Using ecological thresholds to evaluate the costs and benefits of set-asides in a biodiversity hotspot

See allHide authors and affiliations

Science  29 Aug 2014:
Vol. 345, Issue 6200, pp. 1041-1045
DOI: 10.1126/science.1255768

Cost-effective conservation on private land

How affordable is biodiversity conservation in a fragmented landscape? Banks-Leite et al. asked this question for the biodiversity hotspot of the Brazilian Atlantic Forest. An annual investment of <10% of Brazil's agricultural subsidies could support effective ecological restoration on private lands. This would increase biodiversity in set-aside land to the same level observed in protected areas. The cost-effectiveness of this scheme suggests a path forward for conservation strategies in other similarly mixed landscapes, too.

Science, this issue p. 1041


Ecological set-asides are a promising strategy for conserving biodiversity in human-modified landscapes; however, landowner participation is often precluded by financial constraints. We assessed the ecological benefits and economic costs of paying landowners to set aside private land for restoration. Benefits were calculated from data on nearly 25,000 captures of Brazilian Atlantic Forest vertebrates, and economic costs were estimated for several restoration scenarios and values of payment for ecosystem services. We show that an annual investment equivalent to 6.5% of what Brazil spends on agricultural subsidies would revert species composition and ecological functions across farmlands to levels found inside protected areas, thereby benefiting local people. Hence, efforts to secure the future of this and other biodiversity hotspots may be cost-effective.

The combined effects of environmental change are driving species to the brink of extinction across the world’s biodiversity hotspots (1). If species disappear, the ecological functions they perform will also decline (2), with potential consequences including increased pest outbreaks and reduced food security (3, 4). Although the role of large protected areas in preserving species is unquestionable (5), people will benefit more widely from the ecological functions they perform if species occur throughout the biome, not just inside protected areas (6). It is in this context that ecological set-asides on private land have emerged as a promising strategy to preserve species and ecological functions across farmlands (2, 7).

Setting aside private land for conservation nonetheless comes with financial costs to the landowner. For instance, in the Atlantic Forest of Brazil, the median yearly gross profit per hectare of agricultural land is $467 (interquartile range $199 to $868; U.S. dollars) (8), which is more than the Brazilian minimum wage. Conservation schemes that involve payment for ecosystem services (PES) provide a mechanism to increase landowner participation on set-asides (9), and its feasibility is reflected in the ever-increasing number of PES projects across the globe (7, 10). However, most PES projects involving set-asides are relatively local initiatives that do not match the extensive scale of conservation needs and societal issues observed across biodiversity hotspots (10). Because large-scale problems require large-scale solutions, we here advocate and provide a realistic plan for a biome-wide set-aside program that concurrently maximizes the provision of habitat to biodiversity, and of ecological functions across rural areas, while minimizing the costs to society. Our aim is to calculate the overall economic costs of a set-aside program tailored to one of the most threatened hotspots of the world (11)—the Atlantic Forest of Brazil.

First, we estimated the minimum amount of habitat required to maintain biodiversity. This information is crucial (12); if too much area is set aside for conservation, economic costs become an impediment, but if not enough area is set aside, ecological gains are minor. To investigate the relationship between forest cover and species composition, we used data from a large field study conducted in the Atlantic Forest (13, 14). The data set consisted of nearly 25,000 captures of 43 species of mammals, 140 species of birds, and 29 species of amphibians collected in 79 landscapes (200 ha each) varying from 5 to 100% of forest cover (including both old-growth and regrowth) and distributed across a 150-km-wide region (fig. S1) (8). For all groups, reduction in forest cover had a similar effect on community integrity, defined here as the similarity in community composition between fragmented and continuously forested landscapes (Fig. 1, I to L). Results indicate that community integrity of all three groups is maintained until 24 to 33% of forest cover, below which integrity declines sharply with further reductions in forest cover. When all three taxa were combined, analyses indicated the existence of a threshold at 28.5% of forest cover [95% confidence interval (CI) 24.0 to 33.1, N = 56], hence approximately 30% of native habitat is needed to preserve the integrity of vertebrate communities within each landscape (Fig. 1). Similar results have been observed previously in the northern Atlantic Forest, indicating that the patterns we discern are not exclusive to our sites, to vertebrates, or to a particular landscape scale (8, 15).

Fig. 1 Community and phylogenetic responses to forest cover in the Atlantic Forest.

(A to D) Species × site matrix of mammals, birds, amphibians, and vertebrates. Sites (columns) are ordered by forest cover, and species (lines) are ordered by their preference to forest cover. Red indicates forest specialists; blue indicates disturbance-adapted species. Columns to the right of each panel represent endemism to the Atlantic Forest, with black cells showing endemic species and gray cells showing nonendemics. (E to H) Proportional change of forest specialists and disturbance-adapted species in each community. (I to L) Variation in community integrity. (M to P) Variation in phylogenetic integrity. Shaded areas in (E) to (P) indicate 95% CIs of threshold estimates obtained from community integrity. In (I) to (P), points are colored to represent the proportion of forest specialists and disturbance-adapted species. Area represented by two dashed lines in (M) to (P) indicates the CI of threshold obtained from phylogenetic integrity.

Theoretical predictions have suggested that species become more sensitive to habitat alterations below 30% of remaining habitat (16). Our results, however, show that community responses are not solely restricted to species loss, as we observed a replacement of forest specialists by species adapted to disturbance (Fig. 1, A to D) (8, 17). Forest specialists make up 75.9% of all species found in areas with 100% forest cover, whereas disturbance-adapted species represent 73% of species in landscapes with ≤10% forest cover (Fig. 1, A to D). Note that the threshold observed for community integrity is caused by a change in dominance; communities above the 30% cover threshold tend to be dominated (i.e., ≥50%) by forest specialists, and communities below the threshold become dominated by disturbance-adapted species (Fig. 1, E to H).

Although the causes and consequences of local species extinction are unknown, we found a strong correlation between species sensitivity and endemism (Pearson’s r = 0.87, N = 79; fig. S2) (8), with 70.1% of forest specialists found to be endemic to the Atlantic Forest, whereas 68.5% of disturbance-adapted species occur in other biomes (Fig. 1, A to H). Previous studies have shown that forest specialists present narrow niche breadths and high efficiency in resource exploitation, so the loss of forest specialists potentially affects trophic cascades and ecological functions, such as seed dispersal and pest control (1820). Detailed knowledge about the functional role of many Atlantic Forest vertebrates is not available (8), but if we assume niche conservatism (21), it is possible to infer that changes in lineage composition should broadly correspond to changes across functionally important traits (22). Hence, we estimated the phylogenetic integrity of communities, measured as the phylogenetic similarity or fraction of branch length shared between communities present in fragmented and continuously forested landscapes (8, 23). We observed that phylogenetic integrity also decreases below the 30% cover threshold (Fig. 1, M to P), which suggests that the suite of ecological functions performed by these taxa are different above and below the threshold. Our findings highlight the importance of maintaining or restoring forest cover in the Atlantic Forest above the 30% threshold, as extinction of endemic species will lead to the loss of ecological functions that only these species provide.

Upscaling our results to 143 Mha of Atlantic Forest shows that although community integrity is high inside and around protected areas, it is much reduced across the 88.2% of the biome where less than 30% forest cover remains (using 200-ha landscape units; Fig. 2) (8). These results reinforce the need to preserve large existing remnants, as they provide a refuge to many species (5) (Figs. 1 and 2), and show that the Atlantic Forest is in urgent need of restoration to ensure that biodiversity-derived ecological functions are provided across the biome.

Fig. 2 Upscaling community integrity results for the Atlantic Forest of Brazil.

From left to right, map of Atlantic Forest’s historic distribution (gray) and current forest remnants (black), map of predicted community integrity of vertebrates across the biome, and map of priority areas for restoration (as well as current network of protected areas in purple). Large values (red) of community integrity indicate where communities have similar species composition relative to protected areas.

An area of 32.11 Mha would have to be restored to increase every 200-ha landscape of Atlantic Forest to a minimum of 30% forest cover. This is arguably an unfeasible goal, given the livelihoods and needs of 130 million Brazilians living in this biome. However, if we select landscapes with forest cover approaching the threshold, restoration in rural areas becomes better value for money (8). This is because highly deforested areas are typically depauperate of forest specialists (13, 15, 24), and the chances of recolonization after restoration would depend on money- and time-consuming interventions such as translocation. Thus, on the basis of our results and previous prioritization frameworks (8, 25), we selected all landscapes with more than 20% forest cover and calculated how much area would be required to increase these target landscapes to 30% forest cover. Following this framework, we estimate that restoring just 424,000 ha across the biome will push forest cover to 30% in 37,000 landscapes, which together encompass a total area of 7.8 Mha (table S1). Restoring to 30% forest cover will provide benefits to biodiversity but is unlikely to provide a safety net, as it will bring communities just up to the threshold. Preferably, landscapes would be restored to a level of forest cover above the threshold, but this brings additional issues because the amount of area needed for restoration increases at an average rate of 114,000 ha (SE ±2300 ha) for every 1% increase in targeted forest cover (Fig. 3 and table S1) (8).

Fig. 3 Trade-offs in economic costs and ecological gains for a biome-wide set-aside program.

Gains in phylogenetic integrity (measured as changes in PhyloSor index) are dependent on current percentage of forest cover and on amount of restoration targeted. Dark red indicates gains of up to 15% in phylogenetic integrity; white indicates decrease or no increase in integrity. Left y axis indicates target forest cover (in bold), costs relative to Brazil’s GDP for the first 3 years (including PES and active restoration costs in priority landscapes), and costs for maintaining set-asides after the first 3 years (in italics). Black line depicts a histogram showing the number of existing priority landscapes in each class of forest cover.

The Brazilian government and local NGOs have been steadily implementing PES schemes for set-asides on private land over the years (10). Taking the average PES values across the biome, $132.73 ha−1 per year (10), it would cost $56.3 million per year (range $25.7 million to $91.6 million; table S1) to target enough set-asides to reach the restoration target of 30% cover in the priority landscapes, and that cost would increase by $15.1 million (SE ±$300,000) per year for every 1% increase in targeted forest cover (table S1) (8). Although most areas would follow natural regeneration simply by ceasing the drivers of disturbance (26), it has been estimated that 20% of the area considered for restoration would require active reforestation practices, with associated costs of up to $5000/ha during the first 3 years (26). Thus, the overall budget to reach 30% forest cover in priority landscapes is $198 million per year, or 6.5% of Brazil’s annual expenditure on agricultural subsidies for the first 3 years. This investment represents 0.0092% of Brazilian annual GDP, and the cost would reduce further to 0.0026% of Brazilian annual GDP once regeneration is under way (Fig. 3).

Creating set-asides to restore priority landscapes back to 30% forest cover may not save the most threatened species from extinction, but it would increase biodiversity and the ecological functions that species provide across rural areas (e.g., pest control and pollination) to a level similar to what is observed in protected areas (Figs. 1 and 3). Set-asides in priority landscapes would incur a loss of only 0.61% of the agricultural GDP produced in these municipalities (8), and payments would be targeted directly toward the very same people that would spare their land for conservation (9). Only rarely are the trade-offs between ecological gains and economic costs this simple; the costs relative to Brazil’s GDP are low, those losing productive land are the ones that will directly benefit from PES, ecosystem services (e.g., air and soil quality, carbon storage) would be indirectly enhanced, and the ecosystem itself is one of the most charismatic in the world. At present, the Brazilian Atlantic Forest and other biodiversity hotspots are in serious danger of suffering large losses of endemic species. Our results show that this need not be the case.

Supplementary Materials

Materials and Methods

Figs. S1 to S3

Tables S1 and S2

References (2740)


References and Notes

  1. See supplementary materials on Science Online.
  2. Acknowledgments: We thank R. Didham, T. Gardner, R. Gill, and R. Ewers for comments on the manuscript. Supported by NERC grant NE/H016228/1 (C.B.L.), CNPq research fellowship 306715/2011-2 (R.P.), FAPESP grant 05/56555-4, CNPq/BMBF grant 690144/01-6, and a Marie Curie International Incoming Fellowship within the 7th European Community Framework Programme. This article is a contribution to Imperial College’s Grand Challenges in Ecosystems and the Environment initiative. The data reported in this paper can be obtained by contacting the corresponding author, and the R code used in analyses is provided in the supplementary materials. C.B.-L., R.P., L.R.T., A.A.B., R.T.B., T.H.C., M.D., A.T.I., and A.C.M. carried out the data collection; C.B.-L., L.R.T., and W.D.P. analyzed the data; C.B.-L., R.P., L.R.T., W.D.P., and J.P.M. wrote the paper.

Stay Connected to Science

Navigate This Article