Executive Summary
Economic Instruments for Habitat Conservation
TAMARIN windowFinal Report to The World Bank, Washington, DC

David M. Stoms and Frank W. Davis

Institute for Computation Earth System Science


Donald Bren School of Environmental Science & Management


Richard L. Church and Ross A. Gerrard

Department of Geography

University of California, Santa Barbara


Report Date: June 14, 2002

All findings, interpretations, and conclusions are the authors’, and do not necessarily reflect the views of the World Bank, its Executive Board of Directors, or the countries they represent.

Table of Contents

Application of TAMARIN

Optimization Model




Link to full report (1.2 Mb Acrobat format)
Link to TAMARIN users manual (1.1 Mb Acrobat format)



Chomitz et al. (1999) recommended a framework to guide the application of incentive-based policies for encouraging the provision of environmental services – including biodiversity conservation – by landholders.  This framework, they propose, would enable the simulation of alternative schemes that would “encourage clarity in the definition of goals and permit the development of simple, implementable strategies to reach those goals” (Chomitz et al. 1999, p. 168).  Because land use, opportunity costs, and biodiversity value vary spatially, the framework was envisioned as a means to exploit that variation to craft simple, implementable policy options.

In this report we describe an implementation of an analytical framework based upon the suggestions in Chomitz et al. (1999).  Specifically, we outline the framework and its basis in current conservation science and economic theory, which are used to define the desired landscape configuration for the conservation objectives.  We then present a case study in the “Mata Atlântica,” or Atlantic Forest, region of south Bahia state in Brazil to illustrate an application of the framework, named TAMARIN (Toolbox for Analysis of Mata Atlântica Restoration Incentives).  We demonstrate TAMARIN by comparing a series of scenarios, including the present situation and the likely future trend.  The results indicate the tradeoff between the suitability of sites for conservation with their value for other purposes. 

TAMARIN flowchartTAMARIN performs two sets of GIS-based procedures, using a representation of the study area divided into 98.01 ha planning units (990 x 990 m).  First, it assists planning teams to design scenarios and second to evaluate their economic and ecological consequences.  Scenarios can be created by drawing on an electronic map, by defining rules for selection based on conservation and/or economic criteria, or by an external optimization model developed for the project.  Scenarios can be constrained by a maximum budget limit or can be unconstrained with the total costs being calculated as a consequence of the plan.  The framework can then calculate the effects of the scenario and create a series of GIS themes, tables, graphics, and reports that summarize the salient features for comparison with the present situation and other scenarios.

In addition to TAMARINthe project developed an external optimizing land allocation model (Optimal Habitat Patch Selection or OHPAS) that selects the most cost efficient set of areas for conservation action that satisfies the desired future landscape configuration, if feasible within the budget constraint.  The optimal solution is then evaluated in terms of the same socioeconomic and environmental factors as other scenarios inside TAMARIN.  The optimal scenarios therefore set benchmarks against which scenarios for various policy instruments can be compared.

Top of the Document

Central Atlantic Forest Corridor

The Atlantic Forest, Mata Atlântica, is by far the most threatened major ecosystem in Brazil, with less than 8% of its original area remaining.  Conservation International places it third on its list of the 19 highest-priority habitats for conservation on the planet (based on the combination of threat and uniqueness).  Within the Mata Atlântica, the area in southern Bahia is considered one of the highest priorities.  In response to this global importance, the Programa Estadual para a Conservação da Biodiversidade (PROBIO) has funded a major biodiversity assessment and planning project for south Bahia.  The project area is named the Central Atlantic Forest Corridor (hereafter referred to as the “Corridor”) (see location map below).  The dimensions of the Corridor are approximately 580 km (north-south) by 150-250 km (east-west). 

Central Atlantic Forest Corridor location map
Location map of Central Atlantic Forest Corridor.

Central Atlantic Forest Corridor bioregions

Bioregions in Central Atlantic Forest Corridor.

Within a tropical area this large, there is naturally a significant variability of landscapes and land uses.  Nine distinct bioregions are recognized in the Corridor.  Along the immediate coast lies a narrow band of coastal, wetland, and riverine ecosystems.  From east to west, the wetter bioregions support a tropical wet forest that grades into moist or semi-deciduous forest in the interior.  Field surveys conducted by the PROBIO project have revealed a remarkable species replacement in both flora and fauna across the north-south gradient of the corridor.  In particular, the Rio de Contas and Rio Jequitinhonhia form biogeographic barriers that foster speciation.  Consequently, the vegetation types and rivers divide the Corridor into nine bioregions.

Besides the biological differences between bioregions, there are related differences in land use and land tenure.  The northern and central lowland forest bioregions are dominated by cabruca, a form of shade plantation for cocoa.  Cabruca provides better habitat for forest canopy species than most other forms of agriculture but is not as suitable for understory birds as primary forest.  Unfortunately, global market forces, government policies, and invasion of a devastating crop disease, are inducing farmers to convert cabruca to pasture.  Farm size tends to be smallest and population density is greatest in this part of the Corridor.  The southern tabuleiro forest bioregion retains some of the largest forest fragments in the corridor region.  Farm size tends to be larger, and as a result, population density tends to be lower than in the northern bioregions.  The interior semi-deciduous forest bioregions have been largely converted from forest to pasture, with only clusters of very small forest fragments remaining.  Of the 74, 219 km² in the Corridor, only 8.8% remains as primary forest.

Top of the Document


TAMARIN is based on conservation principles of representation, resilience, and redundancy.  These are translated into the following specific criteria used in designing TAMARIN: 1) entire environmental/species gradients should be represented, 2) at the biogeographical scale, representation should only be counted if a forest fragment is larger than the area needed to maintain viable populations of focal species, 3) edge habitat is less suitable for interior forest species and should not count toward the representation goals, 4) several such forest fragments are necessary in each region as backups in case of catastrophic loss of any single fragment, and 5) restoration is most likely to be successful in close proximity to primary forest fragments. 

From theoretical and empirical economics studies, we incorporated the following economic principles into TAMARIN: 1) opportunity costs vary spatially in response to relatively predictable biophysical and socioeconomic factors; 2) recognition of the variability may lead to more cost efficient conservation strategies; 3) economic incentives will more likely elicit the desired behavior from private landholders than command-and-control strategies; and 4) policy instruments derived from these principles must be based on a relatively simple set of rules that address the conservation objectives, yet are understandable and equitable to all stakeholders.  Part of the challenge is that policies are directed towards changing land uses with the hope that those changes will produce the desired outcomes in landscape structure/function/composition.

Top of the Document


We designed TAMARIN to be extremely flexible in allowing users to change many of the assumptions underlying it.  Based on discussions with local experts, we made a number of basic baseline assumptions about future land use trends, the ability of the land to be restored to forest, the role of reserves, the minimum size forest patch needed to maintain viable populations for focal species, and other landscape ecological factors.  We assumed that recent land use trends will continue over the next one or two decades if conservation interventions are not applied:

·Primary forest will no longer be converted to other uses because it primarily remains on marginal lands and has legal protection, but it will be degraded into secondary forest.  Secondary forest will be permanently converted to pasture or agriculture except where adjacent to primary forest.  Cabruca will be entirely replaced by other forms of agriculture.  Pasture and agricultural lands and eucalyptus plantations will persist.

·Agricultural land and pasture that is abandoned will be recolonized by forest within a relatively short period (circa two decades).  Cabruca, because it retains most of the native canopy trees, will recover more quickly.

·Primary forest in existing reserves will be protected from serious degradation, and disturbed sites in reserves will gradually recover.

·Cebus xanthosternosA conservation program for the corridor can either purchase land or purchase easements on the land.  Purchase of the land will cost the market value of the property.  We assume that to purchase a conservation easement, an owner must be paid an incentive at least equal to the foregone opportunity cost of the land in addition to the management costs.

·We defined the desired forest landscape configuration in terms of representation, redundancy, and resilience parameters.  It was decided to require representation of at least two forest habitat units in all seven forest bioregions.  A contiguous area of forest (primary and, if necessary, restored forest) had to be at least 10,000 ha, based on unpublished analysis of extinction probabilities for the Cebus xanthosternos, the yellow-breasted capuchin.  A contiguous habitat unit of this size is expected to give 95% probability of survival for 100 years.  Because C.xanthosternos can traverse nonforest habitats to reach other forest fragments, we relaxed the contiguity requirement by allowing fragments within 1000 m to be considered part of the same functional habitat unit.  (Photo by Russell Mittermeier).

·Edge effects will degrade small forest fragments over time, rendering them of low biodiversity value.  We assume a depth-of-edge-influence extending 300 meters from edges into forest fragments (Gascon et al. 2000).  Core forest was defined for this study as primary or restored forest greater than 300 m from an edge of agriculture/pasture or urban area. 

Top of the Document

Application of TAMARIN

We used TAMARIN to design and evaluate five basic scenarios, including three stylized approaches that conservationists in the corridor had previously suggested.  This provided an opportunity to demonstrate TAMARIN to a workshop in Salvador, Bahia, in June 2001, illustrating the tradeoffs inherent in corridor planning and the value of a formal framework for doing it.  We also include one of the optimal scenarios here as a benchmark for comparison with other designs.

The scenarios were:

·Current—an evaluation of the present situation according to the landcover map, which was derived from 1997 satellite imagery.  This defines a benchmark of how the landscape is currently configured.

·Business-as-usual—our assumption about the likely future if no conservation interventions were applied except for existing reserves.

·Cabruca—a restoration of cabruca (i.e., select all cells with greater than 50 ha of cabruca).

·Link reserves—linking the existing reserves with a series of connecting swaths of habitat of 1-2 km wide.  The highest cumulative path of habitat suitability, analogous to a least cost path, determined linkages.

·Viable islands—manually selecting blocks of planning units to meet the conservation objectives in high suitability/low cost areas.  This scenario was an attempt to manually emulate an optimization approach to show, to a first approximation, how efficiently the objectives might be met.

·Optimal benchmark—a least cost scenario (based on OHPAS) that ensures that at least two habitat blocks are protected per bioregion, and that the blocks are at least the minimum size and contain at least 1,000 ha of primary forest.

Scenarios were evaluated according to ecological and socioeconomic criteria.  The ecological goals of representation, resilience, and redundancy were deemed met if each of seven distinct bioregions encompassed at least two protected ‘viable habitat units’ of 10,000 connected hectares each.  Habitat in the viable habitat units could consist of primary forest or abandoned land (secondary forest, cabruca, or agriculture/pasture) assumed to regenerate into forest.  Additional ecological criteria included area of primary forest placed under protection, the number of habitat units with at least 1,000 ha or primary forest as a source of propagules, and proportion of total forest exposed to edge effects.  Socioeconomic criteria included affected population and opportunity cost of conservation.  We considered two alternative assumptions about the opportunity cost of conservation.  The high assumption used the full market value of the land; the low assumption assumed that landholders derived some benefits from the land even if agricultural uses were restricted, and hence that a conservation easement could be purchased for less than the full market value.

The following table summarizes key results for the six scenarios.  There are several striking results worth noting in the table.  First, in the current scenario the conservation objectives are met in six of the seven bioregions, but the majority of remaining forest consists of small patches (mostly edge forest) within the gap crossing distance.  That is, very few of the fragments are large, relatively round blocks of core habitat.  As a consequence of our business-as-usual assumptions, the percentage of primary forest significantly declines in the other four scenarios (from 8.8% to less than 2% in most cases).  Much of this decline is in the loss of small, scattered fragments consisting of mostly edge forest (note that ¾ of remaining forest is in the edge zone in the Current scenario).  This lack of remaining large fragments required the restoration of nearly half of the planning units selected in both the Viable islands and the optimal benchmark scenarios.  In addition, the habitat-friendly cabruca land use disappears by assumption in all future scenarios, thus lowering the habitat quality of the matrix (the nonforest agricultural lands surrounding natural habitat).  Maintaining cabruca as a strategy (while failing to conserve primary forest) entails very high loss of primary forest and extremely high financial cost, while still not achieving the conservation objectives in two bioregions where cabruca is not found.  By focusing directly on the stated conservation objectives, a corridor can apparently be designed at a remarkably low financial cost, as low as $6 million US (R$12 Brazilian) for easements or environmental compensation, by our initial estimates for the optimal benchmark scenario.  However, the optimal benchmark was unable to find a second planning patch or block in the Northern Semi-deciduous bioregion that was of minimum size and also contained enough primary forest.

Baseline Scenarios

Restore cabruca
Link reserves
Viable islands
Optimal benchmark
# planning units selected
# regions with 2 habitat units > 10,000 ha
# viable habitat units with > 1000 ha of primary forest
12 of 12
4 of 4
5 of 19
10 of 12
13 of 15
13 of 13
% of corridor in forest
% of forest in depth-of-edge-influence zone
Total easement value of selected units (million reais)

Top of the Document

Optimization Model

TAMARIN assists users to craft conservation strategies and evaluate them but does not enforce the desired landscape configuration objectives.  As such the process is essentially rule-based, where the rules are the criteria a planner thinks might create a plausible policy option.  We also wanted to have a tool that would select planning units for conservation that did ensure that the landscape objectives were achieved and do so at least cost.  Such optimal solutions may not always be feasible to implement for other reasons, but they provide a benchmark to measure how much more it will cost to select a suboptimal solution.  The optimization reserve design process involves the following three steps:

1.Determine for each planning unit the environmental compensation costs and suitability in terms of amount of primary (and potentially restorable) forest.  This was easily done through queries of TAMARIN’s database.

2.The size of the basic planning unit is small in regard to what would be minimally acceptable as a viable conservation area.  A Planning Patch is defined as a set of connected planning units, which in concert meet the conditions of resilience.  In this second step, we generate a large number of planning patches, from which to design a reserve system.  This patch growing process (PGP) is based upon a computer algorithm that systematically generates possible patches for consideration that contain a minimum of 10,000 ha of primary and/or restored forest. Further, each Planning Patch must contain at least 1,000 ha of primary forest.

Example planning patch

3.The third step involves a large-scale optimization model (OHPAS) that selects a set of Planning Patches that together optimize a set of objectives and constraints.  The idea is that the set of possible planning patches spans the set of representative design alternatives.  The selection of the optimal set involves minimizing cost as well as meeting all of the desired constraints.  The selection of Planning Patches is also subject to the need to represent different bioregions.  Regional distribution constraints maintain that at least a minimal set of Planning Patches is selected in each region. 

Top of the Document


Salvador workshopTAMARIN was developed to facilitate the design and evaluation of alternative scenarios for application of economic incentives to identify priority sites within a large region for rainforest conservation or restoration.  In particular, our goal was to integrate current principles in conservation biology with economic theory in a GIS framework that makes explicit the costs and benefits of each incentive option.  Compensation of the opportunity costs of conservation is employed here as an incentive to modify behavior as an alternative to outright acquisition or command and control strategies such as agroecological zoning.  Although we made many assumptions for our example scenario concerning the desired landscape configuration, opportunity costs, and trends in land use change, the framework is very flexible in allowing stakeholders and planners to substitute competing assumptions and objectives.  These substitutions can be made as new GIS themes or simply as parameters to be entered when defining a scenario.  We also discovered that users of the TAMARIN framework invented a mode of flexibility we had not foreseen.  At a workshop held in Salvador, Brazil, in June, 2001, planning groups spontaneously began applying different economic incentives in different locations as they allocated a hypothetical budget.

The suite of models for optimization of the conservation and economic objectives constitute another significant contribution of the project.  The patch growing process, PGP, used GIS data from TAMARIN to generate a set of sample planning patches that were guaranteed to satisfy the landscape criteria for minimum patch size and minimum amount of primary forest.  Thus any patches selected automatically achieved the conservation objectives.  The OHPAS optimization model then did the actual selection of the desired number of patches, controlled the amount of overlap allowed among them, and minimized the cost.  This report presents only the initial application of this set of models for the corridor.  We have only begun research on the implications of varying assumptions and objectives.

The real value of TAMARIN is not in assisting with making better decisions per se but in facilitating the planning process by interpreting spatial information to understand the tradeoffs between conservation and other social goals.  Stakeholders and planners are forced to be explicit and quantitative in defining the desired future landscape configuration, to think not just about the current landscape but how it is likely to change, and to be creative in formulating equitable and affordable economic policies that can achieve the desired landscape with minimal disruption to the social fabric.  The details of designing an incentive program, including the identification of exactly which parcels are eligible, would still need to be developed at a finer scale with more stakeholder involvement.  Although TAMARIN was tailored to the planning issues and data sources of south Bahia, the ecological and economic underpinnings make it adaptable to many other locations.

There are many potential enhancements to TAMARIN that would increase its utility for regional conservation planning.  Several that have immediate value include:

·Allow users to identify a nucleus of a forest block and have an option where the software would automatically expand each nucleus to a viable fragment, much like the planning patch program does.

·Developing a formal land use change model was beyond the scope of this project, but could be a valuable addition to TAMARIN.

·Add an option to design feasible alternative policy instruments for a range of environmental services (carbon sequestration and watershed services) and natural capital (biodiversity) and explore the tradeoffs between them.

This project focused on tool development rather than analysis.  Thus there are many opportunities for additional analysis of alternative assumptions, trade-offs, and weights in both the TAMARIN framework and the optimization tools.  We have not yet begun to formally study the implications of alternative policy options for economic incentives.  The incorporation of other conservation benefits, such as carbon sequestration and watershed services, would make a richer research and planning environment for determining the relationship between biodiversity conservation and incentives for other conservation issues.

In the near future, TAMARIN will be made publicly accessible along with the GIS data from our project collaborators for use in conservation planning in south Bahia.  Check our web site at http://www.biogeog.ucsb.edu/projects/wb/wb.html for updates on its status and applications.

Presentations of the results of this project have been made at two scientific conferences:

TAMARIN: A landscape framework for evaluating economic incentives to foster rainforest restoration.  David Stoms.  Presented at the 17th Annual Symposium of the International Association for Landscape Ecology, US Regional Association, Lincoln, Nebraska, April 2002.

Solving a large scale reserve design problem in Bahia, Brazil.  Richard Church, Ross Gerrard, and David Stoms.  To be presented at the INFORMS Annual Meeting, San Jose, California, November 2002.

Top of the Document


Chomitz, K. M., E. Brenes and L. Constantino.  1999.  Financing environmental services: The Costa Rican experience and its implications.  Science of the Total Environment  240: 157-169.


Although staff of the University of California developed the TAMARIN software, Santa Barbara, many people and institutions contributed to its development and realization.  The original concept is due to Kenneth Chomitz, who contributed to the design of the economic aspects of the framework.  Gustavo Fonseca and Keith Alger (Conservation International (CI) /Center for Applied Biodiversity Science (CABS)) also made important contributions to the framework design and to its application to south Bahia.  Crucial data components included: the land cover characterization undertaken by Charlotte Landau (Universidade Federal de Minas Gerais); the vegetation classification by Andre Carvalho (Centro de Pesquisas do Cacau) and Wayt Thomas (New York Botanical Garden); the land value survey data gathered by IESB under the supervision of Heloisa Orlando and interpolated by Timothy Thomas; the RADAMBRASIL land characteristics maps provided by the Instituto Brasileiro de Geografia e Estatística (IBGE).  This project benefited greatly from additional contributions of data, expertise, and local knowledge of the ecological and cultural environments of south Bahia provided by participants in a sister project funded by Programa Estadual para a Conservação da Biodiversidade (PROBIO) and administered by IESB and CABS.  We are grateful also to participants in three workshops (two in Ilheus and one in Salvador) who provided in-depth feedback on TAMARIN during its development.  We regret any inadvertent omissions from this list of contributors. 

This project was funded through a contract between The World Bank (Contract Manager Kenneth Chomitz, Development Research Group, World Bank) and the University of California at Santa Barbara (Principal Investigator, Frank Davis; Co-PI, David Stoms).  Funding was provided by World Bank Research Support Board grant no. 683-42 and by the Rain Forest Trust.  In-kind contributions were made by the PROBIO project.  We also thank Kathy Scheidemen, John Sanchez and other staff at UCSB’s Institute for Computational Earth System Science for contract accounting and administration. Mike Colee provided computing support at the UCSB Biogeography Lab.  Doug Fischer of the UCSB Geography Department helped with some of the programming for the optimization-modeling component of the project.  Thanks also to Peter Choi for help in testing the TAMARIN software and reviewing the users manual.

Top of the Document

Return to project home page