Don L. Dycus, Technical Specialist

Tennessee Valley Authority

The Tennessee Valley Authority (TVA) began a program to systematically monitor the ecological condition of its reservoirs in 1990. Previously, reservoir studies had been confined to reservoir specific assessments to meet specific needs as they arose.

Objectives of TVA’s monitoring efforts, termed Vital Signs Monitoring, are to provide information on the "health" or integrity of the aquatic ecosystem in major Tennessee Valley reservoirs. Ecological monitoring activities provide the necessary information from key physical, chemical, and biological indicators to evaluate conditions in reservoirs and to target detailed assessment studies if significant problems are found. In addition, this information establishes a baseline for comparing future water quality conditions.

This paper describes the monitoring and data evaluation processes used to evaluate the overall ecological health of reservoirs. It summarizes 1997 data as an example of the mechanics of the ecological health scoring system used in the process.

The reservoir ecological health evaluation system is reviewed each year seeking areas in need of improvements. Initially, numerous improvements were made based on experienced gained from working with this new system and input from other professionals. Each year, progressively fewer changes have been needed.

This monitoring program was designed based on several fundamental premises.

With these premises in mind, our challenge has been to develop a sustainable monitoring effort that collects the right kinds of physical, chemical, and biological data to provide enough information to reliably characterize ecological health. Study design must carefully consider selection of important ecological indicators, representative sampling locations, and frequency of sampling, all in light of available resources. Following are some of the basic study design decisions made in developing this program.

Ecological Indicators—Physical, chemical, and biological indicators (dissolved oxygen, chlorophyll, sediments, benthos, and fish) were selected to provide information from various habitats or ecological "compartments". For example, the open water or pelagic area in reservoirs is represented by chlorophyll and dissolved oxygen (DO) in midchannel. The shoreline or littoral area is evaluated by sampling the fish assemblage. The bottom or benthic compartment is evaluated using two indicators: quality of surface sediments in midchannel (determined by chemical analysis of sediments) and examination of benthic macroinvertebrates from a transect across the full width of the sample area (including overbanks if present).

Sampling Locations—Three areas were selected for monitoring: the inflow area, generally riverine in nature; the transition zone or mid-reservoir area where water velocity decreases due to increased cross-sectional area, suspended materials begin to settle, and algal productivity increases due to increased water clarity; and the forebay, the lacustrine area near the dam. Overbanks, basically the floodplain which was inundated when the dam was built, are included in transition zone and forebay areas. Embayments, another important type of reservoir area, also were considered. Previous studies (Meinert et.al., 1992) have shown that ecosystem interactions within an embayment are mostly controlled by activities and characteristics within the embayment watershed, usually with little influence from the main body of the reservoir. Although these are important areas, monitoring of hundreds of embayments is beyond the scope of this program. As a result, only four, large embayments (all with drainage areas greater than 500 square miles and surface areas greater than 4500 acres) are included in this monitoring effort.

Sampling Frequency—Sampling frequencies (indexing periods) must consider the expected temporal variation for each indicator. Physical and chemical components vary in the short term so they are monitored monthly from spring to fall. Biological indicators better integrate long-term variations and are sampled once each year. Fish assemblage sampling is conducted in autumn (September-November). From 1990 through 1994 benthic macroinvertebrate sampling was conducted in early spring (February-April) to avoid aquatic insect emergence. Beginning in 1995, sampling was switched to late autumn/early winter (November and December). The problem with spring benthos sampling was that results were reflective of conditions from the previous year. This caused results for this indicator to be out of synch with those from the other indicators. This change is more thoroughly discussed in Dycus and Meinert (1996).

Another design issue dealing with sampling frequency is year-to-year variation. Meteorological conditions (particularly runoff from rainfall and its influence on flows) have a great effect on reservoirs and can vary substantially from year-to-year. To account for this variation, our design specifies that a reservoir be sampled for five consecutive years. Following that, sampling occurs on an every other year basis.

Like most evaluations, results for ecological integrity studies must be compared to some reference or yard stick to determine if monitoring results are indicative of good, fair, or poor conditions. In streams this is usually accomplished by studying a site that has had little or preferably no alterations due to human activities. Observations at that site provide the reference conditions or expectations of what represents a site with good/excellent ecological health. Given that reservoirs are not natural systems, this approach is inappropriate. Other potential approaches include historical or preimpoundment conditions, predictive models, best observed conditions, or professional judgment. Preimpoundment conditions are inappropriate because of significant habitat alterations. For the most part, models are of limited value for many indicators because of spatial and temporal variations within and among reservoirs. Spatial variation exists within in the multiple zones (e.g., forebay, transition zone, inflow, and embayments) of a reservoir. Further, each zone responds differently to different stimuli. Temporal variations are introduced because reservoirs are controlled systems with planned annual drawdowns in elevations ranging from only a few feet to close to a hundred feet. This leaves best observed conditions and professional judgment as the most viable alternatives for establishing appropriate reference conditions or expectations for reservoirs. Our process uses a combination of these two approaches.

A preliminary step to developing reference conditions is to examine the need to separate the reservoirs under study into separate classes so that appropriate , "apples-to-apples" comparisons can be made. Like streams, important considerations for classing reservoirs include size, gradient/depth, ecoregion, etc. In addition, reservoirs are managed systems and management objectives must considered.

A lesson we learned early in this process was that the issue of classification and its influence on determining reference conditions differed among the environmental indicators. A fundamental question that had to be addressed separately for each indicator was—Should reservoir ecological health evaluations be based on:

Our response (opinion) was that ideal conditions should be expected for DO and Sediment Quality. That is, poor DO is unacceptable regardless of type of reservoir or dam operation. Sediments should not have high concentrations of metals, should have no or at most very low concentrations of pesticides, and should not pose a toxic threat to biota. In this situation, there is no need for classification because the same conditions are desired for all reservoirs.

For chlorophyll, the classification scheme that has evolved is somewhat of a combination of the two approaches. First the geological characteristics (primarily erodablility and nutrient level of soils) of the watershed were examination. Then a conceptual/subjective decision made as to the concentrations indicative of good, fair, and poor conditions. Two classes of reservoirs were developed – reservoirs in watersheds draining nutrient poor soils, basically those in the Blue Ridge Ecoregion (i.e., expected oligotrophic reservoirs); and reservoirs in watersheds draining soils which are not nutrient poor (i.e., expected mesotrophic reservoirs).

For the benthic macroinvertebrate community and fish assemblage, the "best expected/observed conditions" approach was selected initially. Basically, this means the data base from the existing population of reservoirs is examined to determine the range of conditions for each community characteristic or metric (e.g., number of taxa). The process is to first omit outliers (defined as more than three standard deviations from the mean), then trisect the range of remaining values. These three ranges represent good, fair, and poor conditions and form the reference conditions or expectations for each metric. This is still the basic approach used for these two indicators, but experience has shown best results can be obtained by including professional judgment in the process. Cutoff points are examined closely and adjusted, if appropriate, based on professional judgment. This approach is discussed in detail in Dycus and Meinert (1998).

Reservoirs were divided into four classes to evaluate the benthos and fish. One class includes the reservoirs on the Tennessee River plus the two navigable reservoirs on tributaries to the Tennessee River (loosely termed run-of-river reservoir). This group of reservoirs has relatively short retention times and little winter drawdown. The remaining tributary reservoirs were separated into three classes: those in the Blue Ridge Ecoregion, those in the Ridge and Valley Ecoregion, and those on the Interior Plateau Ecoregion. The run-of-the-river reservoirs were not subdivided by ecoregion because most of the water flowing through each reservoir comes from upstream and does not originate within the ecoregion where the reservoir is physically located.

In absence of universally accepted guidelines to evaluate reservoir ecological health, we had to develop an evaluation methodology for reservoirs included in this program. The ecological health evaluation system examines each indicator separately and then combines those ratings a single, composite score for each reservoir.

Dissolved oxygen—The rating criteria represent a multidimensional approach that includes dissolved oxygen levels both throughout the water column (WC_DO) and near the bottom (B_DO) of the reservoir. The DO rating (ranging from 1 "poor" to 5 "good") at each sampling location is based on monthly measurements during April through September for the run-of-the-river reservoirs and May through October for the tributary reservoirs. This is the period when maximum thermal stratification and maximum hypolimnetic anoxia are expected. The WC_DO Rating is the six-month average of the proportion of the reservoir cross-sectional area at the sample location that has a DO concentration less than 2.0 mg/L. The B_DO Rating is the six month average of the proportion of the reservoir cross-sectional bottom length that has a DO concentration less than 2.0 mg/. The final DO rating is the average of these results.

Chlorophyll—Scoring criteria were developed separately for each of the two classes of reservoirs. Reservoirs expected to be oligotrophic receive highest ratings at low chlorophyll concentrations. Reservoirs expected to be mesotrophic receive highest ratings for an intermediate range of concentrations. Figure 1 shows the sliding scale used to evaluate the seasonal average chlorophyll concentration for each reservoir class. For reservoirs expected to be mesotrophic, the rating is reduced at low chlorophyll concentrations because some environmental factor (e.g., turbidity, toxicity, retention tine) must be inhibiting primary production.

Sediment quality—Initially, the scoring criteria for sediment quality was based two components: sediment toxicity tests and sediment chemical analyses for ammonia, heavy metals, pesticides, and PCBs. Since 1995, the sediment quality scoring criteria have been based only on sediment analyses for metals (As, Cd, Cr, Cu, Pb, Hg, Ni, and Zn), organochlorine pesticides, and PCBs. Sediment toxicity tests were discontinued primarily because of budget reductions, but also because frequent changes in toxicity testing methods made year-to-year comparisons difficult. The sediment quality rating compares results for metals analyses to sediment guidelines we adapted from EPA Region 5 (EPA, 1977). Presence of any of the organic analytes is deemed undesirable so results are compared to laboratory detection limits. If none of the metals exceed these guidelines and no PCBs or pesticides are detected, the site would receive the highest sediment quality rating. Occurrences of analytes above these standards lowers the rating.

Benthic Macroinvertebrate Communities—Scoring criteria were developed from the data base on TVA reservoirs, as described above. Seven metrics or characteristics are used to evaluate the benthic macroinvertebrate community. The benthic macroinvertebrate rating is derived from the total of these metrics.

Fish Assemblage—Twelve metrics or characteristics are used to derived the Reservoir Fish Assemblage Index (RFAI) which forms the basis for evaluating the fish assemblage (Hickman and McDonough, 1995).

Species Richness and Composition Metrics

The ecological health scoring process is designed such that four of the indicators (DO, chlorophyll-a, benthos, and fish) are given equal weights and assigned a rating ranging from 1 (poor) to 5 (excellent). The other indicator, sediment quality, is given only half the weight of the other indicators and assigned a rating ranging from 0.5 (poor) to 2.5 (excellent). (Note: Prior to 1995, sediment quality had been rated on the full 1 to 5 range, same as the other indicators. But, discontinuance of sediment toxicity testing, which had contributed half the sediment quality rating, resulted in the rating for this indicator being reduced by one half). Ratings for the five indicators are summed for each site. Thus, the maximum total rating for a sample site would be 22.5 (all indicators excellent) and the minimum 4.5 (all indicators poor).

To arrive at an overall health evaluation for a reservoir, the sum of the ratings from all sites are totaled, divided by the maximum possible rating for that reservoir, and expressed as a percentage. It is necessary to use a percentage basis because the number of sites monitored varies according to reservoir size and configuration. Only one site, the forebay, is sampled in small tributary reservoirs, and up to four sites (forebay, transition zone, inflow, and embayment) are sampled in selected run-of-the-river reservoirs. Also, the number of indicators varies from three to five at different sites. Chlorophyll and sediment quality are excluded at the inflows on run-of-the-river reservoirs because in situ plankton production of chlorophyll does not occur significantly in that part of a reservoir and because sediments do not accumulate there. As a result, the number of scoring possibilities may be as few as 5 indicator ratings for a small reservoir sampled only at the forebay. Or, as many as 18 indicator ratings for a large reservoir sampled at the forebay, transition zone, inflow, and embayment. The total score for the small reservoir would be 22.5 if all indicators rated excellent, whereas, the total score for the large reservoir would be 82.5 if all indicators rated excellent. Hence, using a percentage basis allows easier comparison among reservoirs.

This approach provides a potential range of scores from 20 to 100 percent and applies to all reservoirs regardless of the number of indicators or sample sites. To complete the ecological health scoring process, the 20-100 percent scoring range must be divided into categories representing good, fair, and poor ecological health conditions. This was achieved as follows:

Currently used reservoir scoring criteria are:

This ecological health scoring process has been in use for several years. Each year, slight modifications may be made in the overall evaluation process or in the numerical rating criteria for the five ecological health indicators based on experience gained from working with this process, review of the evaluation scheme by other state and federal professionals, and results of another year of monitoring.

The difference in the poor scoring range between the two types of reservoirs exists because two storage reservoirs with known poor conditions rated slightly higher than the boundary for the lower (poor) grouping on the run-of-the-river reservoirs. Hence, the high end of the lower scoring range for storage reservoirs was shifted upward from 52 to 56 percent to accommodate these reservoirs with known poor conditions.

An example that illustrates the overall reservoir health evaluation methodology is presented in Table 1. Fort Loudoun Reservoir, the example used, has five aquatic health indicators at two locations and two indicators at another location.

Combining all the aquatic ecosystem indicator ratings to determine the overall ecological health for each of the 17 reservoirs sampled in 1997 shows the following:

• 6 of the 17 rated good (4 mainstream reservoirs and 2 tributary reservoirs);
• 6 of the 17 rated fair (2 mainstream reservoirs and 4 tributary reservoirs); and
• 5 of the 17 rated poor (all tributary reservoirs).

The ecological health ratings for all reservoirs during 1997 and earlier years are presented by classification unit in Table 2. Comparisons show that 10 of the 17 reservoirs sampled in 1997 scored within two points of their long term average, 3 scored higher, and 4 scored lower than their long term average. Meteorological (rainfall and runoff) and hydrological (retention time) condition contribute greatly to year-to-year variations.