known as an enabler of deforestation, it is assumed here that protected areas will limit deforestation of Amazon rainforest to a maximum of 70% (we consider that protected areas represent 30% of the Pan-Amazon region, similar to Walker et al. [2009]). Starting from 80% of Amazon deforestation, the presence of protected areas does not represent a limitation, so the deforestation is allowed everywhere. These scenarios are drastic and have a lower probability of occurrence. The Cerrado deforestation scenarios were developed based on the total deforestation closest to the current value, near 60% [Leite et al., 2012].
[5] Deforestation here is defined as the substitution of natural vegetation (tropical forest or savanna) with pasture in the entire grid cell, which is parameterized according to Costa et al. [2007]. For Amazonian deforestation scenarios, natural vegetation was substituted by pasture in increasing amounts, from 10% (current level of deforestation in Amazonian rainforest is close to 20%, [PRODES, 2011]) to 100% (Figures 1b–1k). For “Amazonia + Cerrado” deforestation scenarios, the previous scenarios were augmented by Cerrado deforestation, from 60% (based on the fact that the current level of deforestation in the Central Brazilian Cerrado is approximately 55%) to 100% (Figures 1l–1u). There is no superposition between the areas of the Amazon and Cerrado.
2.1 Subregional Analysis
[6] Based on spatial patterns of bioclimatic transition, we evaluated not only the precipitation of the entire Amazon area (Figure 2a) but also identified and evaluated eight subregions. Subregions B, C, and D (Figure 2b) are subdivisions of the arc of deforestation in the transition zone between forest and savanna. This region contains the highest rates of deforestation and has the longest dry season in the Amazon. Subregions E, F, G, H, and I (Figure 2c) were identified a posteriori (supporting information—section S4) and show, among all of our simulations, a uniform bioclimatic transition, either from rainforest to savanna (Figure 2c, brown subregions) or toward a bioclimatic envelope typical of seasonal forest (Figure 2c, yellow subregions).
3 Results
[7] For the entire Amazon (region A) and for each subregion (subregions B to I), AP generally decreases from the control simulation to the most pessimistic deforestation scenario (Figure 3). Furthermore, the precipitation of “Amazonia + Cerrado” deforestation scenarios is usually reduced in comparison with the scenario that only considers Amazonia deforestation, mainly due to lower levels of moisture advection from the Cerrado to southern and southeastern Amazonia in the beginning of the rainy season (September to November), when Cerrado removal causes a reduction of moisture convergence to the arc of deforestation by 50%. [Costa and Pires, 2010, Figure 7, p. 1977]. Thus, the precipitation decrease is a consequence of changes in regional moisture budget components, especially evapotranspiration and moisture convergence. In particular, each subregion climate responds differently, and three distinct curves of precipitation versus deforestation can be identified, corroborating the three hypothetical patterns previously suggested by Avissar et al. [2002]: a near-constant rate of reduction (group I), a decrease in precipitation after intermediate levels of deforestation (group II), and a rapid precipitation drop for low levels of deforestation (group III).
- [9]
For region A (entire Amazon) and for subregions F (northern Mato Grosso state and southwestern Amazonas state) and I (central Amazonia), AP decreases at a near-constant rate as deforestation increases (a straight line with negative slope, in average (Figures 3a, 3f, and 3i). Although there is a substantial reduction of rainfall (20–25% reduction from control run to the most pessimistic scenario), regions A and I do not cross a bioclimatic tipping point due to their typically high levels of AP and small MCWD (Figures 4a and 4i). Region F approximates to a region in the bioclimatic diagram where Malhi et al. [2009] identified dominantly seasonal forest pixels, due to both AP reduction and increase in MCWD (Figure 4f). Although it is not possible to definitively identify this area as the location of a tipping point transposition, there is clearly a tendency for a bioclimatic seasonalization transition.
- [10]
For subregions E (southwestern Amazonia, over Acre state, Peru, Ecuador, and Bolivia) and H (northeastern Amazonia, over Venezuela, Guyanas, and Suriname), AP remains constant or even slightly increases up to intermediate levels of deforestation (30–50%). However, there is a rapid drop in AP at higher deforestation levels (Figures 3e and 3h) until it declines to 80–88% of the control run precipitation. Region E (southwestern Amazonia) crosses the forest-savanna tipping point due to the combination of a large reduction of AP and increase in MCWD (Figures 3e and 4e). The tipping point occurs once deforestation exceeds 70% of the total forest area, whether Cerrado deforestation is considered or not. For subregion H (northeastern Amazonia), as in region F, the reduction in AP pushes the forest into a region of Malhi et al.'s [2009] bioclimatic diagram where seasonal forest pixels dominate, even though there is little increase of MCWD (Figures 3h and 4h).
- [11]
For subregions B (northern Bolivia), C (northern Mato Grosso state and southern Pará state), D, and G (eastern Pará state), there is a maximum rate of AP reduction in the lowest levels of deforestation. As the total deforested area increases, AP decreases at smaller rates (Figures 3b, 3c, 3d, and 3g). Since these regions are mainly located at the transitional zone between rainforest and Cerrado, this is strong evidence of how they are dependent on vegetation cover to provide sufficient levels of precipitation. For intermediate and pessimistic scenarios, there is no additional reduction in AP, although in region D, precipitation increases for high levels of deforestation. Given that the air near the surface is warmer over deforested areas than on adjacent areas, colder and wetter air from the surrounding forest (or ocean, specifically for region D) may converge and rise over the deforested area forming clouds [Roy and Avissar, 2002; da Silva et al., 2008] potentially leading to increased rainfall if enough moisture is present. Region D approaches a bioclimatic transition (with a reduction in AP of nearly 12%), although it does not cross the AP ≤ 1500 mm threshold. Regions B and G cross the forest-savanna tipping point due to the combination of reduced AP and increased MCWD (Figures 4b and 4g). Even though there is modest proportional reduction in AP (8% from control run to the most pessimistic deforestation scenario for region B; 20% from control run to the most pessimistic deforestation scenario for region G), the reduction is sufficient to affect the bioclimatic equilibrium of both regions. The transition for region B occurs when deforestation exceeds 40% of entire forest area if Cerrado deforestation is not considered and 10% if we consider the Cerrado deforestation effect. Region G's tipping point occurs once deforestation exceeds 20% of the forest total area, either with or without Cerrado deforestation. One should note that, even though CCM3-IBIS underestimates precipitation over region G, the bioclimatic transition is not influenced by the simulated climatology but by the quick precipitation drop after the initial deforestation levels.
[12] Cerrado removal causes a decrease in moisture advection for the arc of deforestation, and, consequently, there is additional reduction in rainfall in these regions—in particular, in those periods when the moisture supply from the surface is important, i.e., the beginning and end of the rainy season [Costa and Pires, 2010]. This process extends the dry season and increases MCWD, strongly indicating that Cerrado removal may also influence Amazon bioclimatic equilibrium [Malhado et al., 2010].
4 Conclusions
[13] The results of our evaluation strongly indicate that subregions inside the Amazon forest respond distinctly to progressive deforestation and vary considerably in their inherent resilience. In our simulations, inner regions of the forest do not cross a tipping point induced by deforestation, while outer regions either have a tendency to bioclimatic seasonalization or cross the bioclimatic savannization tipping point (the term “bioclimatic savannization” refers to a modification of the regional climatic envelope that hinders, in the long term, the existence of other biomes rather than savanna). Cerrado deforestation causes increased water deficit in the arc of deforestation, and this is decisive in causing a transition. The regions most at risk of savannization are the Bolivian Amazon and eastern Pará state. At these regions, a drier climate typical of savannas (longer dry season) could favor forest fragmentation and degradation and increase the frequency of forest fires in a positive feedback [Cochrane et al., 1999]. In an abrupt timescale when compared to previous natural ecosystem changes, complete ecological savannization would take place at a very short time scale, and a new vegetation-climate equilibrium [Lenton et al., 2008; Nobre and Borma, 2009] could be established in Bolivian Amazon and eastern Pará state.
[14] In order to protect the whole forest bioclimatic equilibrium and to avoid subregional savannization, at least 90% of Amazonia and 40% of Cerrado should be sustained. This information is critical for the ongoing debate about the role of Amazonia and Cerrado protected areas in maintaining regional and Amazonian climatic stability. Walker et al. [2009] concluded that Brazilian federal and state governments have created a sustainable core of protected areas in Amazonia (around 40% of Brazilian Amazon is now under some type of protection) that effectively buffers against potential climatic tipping points and protects the drier ecosystems of the basin. They argued that even with high levels of deforestation, desiccation-driven savannization would not be an immediate threat to Amazonia. However, two important factors were not considered in their analysis: First, they analyzed only Brazilian legal Amazonia (thereby excluding the effects on Bolivia, Peru, and Ecuador). Second, they did not consider the potential role of Cerrado in modifying the Amazon's climate, causing an increased water deficit in the arc of deforestation and altering bioclimatic equilibria in different subregions.
[15] Current Amazon and Cerrado deforestation levels—approximately 20% [PRODES, 2011] and 55% [Leite et al., 2012], respectively—are near the threshold for causing bioclimatic tipping points in the Bolivian Amazonia and eastern Pará. At least in these two regions, two measures are urgently needed: documentation of historical trends in climate and ecosystem structure and the rapid development of regional-scale conservation strategies to mitigate the consequences of deforestation in Amazonia and central Brazil.
[16] The Amazon is already experiencing some of the changes predicted by previous climate models [Butt et al., 2011; Spracklen et al., 2012]. Climate modeling provides strategic conservation information and should not be further ignored by decision makers and conservationists. It anticipates possible damages, enables precautionary actions, and provides a rational basis for the optimal allocation of financial, physical, and human resources.
Acknowledgments
[17] This research has been supported by CNPq (Brazil) and the Gordon and Betty Moore Foundation (USA). Thanks to Ana Cláudia Malhado, Richard J. Ladle, and Santiago V. Cuadra for comments on this research.
[18] The Editor thanks two anonymous reviewers for their assistance in evaluating this paper.
