Example 4 (continued)

As in the previous examples, model selection and model strategy are determined by a number of considerations, including stressors, and required treatment of time, space, and the level of study/complexity.

Stressors

Depletion of DO in the river segment results from the interaction of a number of different sources of oxygen demand, including both point sources and diffuse watershed sources. These sources include:

1. A wastewater treatment plant, which is a permitted point source discharging a relatively steady load of oxygen-demanding waste;

2. Urban runoff routed through a separate storm sewer system, which is also a permitted point source, but discharges a highly episodic load of oxygen demanding waste;

3. Runoff from agricultural crop farming and animal operations, which constitute a diffuse nonpoint source of oxygen demanding waste; and

4. Sediment oxygen demand, exerted by organic material stored in the river sediment.

   The stressors have a variety of different spatial and temporal characteristics, which will affect the choice of a model strategy.

Time considerations

The objective of management is to prevent excursions of the DO water quality standard. This should be protected at all times, except under extraordinary conditions. Equivalently, the objective could be stated as maintaining a frequency of excursions of the standard that is below a certain acceptable low frequency, such as once in three years.

Not all the sources are constant in time, and we are concerned that the standard be met at almost all times. For these reasons, long-term average predictions of DO concentrations in the river segment do not provide us what we need to know. Instead, we need a model that will capture transient depression of DO concentrations in the river segment.

There are two ways in which we could try to capture the time-varying nature of impacts. The direct approach would be to implement a continuous model with adequate temporal resolution to predict the actual time-series of DO concentrations. This would typically require a substantial effort. The alternative is to apply a "worst-case" type of model that predicts only what happens at critical conditions, which are those conditions at which the greatest impact is expected.

In modeling the impacts of point sources a worst-case approach is typically used. This consists of assuming a minimal instream dilution capacity or design flow and applying a steady-state water quality model such as QUAL2K- which is much simpler, and less expensive to implement than a continuous model. By choosing a conservative design flow and other conservative design conditions (such as the high end of the expected water temperature range, which increases oxygen demand), a wasteload allocation can be assigned to the point source which is protective of the waterbody under most conditions. Typically, the design flow is assumed to be the 7Q10 flow which is the 7-day average low flow which recurs, on average, once every 10 years. Note that this still potentially allows occasional excursions of the DO standard, during those time periods when the instream flow is less than the 7Q10 flow.

Thus, a relatively simple method is available for the analysis of point source impacts. The addition of episodic and nonpoint sources of load (typical in watershed assessment) complicates this analysis. For these sources, lowest dilution flows and highest source loads often do not coincide, particularly when there are significant precipitation-driven sources. In these cases, the "worst case" may be at some flow higher than the 7Q10 low flow.

The simplest and most conservative approach to the watershed-scale analysis in this example would be to apply a steady-state model at the design low flow, as was done for the point source, and assume a worst-case loading from the storm sewers, agricultural areas, and other precipitation-driven sources. This would certainly be protective of water quality, but is likely to be unrealistic, since maximum loading from the runoff would normally be associated with higher than 7Q10 flows in the river. Indeed, such an ultra-conservative approach is likely to suggest that there is no assimilative capacity available for the point source, even when observations indicate that this is clearly not the case.

To make the modeling assessment more realistic, while continuing to use a simple, conservative approach we could run two steady-state model applications, intended to represent the range of expected impacts in the receiving water segment. The first application would be intended to reflect the impact of the point source at drought condition flows, with no contribution from episodic precipitation-driven sources. The second model application would be designed to reflect the impact of nonpoint loading during a large runoff event, plus the point source. To make this application conservative (but not completely unrealistic) we would combine the large runoff loads with an unusually low dilution flow for such an event in the receiving water. One way to do this is to analyze historical records of flows during large rain events, and choose the lowest instream flow (or once-in-10-years recurrence flow) observed in association with large rain events.

Alternatively, the results can be made more realistic by running a continuous simulation model that will provide a full representation of the correlation of runoff events and instream flows. Continuous modeling allows us to assess the interaction of point sources and nonpoint sources over all flow conditions, including the period following a runoff loading event when flow drops off, but concentrations may be high. The main drawback of using a continuous modeling analysis is the additional data, time, and user expertise required to implement it.

The choice of which method to use will depend on the level of detail required to answer the question at hand, and the resources available for the assessment.