Making a market in environmental credits I: Streams of value

Abstract

This pair of papers examines and describes the state action necessary to make markets function as environmental policy instruments and as strategies of governance. They do this through a detailed look at the mechanics of environmental credit compliance markets in the US states of Oregon, Ohio, and North Carolina in which stream credits are privately created and sold to developers who have impacted protected stream systems. In this paper, we examine the tools, techniques, and people involved in the creation of a value-bearing stream credit out of a physical stream or river site. These observations reveal important principles of how science functions within governance, as well as where gaps and resistances appear that create unforeseen outcomes in market-led policy. We examine the construction and use of instruments that define natural processes as objects with value; these techniques and tools include databases and spreadsheets, algorithms, and field scoring tools that have been scavenged from a wide range of scientific and governance practices and are not themselves inherently capitalist or developed for capitalist purposes. In three different state settings, the move from measure to value is made in different ways that depend on the local institutional and social context. However, they all act to render a network of interacting ecological forces as a field of discrete ecosystem objects amenable to governance with markets.

Keywords

Neoliberal natures environmental governance water pollution trading water governance science studies

Introduction

This pair of papers examine and describe the state action necessary to make markets function as environmental policy instruments and strategies of governance. We consider the details of how such markets are created, examining closely the way that the US government created markets for environmental credits in stream ecosystems. The idea that market-led policies flourish through the reduction of state capacity is long since dispelled (Jessop and Sum, 2006; Brenner et al., 2010), and the active roles of the state require careful empirical study. We believe that looking closely at the mechanics of markets reveals how science functions within environmental governance, as well as where gaps and resistances appear that create unforeseen outcomes in market-led policy. We first examine the construction and use of instruments that define natural processes as objects with value, and then the variety of scales at which governance acts on ecosystems. We find that the use of reduction, fixing, and simplification operating on ecological flows and processes aims to produce a nature composed of discrete objects arrayed at discrete scales. Seen this way, the current efforts to govern nature with markets recapitulates the development of more conventional strategies of accumulation across capitalist economies.

Much is already known about how environmental credit markets work. We know that they require the state to define new objects to bear nature's value: ecosystem services, tons of CO²-equivalent, and nature-based financial derivatives have been explored for years in the critical and mainstream literature on environmental economics (Lohmann, 2006; Pryke, 2007; Randalls, 2010; Bonneuil, 2015; zu Ermgassen et al., 2019). We know that the state makes these markets by creating standard measures that translate between ecosystem science and the needs of policy (Robertson, 2012; Johnson, 2013; Nelson, 2014). This much and more has been established by the literature on “neoliberal natures” over the past 20 years. But in all such work, there is the danger that the power of the logic of capital or the seamlessness of its integration with state strategy has been overstated or under-contextualized. New questions have arisen about the finer-grained elements in which uncertainty and resistance are empirically observed. What is the role of the ecological practitioner in valuation and assessment? Are the tools of valuation and commodification essentially capitalist, or can they serve unforeseen ends? What scales of governance are activated to manage ecological systems, and who can act at these scales? Can modeling and algorithms replace field science, as they have done in climate modeling and hazards insurance? We use three cases of environmental market-making to describe how these questions are answered in practice.

In this paper, we focus on the tools used to quantify and assess streams and rivers for state and market, and which are used to define the stream credit as a commodity. These tools are not new valuation techniques developed for a novel project of governance; rather, they are mundane and long-standing scientific techniques. Further, these tools are used by individuals whose experiences with streams vary considerably. In exploring the relationship between people and technical practices of measure, we draw from the established literature on the social constitution of value through measurement in governance (Mutersbaugh, 2002; Tadaki and Sinner, 2014; Kitchin et al., 2016; Machen, 2018).

We begin our analysis in 2012, in Oregon, where the US government needed a new method to define stream systems as stream credits that can be bought and sold. The US Environmental Protection Agency (EPA) works with the US Army Corps of Engineers (hereinafter, “the Corps”) in administering the Clean Water Act, which regulates impacts on wetlands, streams, and other waters. EPA's research office in Corvallis began to look for experts to propose a satisfactory way of measuring stream impact and restoration sites in the State of Oregon.

The Corvallis office issued a memo to this effect (EPA, 2012); then, through a nonprofit, they issued a “Request for Informal Proposals” (RFIP) to contract the development of the desired stream assessment methodology (CWS, 2012). The ideal assessment method, they said, should produce numeric values for stream quality; these numbers would then form the basis of an accounting system from which the monetary value of the stream's functions could be calculated. Finally, this method would be used to calculate debits at sites in Oregon where EPA and the Corps allowed a stream impact, and to calculate credits at sites where developers restored streams as compensation for impacts:

Assessment protocols will describe how numerical values are derived for each function attribute—which is necessary to quantify each attribute such that it can be used in a mitigation ‘accounting’ system that measures impact (permitted action) and benefit (mitigation actions).

EPA Draft Functional Assessment Framework D, p6 (EPA, 2012)

The EPA's goal was thus to move from stream to stream credit, from a physical ecosystem to an abstract object that can bear value or serve as an object of governance (see Robertson, 2006, Poon, 2007, Nost, 2015, Amoore and Raley, 2017 on the creation of abstract measures for use in governance). Successful proposals would include familiar elements such as databases and spreadsheets, algorithms, and field scoring tools. This charge to develop a way of measuring stream systems within an accounting framework in Oregon provides our starting point for analysis, which we augment with consideration of similar policy efforts in two other US states.

We spent five years observing how practices of environmental valuation for streams are developed and tested by environmental regulators in Oregon, Ohio, and North Carolina (reported also in Lave & Doyle 2021). After a preliminary round of interviews with regulators across the U.S., we chose these states to conduct approximately 60 interviews with state and federal regulators, stream scientists, and environmental consultants. These three states illustrate different important moments in the development of stream credits as value-bearing abstractions: early policy development in Oregon, initial implementation in Ohio, and relatively mature use in markets in North Carolina.

We find, below, that the journey from stream to stream credit is not a single, simple step. It is a deeply technical and personal process that passes through many hands and lays open many points of engagement for critics and advocates alike. In this, it is perhaps more open than the process of abstraction by which climate models turn both coffeemakers and cookfires into Global Warming Potential or the measures of time and effort which turn a person's working day into a general measure of labor and value. But it is not essentially different.

Value from measure

We are observing a moment of regulatory transformation which has often been called “the neoliberalization of nature” (Bakker, 2005; Mansfield, 2008; Castree, 2008). The emergence of market-based policies that frame the environment as a set of commodity-like “services” or “credits” began in the mid-1970s (see Robertson, 2018). Early examples included airborne emissions and wetlands, while more recent efforts include weather derivatives markets and compliance mechanisms for global conservation treaties. In all of these instances, an object—be it a carbon credit, a drought forecast, or a stream credit—must be defined in a way that allows it to circulate and bear value (Marx, 1976). To do so it must have a stable identity over time and space: the part that bears value must not change as it transacts. We are thus discussing both physical streams and something else: the value-bearing object that is created through a social process. That object is related to, but is not the same as, those physical streams.

The word “value” is often thought to be vague and hard to use analytically, but our use of “value” here is simple and follows the work of Graeber (2002, 2013). A thing's value, in the most general sense, is its ability to be compared to something else. Thus value for a capitalist is determined by comparing something by price in a market context, and one of the tasks of governance in a capitalist state is to create the appropriate technical tools and social contexts for this to occur. But capitalist valuation sits alongside myriad other valuation regimes. Although values arrived at through ethical (“virtuous”), scientific (“1.34 kcal”), and capitalist (“$540”) regimes are incompatible in that they cannot exist within the same accounting scheme, they all still involve the comparison of objects. Only things that cannot be compared are considered “invaluable” (see Bigger and Robertson, 2017).

We frequently create value-bearing abstractions from biophysical nature, for all kinds of reasons beyond or alongside the interests of capital. Examples in the geographic literature include the Normalbaum, the “standard tree” of scientific German forestry in the 1800s (Prudham, 2004) or the concepts of species or scale as a tool for environmental governance (Sayre, 2002, 2017). Abstractions like “the gallon” (say, of gasoline) can circulate without being constrained by the complex material relations which create them: the energy infrastructure, the geology of the shale oil deposit, and the politics of ethanol production. This kind of value measure has been an integral part of US environmental management since the classification of ecosystem qualities and assessment of impacts began in earnest in the 1950s (USFWS, 1953; USGS, 1971). Observing the development of new techniques for creating such abstractions gives us a chance to see this often-hidden work of environmental governance on display.

Our analysis of stream markets follows on a large body of literature that has analyzed how an environmental feature like biodiversity or the climate becomes a commodity through practices that distill a metaphysical value from a biophysical object (McAfee, 1999; Bakker, 2005; Robertson, 2006; Johnson, 2013). This is often presented as a neoliberal or capitalist process, and in the policy-advocacy literature we see a variety of hyperbolic claims that market-led environmental policy will reveal the “true value” of nature (cf. Juniper, 2014; Whitworth, 2015). And indeed, working markets for stream credits do exist in Oregon and elsewhere. However, the purposes at work in creating a system of stream assessment are not always clearly capitalist (Lave et al., 2010; Lave, 2012: Doyle et al., 2015), and the techniques of valuation are borrowed from long-established traditions of measuring values in the context of ecological or hydrological science. The interests of capital may be what Jessop (2000) calls “tendentially dominant,” meaning that capitalists are quite good at turning a variety of tools and settings to the tasks of governance and accumulation. But in this paper, we see this done more in the way of a scavenger than a creator. In these three states, we look at how regulators, entrepreneurs, and scientists respond to requests such as EPA's in the RFIP and build a stream of value from a stream of water.

The problem with streams

Streams present special challenges for the process of creating abstract measures of value, a process that has been arguably simpler in carbon and wetlands. Development projects that damage stream systems are very common and usually require a permit under the US Clean Water Act. Clean Water Act regulators encourage developers to avoid floodplains and channels, but “unavoidable” impacts often remain. For much of the 1970s, ‘80s and ‘90s, the general uncertainty concerning how to assess and restore a damaged stream system meant that regulators would typically approve development projects without requiring any stream restoration activities in return, deeming stream restoration to be “impracticable” (Kruczynski, 1990; Owen, 2017).

By contrast, for projects proposing impacts to wetland ecosystems, the practice of compensating for unavoidable impacts by restoring a drained or filled wetland to its natural condition had become well-established by the mid-1980s (Hough and Robertson, 2009). Although wetlands are often treated as discrete parcels with clearly delineated boundaries, streams are conceived of as networks of material and energy that integrate entire watersheds (Wohl, 2012) such that one length of a network is not easily substituted for another; the questions about how to do so usually led to inaction. Ongoing, uncompensated damage to streams spurred the creation of a market in stream credits that could be used as compensation, mimicking the market in wetlands credits that had become established by 1995.

The practice of stream restoration originated at elite fishing clubs in the Catskill Mountains in the 1870s and spread widely through the US courtesy of New Deal work programs in the 1930s (Thompson, 2013). This initial work focused entirely on creating fish habitat within existing stream channels, but in the 1980s practitioners began to experiment with reconfiguring the entire channel in cases where it had been straightened, or where upstream land use changes had resulted in rapid erosion (Doyle 2018). By the mid-1990s, this approach was sufficiently well-established that a handful of restoration practitioners began to argue that regulators could require meaningful improvements in stream systems when granting development permits (Rosgen, 1996). By the end of the 1990s, regulators began to experiment with requiring stream restoration as compensation for permitted impacts to streams. They began to develop criteria for assessing the quality of stream restoration projects (Owen, 2017).

States, tribal governments, and the 38 Corps and 10 EPA regional offices were all involved with developing both stream assessment protocols and stream mitigation policies on how to compensate for lost streams. Some did so diligently, others haphazardly, and many not at all; the result was a poorly coordinated and geographically diverse proliferation of methods and policies (Somerville and Pruitt, 2004; Somerville, 2010). The three cases we examine reveal substantially different experiences:

The state of Ohio began creating stream assessment criteria by adapting from the Savannah (Georgia) Corps District policy around 2002. An interim set of criteria issued in 2004 was due to be finalized in 2010, but the incoming Kasich Administration retracted the draft document and issued a substantially changed assessment method (Ohio EPA, 2012). Investment in assessing Ohio's streams, however, goes back to the 1970s when Ohio used large grants from the US EPA to conduct extensive biological assessments of every stream in the state, resulting in one of the richest datasets of stream biology anywhere in the U.S. (see following paper). The development of assessment criteria in Ohio thus occurs against the backdrop of this wealth of public data on stream ecology.

Although the state of North Carolina did not develop formal stream assessment criteria until 2003, regulators there had been using ad hoc approaches to assess streams since the mid-1990s. In 1998, North Carolina became the first state to require that the Department of Transportation mitigate impacts to streams by restoring other streams rather than by restoring wetlands or doing nothing; this introduced a source of demand for stream restoration credits (Owen, 2017). Since that time, North Carolina has revised its stream assessment protocols several times, most recently via a method known as NCSAM (North Carolina Stream Assessment Method) (NCDENR, 2015), a rapid assessment technique based primarily on visual assessments of stream conditions as proxies for ecosystem function.

In the state of Oregon, a statewide stream assessment methodology began to be assembled in 2011 after the passage of a state law requiring all compensation for environmental impacts to be accounted for in terms of ecosystem services. The product of the US EPAs RFIP, discussed above, is version 1.0 of Oregon's Stream Functions Assessment Method (SFAM) (Nadeau et al., 2018). However, Oregon's history with stream assessment pre-dates this RFIP by decades: since the listing of several salmon runs under the Endangered Species Act in the 1990s, Oregon regulators have become adept at creating standardized measures for impacts to streams, but these were not used to calculate compensation credit requirements.

In each of these regulatory contexts, the state government had to develop a set of criteria that allows the standardized measure of a stream's qualities. But, in addition to satisfying state policies and protocols, the measures also had to define a credit: a representation of the stream able to bear economic value. This is because, in each state in our study, entrepreneurs wanted to build stream restoration sites that generated stream “credits” to be sold to developers who now compensate for impacts to streams. These entrepreneurs were known as “stream mitigation bankers.”

This task was accomplished differently in each state we examined, but in all cases, scientists, regulators, and entrepreneurs worked together to build a valuation system. Through the remainder of this paper, we describe how they used tools such as algorithms, concept curves, ratio multipliers, and spreadsheets. These are tools that are not, in their origin, capitalist, and yet they are now central elements of attempts to describe streams as something that can bear value.

Governing with algorithms

As a scientist there just doesn’t seem to be a ton of opportunity for learning [through developing stream assessment tools]. It seems like it's more of an accounting problem.

Oregon stream scientist

Imagine a stream scientist responding to the EPA's request; they must attempt to create a way of describing a stream site that will be usable by people who are not standing at the bank of the stream. Nor will they usually be stream scientists. The scientist wants their description to become the accepted way of attesting how streams bear value: their way of understanding what is important about the stream must become socialized, accepted by people who do not understand streams as the scientist does. Fortunately, they have many tools at their disposal. One such tool is an algorithm showing how a stream's water quality or biodiversity is related to underlying measurable variables.

An algorithm is a set of instructions for relating observations of a system, usually based on a model of how the system is supposed to behave, a model embodied in concept curves (about which more below). Here we will use an example from the Willamette Partnership's Salmon Calculator (Parametrix, 2009), which was developed by a Portland-based consulting firm called Parametrix for use by the State of Oregon. Parametrix's method defines the “salmon habitat” credit as a stream function derived from the measurement of many stream parameters. The overall “salmon habitat” score for the stream reach being assessed (MUfp) is given by this equation¹:

M U f p = (\frac{A F + I I + H F + T R + S S + V V + C D}{n F})

In this equation, we see variables representing seven subfunctions that are added, and an eighth variable that acts as a divisor (the meanings of all variables are provided in Footnote 1). Each of these eight subfunctions is defined in turn by its own algorithm. For example, the ability of the site to provide anadromous fish habitat, the variable AF, is scored using this equation:

A F = (\frac{C / R + F + N / S + C}{4})

In this equation, there are six variables defining AF, one of which is F, a variable representing the “foraging” score for the site. It, too, is defined by an equation, one composed of seven different variables, two of which are used as a multiplier or ratio:

F = (\frac{L W V + (T A q V e g \times N N v e g) + P D p t h + U c B + A q S u b}{5}) * H mod

H mod and NNVeg are multipliers between 0 and 1, but the other variables are scores derived from data gathered in the field. One of these variables is LWV, Large Wood Volume, expressed in units of cubic feet of wood per linear stream foot (dead branches and trunks in streams can be an important element of salmon habitat). The method does not require the field assessor to directly measure the volume of wood in a stream. Rather, a visual estimate is performed in which the assessor only has to select between six categories of large wood volume cover:

Medium Stream Large Wood Volume (ft³/ft²) Score

Not Present	0.02-0.061	0.062-0.110	0.111-0.202	0.203-0.452	>0.452
0	6	7	8	9	10

The resulting numeric score is then inserted into the equation. Is it reasonable to expect an assessor to distinguish between 0.061 and 0.062 ft³/ft² of wood in a stream? Reasonable or not, it is easily more practical than training them to understand and describe the many ecological, geological, and hydrological complexities of a stream system. And it grounds the value of the stream credit in an observation at the physical site of the stream.

If, for example, the assessor observed 0.1 cubic feet of wood per linear foot of stream, they would insert a score of 7 for LWV. But LWV is then added to UcB, the Undercut Banks Score, which is a score between 1 and 10 based on an observation of distance in inches. These observations involve different dimensional measures; ft³/ft is added to inches, but are both expressed as dimensionless categorical variables.

The same operation is occurring with all of the other equations all the way up to the top-level equation for MUfp, the Salmon Habitat Score for a given map unit. Because these numbers are ordinal and dimensionless, the algorithm that associates them violates a principle of mathematics by which one must not conduct arithmetic on ordinal data (thus, while 5 + 8 = 13, “fifth plus eighth” is meaningless). Among practitioners, this violation is regarded with both humor and derision, but it occurs, and is allowed to stand as an adequate numeric abstraction of the site. Assessment tools like the Salmon Calculator are all patterned generally after the first widely used state-sponsored habitat assessment, the 1980 US Fish and Wildlife Service's Habitat Evaluation Protocol (HEP). The HEP authors were very clear that using their algorithms, one plus one may not equal two: “The rules of ratio mathematics will not necessarily be upheld by this approach. However, with some interpretation the resource manager should be able to develop a reasonably sound set of scores” (USFWS, 1980, 6-6, emphasis ours).

One reasonable response is to say, like this Oregon stream scientist discussing algorithms, that the use of such an algorithm can only create error:

I think in the effort to try to bring all the things you think are important … you end up with a mash-up of these things that really is not terribly helpful. … we'll do a bunch of mashing and then we'll relate the mash-up values to some other thing we can actually measure like sediment.

Oregon stream scientist

The algorithm's “mash-up” has allowed its users to make simple value statements about a stream from real-world measurements and visual assessments. Its arithmetic deficiencies provide a practical way to include the relationship between woody debris and salmon health in stream assessment. The valuation techniques we observed begin with measurements in the field but immediately combine those measures in ways that are, from a strictly scientific standpoint, impossible to defend. Once they leave the context of science, however, they stand as a measure of value. Let us now look at how the purported relationships between measures in algorithms are defended.

Conceptual curves

As we noted above, algorithms are based on models of how a system is supposed to behave. When a measure of large wood volume is used to quantify the value of salmon habitat, the user might assume that this relationship is based on a scientifically documented understanding of the effect of wood removal on fish populations. But in many cases, the underlying relationship is educated guesswork. Conceptual curves—the tracing of a curve showing the hypothesized response of a dependent variable to changes in an independent variable—have long been used to show relationships between phenomena that researchers think might exist. Effectively, these curves are a visual depiction of a hypothesis. They can be a key, even essential, part of developing a research program and framing a question that can be investigated.

The curve shown here (Figure 1) comes from reports submitted in support of the 1994 Northwest Forest Plan (NWFP), a high-profile management strategy emerging from the “forest summit” organized by President Clinton in 1993 to resolve conflicts between the logging industry and environmental regulations in the Pacific Northwest. The NWFP assembled 97 scientists in the Forest Ecosystem Management Assessment Team (FEMAT) to recommend stream assessment techniques, for the purpose of providing a science basis for difficult policy decisions concerning timber-based livelihoods and protected species habitat. FEMAT's report was not crafted as a tool for describing a stream credit commodity, but it has nevertheless been scavenged for that purpose. As an important policy document authored by a large team of scientists—including 18 river scientists—it has the credibility of a scientific report and uses the tools of ecosystem science in recommending forest management strategies. The concept curve shown above was used to suggest a relationship between the width of buffer habitat—that is, trees left standing on the banks of a river—and the improvement of “riparian ecological functions” such as contributions of leaf litter, shade, and large wood. It suggests that if a wooded buffer extends half a tree-height from a stream channel, the effectiveness of the buffer in shading the stream is approximately 70% of the maximum. In the context of the NWFP, it was used to make a policy argument: logging practices should be changed to make sure a wooded buffer is left standing alongside waterways.

Figure 1.

A curve relating various “ecological functions” to the width of a riparian buffer, taken from the report submitted as part of the draft environmental impact statement for the 1994 Northwest Forest Plan (FEMAT, 1993).

The chart is conceptual: the curves representing “percent of root strength effectiveness” or “percent litter fall” do not come from field data and are not rigorously quantified. To a scientist, it is intended to be suggestive, indicative of important questions that can be answered empirically. This particular diagram was identified as particularly influential by a stream scientist we interviewed, who described it as “one of the most famous and least quantitative things that ever showed up” and “the diagram that launched a thousand concept curves”:

So this is the entire riparian protection scheme for 24 million acres, is what is driven out of this graph. … This entire thing is ginned up… there's no data to support this, there still isn't. This is an instinct. … [I]t's not an uninformed instinct. But it's an instinct nonetheless. But I suspect a lot of your functional relationships are of this sort.

Oregon stream scientist

This graph can stand as the type specimen of the many concept curves that appear in environmental assessment protocols, in which the x-axis is a physical measure, and the y-axis is a dimensionless ordinal or category variable. One stream scientist referred to such diagrams as “conceptual curves that are used as if they were rigorously constrained by data.” A North Carolina stream scientist bemoaned how common it is to see such measures in stream assessment: “…in almost every case data were not collected. We just assumed that if you cut back the bank it is not going to erode, or the sediment's suspended solid concentrations will decrease. Well if you don't have the data you don't know that.”

Because of its importance in Oregon, the NWFP gave a license to graph measurable physical phenomena against a y-axis of service “effectiveness” without any (or very little) data to support the curve. One developer of stream metrics in the Northwest described at length their process of drafting such conceptual models:

Once the person developing the function is comfortable with the ranges that we're collecting the data in, then we ask them to just draw a graph that's saying, as that attribute increases, how do you see the performance of the function go? Is it a steady increase? Is it some kind of bell curve? You know is it, kind of, comes up a little bit then levels off? Just have them draw a general curve without any numbers associated with it. And then, once we do a round of review on that with several other people, then we ask,… now can you put some number between 0 and 10 on the y-axis?

Oregon environmental consultant

Although these examples are from the Pacific Northwest, this is a widespread phenomenon: one can find graphic or textual varieties of this model in most techniques that translate physical features into “services.” A generalized model for doing so was presented in the webinar that introduced the final Oregon SFAM in 2018 (Figure 2).

These graphs may be step-functions, linear or sigmoidal, but they are all conceptual models in which the relationship between the measurable phenomenon and the service provided has not been derived from field data. The movement from the x to the y axes is a key movement from physical measure to value-bearing abstraction.

Figure 2.

Slide 15 from March 2018 webinar from USEPA and ODSL introducing the new Oregon Stream Functions Assessment Method (Nadeau and Hicks, 2018).

Departing from accuracy

The point of using algorithms and concept curves, or any other sort of rapid assessment tool, is to take complex biophysical relationships and simplify them to make them easier to understand, estimate, and compare. Our stream scientist is under no illusions that their algorithms and concept curves capture everything about the stream system: they must work for the purposes of the people using them, and the scientist knows they must leave much out. Mark Brinson, a giant in wetland policy and developer of the first systematic method for assessing wetland functions individually (known as the Hydrogeomorphic Method, or HGM) in the 1990s, was frank and clear on these limits:

While some extrapolation is required, it is minor relative to the many extrapolations and inferences that we use on a daily basis in natural resources management… A wetland scientist working in the Pacific Northwest who cannot convincingly make this transition from science to policy for coarse woody debris and fisheries (and thus the societal value arena) should be advised to seek an alternative discipline.

Brinson (1995), 12

In short, there are so many uncertainties in the policy arena that a lack of clarity on the exact relationship between wood and salmon does not rise to the level of a blocking concern. Those for whom it does, says Brinson, are not serious about informing policy through science.

Thus, in the building of abstract representations of stream systems for regulators, accuracy is not a priority. In the Oregon RFIP, EPA's main concern is whether measurable variables (like volume of wood) have a demonstrable relationship to stream quality, rather than a fully characterized one. The RFIP specifies that “the successful assessment method” will rely on attributes with four qualities, none of which is “Accuracy”:

D—Direct Measure—attribute assessed is relevant to anticipated impacts and to proposed restoration objectives for mitigation.

Q—Quantifiable—attribute is measurable, can be repeated consistently by different people with the same results, and with minimal observer bias and sampling error.

S—Sensitive—attribute is responsive to both impacts and mitigation/restoration actions; response can be correlated with actions and can indicate a defensible trend within a practical timeframe (5–10 years) using available technology and tools.

P—Practicable—attribute can be measured within a regulatory context given budget, effort, and time consideration (EPA, 2012, 5)

These qualities depend on the assessors and their social context rather than on stream science: a stream attribute only becomes part of the stream credit's value if it is practicable to measure it. Like wood in streams, any attribute used to measure stream health should be “measurable,” “relevant to anticipated impacts,” able to be “correlated to actions,” and “indicate a defensible trend.” What is missing here is also missing from the NWFP conceptual curve: a relationship between observation and outcome explicitly based on data and field study.

It is actually very helpful, we argue, for this relationship to be obscure or missing in the project of successfully bridging science and policy. For the regulator, accuracy becomes a liability if pursued too assiduously in assessing the resource being regulated. Like all measures, scientific or otherwise, concept curves and algorithms address elements of the stream that make it comparable to other streams. But they create valuations that are specifically unconstrained in the way that scientific claims usually are constrained: by the rules of mathematics and the conventions of validation and proof. Constraint on the legitimacy of the value is, instead, political or economic. Scientific measures provide the tools of valuation, but scientific principles do not constrain the uses to which the value is put. This can be seen in the use of bank height ratio as a metric in North Carolina's stream assessment:

Ideally your bank height ratio is a 1. But If your bank height ratio is a 1.2 or a 1.15, what does that mean in terms of the function of that stream? How much is it impaired? And I don't know that there's anybody who has kind of been able to come up with a way to say that that stream is 15% impaired or 20% impaired, and how do you then go back and tie that to the amount of credit that's required for impacting that stream? … One thing we're not doing with those parameters, at least not yet, is to tie them to performance standards, because that really wouldn't be fair at this point to require our providers to meet standards that may not be achievable, especially if we don't know that they're achievable.

North Carolina regulator

That is, it may be possible to create a measure for stream impairment using bank height ratio, but the regulator resists using that value directly to quantify the regulatory category of “impairment,” much less to quantify the amount of credit that can remedy that impairment. Here, “fairness,” not scientific “accuracy” or ecological recovery, is the criteria that mediates between measurement and value.

Ohio provides an instructive counterexample concerning the role of accuracy. Acquiring the field data to validate concept curves is an arduous and expensive task; whether or not this is done depends on the commitments of regulatory personnel. Ohio regulators prioritized showing a provable response relationship between attribute and outcome, between x axis and y axis. They considered it important that Ohio's assessment systems avoid the “untested simple logic models” (Ohio regulator, 21 June 2012) that characterize many functional assessments of streams and other ecosystems. In Ohio, those personnel would not consent to the legitimacy of the value measure unless it was grounded in the state's rich legacy of biological data (see following paper).

Likewise, in North Carolina, the state Division of Environmental Quality asked scientists to develop a system for measuring the success or failure of a restoration project, but testing showed the state's ecological metrics were not sensitive enough, so the project was eventually abandoned (Tullos et al., 2006; Tullos et al., 2009). That is, in some cases, conceptual models were scientifically tested for data-supported linkages between site conditions and stream system, and sometimes they failed these tests of validation. This challenged the overall effort to develop a valuation system to the point where it either collapsed (as in North Carolina) or began to prioritize accuracy (as in Ohio).

Departing from precision

Another technique commonly seen in stream credits policy is the use of ratios. It is common for regulators to impose a compensation ratio in issuing a permit, as a form of insurance against uncertainty and project failure. Thus with a compensation ratio of 1:2, the loss of 1 unit of stream quality in a development project may require the purchase of two units of compensation (King and Price, 2004; Bendor, 2009). For regulators, this hedges against the possibility that the restored stream credit project may not be providing all the functions that it is supposed to. In cases where the success of the compensation is in doubt, where long-term funding is not provided, or where the compensation is unusually distant from the impact, the compensation ratio might go even higher. A ratio multiplier is used as a kind of universal hedge to manage many sources of risk and uncertainty that may affect the compensation site. The effect is to further separate the value of the site from observations and measurements taken at the site. The precision of measurements in the Oregon salmon metric so carefully defined above is washed out by a coarse multiplier:

“Size of gravel” is an important component of overall stream function. But coho [salmon] need a certain type of gravel. All of them have very, very slight different preferences it seems. But when you start adding those in the rule curves, you're changing by like .003 percent on individual rule curves. I mean… you can spend the next ten years making sure you get the decimal point in the right place, but you're gonna swamp it with a 2:1 ratio.

Oregon stream credit developer 1

Once you apply a 2:1 or 3:1 trading ratio on top of a metric output you swamp any variation that might be caused by measurement error. And people are like, “Oh, ok, then I'm not gonna worry about it.”

Oregon stream credit developer 2

Algorithms on the relationship between gravel and salmon produce an answer that is often impressively precise: the Salmon Calculator may find that 905.34 square yards of gravel-bedded spawning ground are needed to compensate for a proposed impact on coho salmon habitat. But when this square yardage of required compensation is multiplied by 3, this precision gives way to overdesign. The state seeks an overabundance of certainty that compliance with the law has been achieved by requiring a high compensation ratio. Likewise, the producers of stream credits prefer a high compensation ratio because it allows them to sell more credits. Increasing compensation ratios (which can range from 1:1 to 10:1, or more) increase the total credits in circulation, thus thickening a market and increasing potential profit by increasing demand and reducing the turnover time of capital from a restoration project site. Regulators and entrepreneurs thus both have strong incentives to increase compensation ratios.

States routinely exercise this kind of arithmetic power on the values provided by stream assessment methods; in fact, they can be said to have an “appetite for imprecision” (Ghosh, 2019). For example, in Ohio, an impact on a high-quality stream will trigger a 3:1 compensation ratio. In North Carolina, the use of compensation ratios is widespread but more ad hoc:

…the first step is to do this assessment [of the impact site], and that gives you a feel for what the condition of the channel is and the overall functional value of the channel. But what it doesn't do is it doesn't tie into a specific ratio for mitigation. That's where the implementation of the method comes into play, and those are still connections that we're trying to make. So that's to say that we haven't finalized any kind of method for turning those scores into a specific ratio.

North Carolina regulator

As a technique used to build a regulatory market for stream credits, the use of ratios has obvious merits: ratios buffer state regulators against the accusation that compensation does not fully replace what is lost, or that they are being too lenient on developers, and they provide stream credit entrepreneurs with a market demand multiplier. But their use stands in stark opposition to the painstaking work of creating and refining response curves and algorithms. The state is managing the creation of value-bearing streams, but it is doing so using tools that work against each other and answer to different (and contradictory) requirements.

Value faster

At the 2012 River Restoration Northwest Conference, one participant gave a workshop on a software program meant to help the US Army Corps of Engineers make decisions about which stream restoration projects to pursue (Wilcock and Baker, 2012). The Corps had funded its development, recognizing the need for “decision analysis and design guidance” in the form of a simple-to-use system, built on the best stream assessment science, that could indicate which project design options would produce the most benefits. The program was developed in consultation with Craig Fischenich, the same Corps stream scientist whose stream functions classification system (Fischenich, 2006) served as the basis for the RFIP discussed above. It was an Excel-based program that required only the ability to use a spreadsheet and to take simple physical measures. The user simply inputs measurements into blank fields, and an array of lookup tables did the work of interpreting the meaning of those measures. About 95% of the cells in the Excel forms autopopulated based on the calculations (algorithms) embedded in the lookup tables.

Although the formal name of the product was Stream Project, one user attending the workshop said he thought of it as “TurboStream,” a name that played off the popular tax-preparation software TurboTax. Just as you don’t need to know the intricacies of the US tax system to calculate your refund with TurboTax, “TurboStream” implies that you don’t need to know the intricacies of a river system to produce a meaningful stream assessment value. The data inputs were designed to be, as one Oregon stream scientist said of such assessments in general, “the kind of thing you could send a GS7 out with a scoresheet” to fill out (a GS7 is a low/mid-ranking Federal pay grade). Thus, Stream Project allowed the Corps to accelerate and routinize their production of actionable information about stream credit sites² (Figure 3).

Figure 3.

A cell referencing a lookup table in the Willamette Partnership's Shade-a-lator, an excel program similar to Stream Project that generates a score indicating how beneficial a stream design will be at lowering stream temperature (TWP, 2014).

As Stream Project and similar Excel programs such as The Willamette Partnership's Shade-a-Lator in Figure 2 shows, Excel lookup tables encode knowledge, automating the calculation of credit values from the input of simple field observations and requiring minimal training.

The tools that abstract value from ecosystems are thus fast relative to the normal scientific practices of field ecology (Lave, 2012). As one Oregon regulator put it, when pressed about whether Oregon's stream assessment tools were sufficiently scientific,

… we've always used the term science-based. It may not be what a researcher would go out and measure but these are things that give us a clue as to what we're dealing with on a fairly rapid basis.

Oregon regulator, 7 February 2012

The term science-based acknowledges the distance being traveled from the measurement of the stream to the value of the stream credit. This move was not only about unvalidated concept curves or arbitrary ratios, but about the practical limits of regulators’ time and expertise. In North Carolina, data collection at a stream site is designed to be breathtakingly brief from a scientific perspective: “We defined rapid as taking no more than fifteen minutes. Because that is how much time we have.” (North Carolina regulator). By comparison, users of the ultimate product of the Oregon RFIP will have a comparatively luxurious stretch of time: “We also state—unambiguously—in there that it needs to be something that can be done by a two-person team in two days for one reach with two days of training” (Oregon environmental consultant). The Corps regulators who oversee the CWA compensation program are overstretched (GAO, 2005) and lack the personnel to effectively oversee the program at current staffing levels:

There's only so much time we can spend on each project. Our project managers may be looking at four or five sites in a day, and each site may have eight to ten streams that they're having to evaluate, you know what I mean? You try to imagine running an assessment that requires you to take out rods and tape measurers and it's just not even—it's not going to happen.

North Carolina regulator

The necessary rapidity and simplicity of data collection and analysis creates a bias toward technologies and tools that encode knowledge and are designed to provide quick answers. An Ohio regulator describes the continuing pressure for Ohio's assessment system to be accelerated and simplified:

The regulated community continues to complain that it's so hard, it's so difficult. … So then I sat down with Excel and made it as easy as you could get it. You just have to literally type the data in and it gives you the score. … So at the next meeting I came in and I put it up on the screen and said “Okay, you just type the data in here and it propagates this table, and all we want is the graph and the summary table submitted to us.” And they were sort of like, [pause] “Hunh”. … it actually defused a lot of the criticism.

Ohio regulator

Similar pressures apply to those in the private and non-profit sectors developing metrics for use in Oregon and elsewhere:

For us the data collection is pretty damn quick, you know, 15 min an acre, if it's a fairly homogenous site. … everything that we do, it's very much, “Ok, what can we get from a reliable visual assessment?”

Oregon environmental consultant

We always ought to be looking for the simplest possible way to measure, to find that indicator that we measure to, so that you're not spending more money on the measurement than you are on the conservation.

Oregon environmental non-profit director

The limits on time and expertise that actual users bring to the task of stream assessment are not simply a deficiency. The specifics of database programming and the labor costs of consulting firms are now creative forces in shaping what counts as an adequate “abstract stream” and thus a marketable stream credit.

Mad at the metric

Algorithms, ratios, concept curves, and Excel sheets are all key technologies by which we move from the physical stream to the abstract stream transacted in a market or enumerated in a federal agency report. These technologies should make the labor of valuation more efficient. The very point and value of “conservation by algorithm” (Adams, 2019) for management is that it saves labor costs and reduces the vagaries of a fieldworker's professional judgment: “Technical advance means that conservation data collection is increasingly being automated, bypassing (or making redundant) scientifically skilled conservation workers” (343). However, we find that valuation still depends on the work of field technicians, who often find that these technologies mediate and constrain their experience of assessing a site. Although many assessment tools are developed to be used by nonexpert end-users, those end-users are often degree-holding experts. Valuation, in each case, still begins in their hands in a field context, and they often closely guard that role (see Hirsch [2019, 11], in the context of salmon habitat assessment).

The experience of using an assessment tool matters in a way that can be consequential for valuation. One credit entrepreneur noted that, whether simple or complex, an assessment system that annoys the assessor is a bad system:

If you could divide a site into homogenous habitat areas or cover, it'd be a lot faster to go through a metric. You could ask fewer questions and because you had such little diversity within that map unit, you could fly through an answer sheet. Generally you can, you can go faster through an individual data sheet but you have to do so many data sheets it takes way too long. And it really annoys people and so it makes them mad at the metric… So when you're mad at a metric you don't trust the outcome. Right?

Oregon environmental consultant

In fact, making sure people are not “mad at the metric,” managing user frustration, turns out to be an acknowledged driver of assessment design. One assessment designer describes a tradeoff between precision, frustration, and respect for the user in good design:

… you want to have a method that instills trust in the users, it's got to be transparent. … you’re trading off that precision of measurement for “eyeball” kinds of stuff [i.e., informal measures]. But the eyeball allows you to cover more indicators and to reduce user frustration.

Oregon stream scientist

When field scientists and technicians get mad at the metric, they will sometimes refuse to vouch for the credibility of the metric. Metrics based on models that were not built from site data but were rather derived from principles, regional trends, or legacy data, face a legitimacy challenge from field-oriented scientists. Consider a dispute over the effect of a proposed dam on stream temperatures:

They were fighting over the interpretation of some model results that were basically something that you could not actually measure in the river. The only way you could show the effect of the dam [on stream temperature] was through a model, it wasn't through a thermometer. So it was all about modeling and the magnitude of the changes that people were talking about were 1 to 1.5 degree differences that were entirely subsumed within this much larger variance cloud. And so [I was] watching what people did, faced with this thing which you couldn't see, smell, taste, or measure…

Oregon stream scientist

When the effect of a dam was assessed using a stream temperature model derived from hydrologic principles, the dam was shown to have a small but perhaps meaningful effect on stream temperature. However, no assessment technician in midstream with a temperature probe could have said any difference they detected in situ was caused by the dam. The effect was only visible to the model. This created a moment of absurdity for the field-facing participants that threw the entire valuation exercise into doubt—another moment of vulnerability for the entire project of valuation.

Another field scientist objected to reliance on models as a “top-down” method of environmental assessment: he explained that most model-building exercises end up focusing on variables for which no possible field-usable indicators exist:

I’ve been in too many rooms with academics where they’ve spent all day on the conceptual model: “Now how do we find indicators for this?” “Gee, I don’t know!” And 95% of the boxes and arrows have no indicators at all! They’ve all congratulated themselves on how well they know the science and the connections, but there's no way to translate it into a real tool in the real world.

Oregon stream scientist

It is easy to imagine the state developing a system of valuing streams that depend only on modeling, discarding the field technician's observation. This could be done through hydrological or population models or predictive mapping and would have obvious benefits in time and cost. But in practice this didn’t happen in our sites: the value-bearing, abstracted stream remains tethered to observations in the physical stream in all three states. The field technician remains central to valuation, and the power of the assessment is still dependent, in part, on their sense that their measurements are meaningful (see Nost, 2015). The assessor's judgment thus has a certain amount of power that ratios and algorithms can constrain and discipline but not entirely eliminate. This dependence on empiricism may hobble the commodification of stream credits. This is obviously not true in all similar settings: in Leigh Johnson's (2013) work on catastrophe insurance, field observations of hurricane damage are irrelevant and damages are awarded based on the presence of sufficient wind speeds in the area in a synoptic climate model.

This tension between relying on individual discretion by regulators and enforcing consistent practice among them is well-recognized in the sociological literature on the “street-level bureaucrat” (Lipsky, 1980), and we want to call attention to the role the street-level (or in this case, the stream-level) assessor plays in creating the abstractions that define the commodity. Like ratios and concept curves, the temperament and expertise of the assessor ensure that social and even psychological factors can intervene between the act of measurement and the constitution of value.

Conclusion

It is not that ‘government’ exists first and technologies are developed in order to achieve its goals, but rather the opposite: technologies present themselves as potent sites for introducing ‘economy’ or ‘administration’ into everyday life.

Braun, 2014, 55

We have examined in detail the process of moving from an environmental object to an environmental credit and found that, generally, scientific understandings of streams are too complex, accurate, precise, slow, and frustrating to govern the entire process. On the face of it, this sounds like a bad thing. But the task of all governance is to reduce the complexity and dynamism of the social and natural world into stable objects that can be disciplined, transacted, accounted for, or surveyed (Scott, 1998). The methods for doing so are based in environmental science and must have the sanction of some environmental scientists, but must simultaneously create measures and objects capable of circulating outside of scientific knowledge-systems.

We conclude with three observations about the creation of the objects which circulate in environmental credit markets. First, empirically we observe that this process recruits conventional, long-standing scientific tools such as algorithms and conceptual curves that allow participants to derive value from measurements. In using these tools, however, they depart from science as a system of knowledge in several key ways. Scientists may see these departures as failures, but without them, the successful social abstraction of a stream credit cannot be built. Recent work on neoliberal nature supports the idea that governance will readily adapt the tools of science but will tend to leave aside the practices which validate scientific data as accurate or precise (Machen, 2018, Ghosh, 2019). Accuracy and precision are simply not always relevant to the need for a stable commodity to produce or transact, or the need to verify compliance with state conservation goals. As Amoore and Piotukh (2015, 355) observe with regard to “big data,” there is power in the ability to decontextualize data from the principles which gave it meaning at its site of generation.

Second, we find that there is still a degree of power vested in field observation by human assessors. Their experience in field settings appears to be, so far, an indispensable part of the stream valuation process, which may in fact be designed around their concerns. This is contrary to some expectations in the study of data-driven conservation (Adams, 2019; Amoore and Piotukh, 2015), in which researchers have observed a drive to reduce or eliminate site-based data in favor of modeling or remote observation. Here, modeling can supplement field observations in measuring environmental attributes, but it has not been allowed to displace assessor judgment in the cases we examine. Although the use of algorithms invites us to forget or discount the experience of assessors, their subjectivities form an essential part of the valuation and a place from which to contest it. Amoore and Raley (2017, 8) urge us to consider that “all action at the human-algorithm interface is embodied and to attend to precisely how the putatively immaterial algorithm becomes materialized in the moment of decision.” Measures of stream value are vulnerable when they are strictly meaningless to, or unobservable by, a person standing on the streambank, no matter how elegantly modeled.

Finally, the novelty of these environmental governance policies should not distract us from the fact that they often rely on tools that long predate neoliberalism (see Dempsey and Bigger, 2019, 528), and that the use of these tools can be a site for friction or contestation with earlier and ongoing uses. The persistent moment of field assessment means that the state's valuations still pass through the hands of individuals who design and execute fieldwork, and whose concerns about accuracy, precision, and labor-time are often accommodated. In developing valuation tools for streams, the tools and techniques of science are scavenged in a way that can disconnect the measurement activity from the conventions of science. However, this disconnection is not always successful because the meaning and validation of measured data is a discussion open to many participants. A wide range of regulators and entrepreneurs and scientists—some of them pointedly uninterested in the success of stream credit markets—must all agree that the output of the assessment procedure describes something capable of bearing value.

And this brings us to a concluding consequence of all this friction in scavenging and translation. If constructing value really is that difficult, it should sometimes fail, and it should fail for reasons that have to do with one of these parties withholding their consent to the adequacy of the translation between the physical stream and its capitalist value. Because these tools are pluripotent and more-than-capitalist, we should look for them to be used for a variety of ends (see Braun, 2014; Kitchin et al., 2016), some of which resist the abstraction of stream systems. And indeed, the leap between measure and value is not always made. In Ohio, the state's insistence on accuracy with reference to a biotic dataset was an obstacle to issuing a final assessment method, delaying it from 2010 until 2016. In the Oregon market, very few stream credit producers used the salmon metric to generate salmon credits:

There's two barriers: one, there's nobody buying salmon credits. It's not worth it to be that precise; but second, [the developers of the salmon metric] have worked very hard on a very, very, very complicated system and they don't get a lot of credit for it, partly because nobody trusts it because it's such a black box and it's proprietary.

Oregon environmental consultant

It may also be that overly simplified assessment metrics may act to drive out the kind of experts who currently provide the assurance that the assessment of stream credits is “science-based”:

At meetings where somebody is talking about concentrations of total mercury or nitrogen changing form or biomass accretion in Atlantic white cedar; to somebody who spent their life studying that, that's amazingly interesting, but to somebody who is a people manager that's just boring beyond human belief. … You need to have a mixture of scientists and people that are interested in the actual restoration and people that are good people managers. Right now it's overwhelmingly administratively orchestrated and most of the resource people have fled, because they don't enjoy thinking about “if the [applicant] worked hard, should he get [the permit]?”

North Carolina regulator

In North Carolina, the years of investment since 2003 in the NCSAM resulted in a valuation system that could not successfully distinguish degraded from restored streams (Tullos et al., 2006), and informants indicate that since its release in 2015, it has lain mainly unused in spite of the time and resources devoted to its creation. Oregon's salmon metric has faltered in its attempt to describe a commodity that entrepreneurs want to produce and that scientists trust and “enjoy thinking about.”

These three states are not unique: throughout the US and the world, governments are developing methods to assess the value of natural systems. Reasons for the success or failure of neoliberal environmental policy can be sought at the level of the design and use of valuation methods, where the abstract logics of state or capital come to ground. Although environmental credit markets, green GDP, ecosystem services accounting, and other such strategies are still relatively new, it is important not to assume that because the needs of capital may be primary, the tools of valuation will be deployed with ease. The work of moving from measure to value has always pulled, and continues to pull, the tools and techniques of science into new and precarious relationships with capital, with governance, and with the people who perform that work.

Highlights

The translation from physical site to value-bearing commodity is achieved with the techniques and tools of environmental assessment and quantification

These techniques and tools are not inherently capitalist but rather scavenged from other arenas of practice

Concept models and algorithms unconstrained by field data are essential to defining environmental credit commodities

Accuracy and precision in environmental science may be liabilities in crafting environmental credit markets

User experience in the field constrains and shapes the tools and techniques used to assess and quantify environmental credits

Footnotes

Acknowledgments

The authors are deeply grateful for the insightful comments of Eric Nost, Jen Rose Smith, and a host of anonymous reviewers. The authors are profoundly indebted to the community of stream regulators and restorationists who shared their wisdom and expertise. All errors are the responsibility of the authors.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the NSF Grant BCS-0961551.

Notes

References

Adams

(2019) Geographies of conservation II: Technology, surveillance and conservation by algorithm. Progress in Human Geography 43(2): 337–350.

Amoore

Piotukh

(2015) Life beyond big data: Governing with little analytics. Economy and Society 44(3): 341–366.

Amoore

Raley

(2017) Securing with algorithms: Knowledge, decision, sovereignty. Security Dialogue 48(1): 3–10.

Bakker

(2005) Neoliberalizing nature? Market environmentalism in water supply in England and Wales. Annals of the American Association of Geographers 95(3): 542–565.

Bendor

(2009) A dynamic analysis of the wetland mitigation process and its effects on no net loss policy. Landscape and Urban Planning 89(1–2): 17–27.

Bigger

Robertson

(2017) Value is simple. Valuation is Complex. Capitalism, Nature, Socialism 28(1): 68–77.

Bonneuil

(2015) Tell me where you come from, I will tell you who you are: A genealogy of biodiversity offsetting mechanisms in historical context. Biological Conservation 192: 485–491.

Braun

(2014) A new urban dispositif? Governing life in an age of climate change. . Environment and Planning D: Society and Space 32(1): 49–64.

Brenner

Peck

Theodore

(2010) Variegated neoliberalization: Geographies, modalities, pathways. Global Networks 10(2): 182–222.

10.

Brinson

(1995) The HGM approach explained. National Wetlands Newsletter 17: 7–13.

11.

Castree

(2008) Neoliberalising nature: Processes, effects, and evaluations. Environment and Planning A 40(1): 153–173.

12.

CWS [Clean Water Services] (2012) Request for Informal Proposals for Stream Assessment Services for the EPA Streams Project No. 6565. Hillsboro, OR: Clean Water Services.

13.

Dempsey

Bigger

(2019) Intimate Mediations of For-Profit Conservation Finance: Waste, Improvement, and Accumulation. Antipode.

14.

Doyle

(2018) The Source: How America Made and Remade its Rivers. New York: W.W. Norton.

15.

Doyle

Singh

Lave

, et al. (2015) The morphology of streams restored for market and nonmarket purposes: Insights from a mixed natural-social science approach. Water Resources Research 51(7): 5603–5622.

16.

EPA [US Environmental Protection Agency] (2012) Draft Functional Assessment Framework Excerpt: Attributes, Considerations, Criteria. Seattle: US EPA Region 10. January 2012. Available at: http://willamettepartnership.org/wp-content/uploads/2014/09/EPA-Draft-Functional-Assessment-Framework-Excerpt_2012.pdf (accessed 26 March 2017).

17.

FEMAT [Forest Ecosystem Management Assessment Team] (1993) Forest Ecosystem Management: An Ecological, Economic and Social Assessment. Washington DC: Forest Service, US Department of Agriculture. July 1993.

18.

Fischenich

(2006) Functional Objectives for Stream Restoration. ERDC TN-EMRPP SR-55. Vicksburg, MS: USACE Research and Development Center.

19.

GAO (US Government Accountability Office) (2005) Corps of Engineers Does Not Have an Effective Oversight Approach to Ensure That Compensatory Mitigation Is Occurring. Report to the Ranking Democratic Member, Committee on Transportation and Infrastructure, House of Representatives. Washington, DC: GAO.

20.

Ghosh

(2019) Appetite for imprecision: The role of bureaucracy in implementing a pay-for-performance program. Annals of the American Association of Geographers 109(4): 1208–1225.

21.

Graeber

(2002) Toward an Anthropological Theory of Value: The False Coin of our own Dreams. New York: Palgrave.

22.

Graeber

(2013) It is value that brings universes into being. HAU: Journal of Ethnographic Theory 3(2): 219–243.

23.

Hirsch

(2019) Anticipatory practices: Shifting baselines and environmental imaginaries of ecological restoration in the Columbia River Basin. Environment and Planning E: Nature and Space 3(1): 40–57.

24.

Hough

Robertson

(2009) Mitigation under section 404 of the clean water act: Where it comes from, what it means. Wetland Ecology and Management 17(1): 15–33.

25.

Jessop

(2000) The crisis of the national spatio-temporal fix and the tendential ecological dominance of globalizing capitalism. International Journal of Urban and Regional Research 24(2): 323–360.

26.

Jessop

Sum

N-L

(2006) Beyond the Regulation Approach: Putting Capitalist Economies in their Place. Northampton, MA: Edward Elgar.

27.

Johnson

(2013) Catastrophe bonds and financial risk: Securing capital and rule through contingency. Geoforum; Journal of Physical, Human, and Regional Geosciences 45: 30–40.

28.

Juniper

(2014) What Has Nature Ever Done for Us?: How Money Really Does Grow on Trees. New York: Synergetic Press.

29.

King

Price

(2004) Developing Defensible Wetland Mitigation Ratios: A Companion to “The Five-Step Wetland Mitigation Ratio Calculator”. Solomon’s Island, MD: King and Associates, Inc, 41.

30.

Kitchin

Maalsen

McArdle

(2016) The praxis and politics of building urban dashboards. Geoforum; Journal of Physical, Human, and Regional Geosciences 77: 93–101.

31.

Kruczynski

(1990) Mitigation and the section 404 program: A perspective. In: Kusler

Kentula

(eds) Wetlands Creation and Restoration: The Status of the Science. Washington, D.C.: Island Press, 549–554.

32.

Lave

(2012) Fields and Streams: Stream Restoration, Neoliberalism, and the Future of Environmental Science. Athens, GA: University of Georgia Press.

33.

Lave

Doyle

(2021) Streams of Revenue: The Restoration Economy and the Ecosystems it Creates. Cambridge, MA: MIT Press.

34.

Lave

Doyle

Robertson

(2010) Privatizing stream restoration in the US. Social Studies of Science 40(5): 677–703.

35.

Lipsky

(1980) Street-level Bureaucracy; Dilemmas of the Individual in Public Services. New York: Russell Sage Foundation.

36.

Lohmann

(2006) Carbon trading: A critical conversation on climate change, privatisation and power. Development Dialogue 48: 1–359.

37.

Machen

(2018) Towards a critical politics of translation: (Re)Producing hegemonic climate governance. Environment and Planning E: Nature and Space 1(4): 494–515.

38.

Mansfield

(2008) Introduction: Property and the remaking of nature-society relations. In: Mansfield

(eds) Privatization: Property and the Remaking of Nature-Society Relations. Malden, MA: Blackwell, 1–13.

39.

Marx

(1976) Capital: A Critique of Political Economy. New York: Penguin Books.

40.

McAfee

(1999) Selling nature to save it? Biodiversity and green developmentalism. Environment and Planning D: Society and Space 17: 133–154.

41.

Mutersbaugh

(2002) The number is the beast: The political economy of organic-coffee certification and producer unionism. Environment and Planning A 34(7): 1165–1184.

42.

Nadeau

Hicks

. (2018) A Functions-Based Stream Assessment Method To Support Compensatory Mitigation Decisions in Oregon. Webinar presentation, March 2018. Oregon Department of State Lands and US Environmental Protection Agency. Available at: https://www.oregon.gov/dsl/WW/Documents/SFAM_Phase_II_pre-field_webinar_FINAL.pdf (accessed 26 September 2019).

43.

Nadeau

Hicks

Trowbridge

, et al. (2018) Stream Function Assessment Method for Oregon (SFAM, Version 1.0) Oregon Dept. of State Lands, Salem, OR, EPA 910-D-18-001, U.S. Environmental Protection Agency, Region 10, Seattle, WA.

44.

NCDENR [North Carolina Department of Environment and Natural Resources] (2015) NC Stream Assessment Method, User Manual. Version 2.1. Raleigh, NC: NCDENR. August 2015.

45.

Nelson

(2014) Beyond the limits to growth: Ecology and the neoliberal counterrevolution. Antipode 47: 467–480.

46.

Nost

(2015) Performing nature’s value: Software and the making of Oregon's ecosystem services markets. Environment and Planning A 47(12): 2573–2590.

47.

Ohio EPA (2012) Field Evaluation Manual for Ohio’s Primary Headwater Habitat Streams. Version 3.0. Ohio EPA Division of Surface Water, Columbus, Ohio. 117 pp.

48.

Owen

(2017) Little streams and legal transformations. Utah Law Review 1: 1–55.

49.

Parametrix (2009) Salmon Credit Calculation Rules and Equations. Portland, OR: Parametrix, Inc.

50.

Poon

(2007) Scorecards as devices for consumer credit: The case of fair, Isaac & company incorporated. The Sociological Review 55(s2): 284–306.

51.

Prudham

(2004) Knock on Wood: Nature as a Commodity in Douglas-fir country. New York: Routledge.

52.

Pryke

(2007) Geomoney: An option on frost, going long on clouds. Geoforum; Journal of Physical, Human, and Regional Geosciences 38: 576–588.

53.

Randalls

(2010) Weather profits: Weather derivatives and the commercialization of meteorology. Social Studies of Science 40(5): 705–730.

54.

Robertson

(2006) The nature that capital can see: Science, state and market in the commodification of ecosystem services. Environment and Planning D: Society and Space 24(3): 367–387.

55.

Robertson

(2012) Measurement and alienation: Making a world of ecosystem services. Transactions of the Institute of British Geographers 37(3): 386–401.

56.

Robertson

(2018) Flexible nature: Governing with the environment in the development of US neoliberalism. Annals of the American Association of Geographers 108(6): 1601–1619.

57.

Rosgen

(1996) Applied River Morphology. Pagosa Springs, CO: Wildland Hydrology.

58.

Sayre

(2002) Ranching, Endangered Species, and Urbanization in the Southwest: Species of Capital. Tucson, AZ: University of Arizona Press.

59.

Sayre

(2017) The Politics of Scale: A History of Rangeland Science. Chicago, IL: University of Chicago Press.

60.

Scott

(1998) Seeing Like a State: How certain schemes to improve the human condition have failed. New Haven, CT: Yale University Press.

61.

Somerville

(2010) Stream Assessment and Mitigation Protocols: A Review of Commonalities and Differences. U.S. Environmental Protection Agency, Office of Wetlands, Oceans, and Watersheds (Contract No. GS-00F- 0032M). Washington, D.C.

62.

Somerville

Pruitt

(2004) Physical Stream Assessment: A Review of Selected Protocols for Use in the Clean Water Act Section 404 Program. U.S. Environmental Protection Agency, Office of Wetlands, Oceans, and Watersheds, Wetlands Division (Order No. 3W-0503- NATX). Washington, D.C. 213 pp.

63.

Tadaki

Sinner

(2014) Measure, model, optimise: Understanding reductionist concepts of value in freshwater governance. Geoforum; Journal of Physical, Human, and Regional Geosciences 51: 140–151.

64.

Thompson

(2013) The Quest for the Golden Trout: Environmental Loss and America's Iconic Fish. Lebanon, NH: University Press of New England.

65.

Tullos

Penrose

Jennings

(2006) Development and application of a bioindicator for benthic habitat enhancement in the North Carolina piedmont. Ecological Engineering 27: 228–241.

66.

Tullos

Penrose

Jennings

, et al. (2009) Analysis of functional traits in reconfigured channels: Implications for the bioassessment and disturbance of river restoration. Journal of the North American Benthological Society 28: 80–92.

67.

TWP [The Willamette Partnership] (2014) Protocol for quantifying the thermal benefits of riparian shade. Available at: http://willamettepartnership.org/wp-content/uploads/2015/02/Shade-a-lator-v8.0.5-8.0.8-Protocol-2014-03-20.pdf (accessed 26 March 2018).

68.

USFWS [United States Fish and Wildlife Service] (1953) Classification of Wetlands of the United States. Special Scientific Report: Wildlife No. 20. Washington DC: USFWS June 1953.

69.

USFWS [United States Fish and Wildlife Service] (1980) Ecological Services Manual: Habitat as a Basis for Environmental Assessment. Washington, DC: USFWS.

70.

USGS [United States Geological Survey] (1971) A Procedure for Evaluating Environmental Impact. Geological Survey Circular 645. Washington DC: US Department of the Interior.

71.

Whitworth

(2015) Quantified: Redefining Conservation for the Next Economy. Washington DC: Island Press.

72.

Wilcock

Baker

(2012) Introduction to Stream Project: Decision Analysis and Design Guidance for Stream Restoration. Short Course at River Restoration Northwest Conference. Stevenson, WA. 30 January 2012.

73.

Wohl

(2012) Disconnected Rivers. New Haven, CT: Yale University Press.

74.

zu Ermgassen

SOSE

Baker

Griffiths

, et al. (2019) The ecological outcomes of biodiversity offsets under “no net loss” policies: A global review.” Conservation Letters, online.