One of the first questions to answer is how much money is available to the project. Certainly a project scope can be formulated without regard to funds available, but very few, if any, projects have no funding restriction. Therefore, a scope must be carefully considered to ensure the assessment will serve the needs of the client within the required budget constraints.

Watershed Sampling

One of the first considerations is whether additional field sampling is necessary. In very rare instances a watershed will have an extensive amount of data already generated. Under these circumstances a comprehensive resampling might not be needed. In most cases, however, some field sampling makes sense, either to verify existing data or to generate enough data to make a reasonable assessment of the watershed. Many streams on the 303(d) list may have questionable or limited data to justify their inclusion on the list. The assessment should make an effort to confirm the validity of inclusion on the list.

The sampling program ought to be carefully formulated to account for cost and the negotiating position of the client. The project should be able to quantify the client’s contribution to the water quality of the watershed. For example, if a municipal client performs a watershed assessment in association with an upgrade to its discharge permit, the water-quality contribution of the municipality requires quantification. The sample locations should be placed such that water quality entering the municipality and water quality leaving the municipality are measured. This is not a matter of being able to lay blame for the watersheds problems. Rather, this is an effort to provide an "action position" for the municipality. By this we mean the municipality will have a clear understanding of its contribution to water quality and therefore a solid foundation for developing and implementing a watershed management plan. Without this clear understanding, the municipality might attempt to address problems over which it has no control.

Sample points should also be placed to potentially reflect land-use categories. If samples can be taken at points with a relatively homogenous upstream land use, the constituent runoff characteristics can be more easily defined. In the tradeoff between cost and data, however, it is often very difficult to sample specific to a land use. In medium to large watersheds, there is often a heterogeneous collection of land uses, each contributing a specific mix of constituent runoff.

Another approach to selecting sample points involves selecting points reflecting subwatersheds within the study area. Even if the subwatersheds have heterogeneous land-use mixtures, the sample points identify potential water-quality problems specific to each subwatershed. The sample data help formulate watershed management options that can be tailored to a subwatershed.

Existing Assessment Study

With sample locations selected, the question becomes how many sample events will be completed? The constituent sampling program is the most labor-intensive part of the assessment process and also has a substantial subcontractor cost associated with the laboratory analysis. Generally, you want to sample at least several "dry" events and several "wet" events. No measurable runoff in the days prior to sampling defines a dry event. A wet event is when substantial measurable runoff occurred prior to sampling. There are various ways to define the time frames and runoff volume for either a wet or a dry event. The point is that runoff constituents exist at different levels in the water depending on whether rainfall–and therefore runoff–occurred immediately before a sampling event. There should be some effort to quantify those differences.

How many sample events to complete also depends on the specific constituents for which you are sampling. To reiterate, laboratory costs are a major part of the assessment budget. Judicious application of sample parameters will provide the most data for the dollar. Certainly the constituent causing the stream to be listed should be analyzed. Typically this is a bacterial, organic, or metal constituent. A comprehensive and generally cost-effective constituent sampling list includes the following parameters:

fecal coliform
nitrite (NO₂)
nitrate (NO₃)
total Kjeldahl nitrogen
ammonia nitrogen
orthophosphate
total phosphorus
total suspended solids (TSS)
chemical oxygen demand
carbonaceous biochemical oxygen demand
total calcium
total magnesium
total lead
total zinc
total copper
total cadmium

Of course, there can be some "mix and match" depending on any previous knowledge of the watershed. Sampling five to 10 sample points approximately four times can cost between $10,000 and $30,000, including field labor, for the constituents listed.

In-situ measurements are less expensive to take and can yield important trends in water quality. A good-quality instrument can measure dissolved oxygen, pH, specific conductivity, temperature, and turbidity in real time. These measurements, in many instances, can be correlated to the laboratory constituents, thus providing an extrapolation of laboratory results to in-situ sampled events. However, this technique requires a working knowledge of water-quality chemistry and statistical analysis to produce valid and meaningful correlations. In-situ measurements can also potentially be used to screen for constituents that might have the greatest impact to the watershed. For example, if turbidity measurements are consistently low, even after rainfall, you might not need to analyze as frequently for TSS as you would for a constituent found in greater relative concentrations.

Assessment sampling

Laboratory costs increase dramatically when dealing with additional constituent parameters. Priority pollutant scans are relatively expensive and best applied as screening tests at a limited number of sample points and for a limited number of events. Clean-sampling techniques are also expensive, and a limited number of laboratories conduct these kinds of analyses. These analyses are best applied where there is some indication of contamination because of sampling materials or techniques and serve as confirmation of the constituent levels found in standard sampling procedures.

One often-ignored field task is the biological assessment. Streams end up on the impaired list not only for a specific constituent, but also for the habitat quality. Good water quality does not necessarily mean an existing healthy stream. Many urban streams have been degraded through years of unimpeded runoff from new development. Even if the water proves relatively clean exiting developed sites, increased velocities and volumes work to change the habitat dynamics of a stream. Many organisms that thrive in the predevelopment stream have a difficult time adapting to the altered habitat and in many cases cannot sustain reproductive life. It is prudent to employ a qualified stream biologist to assess the habitat quality. The existence of altered habitat without any water-quality concerns will focus watershed management options where they are most needed.

Another aspect of the assessment that receives variable treatment is the reference site. A necessary requirement of the biological assessment, a reference site usually exists outside the watershed being studied and provides a baseline quality level for comparison to the watershed sample sites. The reference site is in an undeveloped and ideally pristine watershed not subjected to the human-use pressures found within the watershed under study. Often the reference site remains ignored with regard to other aspects of the field effort and is not sampled except under the biological assessment work. It can, however, provide a frame of reference for the sampling results, especially for constituents over which there is disagreement as to their impact on water quality. Fecal coliform is one example. In many pristine and undeveloped watersheds, fecal coliform can exceed regulatory levels. Obviously, it would be unwise to mitigate a constituent at levels representative of a healthy system, regardless of the regulatory level. Some sampling of the reference site would provide perspective into this aspect of an assessment study.

Watershed Modeling

Collecting and analyzing data from the watershed is only the first step. Next comes an attempt to produce a computer model describing the relationships among rainfall, runoff, and constituents throughout the watershed. These models can range from the simplest spreadsheet models to the most complex models in use today. A model of the watershed is necessary to meet the requirement to quantify potential improvements or degradations to the watershed that occur as various watershed management options are applied. A model extrapolates constituent loadings developed from the sample data into maximum daily loadings. In fact, the whole assessment process is predicated on the development of total maximum daily loads (TMDLs), the biological buzz phrase of the new millennium. Developing these numbers requires some sort of mathematical process. The models in the following discussion are some of the most widely used methods.

If the watershed is relatively small and simple, a spreadsheet model may be the most cost-efficient tool. There are many sources for rainfall/runoff/constituent formulas that work well in a spreadsheet environment. For small, uncomplicated watersheds, this is probably the most cost-efficient method.

EPA sponsors a suite of models incorporated into a geographical information system (GIS) platform. The name of the overall programming is BASINS (for Better Assessment Science Integrating Point and Nonpoint Sources). The core of the system is a nationwide database of various types of data useful in watershed assessments. The database is broken down geographically into Hydrologic Unit Code categories developed by the United States Geological Survey (USGS). The database contains constituent sample results, somewhat coarsely defined land-use data, and in-stream discharge and intake permittees, to name a few of the database types. The data are by no means exhaustive, and care should be taken when using the database, as mistakes crop up throughout.

The BASINS suite incorporates several interfaces with water-quality models developed by governmental agencies. The primary model for use in watershed assessments is the Non-Point Source Model (NPSM). This is a modified version of the USGS model Hydrologic Simulation Program—Fortran (also known as HSPF), a complex runoff model that requires a great deal of effort and expertise. The NPSM version within BASINS, somewhat easier to use, still requires a fair amount of understanding and expertise to obtain valid results. The advantage of the BASINS system is that it integrates GIS with model formulation. Nevertheless, software bugs still need to be worked out. Once that is done, incorporating new data into the databases can be readily accomplished, thus facilitating a modeling process with some measure of standardization, very similar to the standardization of floodplain modeling software. EPA’s primary goal is to establish a standardized methodology for evaluating watersheds and calculating TMDLs.

BASINS can be downloaded free from the Internet; however, it does require the ESRI ArcView software to operate. ArcView, currently priced at around $1,000, can also be used for applications other than watershed assessments.

Another EPA-sponsored model with widespread use and application to watershed assessments is the Stormwater Management Model (SWMM). It is also a complex model requiring a great deal of expertise to operate. SWMM has several modules; the one for use with assessments is the Runoff module.

There are several methods for generating constituents by applying build-up parameters specific to a particular land use, then washing the constituents into the streams through additional runoff parameters, again specific to a land use. There is a fairly large body of knowledge on constituent runoff generation in SWMM, something that does not currently exist for the BASINS system.

The major disadvantage to using SWMM is the cost of a graphical interface for data input, which can exceed $4,000. Without a graphical interface, a SWMM input deck is tedious and time-consuming to develop. The solving algorithms can also be unstable, thus requiring trial and error to get the initial computer run to complete. The same vendors offering the graphical interfaces have also improved the algorithmic stability, so blow-ups occur less frequently.

SWMM’s major advantage is that it can accept future revision to serve as a hydraulic model as well. The value added to the project can be substantial, because SWMM can serve as a hydrologic/constituent runoff model and later as a hydraulic routing model, routing both flows and constituents. If future efforts combine infrastructure improvement with watershed management options in the watershed under study, SWMM might be the appropriate selection. It is particularly well suited for urban areas where stormwater pipe conveyance systems predominate and for systems where backwater and reverse flow are of concern.

The model, whatever form it takes, will produce "what-if" scenarios of the watershed under study. First the model calibrates to existing conditions through the sampling data, then it formulates scenarios to predict what will happen in the watershed as a result of future development. This future scenario usually relates to a future land-use plan, typically 20 years in the future. With the increased development described by the future land-use plan, there will likely be increases in some pollutant loadings. The what-if scenario shows what will happen in the future with current land-use development practices.

It is very difficult to apply structural best management practices (BMPs) to a calibrated watershed model, particularly one that is based on specific land-use runoff. The database used to establish a quantifiable constituent control reduction for a land use through the implementation of a BMP is not extensive. SWMM’s Transport module, however, has tools for some structural BMP implementation analysis.

Nonstructural BMPs, such as changing the land use on a portion of the watershed, might be modeled implicitly. The assumption is that the watershed sampling and subsequent model calibration would define an improved runoff water quality for, say, a forested land use. Replacing portions of another land use with forest would then show an improved condition as a result of application of the nonstructural BMP.

With land use being such an important part of watershed assessment modeling, integration with GIS becomes a great efficiency in analysis. This is one of the great advantages of the BASINS system. The impacts from a variety of land-use options can be examined rather easily and efficiently once the model is established.

Watershed Management

The last component of a watershed assessment, and probably the most difficult, is formulating watershed management options, generally in the form of structural and nonstructural BMPs. The difficulties lie not in defining important actions to take in the watershed but in implementing those actions and getting the variety of stakeholders to agree on the details. Bringing together a consensus for action in the watershed can be a monumental task, which is even more difficult if the watershed covers several distinct municipal jurisdictions. Typically, levels of development control vary from jurisdiction to jurisdiction. Political realities and pressures sometimes illogically control development decisions with regard to new and innovative site designs–designs that improve the quality parameters of stormwater runoff.

It is clear that comprehensive watershed management will not improve water quality without input from a broad range of stakeholders. To help formulate options, a diverse peer-review group should be formed, not only representing the affected jurisdictions but also drawing from various groups with a vested interest in the watershed under study. Participants may include homeowners, developers, environmentalists, business leaders, agricultural interests, watershed management experts, and any others with legitimate viewpoints who can bring to the table an earnest solution to the water-quality problem.

The fact that there is still disagreement on the efficacies of specific remedies does not improve matters. The science of nonpoint-source pollution control continues to develop. Controlling nonpoint-source pollution is more difficult than controlling point-source pollution. A point-source discharge is easy to quantify in terms of impact to overall water quality. It is then simply (although some engineers would disagree) an engineering exercise to place the appropriate control on the pollutant source. Nonpoint-source pollution is, in many ways, an abstract problem. One cannot generally point to a specific location as the source of the pollution. The phrase "holistic approach" is certainly valid when dealing with nonpoint sources. Controlling nonpoint sources not only requires establishing appropriate engineering controls but also sometimes requires a shift in the way we think and live our lives. There is a necessary paradigm shift in the way we handle economic growth and development, as in the way we handle our existing land uses to eventually achieve the goal of improved water quality and habitat.

After all, it is not hard to agree on the fact that we want clean water. There is clear evidence that most people are environmentalists at heart. It is not a question of improving the quality of our natural resources but of doing it fairly within our democracy. The assessment process offers a generally sound methodology, albeit still a developing one, for improving our water resources. Future generations demand it and will judge us harshly if we cannot take at least the initial steps to accomplish it.