This section includes excerpts adapted from the paper “Best Practices for Integrating Ecosystem Services into Federal Decision Making.”
A large body of literature explains how to quantify changes in ecological conditions.1 But these analyses alone are not sufficient for an ecosystem services assessment because they typically focus on ecosystem processes or features (e.g., net primary productivity) rather than on benefit-relevant endpoints. In fact, there is far more literature on ecological assessments than on changes in ecosystem services. Once agencies and other organizations start using ecosystem services assessment, the literature on such assessments should expand and mature.
When assessing or monitoring the ecosystem service outcomes of an action, perhaps as a performance metric, a direct measure of a benefit-relevant indicator (BRI) can be used. In contrast, predicting changes in the provision of services resulting from management or policy actions (a necessary step for preference evaluation) involves converting the conceptual model depicted as a causal chain into an operational empirical model. There are several ways to measure the relationship between an action (policy, project, management) and its effect on the production of services. These methods differ in the time, resources, and capacity required. A narrative description of changes in ecosystem services could take the least time and resources, but it would not meet the minimum best practice requirement proposed for ecosystem services assessment because it is neither repeatable nor comparable, nor is it readily used in valuation or decision analysis methods.
Informal and formal methods of expert elicitation (e.g., Bayesian belief networks) can be used to generate quantifiable causal chains, including estimates of uncertainty. Alternatively empirical methods can be used. The empirical method that is likely to take the least time and resources and that meets the proposed best practice is to use existing models or to derive new models that use available data (collected by the agency or by others) or well-established relationships from the literature. For example, in the wetland restoration example (Figure 2), a study of fish mortality and reproduction that collected data on the effects of wetland restoration in a similar region could be used to estimate the proposed project’s effect on services. Likewise, the health effects of smoke from fires might be estimated using a concatenation of several models (fire intensity from a fire behavior model, smoke production from fire intensity, a plume model for the airshed, and so on). Again, the overall uncertainty of the full model would reflect the concatenation of models (and error propagation) as well as the uncertainty arising because the models would likely not reflect conditions at the study site.
It will often be the case that an ecosystem services assessment will be based on models derived from secondary data because primary data collection is not always possible.2 Clearly, these models could be improved if they were based on data collected in the study area. The gold standard for these assessments would be a model generated on-site or within the study region, based on manipulative experiments using the management actions being evaluated and explicitly measuring outcomes in terms of the BRI (and any intermediate variables needed to build the model). This approach works well with adaptive management, in which management treatments are implemented as experiments (with controls) and outcomes are monitored over time. In this case, the measured outcomes would empirically support the BRI, and the result would be a local model that explicitly translates the management action into its ecosystem services outcomes. Clearly, this approach is ambitious. But because adaptive management is a stated ambition of most, if not all, federal land management agencies, this aspiration is consistent with agency missions.
Many federal decisions use descriptive or narrative information to describe changes in ecological conditions and ecosystem services resulting from possible actions (e.g., Environmental Assessments and Environmental Impact Statements for NEPA). Narrative information can provide context for creating well-defined measurement scales. But narrative information is difficult to evaluate and cannot be used in preference evaluation (e.g., economic valuation) or tradeoff analyses. Narrative information is also not easily reproducible or testable in the same ways as information expressed using a well-defined scale. Given these limitations, descriptive narratives alone do not qualify as minimum best practice for ecosystem services assessments.
In most cases, however, it is relatively easy to transform descriptive narrative data into well-defined categorical or quantitative data that can satisfy minimum best practice. For example, descriptive measures can be transformed into a binary measure of presence and absence, or a categorical measure, or a continuous quantitative measure. Quantitative and categorical measures of ecosystem services will make the services easier to evaluate intuitively and to incorporate into formal valuation or tradeoff analysis, making the services more likely to be fully considered in decisions.
For measurements to be effective, their scales must be defined clearly enough to be applied by different users and to different decision contexts with consistent results (e.g., they must be repeatable). Numerical measurement scales, whether continuous (e.g., board feet of merchantable timber available from a specified land parcel) or discrete (e.g., numbers of deer taken by recreational hunters during a specified period of time from a specified geographic region), are the most obvious scales, but some categorical measurement scales can also meet these standards.
Categorical measurement scales can be used when numerical scales would be inappropriate or when estimation using numerical scales is too difficult or too expensive. An example is a scale describing degree of preservation of a tribal cultural site (e.g., “destroys a specified cultural site,” “preserves the site but prevents access by tribal members,” and “preserves the site and permits access on specified days”). In another example, categorical data may simply reflect presence or absence, such as the presence/absence of a particular listed species in a specific geographic area during a specific period of time, as determined by an agreed-on detection method. Other types of categories might reflect key thresholds or officially defined categories—for example, whether a population is considered endangered or threatened according to established guidelines. Thresholds between categories need to be defined clearly to provide reliable results. Scales such as “low,” “medium,” and “high” fail to meet this standard of clarity, unless such terms are clearly linked to well-defined thresholds. Categorical measures of BRIs must be defined using a scale that is unambiguous, measurable, and replicable to meet best practice guidelines.3
Identification and quantification of those people who could benefit from an ecosystem service—beneficiaries—involves defining the serviceshed and flows of services (Figure 1).4 For a locally used service like municipal water supply, the serviceshed is easily drawn around those using water within the watershed downstream of the policy or project action. For a service used or appreciated by a broader or spatially distributed group of people, like recreational use or cultural appreciation of a particular location, the serviceshed would include the area providing the service and its connections to those using or appreciating the service even if they live scattered about the region. Decision makers need to know not only where these people are but who and how many they are and whether they are affected by potential changes in the provision of services (e.g., reduction in flood or fire frequency or intensity). In the absence of a primary study or other direct means to identify the distribution of affected individuals (e.g., a survey conducted using a random sample over the potentially relevant area), indirect methods may be used.
Figure 1. Hypothetical serviceshed boundaries
Source: Tallis, H., C.M. Kennedy, M. Ruckelshaus, J. Goldstein, J.M. Kiesecker. 2015. “Mitigation for One and All: An Integrated Framework for Mitigation of Development Impact of Biodiversity and Ecosystem Services.” Environmental Impact Assessment Review 55: 21–34.
Note: The serviceshed for recreational fisheries is determine by the accessible lakes (or rivers) with harvestable recreational fish species that are within an acceptable travel time of people. Lakes 4 and 5 are outside the example serviceshed because they lack physical access or are too far away, respectively. Lake 3 is within the potential serviceshed area but is protected, so lacks legal access.
Although indirect methods are almost always less accurate than direct methods of identifying affected individuals, they can provide sufficient insight for many purposes, particularly when direct methods are infeasible. For example, data from the U.S. census or large-scale surveys like the National Survey on Recreation and the Environment, and perhaps information on what people purchase (e.g., fishing gear or bird identification guides), can help identify and quantify affected people.5 Small reductions in nitrogen oxide and sulfur oxide pollution can have significant health benefits over large areas, which can be characterized from air movement patterns. Direct engagement and outreach with communities and community groups, along with social media and surveys, can also help identify and determine the size of affected communities. A considerable economic literature is devoted to determining the “extent of the market” for ecological benefits (or where benefits occur); this literature details approaches that can be used for various types of applications.6
Best Practice Questions: Measuring BRIs
To follow best practice, the assessor should be able to answer yes to ALL of these questions:
Benefit-relevant indicators (BRIs) need to meet two criteria—(1) reflect changes that are relevant to beneficiaries and (2) capture some aspect of intensity of use and physical and institutional access where relevant. However, some BRIs are better at reflecting the most relevant information about an ecosystem service than others. The best BRIs will indicate a highly certain link between the environment and a human benefit and will also indicate the intensity of human use or enjoyment.
BRIs that capture biophysical outcomes as close as possible to human use, enjoyment, or appreciation are preferred. As causal relationships are established between management or policy actions and various ecological outcomes, indicators along the chain of ecosystem services production can be distinguished on the basis of their distance or proximity to social outcomes.
Consider the following causal chain arising from restoration of a wetland (the policy action) (Figure 2): (1) wetland restoration affects nitrogen levels in surrounding waters, (2) those nitrogen levels affect the water’s oxygen levels through algal blooms, (3) oxygen levels affect fish mortality and reproduction, and (4) fish mortality and reproduction affect fish abundance in waters used by anglers. Measuring fish abundance in waters used by anglers is a BRI. Measuring wetland restoration is not, unless a tight relationship has already been firmly established between wetland restoration and fish abundance. BRIs that capture intermediate outcomes “earlier” in the causal chain are less desirable than BRIs that capture more final outcomes “later” in the causal chain because the earlier BRIs increase the number of links to be established to firmly anchor the measure to benefits. All else equal, therefore, it is preferable to develop indicators that capture “final” biophysical outcomes rather than “intermediate” outcomes.
Figure 2. Use of BRIs to assess the fishing benefits derived from wetland restoration
Note: Black text indicates an ecological assessment and indicators; red text indicates extension to an ecosystem services assessment and indicators, with ovals illustrating BRIs; and blue text indicates measures of social benefit and value.
The notion of final ecosystem goods and services (FEGS) is a concept used by a number of agencies. FEGS emphasizes the distinction between “final” and “intermediate” ecological goods and services.7 Final ecosystem goods and services are “components of nature, directly enjoyed, consumed, or used to yield human well-being.”8 Intermediate ecosystem goods and services are ecological processes, functions, structures, characteristics, and interactions that are essential to the existence of final ecosystem goods and services but are not directly enjoyed, used, or consumed by beneficiaries.9
In some cases, the links between intermediate ecosystem goods and services to final ecosystem goods and services are well established, and a measure of an intermediate ecosystem good or service can be used as a BRI. Carbon sequestration provides a good example. Carbon sequestration is not a final good or service because it is not directly linked to benefits. Rather, carbon sequestration is an input into climate regulation, which is linked to the severity of future climate change and its associated impacts. A large research effort has gone into establishing the causal links between atmospheric CO2 concentrations and climate change and between climate change and potential future damages (from sea level rise, changes in precipitation patterns, and so on) to derive estimates of the social cost of carbon.10 Insistence on measuring only final ecosystem goods and services would not allow measurement of carbon sequestration and the social cost of carbon as an approach. However, it is only because the work has been done to link carbon sequestration to benefits through the social cost of carbon that carbon sequestration is an acceptable BRI.
Summarizing the above, BRIs are related to but not the same as FEGS. One of the key features characterizing BRIs is that BRIs are clearly and measurably relevant to human welfare. Hence, all FEGS can and should be measured using BRIs. However, some things that are not FEGS may qualify as BRIs, if causal chains are sufficiently well developed to link those things clearly and measurably to welfare. An example is an intermediate service for which links to FEGS are well established. Hence, from a conceptual perspective, all FEGS are BRIs, but not all BRIs are FEGS.
Measurement of BRIs often involves considerable uncertainty. The complexity (length) of the causal chain amplifies uncertainty because information loss occurs at each link of the chain. For example, we will be more confident about the impact of restoration on nitrogen concentrations than we will be about oxygen content and about fish population demography, and still less confident about numbers of catchable fish (fig. 7). Similarly, we might expect some uncertainty about human health impacts of smoke from fires because of the propagation of model uncertainties about fire behavior, smoke production, plume dispersion in the airshed, and human response to smoke exposure. It is worth underscoring that one advantage of using causal chain diagrams is that they facilitate communication of these uncertainties.
Another source of uncertainty in BRIs arises from the difficulty of measuring impacts in relevant terms. For example, an ideal measure of “catchable fish” would be provided by an empirical stock assessment. In the absence of such information, for commercial fishery we might index “catchable fish” directly from commercial landings. This measure is confounded by fishing effort, and is hence not an ideal indicator, although related measures (e.g., catch per unit effort) can at least partially address this problem. In other (noncommercial) instances, we might have to be satisfied with estimates of fishing success derived from fishing permits, visitor days, or some other measure imperfectly related to actual numbers of fish caught. Use of proxies should be accompanied by an estimate of confidence in the accuracy of the proxy estimate, and recognition that some proxies are superior to others. When choosing the most suitable BRIs for any particular application, analysts may need to balance the direct proximity or relevance of the measure to benefits with the ability to obtain accurate information.
Additional information about the importance of a service is added when information about the intensity of use or enjoyment of a service is available. For example, knowing whether the affected waters in the wetland example are the most popular fishing areas in the state given their accessibility (averaging 100 people per day during the season) or are highly prized for their beauty but somewhat isolated and used by fewer people (10 people per day during the season) would be helpful. Data on fish mortality and reproduction can be a sufficient BRI, but number of fish caught would provide some information about the intensity of fishing, making that measure a better BRI. This aspect of a BRI is nicely illustrated by the example of the health impacts of smoke from fire, by developing a BRI explicitly in terms of exposure (how many and which people?) and hazard (how bad is the air?).
Even “better” BRIs leave out information about people’s preferences. For example, people may value a service or good more if it is scarce, if it has no substitutes (other ways to gain goods or services that provide similar benefits), or if it does not require many other inputs (or complements) to produce benefit. Although BRIs often lack measures of these components of benefit, they represent a significant advance beyond ecological assessment alone. When feasible, benefit-relevant factors and benefits assessments can be used to capture these additional components of benefit. Indeed, BRIs might be viewed as nearly ideal inputs into more formal valuation methods because they are already in appropriate units and the relevant stakeholder populations are identified in at least a preliminary way.