Research ArticleAGRICULTURE

# Bigger is better: Improved nature conservation and economic returns from landscape-level mitigation

See allHide authors and affiliations

Vol. 2, no. 7, e1501021

## Abstract

### Expanding the scope and assessments of mitigation policy impacts

Mitigation policies should consider impacts on all stakeholders in an affected region (10) to balance the costs and benefits of development and prevent the displacement or “leakage” of agricultural activities to other locations (42, 43). Although our analysis quantifies the aggregate net benefits of planning at landscape scales, it does not account for human welfare, distributional issues, or the displacement of land conversion (see the Supplementary Materials for further details). Previous studies provide evidence that conservation may also generate benefits in terms of improved water availability (44), reduced disease burden (45, 46), and reduced poverty (47). We provide the first step in highlighting how to improve existing land use policies by showing that planning at larger scales may provide long-term environmental benefits in a cost-effective way. An important future extension of our work is the empirical validation of the predictions after FC compliance is achieved on the ground as well as the modeling and evaluation of the policy in terms of welfare impacts (47). The modeling work presented here can inform the design and data collection of such studies (48).

## CONCLUSION

Many countries, such as Brazil, are at a tipping point, struggling to balance accelerating development pressures with dwindling natural resources. Improving the effectiveness of mitigation to balance economic development with nature conservation is now pivotal (9, 10, 49), given accelerating large-scale development from sectors such as agriculture, energy, mining, and transportation that affect vast lands across the globe (50, 51). Advancing mitigation by adopting multiobjective landscape planning can promote cost-effective conservation and more sustainable development trajectories. It can also complement the growing global and national commitments to the large-scale restoration of degraded lands by directing and consolidating mitigation efforts to restore priority areas for biodiversity and ecosystem services (52, 53). Although recent policies and voluntary commitments by governments, businesses, and financial institutions are positioned to enable and incentivize private developers to adopt the principles of landscape-scale mitigation in their planning and practice (7, 8, 12), this approach is not yet widely implemented (10). Our analysis underscores that LL planning can improve the long-term performance of land use policies for conservation in a cost-effective way. Because the magnitude of benefits may depend on the specific context (for example, scale and type of development, the biophysical linkages between land use change, biodiversity, and ecosystem services), it will be critical to conduct rigorous empirical impact evaluations from multiple contexts to strengthen the evidence base for business, conservation, and people. Empirical assessments, such as those proposed for the landscape-scale infrastructure initiatives under new U.S. mitigation strategies (12, 16) and for restoration efforts on large versus small landholdings under the Atlantic Forest Restoration Pact (38, 54), are essential to quantify on-the-ground outcomes. Information on time and cost savings as well as on the benefits to nature and people from LL mitigation will help to overcome political, institutional, and logistical barriers to seeing its widespread adoption.

## MATERIALS AND METHODS

### Study region, FC requirements, and mitigation scenarios

Our study area encompasses the Ribeirão São Jerônimo watershed, an approximately 400,000-ha area in Minas Gerais State, southeastern Brazil. This region is currently largely composed of pasture that is being converted to sugarcane fields (19, 43). Less than 20% of the natural habitat remains and consists of four dominant vegetation types (4% cerrado, 7% cerradão, 3% semideciduous forests, and 4% wetlands) (Fig. 1, fig. S2, and table S1). All remnant vegetation resides on private land holdings and is regulated by the Brazilian FC (13, 20).

We used a 25% natural vegetation FC requirement at the PL and at the LL. Because FC compliance is generally low for small holders in our study area (13, 20), we assumed that only the commercial sugarcane producer would comply with the FC. All land that was not rented for sugarcane production or FC compliance remained unchanged. Thus, in all scenarios, we assumed that 25% of the farm area rented for sugarcane production must be placed under natural vegetation. This percentage is consistent with the natural vegetation requirements in the region: both in Legal Reserves, which target ~20% natural area set-asides anywhere on farms to protect biodiversity, and in Permanent Protected Areas, which target ~5% of vegetation to be placed along stream banks and steep slopes to protect water quality. Given the resolution of our data (90-m pixels), our models did not distinguish between Legal Reserves and Permanent Protected Areas and instead combined the required natural areas, which could be allocated anywhere within the watershed. See the “FC requirements and mitigation compliance options” section in the Supplementary Materials for further details.

The natural area requirement were met via the protection or restoration of the cerrado habitat types historically found in the region (see the “Physiogeographic characteristics” section in the Supplementary Materials). For PL planning, the choice between restoration and protection depended on the proportion of natural habitat currently on the farm. For LL planning, we considered three cases for compliance: (i) the protection of existing natural remnants and no restoration (LL-P), (ii) the restoration of nonnatural vegetation (for example, pastures) to natural habitat types (LL-R) and no protection, and (iii) the protection and restoration of natural habitats (LL-PR). By assumption, the protection or restoration of habitat holds in perpetuity, as required by federal law. Further, because of the lack of studies on cerrado habitat types that would allow us to differentiate between different types of restoration (55), we assumed one-time investments in active restoration, instantaneous vegetation growth, and perfect and uniform restoration success for all natural habitat types (see the Supplementary Materials for further details). Thus, our results pertain to a steady-state (long-run) equilibrium; we did not model the transitional dynamics.

To calculate the additional impacts of the FC, we used a baseline scenario that modeled agricultural production in the absence of the law (that is, no habitat is protected or restored). In all of our scenarios, planning minimized the costs of environmental compliance and sugarcane production to a large commercial producer while meeting a production target.

We modeled two annual production targets—2.5 million tons (low target) and 8.5 million tons (high target)—to reflect the average current and projected sugarcane processing capacities, respectively, of sugarcane mills in Brazil (table S2). Because the results for the two targets exhibited consistent directional trends (as described in the Supplementary Materials), we present the results only for the larger production target in the main text. We quantified the long-term biodiversity, water quality, and carbon benefits from the landscapes that met the sugarcane production and FC compliance targets at the lowest cost.

### Sugarcane profit modeling

The decision of where to grow sugarcane was based on a static and deterministic model that balances the revenue from growing sugarcane with the costs of production and FC compliance. We modeled the production decisions of a large agricultural producer based on a spatially explicit sugarcane production model that incorporates the revenue from sugarcane (the product of the price of sugarcane and predicted yield) and the costs associated with production (soil preparation, sowing, harvesting, fertilization, transportation, leasing, clearing, management, and transaction costs) and FC compliance (transaction, restoration, and leasing costs) (table S3). The exceptionally detailed cost data were obtained from a local commercial sugarcane producer in our study region. In our economic models, we did not consider the potential benefits to sugarcane production from natural vegetation (for example, pest control, soil fertility and stability, and water availability for irrigation) (56) because these have been traditionally very difficult to quantify and are not typically considered in status quo business decision-making. The differences in production and compliance costs between the different planning scenarios are presented in terms of NPV (in million U.S. dollars) and were calculated for a standard sugarcane production cycle of 6 years, with a discount rate of 10.32%. See the “Sugarcane production model” section in the Supplementary Materials for further details.

### Land-use optimization

The goal of the optimization procedure was to generate landscapes that maximized the net returns from sugarcane production subject to (i) meeting the sugarcane production target and (ii) meeting the requirements of the FC. Under the assumption of exogenous yield, the decision variables were at the extensive margin (that is, which pixels to select for sugarcane production and FC compliance and, having allocated those, which farms to lease). We did not model decisions at the intensive margin (that is, how much sugarcane to produce on a given pixel by varying the production inputs). We used an integer programming branch-and-bound algorithm to select the pixels and farms in a static framework (57). We did not model the potential displacement of cattle ranching due to sugarcane expansion in the study area. See the “Landscape optimization” section in the Supplementary Materials for more details.

The optimization procedure generated partial landscapes that indicated which farms should be leased and, within those farms, which pixels should be allocated to sugarcane production and FC compliance or which should remain under different uses. To produce a final landscape, we used local raster tools in ArcGIS and assigned the land cover/land use currently found in the region if the pixel was not selected for sugarcane production or natural habitat for FC compliance. Where restoration of natural habitat was predicted by the optimization, we used the predicted vegetation layer (fig. S2) to assign a natural habitat type.

### Biodiversity, water quality, and carbon sequestration models

We assessed the profit-maximizing landscapes under the different planning scenarios in terms of their potential to support biodiversity, water quality, and carbon storage. To quantify the expected number of mammal and bird species, we applied the model from Polasky et al. (58) that predicts the probabilities of species persistence based on the habitat area required for a breeding pair, the relative suitability of all land cover types, and the ability of species to disperse among patches in the landscape. Focusing on 407 terrestrial bird and 132 mammal species, we assessed how species richness, composition, and habitat specialization shift across the planning scenarios. Owing to the lack of relevant studies from our study region, we did not consider different habitat successional stages and attributed a single value per habitat type for all parameters. See the “Biodiversity modeling” section in the Supplementary Materials for further details.

To quantify the water quality benefits, we used the terrestrial nutrient and sediment models from InVEST 2.5.6 (59) and calculated the total annual predicted loadings of nitrogen (N), phosphorus (P), and sediment (S) reaching the waterways in the study area. The N, P, and S loadings were then converted into predicted concentrations and combined into a WQI (60). The WQI scale ranges from 0 to 100, with an index change of at least 10 points required for water quality status to shift between categories (very good, good, fair, poor, and very poor), depending on the starting value. See the “Water quality surface models” section in the Supplementary Materials for further details.

To quantify the long-term carbon sequestration benefits, we used values from published studies for the amount of carbon stored in aboveground and belowground biomass and soil in steady-state systems in the Cerrado biome. For each scenario, we determined the additional mean carbon storage potential per planning scenario and monetized the value of the ecosystem services using prices from the voluntary carbon market as a lower bound and recent estimates of the social value of avoiding damages from carbon emissions as an upper bound. See the “Carbon valuation” section in the Supplementary Materials for further details.

Whenever possible, all biophysical models used spatially explicit data and parameters from previous studies conducted in the biome. We found that the predictions from our biophysical models were consistent with those from published studies on the Cerrado or deemed reasonable by expert review when studies were not available (see the Supplementary Materials for further details).

## SUPPLEMENTARY MATERIALS

Supplementary Materials and Methods

Supplementary Results

fig. S1. Study area in relation to major ecological biomes of Brazil.

fig. S2. Distributions of natural vegetation types predicted for our study area.

fig. S3. Distributions of soil types for our study area.

fig. S4. Optimized landscapes corresponding to each mitigation scenario.

fig. S5. Cost savings for LL mitigation relative to PL mitigation.

fig. S6. Sources of cost savings for LL mitigation relative to PL mitigation.

fig. S7. Area of natural habitat across the mitigation scenarios.

fig. S8. Types of natural habitat restored or protected under the PL and LL scenarios.

fig. S9. Changes in habitat fragmentation for LL mitigation relative to PL mitigation.

fig. S10. Patterns of habitat fragmentation for the different mitigation scenarios.

fig. S11. Changes in fragmentation by habitat type for LL mitigation relative to PL mitigation.

fig. S12. Changes in the expected number of bird and mammal species for LL mitigation relative to PL mitigation.

fig. S13. Changes in the expected number of species by habitat specialization for LL mitigation relative to PL mitigation.

fig. S14. Changes in the predicted carbon storage for LL mitigation relative to PL mitigation.

fig. S15. Changes in WQI for LL mitigation relative to PL mitigation.

fig. S16. Changes in the average predicted nitrogen and phosphorus concentrations and total loadings for LL mitigation relative to PL mitigation.

fig. S17. Changes in average predicted turbidity and total sediment loading for LL mitigation relative to PL mitigation.

table S1. Land cover types and definitions for the study area.

table S2. Final yield for each scenario.

table S3. Summary of the parameters used in the agricultural profit optimization models.

table S4. Definitions of the parameters used in the agricultural profit optimization equations.

table S5. Amount of habitat restored or protected under each mitigation scenario.

table S6. Fragmentation metrics for patches of all natural habitat types grouped together.

table S7. Fragmentation metrics for patches by habitat type.

table S8. Data sources used to determine relevant species by taxonomic group.

table S9. Average (±SD) habitat suitability values for land cover types in our study region.

table S10. Average (±SD) parameters by trophic level used in the biodiversity model.

table S11. Expected number of species based on the biodiversity model across mitigation scenarios.

table S12. Expected number of species by habitat specialization for each mitigation scenario.

table S13. Aggregated values for carbon storage per land cover/land use category for our study area.

table S14. Additional carbon storage provided by each mitigation scenario.

table S15. Minimum and maximum values for nitrogen (TN), phosphorus (TP), and turbidity concentrations in pristine areas in the Cerrado biome.

table S16. WQI across mitigation scenarios.

References (61126)

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.

## REFERENCES AND NOTES

Acknowledgments: We are grateful to E. Okumura, E. Garcia, C. Pereira, A. Poloni, K. Souza, and J. Pereira for their input on the business context and sugarcane profit modeling; A. Davidson, C. Sekercioglu, E. M. Vieira, and P. Develey for providing data and input on the biodiversity parameters and results; P. Hamel, L. Azevedo, J. Guimarães, K. Voss, I. Alameddine, S. Thompson, B. Keeler, and R. Griffin for their guidance on the hydrological modeling; B. Murray, T. Kroeger, B. Griscom, M. Borgo, G. Tiepolo, and N. Virgilio for helpful discussions on carbon valuation; M. Matsumoto for help with calculations of FC requirements; S. Baruch-Mordo for help with R coding issues; J. Wilkinson and A. B. Villarroya for input on mitigation policies; and J. Wilkinson, P. Kareiva, and M. Weick for helpful comments on earlier drafts. Funding: This research was supported by The Dow Chemical Company Foundation, The Dow Chemical Company, The Nature Conservancy, Anne Ray Charitable Trust, and the 3M Foundation. Author contributions: C.M.K., D.A.M., and J.K. conceived and designed the study; D.A.M., C.M.K., L.B., and K.S. collected the data; D.A.M., C.M.K., P.L.H., L.B., K.S., J.R.O., and S.P. conducted the modeling and analysis; L.B. assessed Brazilian FC requirements; E.M.U. informed the business context; D.A.M., C.M.K., and K.S. produced the figures and tables; C.M.K. and D.A.M. designed and wrote the paper; and D.A.M., C.M.K., J.K., S.P., L.B., K.S., P.L.H., E.M.U., and J.R.O. revised the paper. Competing interests: The authors declare that they have no competing interests. Data availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper may be requested from the authors. Data for this research are available from The Nature Conservancy and are found at http://nature.org/TNC-Dow-Brazil.
View Abstract