gtr_nrs170.pdf

United States Department of Agriculture Assessment and Valuation of Forest Ecosystem Services: State of the Science Review Seth Binder, Robert G. Haight, Stephen Polasky, Travis Warziniack, Miranda H. Mockrin, Robert L. Deal, and Greg Arthaud Forest Service Northern Research Station General Tech...

United States Department of Agriculture Assessment and Valuation of Forest Ecosystem Services: State of the Science Review Seth Binder, Robert G. Haight, Stephen Polasky, Travis Warziniack, Miranda H. Mockrin, Robert L. Deal, and Greg Arthaud Forest Service Northern Research Station General Technical Report NRS-170 May 2017 Abstract This review focuses on the assessment and economic valuation of ecosystem services from forest ecosystems—that is, our ability to predict changes in the quantity and value of ecosystem services as a result of specific forest management decisions. It is aimed at forest economists and managers and intended to provide a useful reference to those interested in developing the practice of integrated forest modeling and valuation. We review examples of ecosystem services associated with several broad classes of potentially competing forest uses—production of timber, sequestration of carbon, regulation of the quality and quantity of water, provision of residential and recreational amenities, and protection of endangered species. For each example considered, we briefly describe what is known about ecological production functions and economic benefits functions. We also highlight the challenges and best practices in the creation and use of this knowledge. In the final section, we discuss the process, strengths, pitfalls, and limitations of utilizing integrated models for benefit-cost analysis of proposed forest management activities. Quality Assurance This publication conforms to the Northern Research Station’s Quality Assurance Implementation Plan which requires technical and policy review for all scientific publications produced or funded by the Station. The process included a blind technical review by at least two reviewers, who were selected by the Assistant Director for Research and unknown to the author. This review policy promotes the Forest Service guiding principles of using the best scientific knowledge, striving for quality and excellence, maintaining high ethical and professional standards, and being responsible and accountable for what we do. Manuscript received for publication 22 February 2016 Published by U.S. FOREST SERVICE 11 CAMPUS BLVD SUITE 200 NEWTOWN SQUARE PA 19073 May 2017 Assessment and Valuation of Forest Ecosystem Services: State of the Science Review Seth Binder, Robert G. Haight, Stephen Polasky, Travis Warziniack, Miranda H. Mockrin, Robert L. Deal, and Greg Arthaud The Authors SETH BINDER is an Assistant Professor with St. Olaf College, Department of Economics, Department of Environmental Studies, Northfield, MN 55057, [email protected]. ROBERT G. HAIGHT is a Research Forester with the U.S. Forest Service, Northern Research Station, St. Paul, MN 55108, [email protected]. STEPHEN POLASKY is a Fesler-Lampert Professor of Ecological/Environmental Economics with the Department of Applied Economics, University of Minnesota, St. Paul, MN 55108. TRAVIS WARZINIACK is a Research Economist with the U.S. Forest Service, Rocky Mountain Research Station, Fort Collins, CO 80526. MIRANDA H. MOCKRIN is a Research Scientist with the U.S. Forest Service, Northern Research Station, Baltimore, MD 21228. ROBERT L. DEAL is a Research Forester with the U.S. Forest Service, Pacific Northwest Research Station, Portland, OR 97205. GREG ARTHAUD is a Research Social Scientist with the USDA Forest Service, 1400 Independence Avenue, SW, Washington, DC 20250-0003. General Technical Report NRS-170 Page intentionally left blank CONTENTS Introduction............................................................................................................................................................1 Framework for Identifying and Valuing Ecosystem Services...............................................................................2 Timber Production.................................................................................................................................................4 Ecological Production Functions for Timber........................................................................................................4 Economic Benefit Functions for Timber..............................................................................................................6 Carbon Storage.......................................................................................................................................................6 Ecological Production Functions for Carbon Storage...........................................................................................8 Economic Benefit Functions for Carbon Storage.................................................................................................9 Water Regulation.................................................................................................................................................11 Flow Regime....................................................................................................................................................11 Thermal and Light Inputs.................................................................................................................................14 Sediment Flux..................................................................................................................................................15 Chemicals, Nutrients, and Pathogens...............................................................................................................16 Aesthetic Amenities.............................................................................................................................................17 Recreation.............................................................................................................................................................19 Wildlife...................................................................................................................................................................21 Ecological Production Functions to Predict Changes in Wildlife Abundance......................................................21 Economic Benefit Functions for Wildlife Abundance.........................................................................................23 Integrated Modeling and Benefit-Cost Analysis.............................................................................................25 Static Benefits and Costs, and Selection, Among Discrete Alternatives.............................................................25 Temporal Changes in Benefits and Costs, and Optimization, of a Continuous Choice Variable...........................26 Spatial Interdependence of Benefits and Costs.................................................................................................27 Uncertainty......................................................................................................................................................27 Benefits Transfer...............................................................................................................................................28 Non-use Values and Distributional Analysis......................................................................................................29 Conclusions............................................................................................................................................................30 Acknowledgments...............................................................................................................................................30 Literature Cited....................................................................................................................................................31 Appendix: Recreation Valuation Methods........................................................................................................45 Discrete Choice Random Utility Model..............................................................................................................45 Hedonic Travel Cost Method.............................................................................................................................45 Generalized Corner Solution.............................................................................................................................46 Challenges........................................................................................................................................................46 Page intentionally left blank INTRODUCTION Ecosystems provide many goods and services that enable and enrich human life, from traditional natural resources, such as timber, fish, and edible plants, to the aesthetic qualities and characteristics of a place, to clean water and air (Daily 1997). Human ingenuity has enabled people to refine, re-allocate, and intensify the production of many goods and services by combining natural processes with human-created tools and labor. This has led to extraordinary advances in longevity and material well-being. However, it has also led to declines in some forms of natural capital and many nonmarketed ecosystem services (Millennium Ecosystem Assessment 2005). Scientists, policymakers, and land managers increasingly recognize the varied contributions of healthy, multi-functional ecosystems to human wellbeing and seek to develop the tools and knowledge necessary to manage these systems to best meet societal objectives. Within the last decade, several major academic and governmental initiatives related to ecosystem services have emerged. These include the publication of the Millenium Ecosystem Assessment (Millennium Ecosystem Assessment 2005), a National Research Council report on valuing ecosystem services (National Research Council 2004), a report from the EPA Science Advisory Board (U.S. Environmental Protection Agency 2009), a report on the economics of ecosystem and biodiversity (Kumar 2010), and the establishment of the new Intergovernmental SciencePolicy Platform on Biodiversity and Ecosystem Services (IPBES). The Executive Office of the President (EOP) recently released a memorandum directing Federal agencies to factor the value of ecosystem services into Federal planning and decisionmaking. An additional EOP memorandum outlines research needs to assess ecosystem services in coastal green infrastructure. Within the EOP Office of Science and Technology Policy, an interagency Ecosystem Service Working Group was formed to facilitate cooperation among relevant agencies. A consistent finding among these publications is that economic valuation of ecosystem services and comprehensive benefit-cost analyses are important tools to help decisionmakers manage ecosystems. This review focuses on the assessment and valuation of ecosystem services from forest ecosystems—that is, our ability to predict changes in the quantity and economic value of ecosystem services as a result of specific forest management decisions. It is aimed at forest economists and managers of public and private forest land, with the intention of providing a useful reference to those interested in developing the practice of integrated forest modeling and valuation. To this end, we review examples of ecosystem services associated with several broad classes of potentially competing forest uses—production of timber, sequestration of carbon, regulation of the quality and quantity of water, provision of residential and recreational amenities, and protection of endangered species. For each ecosystem service, we review a selection of ecological and economic research related to ecological production functions and economic benefits functions, and highlight challenges and best practices in the creation and use of this knowledge. In the final section, we discuss the strengths, pitfalls, and limitations of utilizing integrated models for benefit-cost analysis of proposed forest management activities. We supplement this discussion with a more quantitative treatment for relatively simple decision problems of optimal land use (i.e., preserve, harvest, or develop a given forest area) and optimal rotation age. The academic literature on ecosystem services is vast and we limit our scope to services of non-urban forests, public or private, that are amenable to economic valuation. We do not cover cultural ecosystem services such as cultural heritage or spiritual significance that are difficult to quantify and whose value is often thought to be antithetical to consideration in monetary terms. These cultural services have value in their own right, and they have played an important role in motivating public support for the protection of ecosystems. Daniel et al. (2012) review research on relationships between ecological structures/ functions and cultural values including landscape aesthetics, cultural heritage, outdoor recreation, and spiritual significance. We also do not cover ecosystem services provided by urban forests, wetlands, lakes, and undeveloped areas (e.g., McPhearson et al. 2014) where the beneficiaries are primarily urban residents. Methods have been developed to estimate the economic value of urban forests based on their effects on air quality (Nowak et al. 2014), water Assessment and Valuation of Forest Ecosystem Services: State of the Science Review 1 Figure 1.—Conceptual diagram of the links among changes in ecosystem management decisions, the production of ecosystem services, and resulting benefits and costs to society. From Polasky and Segerson (2009). quantity and quality (Hobbie et al. 2014, Keeler et al. 2012, McPherson et al. 2005), residential energy consumption (Akbari 2002), and aesthetic amenities (Sander et al. 2010). Non-urban forests may affect aesthetic amenities of nearby residents and so we do discuss hedonic property value studies (e.g., Sander et al. 2010) as a way to estimate the value of forests for the provision of aesthetic amenities. Framework for Identifying and Valuing Ecosystem Services Science-based decision support tools have the potential to provide information to Federal agencies, States, and private landholders about the benefits and tradeoffs among various forest uses. These tools require understanding—and quantitatively modeling of—the chain of relationships that link changes in forest policy or management to changes in human well-being (Fig. 1). Government agencies control some important management decisions directly (e.g., land use within National Forests). In this case, the agency would start the analysis by considering the effect of its management decisions on ecosystems (Link 2). In other cases, government agencies set policies that provide incentives to private landowners (e.g., the 2 Conservation Reserve Program). Here the agency would need to predict how the policy would affect private landowner decisions (Link 1) and then what impact these decisions have on ecosystems (Link 2). From here, analysis must consider how changes in ecosystem structure and function translate into changes in ecosystem services (Link 4). Ecological production functions (NRC 2005, Polasky and Segerson 2009, Swallow 1990) capture these relationships. An example of an ecological production function is an empirically estimated equation that predicts the abundance of a wildlife species that people care about (the ecosystem service) as a function of the age, species composition, slope, and elevation of the forest stand in which the wildlife population lives. Ecological production functions can be used to estimate production potential and identify biophysical tradeoffs between alternative ecosystem services. Some analysts may prefer to base policy decisions on consideration of impacts on the flow of various ecosystem services, recognizing the potential for tradeoffs in those flows (Link 5) rather than assessing them in terms of the public’s preferences. In many cases, inefficiencies in existing management imply the possibility of identifying alternative “winwin” management scenarios that increase all ecosystem Assessment and Valuation of Forest Ecosystem Services: State of the Science Review services. In other cases, however, decisionmakers require additional information to help navigate tradeoffs among ecosystem services. Here, understanding the relative values of increases or decreases in different services can help managers or policymakers select the option that brings the greatest benefits to society. This requires the estimation and use of economic benefits functions, which quantify in monetary terms the relationships between changes in the provision of ecosystem services and changes in human well-being (Link 6). The benefits function for our example above would be an equation that translates changes in wildlife abundance into a dollar value, based perhaps on an economic valuation study of recreational demand for wildlife abundance. Economic valuation of an ecosystem’s goods and services represents an attempt to estimate changes in people’s economic well-being—as measured by their own preferences—due to incremental (marginal) changes in the ecosystem’s components. When ecosystem goods are traded in markets (e.g., timber), the market price (e.g., U.S. dollars/cubic meter) is a measure of the benefit people get from a unit of the good. Since most ecosystem services are not traded in markets, and therefore do not have observable prices, economists estimate the value of changes in ecosystem services by leveraging the information conveyed by individuals’ observable decisions. Information obtained from observable decisions in hypothetical markets created by the analyst is known as stated-preference data. In contrast, revealed preference data is obtained from observable decisions in actual markets for a weak complement to the non-market ecosystem service. In both cases, the choices and tradeoffs people make reflect their willingness to pay (WTP) to access or obtain ecosystem services or their willingness to accept (WTA) some amount in exchange for a reduction in services. Their WTP or WTA is a monetary measure of the benefits they get from a change in the service. Economic benefits functions estimate WTP or WTA based on the nature and extent of changes in ecosystem components, the availability of substitute or complementary goods or services, and beneficiaries’ income and other demographic characteristics. It is important to point out that methods for economic valuation of ecosystem services differ from survey methods, such as public participation geographic information systems (PPGIS), for assessing public preferences for ecosystem services. Analysts using PPGIS methods typically ask selected individuals to locate ecosystem services (e.g., aesthetic, recreation, economic, and ecological services) that they value within a given landscape (e.g., Brown and Reed 2009). The maps of landscape values are then analyzed to determine their relative importance as an estimate of people’s preferences (e.g., Brown and Donovan 2014). Economic valuation methods go further than PPGIS methods by estimating how much people are willing to pay for an incremental change in the level of any given service, based on their stated or revealed preferences in hypothetical or actual markets. Together, ecological production functions and economic benefits functions form an integrated assessment model that links management decisions to their full costs and benefits (Daily et al. 2009, Nelson et al. 2009). This scientific approach to analysis, when transparent, will provide policymakers and managers with valuable information to support their decisions. However, working across disciplines and integrating models that were not necessarily designed to be compatible is no easy task. One of the biggest stumbling blocks can be an understanding and operationalization of the ecosystem service concept. Ecosystem services have been variously defined as the benefits people obtain from nature (e.g., recreational fishing) (Millennium Ecosystem Assessment 2005), the end products of an ecosystem that are directly used or consumed by people (e.g., the fish anglers seek for their recreational benefit) (Boyd and Banzhaf 2007), or the processes by which ecosystems produce resources (e.g., nutrient cycling that enhances fish populations) (Ecological Society of America 2012). While all of these definitions make useful connections between ecology and human well-being, we use the definition advanced by Boyd and Banzhaf (2007) because it facilitates measurement, integrated modeling, and valuation. Ecosystem services are components of nature, directly enjoyed, consumed, or used by people for their well-being. As noted by Boyd and Banzhaf (2007), this definition has several important features. First, ecosystem services are end products of nature that are directly consumed, enjoyed, or used for human benefit. We distinguish between a benefit and an ecosystem service because benefits are often produced with Assessment and Valuation of Forest Ecosystem Services: State of the Science Review 3 capital and labor in addition to biophysical inputs. For example, flood control is a benefit that depends on the construction of levees, canals, and other engineering features in addition to the peak flow of water downstream from the forest. We think it is important to identify the end product of the ecosystem (peak flow of water) so that it can be measured and valued in the context of the benefit to which it contributes. Second, the distinction between end products and intermediate products or processes is important in welfare accounting. Because the value of intermediate goods or processes is embodied in the value of final goods, only the value of the final good need be counted. (However, it is still possible and may sometimes be desirable to estimate the value of a change in an intermediate ecosystem service.) Third, ecosystem services are components of ecosystems, which means they are ecological things or characteristics, not the functions or processes that support or produce the end products. Fourth, ecosystem services are measured by their quantities or physical units, which subsequently can be paired with estimates of the monetary value of changes in these quantities. We emphasize that other definitions and lists of ecosystem services, such as Daily’s (1997), were constructed to illustrate the connection between ecology and human well-being, not to facilitate measurement, integrated modeling, and valuation of services. For each forest use, Table 1 provides a subset of forest benefits and associated ecosystem services that we illustrate in the following sections. Because a forest ecosystem consists of all the biological organisms in a woodland functioning together with all of the nonliving physical components of the woodland, goods and services may be biotic (e.g., timber, trout) or abiotic (e.g., stream water, carbon) components of the ecosystem. The examples given in the table and covered in this review are not exhaustive of either the benefits arising from different forest uses or the ecological services associated with a particular benefit. Rather, they are examples for which integrated biophysical and economic modeling techniques have been used for service valuation. Each row of the template represents a unique forest benefit and beneficiary. For each benefit, the template identifies: 1) the ecosystem service (i.e., the ecological end product) that can be measured or modeled in biophysical assessments and that directly affects human well-being; 2) the ecological 4 production function that models how changes in ecosystem structure and function translate into changes in ecosystem services; and 3) the economic benefit function or method to estimate the monetary value of changes in the ecosystem service that result from changes in forest management. TIMBER PRODUCTION Forests provide timber for the wood products industry (Table 1), and timber production has long been an objective of public and private forest management. We begin with a description of timber management systems and ecological production functions that project timber yields. Then, we describe economic benefit functions, including the market processes that determine timber prices on public and private land and the net present value criterion to evaluate timber management systems. While we focus on timber, we emphasize that forest management is inherently a problem of joint production: decisions intended to produce timber affect the production of other forest ecosystem services. Here, we briefly note how ecological production functions and economic benefit functions for timber can be extended to assess changes in other ecosystem services and their values. In a later section, we give a detailed account of integrated modeling and benefit-cost analysis for the joint production of multiple ecosystem services. Ecological Production Functions for Timber Forests that are managed for timber production are subdivided into stands, which are administrative units bounded by physical and geographic features such as roads and property boundaries and by discontinuities in forest vegetation. Each stand is managed using a particular management system, typically evenage or uneven-age management (Smith 1962). A stand management system is a planned program of silvicultural treatments during the life of the stand. Treatments include regeneration cuttings that are used to establish a new stand by means of natural or artificial regeneration. Intermediate cuttings such as thinnings or selection harvests are used to control the density and growth of existing trees. Stand management systems are classified based on their regeneration method. The most well-known management system uses the clearcutting Assessment and Valuation of Forest Ecosystem Services: State of the Science Review Table 1.—Template for identifying and valuing ecosystem services from forests. For each forest use, we identify one or more benefits, beneficiaries, and ecosystem services. Ecosystem services are the valued end products of the forest ecosystem that contribute to the production of benefits and may be affected by forest management activities. The ecological production functions and economic benefits functions form an integrated assessment model that links management decisions to their full costs and benefits. Forest use Benefit Beneficiary Ecosystem service (valued end product of Ecological production the forest ecosystem) function Timber production Timber for wood products Industrial wood producers Merchantable timber (stumpage) Stand simulation models (e.g., Forest Vegetation Simulator) Forest landscape simulation models Market price of timber (stumpage price) Carbon storage Climate regulation Everyone Sequestered carbon Carbon budget simulation models (e.g., FORCARB2) Stand simulation models Social cost of carbon Water regulation Irrigation water Farmers Flow of water downstream Paired watershed studies Forest hydrological models (e.g., Distributed Hydrology Soil Vegetation Model) Market price for water Shadow price of water Hedonic price model for farmland Flood control Homeowners Peak flow downstream Paired watershed studies Forest hydrological models Avoided damage Coldwater fishing Anglers Trout abundance Energy transfer models for stream temperature coupled with trout population models Hedonic travel cost model Discrete choice RUM Clean drinking water Local water consumers Amount of sediment in water Erosion prediction models (e.g., Water Erosion Prediction Project model) Household demand model for water, avoided costs of treatment, replacement cost Safe navigation Commercial navigators Amount of sediment in water Erosion prediction models Avoided costs of dredging Clean drinking water Local water consumers Amounts nitrate and phosphorus in water Nutrient and chemical movement models (e.g., Soil and Water Assessment Tool) Household demand model for water, avoided costs of treatment, replacement cost Aesthetic amenity Homeowners near forest Forest cover in viewshed Stand simulation models Forest landscape simulation models Hedonic property price model Aesthetic amenity Leisure travelers Forest cover in and commuters viewshed Stand simulation models Forest landscape simulation models Recreational demand model Recreation Recreational hiking, camping, and biking Hikers, campers, Old growth area, bikers forest density, burned area Stand simulation models Forest landscape simulation models Recreational demand model Discrete choice RUM Wildlife Recreational hunting Hunters Game abundance Demographic models of wildlife abundance and viability (e.g., RAMAS) Recreational demand model Discrete choice RUM Hedonic pricing of licenses Protecting rare and endangered species Everyone Species survival probability Demographic models of wildlife abundance and viability Contingent valuation method Aesthetic amentity Assessment and Valuation of Forest Ecosystem Services: State of the Science Review Economic benefits function 5 reproduction method and produces even-aged stands. In this case, even-aged management is a cycle of events (rotation period) that includes clearcutting a mature stand, planting a single cohort of trees, periodically thinning the new crop, and clearcutting the crop at a specified rotation age. Uneven-age management uses a selection harvesting method that involves the periodic harvest of trees in specified size classes. Selection harvests are conducted to control the spacing and growth of the remaining trees and to enhance natural regeneration from seeds produced by the remaining mature trees. Since selection harvest and regeneration take place simultaneously, uneven-aged stands include a mixture of trees in a wide range of size and age classes. predict changes in the tree attributes as a function of stand density. An equation for tree mortality adjusts the expansion factors, and an equation for regeneration creates new records. Like stage-class model construction, equations for changes in tree dimensions are estimated separately and then inserted in a simulation shell for projecting stand growth. Major modeling efforts in different regions of the United States have created individual-tree simulators, and many of those have been incorporated into the Forest Vegetation Simulator, which is a family of forest growth simulation models supported by the USDA Forest Service (see Crookston and Dixon 2005 for review). For both even-age and uneven-age management systems, timber yields are projected using stand simulation models (i.e., ecological production functions), which have been developed since the 1960s to provide forest managers with accurate growth and yield information for planning (see Munro 1974 for review). Stand simulation models include stage-class models or individual-tree simulators, both of which use discrete time difference equations to project tree growth, mortality, and regeneration. The models differ in their characterization of stand structure and their tractability for optimization. An important property of individual-tree simulators is the detail in which a stand and its growth processes are described. Since individual-tree simulators explicitly project the dimensions of hundreds or even thousands of trees in a stand, they provide a detailed picture of stand structure and composition over time. This detail facilitates the simulation of many forest-related processes including wildlife population dynamics, insect and disease dynamics, wildfire intensity and spread, and carbon sequestration (Crookston and Dixon 2005). Recently, progress has been made connecting individual-tree simulators with optimization algorithms to analyze stand treatment and harvest decisions in relation to the ecosystem services produced (e.g., Hyytiäinen et al. 2004, Rämö and Tahvonen 2014). In stage-class models, trees are classified into species and stem-diameter classes. Equations for tree growth, mortality, and regeneration project the changes in the number of trees in each class over time as a function of stand density (see Getz and Haight 1989 for review). Stage-class models are constructed by first estimating the parameters of each of the component equations separately and then inserting the equations into a matrix framework to project stand growth. This matrix framework facilitates the application of linear and nonlinear programming algorithms for optimization of stand treatments and harvests (Getz and Haight 1989, Haight 1987). Individual-tree simulators describe the stand with a list of tree records where each record contains the current tree dimensions (e.g., stem diameter, height, and crown length) and an expansion factor representing the number of trees of its kind in the stand (see Liu and Ashton 1995 for review). Equations for diameter growth, height growth, and crown development 6 Forest landscape simulation models (FLSMs) have been developed to project the effects of forest management options on the spatial configuration, composition, and heterogeneity of vegetation, including community types, tree species age classes, and aboveground biomass (see Scheller and Mladenoff 2007 for review). In contrast to stand simulation models, FLSMs are applied to extensive areas of forest. They subdivide the landscape into cells (or polygons) and project attributes of the vegetation in each cell. Projections are based on vegetation attributes within and across cells and disturbance processes that are endogenous (e.g., wildfire) or exogenous (e.g., land-use change or timber harvesting) to the model. The extent and detail of vegetation attributes that are projected with FLSMs allow for the projection and analysis of many forest benefits including carbon storage, recreation, wildlife abundance, and water yield. Assessment and Valuation of Forest Ecosystem Services: State of the Science Review Economic Benefit Functions for Timber Wood products (e.g., lumber, paper, structural panels, and fuel) have well-defined markets and significant economic value, contributing $280 billion in 2007 to the U.S. economy (United States Census Bureau 2009). The economic benefit of timber for wood products is measured by stumpage value—the amount per unit area that a commercial wood cutter is willing to pay for an area of standing trees (Helms 1998). It is the product of the stumpage price and the amount of timber offered for sale. To understand how stumpage price is determined, it is useful to first describe the ownership of forest lands in the United States. Forest lands are primarily owned by private entities, including nonindustrial owners without processing facilities, forest industry owners with processing facilities, and various types of forest land investment organizations. Nonindustrial and forest land investors own over two-thirds of U.S. forest land; combined with the forest industry, the U.S. private sector owns roughly 75 percent of U.S. forest lands and produces over 90 percent of the industrial wood harvest. The remaining 25 percent of forest lands is owned by government agencies—Federal, State, and local—which produce less than 10 percent of the industrial wood (Sedjo 2006). Stumpage prices for private timber are marketdetermined (Sedjo 2006). They depend on the industry’s aggregate demand for trees (which is based in part on market-determined prices for wood products, which in turn depend on domestic and international trade) and the aggregate supply of industrial wood from private landowners (which is based on the current and expected future stumpage prices and age of the forest). Stumpage prices for a given ownership also depend on wood quality—itself dependent on timber age, species, and condition—and cost considerations associated with timber accessibility, mill distance, terrain, and other factors reflecting extraction, transport, and processing costs. Stumpage prices for most sales of public timber are determined through a competitive auction process, which reflects the market information on stumpage prices of private timber (Sedjo 2006). The decision about which management system to use in a particular stand and the attributes of the management system is often made using a net present value criterion (e.g., Haight 1987). For even-age management, the problem is to determine: the timing and intensity of silvicultural treatments for the current stand; the time when the stand is clearcut and replaced with a plantation, if it is currently under natural forest cover; and the timing and intensity of silvicultural treatments and clearcut age for the plantation. For uneven-age management, the problem involves determining the sequence of selection harvests that converts the current stand to the desired uneven-age steady state. The net present value of each management system is calculated based on the discounted value of timber yields and costs of management (e.g., planting, weeding, and pruning) over all future harvests (Faustmann 1849, Samuelson 1976). We emphasize that forest management systems intended to produce timber affect the production of other forest ecosystem services. For example, stands managed with selection harvests maintain plant understory species richness and abundance, which are important for wildlife habitat (Deal 2001), while maintaining stand growth and providing industrial wood (Deal and Tappeiner 2002). The stand structures that develop after selection harvests create structurally complex, multi-layered forest canopies that were much more similar to old-growth forests than the uniform young-growth stands that develop after clearcutting (Deal et al. 2010). The value of non-timber ecosystem services can be incorporated into the calculation of net present value of forest management systems. Integrated assessment models that account for the production and value of multiple ecosystem services have been developed and used in forest management since the 1980s (e.g., Bowes and Krutilla 1989). CARBON STORAGE Forest ecosystems provide a climate regulation benefit (Table 1) because forests store carbon in the soil and in biomass that might otherwise be released into the atmosphere in the form of CO2. Carbon storage is a valuable ecosystem service because reducing atmospheric carbon reduces the intensity of future climate change. Reducing the likelihood of damage associated with more intense climate change will benefit everyone. In this section, we review biophysical modeling approaches to estimating forest carbon stocks and fluxes, and we discuss economic approaches to estimating the value of storing additional carbon through enhanced sequestration or release prevention. Assessment and Valuation of Forest Ecosystem Services: State of the Science Review 7 Ecological Production Functions for Carbon Storage When discussing regional, national, or global carbon stocks and fluxes, most papers report carbon in teragrams (Mt, megatons, million metric tonnes) or megagrams (Mg, one metric tonne). At the stand or forest levels, megagrams (Mg) per hectare of carbon are used. These measures are carbon mass, not CO2 mass, because carbon is a standard currency and can easily be converted to any other unit. Many reports give stocks and fluxes of the mass of CO2. To convert C mass to CO2 mass, multiply by 3.67 to account for the mass of the O2. The evaluation of forestry opportunities for carbon storage received intense analysis beginning in the 1990s. Birdsey (1992) was the first to provide comprehensive estimates of carbon storage and accumulation in U.S. forests. Carbon was estimated separately for trees, soil, forest floor, and understory vegetation for major forest types and plantation species in eight geographic regions. For trees, carbon estimates were based on merchantable volumes in existing forest inventory data, which were converted to carbon based on conversion factors for total tree volume, total biomass, and carbon as percentage of dry mass. Estimates of carbon stored in soil, forest floor, and understory estimation were obtained from the ecological literature. Birdsey (1992) also estimated changes in carbon storage over time based on changes in live trees. The conversion factors and carbon estimates of Birdsey (1992) were incorporated into carbon budget models to examine the effects of forest management practices on carbon storage in U.S. timberlands (e.g., Adams et al. 1999, Alig et al. 1997). In the early 2000s, work began to replace Birdsey’s (1992) conversion factors for merchantable tree volume with estimators for forest biomass and carbon. Jenkins et al. (2003) developed a consistent set of aboveground tree biomass equations as a function of tree diameter for over 100 species in the United States. These individual-tree equations were then applied to inventory plot data to estimate equations for biomass density (Mg/ha) of live and standing dead trees as a function of merchantable volume (m3/ha) for broad forest types and regions of the coterminous United States. (Smith et al. 2003a). Tree biomass is about 50 8 percent carbon, so carbon estimates can be derived from estimates of biomass by multiplying by 0.5. The equations for forest biomass were incorporated in the U.S. Forest Service carbon budget simulation model (FORCARB2), which provides inventory-based estimates of U.S. forest carbon stocks (Smith et al. 2004). The model includes separate, non-overlapping components of total forest ecosystem carbon pools, including live trees, standing dead trees, understory vegetation, down dead wood, forest floor, and organic carbon in soil. The model is applied to plot-level inventory data, where merchantable tree volume (m3/ ha) and age are used to estimate tree and forest floor carbon, respectively. FORCARB2 was used in U.S. Greenhouse Gas Inventory (United States Department of Agriculture 2008) and Forest Service studies (e.g., Heath et al. 2011), which provided managed forest carbon estimates for 253 million ha of U.S. forest land. In addition to tools like FORCARB2 that estimate carbon storage at the county, state, and national levels, simulation models of forest growth have been developed to predict changes in carbon storage at the national, regional, and stand levels. Further, they are used to estimate net carbon storage under alternative forest policy or management scenarios relative to a baseline scenario to evaluate the carbon impacts of changes in policy (Richards and Stokes 2004). For example, Wear and Coulston (2015) develop a forest carbon projection model based on observations from over 350,000 permanent monitoring plots across the United States that are part of the USDA Forest Service’s Forest Inventory and Analysis Program. Their model, which is developed for regional or national projections, includes estimates of carbon densities by forest age class, forest sequestration rates by age class, areal extent of forest by age class, and age transition probabilities aggregated at the state or regional level, including disturbance and management effects. Forest dynamics are applied as transition probabilities to current estimates of areal extent by age and define changes in forest structure. Carbon sequestration is then estimated using observed carbon stock densities and sequestration rates applied to the new forest structure. Wear and Coulston (2015) project the effects of national-level policies intended to boost forest carbon sequestration, including reducing deforestation and development, increasing afforestation and reforestation, and reducing wildfire. Assessment and Valuation of Forest Ecosystem Services: State of the Science Review At the stand level, carbon projections can be made with the Forest Vegetation Simulator (FVS), an individualtree, distance-independent, growth and yield model that predicts changes in tree diameter, height, crown ratio, and crown width, as well as mortality, over time (Crookston and Dixon 2005). FVS has been calibrated for geographic areas of the United States and can simulate a wide range of silvicultural treatments for most major forest tree species, forest types, and stand conditions. The model includes equations to predict the biomass of live trees, dead trees, down dead wood, understory, and forest floor. Biomass, expressed as dry weight, is assumed to be 50 percent carbon. FVS is used to estimate the potential carbon consequences of forest management actions, including planting densities, thinning regimes, and rotation age (Hoover and Rebain 2011). Economic Benefit Functions for Carbon Storage Evaluating the economic benefits of carbon storage by U.S. forests depends on identifying a monetary value per ton of carbon removed from the atmosphere. Monetary units are especially helpful because they can then be compared with monetary costs of carbon policies and programs. The value to society of sequestering or preventing the release of additional carbon dioxide can be viewed as the avoided economic damages or costs of additional carbon in the atmosphere. The value of carbon is not fully (or even mostly) reflected in market prices. Although both voluntary and compulsory carbon markets exist, the prices in these markets reflect a demand for mitigation that falls far short of estimates of the actual economic damage avoided by preventing the release of additional carbon to the atmosphere. It is this value—the value of avoided damage—that forest managers should use in decision-making if they wish to maximize benefits to society and to treat the benefits of carbon storage on par with the benefits of other ecosystem services, for which valuation methods are designed to capture the full social marginal value. Economists use the term social cost of carbon (dioxide emissions) (SCC) to describe the marginal damages from a ton of carbon emitted to the atmosphere (Tol 2008). Storing an additional ton of carbon (above a given baseline) leads to less intense climate change and damage. The value of permanently storing that additional ton is equivalent to the SCC. This value depends on the specific trajectory of emissions, economic production, and climate change over time (Nordhaus 2008). Only if the world adopted an optimal incentive-based carbon policy would the resulting price of carbon exactly equal the SCC. In the absence of a complete, compulsory market for terrestrial carbon storage, a coherent approach to forest management that accounts for the full value of carbon storage will require coordination across agencies to choose a common methodology for the estimation of the SCC, and to jointly adopt revised estimates and management plans as economic conditions change. The SCC depends fundamentally on estimates of total damages arising from a given change in climate over a specified period of time. As Tol (2009) notes, despite a proliferation of SCC estimates, there exist only 13 different studies of total damage estimates upon which to base estimates of the SCC, of which only nine had been used at the time of Tol’s writing. Estimation of total damages remains an important and active area of research. Given the need for convergence in the selection of IAM approaches, features, parameter values, and underlying damage cost estimates, it is clear that estimates of the SCC will continue to evolve. Several methods have been used to estimate the SCC, with important differences in the choice of model, scope, and parameterization. The academic literature lacks consensus in regard to these choices, and differences in approach have led to wide differences in SCC estimates. In a survey of the literature, Tol (2008) finds more than 200 different estimates of the SCC, with a wide and highly skewed distribution: mean, median, and mode values are 127, 74, and 35 $U.S. 1995 per Mg CO2, respectively. The standard deviation in estimates is $243 per Mg CO2. Perhaps the most widely referenced SCC estimates come from the Dynamic Integrated ClimateEconomy (DICE) model (Nordhaus 1993) and the Regional Integrated Climate-Economy (RICE) model (Nordhaus and Yang 1996) developed and continuously updated by William Nordhaus. Other integrated assessment models (IAMs) include FUND (Tol 1995), PAGE (Hope 2006), and WITCH (Bosetti et al. 2007). Various scholars have adjusted the basic DICE/ Assessment and Valuation of Forest Ecosystem Services: State of the Science Review 9 RICE approach (though not all report changes in SCC): Sohngen & Mendelsohn (2003) incorporate the mitigation potential of forest carbon storage; Buonanno et al. (2003) and Popp (2004) include endogenous technical change; Sterner and Persson (2008) account for relative price changes and the consumption of non-market goods; de Bruin et al. (2009) integrate the costs and benefits of adaptation; Lemoine and Traeger (2012) incorporate tipping points and ambiguity aversion; and Cai et al. (2012) utilize a continuous-time framework to improve the temporal resolution and reliability of analysis. Large differences in SCC estimates can be obtained even from the same model, depending on the choice of international equity weighting (Fankhauser et al. 1997) or intertemporal discount rate. Equity weighting is an attempt to correct differences in estimates of people’s WTP for reductions in damages from climate change (e.g., a reduction in human mortality risk). These differences in WTP may arise because of differences in income or other socio-economic conditions, which may be considered unfair. Equity weighting can significantly increase aggregate (global) damage figures, although some specifications of weighting functions also imply reduced estimates (Fankhauser et al. 1997). Two camps have emerged with regard to the appropriate choice of the intertemporal discount rate. The first insists on a parameterization that is consistent with an ethical framework that values future generations on par with the present (Stern 2007), which implies a very small pure rate of time preference (PRTP) that exceeds zero only to reflect the very small probability of non-existence of future generations. The second camp insists on a parameterization that is consistent with observed behavior and other economic model parameters (Nordhaus 2007), which implies a much larger PRTP. Kaplow et al. (2010) and Goulder and Williams III (2012) argue that these approaches can be reconciled: the PRTP used for evaluative purposes (for example, in a planner’s social welfare function) need not be the same as the PRTP used for predictive purposes (i.e., in a positive economic model). An important point of recent consensus among leading economists is that the discount rate should generally decline over time as a result of uncertainty, regardless of the choice of PRTP (Arrow et al. 2013). This means 10 that the certainty-equivalent value of benefits and costs realized further out in time should be discounted at progressively lower rates. More concretely, in a threeperiod context, the value of consumption in the third period would be discounted back to the second period at a lower rate than that at which consumption in the second period is discounted back to the first. The “term structure” of declining discount rates should apply equally to all costs and benefits relevant to a given decision context, so the forest manager should ensure that the structure underlying any off-the-shelf estimate of the SCC to be incorporated into an integrated assessment model is consistent with the structure applied to other market and non-market benefits. In 2009 the U.S. Government convened an interagency working group to estimate the SCC for use in regulatory analysis (Greenstone et al. 2013, U.S. Interagency Working Group 2010). They assumed a global perspective and used three IAMs (DICE, PAGE, and FUND). The estimates were subsequently updated in 2013 (U.S. Interagency Working Group 2013). Using a 3 percent discount rate, the working group estimated that SCC increased from $44 to $72 per Mg of CO2 from 2015 to 2044 measured in 2016 U.S. dollars. In addition to navigating the complications of SCC estimation and the alignment of discount rates, forest managers must also account for potential leakage and the impermanence of terrestrial carbon storage in calculating carbon values for large-scale policy decisions. Carbon leakage occurs when management actions that successfully lead to greater carbon storage locally indirectly create incentives for greater carbon emissions elsewhere. For example, a prohibition on harvest on all National Forest land would dramatically increase carbon storage in the National Forests. However, on a global timber market, the reduction in timber supply from National Forests may be partially or fully offset by increases in harvest on forest lands elsewhere, leading to additional carbon emissions that would partially, fully, or even more than offset the additional carbon stored locally. In contrast, an afforestation or reforestation project on marginal lands would produce carbon storage benefits with little risk of leakage. Assessment and Valuation of Forest Ecosystem Services: State of the Science Review For any given management plan, the value of carbon should be assessed on the carbon expected to be stored net of leakage. Murray et al. (2004) investigate the extent of leakage in the United States associated with forest set-asides (prohibitions on harvest), avoided deforestation for conversion to agriculture, inducements for afforestation, and an integrated afforestationavoided deforestation program. They find leakage rates that range from minimal (<10 percent) to enormous (>90 percent) depending on the activity and region. Further, for small projects, leakage is usually small in absolute terms but larger in proportion to the direct project benefits compared with the leakage rate of larger projects. They conclude that leakage effects should not be ignored in accounting for the net level of greenhouse gas offsets from land-use change and forestry mitigation activities. Aside from the problem of leakage is the issue of impermanence in assessing the value of stored carbon. In the context of industrial production, the benefits arising from adopting a technology that decreases carbon emissions by one ton for a fixed amount of output are equivalent to the SCC. This equivalence stems from the fact that the ton of carbon that would otherwise have been emitted is permanently kept out of the atmosphere. However, in the context of forest and range land management, additional carbon storage may not be permanent. For instance, increasing timber rotation ages leads to greater carbon storage, but this carbon will eventually be re-emitted to the atmosphere after harvest. Managers can account for impermanence in one of two ways. The first is to value both the carbon sequestered (i.e., new, additional carbon stored) and the carbon emitted using the SCC at the time of sequestration and emission, discounting all costs and benefits back to the present (van Kooten et al. 1995). The second approach is to calculate a carbon rental value (Sohngen and Mendelsohn 2003). Conceptually, the rental value is equal to the interest one could earn on the proceeds obtained by selling the asset (a ton of stored carbon) at its current price (the SCC), minus any expected capital gains due to changes in the SCC. WATER REGULATION Forest structure and composition affect the quality of aquatic ecosystems, which in turn affect many different benefits, from irrigation water to clean drinking water (Table 1). We focus on the effects of forest management within the riparian zone and surrounding hillsides of the riverine system. We divide the section according to four environmental drivers of aquatic ecosystem health: flow regime, thermal/light inputs, sediment flux, and chemicals, nutrients, and pathogens. Flow Regime Flow regime refers to the quantity, rate, timing, and pathways of water through the watershed. It is characterized by base flow, seasonal timing and annual variation, frequent (e.g., 2 year) floods, and rare or extreme (e.g., 100 year) floods. Floods and droughts create a patchiness of riparian landscape important for variation in species and age class of species. In semiarid regions, extreme floods bring large wood into riparian zones and rivers; in wet regions, large floods add wood to rivers by eroding banks and causing trees to fall into the channel (Naiman et al. 2008). After entering the channel, large wood helps retain organic matter, forms deep pools, and promotes nutrient uptake in the river. More frequent small floods serve to flush large wood and sediment down the river and eventually out of the system (Latterell and Naiman 2007). Flow regimes can be affected by diversions, dams, stream channelization, timber harvests, and wildfire. Riparian and aquatic species have adapted to these natural flow regimes, creating locally distinct habitats (Lytle and Poff 2004, MacDougall and Turkington 2005). In the Pacific Northwest, the lives of salmon are lockstep with the hydrograph: high flows in fall cue spawning and create the necessary spawning habitat; baseflows in the dry season maintain juvenile habitat; and elevated flows in spring improve emigration out of the river. In the snowmelt-dominated streams of the Rockies, willows and cottonwoods release their seeds during the recession of spring floods when seeds find scoured ground and moist substrate needed for germination (Scott et al. 1997). Maintaining natural flow regimes on forests is an effective means of Assessment and Valuation of Forest Ecosystem Services: State of the Science Review 11 managing invasive species, providing adequate habitat, and sustaining human uses of water on the forest and in downstream communities. Two main methods are used for studying the relationship between changes in forest management and changes in flow regime: paired watershed studies and forest hydrology models. Paired watershed studies compare the flow regimes of two or more watersheds with similar physical characteristics (climate, soil, etc.). During the study, one set of watersheds undergoes a management action (e.g., clearcutting, prescribed fire) and one set of watersheds remains undisturbed to serve as a control. Ideally, comparisons of changes (or lack of changes) between watersheds allows one to tease out effects of land cover, management actions, and disturbances on changes in the flow regime. Such studies have shown a wide range of effects from clearcuts and partial cuts of forested watersheds on streamflow. Beschta et al. (2000) looked at three small watersheds and six large basins in the western Cascades (H.J. Andrews Experimental Forest). During the period of study, two of the three small watersheds experienced typical forest management actions, including road building, clearcutting, cable logging, and site preparation. The third watershed was left undisturbed and served as the control site. They found peak flow increased 13-16 percent in the treated watersheds for 1-year recurrence interval events and 6-9 percent for 5-year recurrence interval events. Swank et al. (2001) examined changes over a 20-year period in a mixed hardwood covered watershed in the southern Appalachians (Coweeta Hydrologic Laboratory) following clearcutting and cable logging, compared to an untreated control watershed. They found annual flow increased 28 percent during the first year following logging but continued to decrease until virtually no effect was seen after the fifth year. Brown et al. (2005) reviews other paired watershed experiments that look at changes in water yield resulting from alterations in forest vegetation. Most s