NERRS banner

System-wide Monitoring Program

Synthesis of the Water Quality Data
INTRODUCTION

In 1993, Research Coordinators representing the National Estuarine Research Reserve (NERR) system proposed a System-wide Monitoring Program (SWMP) for the purpose of collecting long-term water quality and ecological data in each Reserve in order to define baseline conditions and establish trends for the NERR system. The Research Coordinators hypothesized that the SWMP would demonstrate that the reserves represented estuarine systems with a high degree of environmental quality and, as a group, the NERRs may represent a reference condition for the Nation’s estuaries. The SWMP would provide important information for defining linkages between the environmental quality of estuarine environments and adjacent human activities. This information represented a more meaningful ecological unit of measure (relatively small watersheds) across a broader spatial scale (i.e., regional and national) than was possible using data collected by ongoing regional and national monitoring programs. Although most of these existing programs focused on short-term measurements (days to weeks) collected annually or biannually for only a subset of the Nations coast, the estuaries in the NERR system represented all U.S. coastlines. In addition, the long-term nature of the SWMP data set made it possible to assess intra- and inter-annual patterns in estuarine systems.

Initiated in 1995, the NERR SWMP identifies and tracks short-term variability and long-term changes in representative estuarine ecosystems and coastal watersheds. The first phase of the program involved monitoring a suite of water quality variables or measures that reflected the condition of estuaries. Because it was not possible to measure all environmental variables known to be important in estuaries, a few basic water quality variables (i.e., pH, salinity, temperature, dissolved oxygen, turbidity and water level) that could be measured cost-effectively in a nearly continuous manner over long time periods were chosen for study.

While these variables may or may not directly address the ecological resources at risk, they are proven indicators of environmental stress and habitat quality and are linked to a broad range of anthropogenic stresses. For example, dissolved oxygen (DO) levels are strongly influenced by point and non-point source discharges, and for that reason, the EPA has used DO as a basic measure of water quality for several decades. Increasing DO above 120% saturation and decreasing DO below 28% saturation can indicate declining water quality. The trend of DO values between hypoxia and supersaturation can be a measure of system health. Temperature controls the rate at which living resources use oxygen, and salinity is the main factor controlling the distribution, abundance and composition of biological resources in estuaries. Changes in the salinity range have been reported to be a sensitive measure of the degree to which watershed development has altered hydrographic conditions. A common perception of coastal zone environmental managers is that much is known about the salinity structure of the nation’s estuaries. Reliable data on salinity distributions exist, however, for only a few of the most studied estuaries. Turbidity, a major factor controlling primary productivity in estuaries, varies greatly over short time periods and is linked to both seasonal changes in biological processes and the amount of suspended sediments in freshwater inflows. The pH of the water is an important factor controlling the availability of contaminants and nutrients to living resources. While pH varies with DO levels and salinity, influence of tidal cycles and diel cycles on pH are poorly understood and have been measured in relatively few estuarine systems.

The system-wide water quality monitoring database collected by the NERR program is one of the most intensive and extensive ever collected for estuaries. The database includes two sites in each of 22 Reserves collected over several years at half-hour intervals. The 22 Reserves represent all geographical regions of the country and a broad range of estuarine habitats and types. Considerable effort and expense were required to collect, quality assure, and archive these data. These data represent valuable scientific information for advancing the understanding of how estuaries function and change and may enable scientists to eventually predict how estuarine systems will respond to changes in climate and human-induced perturbations.

Synthesis Goals
The goals of this project were to conduct a detailed analysis of the NERR water quality monitoring data collected between 1996-1998 in order to:
  • Provide a characterization of the water quality for each NERR site (i.e., extent and severity of naturally occurring hypoxia, temperature, and salinity range)
  • Determine the degree to which the SWMP is producing important scientific information about the water quality of the Nation’s estuaries.
  • Determine if the program could be modified to be more effective or efficient or to obtain more ecologically relevant water quality information.

Results obtained during this project provided information for characterizing water quality at each of the NERR sites and for making national comparisons. Additional benefits of this project included assessment of the value of the SWMP water quality monitoring element, modification of the existing sampling design to make it more efficient and effective and plans for the next phases of the SWMP monitoring effort. The specific objectives of the project were to:

  • Determine the frequency, duration, and periodicity in hypoxia and other key water quality indicators at each site and among sites within each NERR.
  • Compare key indicators of water quality stress, with an emphasis on hypoxia, among NERR systems and regions.
  • Classify sample sites and reserves into groups based on their similarity/ dissimilarity in key water quality indicators and site characteristics.
  • Characterize water quality conditions at sites individually, regionally, and nationally.
  • Prepare a synthesis for broad distribution that evaluates the status of hypoxia in the NERR SWMP.
  • Develop a Technical Report for distribution to the Reserve staff and other research scientists that presents the results of statistical analyses and includes recommendations for improving the efficiency or effectiveness of the sampling design.
  • Conduct a workshop for NERR researchers that presents and discusses major findings in the Technical Report, reviews recommendations, and trains them to apply the analytical protocols developed for evaluating water quality data.

Findings of the Expert Panel
A workshop was convened in November 1999 in Charleston, SC, to develop a detailed analytical plan for the synthesis of the NERR monitoring data. An expert panel of scientists, experienced in the collection, analysis, and interpretation of water quality data, was assembled at the workshop. Members of the Expert Panel included Dennis Allen (University of South Carolina/ North Inlet-Winyah Bay NERR), Walter Boynton (University of Maryland Chesapeake Biological Laboratory, Chesapeake Bay Laboratory), Loren Coen (South Carolina Marine Resources Research Institute), Don Edwards (University of South Carolina), Holmes Finch (University of South Carolina), John Grego (University of South Carolina), Fred Holland (South Carolina Marine Resources Research Institute), Todd Hopkins (Florida Department of Environmental Protection/ Rookery Bay NERR), Richard Langan (University of New Hampshire/ Great Bay NERR), Chris Nietch (University of South Carolina), George Riekerk (South Carolina Marine Resources Research Institute), Steve Ross (North Carolina NERR), Denise Sanger (South Carolina Marine Resources Research Institute), Tammy Smalls (Centralized Data Management Office), Kevin Summers (United States Environmental Protection Agency), and Elizabeth Wenner (South Carolina Marine Resources Research Institute/ ACE Basin NERR).

Identification of Key Variables
The panel concluded that five variables were critical for inclusion in the statistical analyses to be conducted for this project: (1) DO (% sat), (2) DO (mg/l), (3) salinity, (4) temperature, and (5) depth. They suggested that turbidity and pH were also potentially important variables that may be used in future analyses following protocols developed by this project. Because of quality assurance concerns about the reliability of the turbidity and pH data, and the recent implementation of turbidity data collection, these two variables were recommended for exclusion from this project at this time.

In addition to prioritizing the water quality variables, the panel discussed other metrics that could be calculated using the existing data. Three calculated metrics that were determined to be particularly important and recommended for inclusion in the analysis were: (1) The percent of time DO values were less than 28% saturation; (2) The percent of time DO values were greater than 120% saturation; and (3) The duration of hypoxic events. The first metric defines the total amount of time each site experienced hypoxia, and the second metric defines the total amount of time each site experienced supersaturation. There has generally been little documentation of extent and severity of supersaturation events in estuarine water or an assessment of the value of this information for the identification of algal blooms or other environmental problems. Supersaturation, however, frequently occurs at NERR sites and is potentially as stressful to biota as low DO levels. Therefore, the panel felt it was important to evaluate occurrence of supersaturation events. The third metric defines the amount of time that each hypoxic event persisted at each site.

Potential Stratification Changes
Stratification of the NERR sites for summarization of the data was discussed. Potential a priori stratification schemes identified included classification by geographic region, salinity regime, tidal regime, and amount of development (i.e., reference vs. impacted). The panel decided that a priori stratification for summarization and analysis of the data was unnecessary. They recommended allowing the analytical results to define any strata that occurred, including conducting specific analyses to classify sites using multivariate methods. The critical step for conducting classification analyses was development of a data matrix (ecological attributes by site) that would provide the input data for these analyses.

Completeness and Quality of the Data
The completeness of the existing water quality data for the 22 NERRs is variable among years and sites. In general, most sites have relatively continuous data for some years and large data gaps for other years. The data gaps are due to equipment failure, staffing problems, or weather (i.e., meters are not deployed at some sites in winter due to ice formation). A major topic of discussion at the workshop was whether data gaps should be corrected by estimating the values for missing data or whether analyses should only focus on available data. The panel recommended that data gaps did not need to be filled to address the goals of the project and that only the data that were collected should be included in statistical analyses.

Another important data quality issue discussed was the apparent systematic downward “drift” in the DO values at many sites 3-5 days after deployment of the meters. The cause of DO “drift” was thought to be fouling by living organisms of the membrane covering the oxygen probe. The DO “drift” biases estimates of central tendency (e.g., mean, median), variance, and the percent of time DO was below 28% saturation. The panel recommended pre- and post-calibration values be obtained and used to correct for DO drift using standard methods and criteria. The panel further recommended that the unreliable DO data resulting from “drift” (i.e., the records most affected by “drift”) should be excluded from analyses. In addition, the extent of drift for each reserve should be determined and evaluated if time allowed.

Summarization/Evaluation of the Existing Data
The panel recommended that the general summarization of data should include scatter plots and frequency distributions of raw data and the percent of time below or above critical levels. In addition, the dynamics of the data should be analyzed using harmonic regression analysis. Fourier analyses were not recommended because of the large data gaps. Harmonic regressions, ANCOVA using trigonometric functions to adjust for known periodicity in the data (12.42 and 24 hour cycles), partition the variance within each data set into known cycles. This procedure allows the relative importance of various sources of periodicity, as well as interactions among sources, to be evaluated and compared for parameters (e.g., DO mg/L) after adjusting for known periodicity among reserves and sites (treatments). Harmonic regressions include an evaluation of interactions among treatments and co-variates as well as a determination of the magnitude of random error in the signal. The net result of harmonic analysis would be to identify systems that tended to be physically dominated by 12.42-hour cycles, 24-hour cycles, or interaction between these two cycles.

The panel recommended that analyses (harmonic and other) should first summarize and compare data by season and month within a reserve. Comparisons between reserves would depend on the quantity (completeness) and quality of the data available and would occur after conditions within each reserve had been defined. The main questions to be addressed included: (1) What are the basic water quality characteristics for each site; (2) How “good” were the data for each site (i.e., do they represent conditions that occur at the site); and (3) How do water quality conditions compare among sites? The panel also recommended that a summarization and analysis of water quality data should utilize readily available and user-friendly software such as Microsoft Excel and that a technical training manual be prepared to illustrate data management and graphical methods. Details about harmonic regression analysis and associated statistical programs should be included in the technical report for the project.

Representativeness/Characterization of NERR Sites
The panel concluded that NERR sampling sites are representative of the processes that occur at each site, but not necessarily representative of conditions in a particular reserve or of the NERR system as a whole. The panel recommended comparisons of NERR sites with other programs that collect similar data (i.e., EMAP, National Coastal Monitoring Program, C-GOOS, National Assessment of Eutrophication) to determine if NERR sites could be potential reference areas for other programs.

The panel concluded the NERR water quality data were likely to be important for defining water quality conditions in relatively undisturbed/undeveloped systems. The panel knows of no other multi-regional, multi-year water-quality monitoring program that collects measurements as frequently as the NERR program. Because of these unique attributes, the NERR SWMP may likely provide baseline information that would provide information for defining baseline conditions at sites.

Metabolic Status of NERR Sites
The panel recommended that the status of the NERRs should be addressed using results of the previously discussed analyses, particularly the harmonic analysis and metrics using percent of time. In addition, determination of the metabolic properties of each site was suggested as a means of providing information for a status assessment of NERR sites. Production/respiration (P/R) analyses determine if a system is autotrophic or heterotrophic by computing net production and nighttime respiration from semi-continuous DO measurements. P/R analyses would be done at a few of the reserve sites having nutrient data that would facilitate the interpretation of results.

Alternative Sampling Approaches
The panel agreed that it was too early in the project to discuss alternative sampling approaches; however, the following topics were recommended for examination after the existing data had been thoroughly analyzed. First, should duration of deployment vary among geographical location and/or among sites within reserves? Second, are there critical time periods when data must be collected which may vary among geographical locations and sites within reserves? Third, what changes to the sampling design would make the SWMP more efficient or effective? Many panel members were of the opinion that if the goal of the SWMP is to understand the long-term dynamics of water quality variables within reserves, then it might be useful to keep sampling methods and the sampling design relatively similar to the approach currently being used.

Trends Assessment
The panel concluded that the amount of data available (i.e., 3 years) was not sufficient to determine trends in water quality for sites. This project may be able to provide recommendations for trend analysis activities in the future, based on the results of the analyses that will be conducted. The panel recommended that the following topics be examined once the 1996-1998 SWMP data have been synthesized: (1) Determine what trends are likely to be important; (2) Consider the use of average values in trend assessment; and (3) Assess the importance of amplitude changes over short time periods when defining trends.