|
System-wide Monitoring Program
Synthesis of the Water Quality Data DISCUSSION Water Quality Synthesis The 44 water quality sites of the NERR SWMP represent multiple geographic regions, climates and salinity regimes. These sites also have varying levels of anthropogenic influence, ranging from no developed land in the immediate watershed of the site to having >20% of the watershed either developed or utilized for agriculture or silviculture. Most of the sites are located in shallow water (<2 m) and the daily fluctuations in temperature, salinity and dissolved oxygen are representative of conditions that occur in shallow water estuarine environments throughout the nation’s estuaries. Location of water quality monitoring sites may have a profound impact on the magnitude and variability of water quality variables; hence data from only a couple of sites may not provide adequate representation of conditions that occur throughout the estuary, particularly with respect to deep areas. Dissolved Oxygen Dissolved oxygen (DO) is a fundamental requirement for maintaining a diverse estuarine ecosystem. Oxygen content of estuarine waters varies with temperature, salinity, turbulence, photosynthetic activity of algae and plants, and atmospheric pressure. Temperature and biological activity can greatly influence variability in DO at daily and seasonal time scales. Biological respiration, including that associated with decomposition, reduces DO. Spatial variability in DO concentrations can occur within estuaries as a result of differing biological and physical processes. Increased loads of organic matter and nutrients can decrease DO concentration as a result of increased microbial respiration that occurs during organic degradation processes (Chapman 1992). In addition to biological and chemical processes, dissolved oxygen concentration is also affected by land use patterns. Among watershed types (forested, industrial, urban and suburban) in the Charleston Harbor estuary, Lerberg et al. (2000) noted that hypoxia was regularly observed in both developed and undeveloped creeks. Wenner et al. (1998) noted that the frequency and duration of low DO events tended to be greater for creeks with developed watersheds, suggesting a relationship between land use and dissolved oxygen concentration. Because low DO concentrations regularly occur in shallow tidal creeks, frequency and duration of these events may be a better indication of degraded water quality in tidal creeks rather than just occurrence of low DO. Increases in the extent and volume of hypoxic and anoxic water in the marine environment have occurred globally over the last several decades and are an important environmental issue that coastal resource managers must address (Diaz and Rosenberg 1995). According to the Environmental Protection Agency (EPA), low dissolved oxygen is one of the most common causes of not achieving designated uses in the Nation’s estuaries (EPA 1990). The EPA found that about 19-25% of the bottom water area in the Virginian Province had DO concentrations <5 ppm and 4-8% had concentrations <2 ppm. These hypoxic conditions occur primarily in the upper portion of Chesapeake Bay where areas continuously experience low dissolved oxygen, rather than just at night, during the summer months (Whitledge 1985). Determination of DO concentrations is an important part of water quality monitoring because oxygen influences nearly all chemical and biological processes within water bodies. A threshold concentration of 4-5 ppm is used by the EPA and by some states (e.g., South Carolina) to set water quality standards (EPA 1990, DHEC 1993). When DO concentrations decline below 2 ppm, generally accepted as the minimum oxygen level necessary for sustaining animal life and reproduction, hypoxia results. It has been suggested that exposure to anoxic and hypoxic DO conditions is a controlling factor in the distribution of aquatic organisms (Diaz and Rosenberg 1995, Lerberg 1997). Subtle effects on physiology, reproduction, and behavior may occur from lowered DO, while severe hypoxia can cause shifts in distribution or abundance of populations (Fry 1971, Summers et al. 1997). Conversely, excessively high levels of dissolved oxygen (supersaturation) may also be indicative of high primary production. Thus, DO concentration provides an index of the balance between production and oxygen consumption. In areas with elevated primary production, supersaturation values > 120% are not uncommon during daylight periods and may reach 500 % in shallow estuarine waters (Kalle 1972). While there is little information available on supersaturation effects on biota, high oxygen concentrations may have a toxic effect on plants and animals through the formation of reactive oxygen species (Turner and Brittain 1962, Dalton 1995). In water that is over-saturated with oxygen, respiration of plants is markedly increased (Gessner 1959). Supersaturation was most prevalent in spring at NERR sites and may be related to increased algal production as temperatures rise during daylight hours. Lentic conditions, in combination with lower suspended solids and increased water clarity, have resulted in high algal biomass, higher pH readings, and frequent dissolved oxygen supersaturation as well as nutrient depletion in slack-water areas of the lower Mississippi River (Beckett and Pennington 1986). A relationship between high levels of chlorophyll a, oxygen supersaturation (>120%) and non-conservative nutrient concentrations have been noted for other estuaries during low-flow conditions in spring and summer (Rendell et al. 1997). Our synthesis of water quality data from the NERR SWMP indicates that water quality in the Reserves has not experienced many of the problems found in other more populated areas of the country. Although hypoxic events occur frequently in the NERRs, these events are generally of short duration. Short periods of hypoxia are generally within the tolerance range of many aquatic animals that take up more oxygen and transport it more effectively to cells. A few sites experienced hypoxic events that lasted for >24 h, a duration which can affect growth and survival. The longer the duration of hypoxia, the greater the impact to juvenile and larval growth and survival (USEPA 2000). While the percent of time that DO was less than 28% saturation (~2 mg/l) varied substantially among reserves and between years, most of the hypoxic events occurred in summer when water temperatures were highest. Seasonal hypoxia, as experienced by many of the Reserve sites, develops at bottom depths when respiration in the water and sediment depletes oxygen faster than it can be replenished. Reserves in the Gulf of Mexico and Caribbean Sea had the highest occurrence of hypoxia events > 24 h duration, which was associated with the high temperatures observed at these Reserves. Hypoxia may also be related to estuarine circulation patterns and whether circulation is adequate to promote flushing of water over one or several tidal cycles. Daily cycles of hypoxia are related to lack of stratification in shallow estuarine habitats where nighttime respiration depletes DO. Although it is clear that both natural factors (i.e., degree of stratification, water temperature and circulation patterns) and anthropogenic factors (i.e., nutrient loading) can contribute to low DO events, further investigation is needed to evaluate long-term trends and inter-annual variability in low DO events recorded from the NERR SWMP. Incorporation of additional information such as weather patterns, rainfall, sea level rise, nutrient loading, habitat differences, and land cover are required to better understand such trends. The large fluctuations in dissolved oxygen which occur over short-time periods in the NERRs demonstrate the need for long-term continuous measurements to estimate the frequency and duration of exposure to low DO. For example, hypoxia events occasionally lasted for greater than 24 hours during the first 48 hours post-deployment; thus, less frequent sampling would underestimate the extent to which natural and anthropogenic hypoxia occur. Similarly, harmonic regression models required enormous amounts of input data in order to fit these models to the observed water quality data. The vast amount of data collected by the NERRs may also prove useful in developing models that coastal zone managers could use to predict oxygen conditions in shallow estuarine systems. The inability to correctly characterize dissolved oxygen conditions in estuarine systems impairs conclusions from waste load allocation models and other water quality evaluations that are used to develop wastewater treatment strategies and requirements (Summers et al. 1997). Salinity Salinity among sites in the NERR SWMP ranged from limnetic to euhaline. Few sites within reserves were statistically similar in mean salinity or salinity range; however, these differences were typically small and may not be biologically significant. Large differences in salinity conditions between sites within a Reserve were observed at several reserves. Seasonal differences in salinity occurred at most reserve sites and were probably a reflection of seasonal changes in precipitation and evaporation. Tidal and diel cycles also appeared to strongly influence salinity at reserves. Salinity may be the most important factor affecting the distribution of estuarine organisms. It is an essential element in determining estuarine habitat and directly affects the distribution, abundance and composition of biological resources. Although variability in salinity influences distributions of organisms, most of the nation’s estuaries experience significant salinity variability both daily and seasonally. The frequency and magnitude of this variability differs in each estuary, largely as a result of fresh water inflow, tides, winds, and coastal shelf processes (Orlando et al. 1994). Depending upon the time scale, variability in salinity can be predictable or highly episodic. Fluctuation in salinity at hourly intervals is most often attributed to the tidal cycles. With a few exceptions (i.e., Weeks Bay, Old Woman Creek, Padilla Bay-Joe Leary Slough), NERR sites experience semi-diurnal tides in which high salinity oceanic waters enter the estuary during the flood tide stage. At the time scale of days to weeks, variability is most often attributable to short-duration fresh water pulses, the spring-neap cycle, and frontal passages. Seasonal effects are responsible for most of the net changes in annual salinity, due to differences in fresh water discharge and prevailing wind speed/direction. Intense, episodic events such as floods also have a dramatic effect on salinity, as indicated by examination of salinity conditions following hurricanes that occurred in the North Carolina and Virginia NERRs during 1998. Tropical storms and high volume releases from control structures can eliminate vertical stratification and suppress tidal influences (Orlando et al. 1994). Anthropogenic factors can affect salinity variability at a range of spatial scales. Changes in salinity at the watershed level have occurred due to alterations of flow regime. Within individual tidal creeks, salinity fluctuates in response to runoff events. In the Charleston Harbor estuary, salinity varies within creeks, among creeks, and among watershed classes (Lerberg et al. 2000). Within this estuary, tidal creeks are much more dynamic than most large estuaries and typically experience salinity fluctuations > 6 ppt over tidal cycles (Lerberg et al. 2000). Variance in salinity, as represented by the salinity range, is greater in creeks surrounding suburban, urban and industrial land uses, than in forested upland creeks, which was attributed to increased amounts of impervious surface at the anthropogenically impacted sites (Lerberg et al. 2000). Increased amount of impervious surface alters hydrodynamic processes, especially the rate at which precipitation runoff reaches receiving waters. As additional data are collected from the NERR SWMP, it would be informative to examine whether extreme and highly variable fluctuations in the salinity pattern are an indicator of hydrodynamic changes to monitoring sites resulting from watershed development or land use changes. Periodicity (Harmonic Regression Analysis) It was abundantly clear, even after only a cursory inspection of only a few of the raw data series, that the periodicity in these data was strong and very complicated. It was also clear that there were many complicating factors present including meter probe accuracy, deployment effects, and weather anomalies. Isolation of periodicity from this diverse web of signals, and subsequent interpretation of the results without oversimplification of the patterns presented, proved challenging. Comprehensive interpretation of these diverse results, especially for salinity and dissolved oxygen, will require active participation of individuals more familiar with the sites and with the physical phenomena driving these measured water quality variables. The standard errors of the estimates in the harmonics figures are relatively small, indicating that estimated annual averages based on these harmonic regressions are fairly accurate; however, many of the sites within reserves are significantly different from each other. Sites within reserves are often as different, or more different, than sites between reserves. It is clear that enough data are being collected (i.e., every 30 minutes) to obtain accurate estimates. Results would be improved greatly if meters delivered more reliable results for longer periods of time, deployment effects were reduced (i.e., carefully deploying new meters to obtain a more continuous signal), and annual sampling regimes were more complete and uniform across sites. In some cases, seemingly anomalous site results may be due to incomplete or non-uniform sampling regimes. Several sites did not collect data at certain seasons of the year (i.e., winter for reserves in the northeast) or contained gaps of more than 45 days between deployments. In these cases, fitted seasonality curves were truncated to correspond only to periods of the year where gaps were less than 45 days. Non-uniformity in seasonal sampling confounded the predicted values for water quality variables. For example, mean predicted water temperatures for East Coast sites increased from north (New England) to south (Mid-Atlantic, Southeast Coast, and Gulf of Mexico/Caribbean). An exception to this trend was the Chesapeake Bay MD NERR, which had an unusually high mean temperature relative to other sites in its region; however, the anomalous mean predicted water temperature for this site was due to the fact that sampling was primarily conducted in summer. The approach taken here, to separately model each deployment's series with a harmonic regression and then analyze summary measures for each deployment across years and sites, is statistically valid and has been to some extent successful. In particular, the 37-term deployment-level harmonic regression model has fit surprisingly well across about 10,000 deployments having a great diversity of patterns. A number of the descriptives calculated from these fitted regressions have not been as useful as hoped; however, descriptives based on first-full 12.42 hour and 24 hour signatures in each deployment tended to have large standard errors, and for some sites showed anomalous results. If similar analyses are to be performed in the future, dropping cross-product terms from the deployment level regressions should be seriously considered. Though the inclusion of cross-product terms led to better-fitting models, especially at extreme values in the data, they also complicated the fitted models substantially, making it difficult to separate the patterns due to 12.42 hour and 24 hour cycles. Omission of cross-product terms would produce 12.42 hour and 24 hour "average" signatures for each deployment, which could then be more easily viewed and described, and which would be more accurately estimated than first-full-cycle profiles. Whether these "average profiles" would be a fair representation of the 12.42 hour and 24 hour cycles for a given deployment should be resolved. Metabolic Properties Production, respiration and net ecosystem metabolism were calculated using dissolved oxygen data (% saturation, mg/L), water temperature (°C) and salinity (ppt) data from 27 sites (14 reserves) encompassing a wide range of estuarine conditions (i.e., freshwater, salt marsh, and mangrove swamp). All sites were subjected to upland runoff and several sites were located on large bodies of water that were well flushed by tidal cycles. In general, results were consistent with our understanding of how estuaries function. At individual sites, an autotrophic versus heterotrophic status appeared to be related to the habitat type, specifically whether it was dominated by submersed aquatic vegetation or not. Where primary producers such as SAV dominated, systems were autotrophic. In marsh-dominated systems, organic material produced in the marshes is exported to tidal creeks and bays where decomposition occurs leading to heterotrophic conditions in the water column. Similarly, in the open water of estuarine bays and rivers, allothchonous inputs of organic carbon can also lead to heterotrophic conditions. Smith and Hollibaugh (1993) found that 17 of 27 marsh, estuarine or coastal ocean locations were net-heterotrophic. This review along with the results of this study suggests that aquatic heterotrophic conditions are a common feature among many estuarine systems. The results from this study suggest that freshwater flow and residence times are critical variables controlling metabolic rates. In order to provide a comprehensive evaluation of production and respiration at the ecosystem level throughout the entire NERR system, metabolic rates should be calculated for the eight Reserve sites not included in this synthesis. Furthermore, additional metabolic calculations should be made for all Reserve sites for the years not included in this study (1995, 1999, and 2000) to better assess inter-annual variation at NERR sites. Analysis of this longer data record would also permit examination of how inter-annual changes in the timing and amounts of freshwater (e.g., El Niño and La Niña climates) affect metabolic rates at each Reserve. Metabolism results suggest that nutrient (mainly nitrogen) loading to these systems may be a critical factor in determining whether systems are autotrophic or heterotrophic. While nutrient concentration data can be helpful, many factors (i.e., uptake, regeneration, burial, advection) affect concentrations making it difficult to interpret the results. Thus, annual estimates of nitrogen and phosphorus loadings to the systems would be the most useful data for interpreting metabolism results. Additional ancillary data such as watershed area, detailed bathymetry and hypsography (i.e., mean depth, estuarine area at different depths, estuarine volume), freshwater flow, and residence time (including seasonal and spatial variability) are also needed for each Reserve to better interpret metabolic data. We believe that estimating metabolic rates provides valuable information for the individual Reserves and the Reserve system as a whole. Additional deployment sites should be selected to ensure good estimates of metabolic rates. Sites should be chosen so that biological processes dominate over physical processes. Two of the sites examined, Saw Kill Creek (Hudson River NERR) and Joe Leary Slough (Padilla Bay NERR), are good examples of sites where physical processes, rather than biological processes, dominated oxygen exchange; thus, the methods for calculating metabolic measurements were not appropriate at these sites. Program Overview and Recommendations Instruments and Approach With the exception of dissolved oxygen drift due to instrument fouling (pp. 18-19), data sondes provided reliable and comparable water quality data for the range of geographical areas and environmental settings sampled. This instrumentation was well suited for collecting the data needed by a national program that has the objective of characterizing short-term variability and long term changes in aquatic environments. The instruments were relatively easy to calibrate and maintain and generally hold calibration for the parameters measured for the 10-14 day deployment periods. The only problem noted was that at some sites, mainly those located at southern latitudes, the dissolved oxygen probe frequently showed evidence of fouling after about 48 hours. Fouled probes resulted in an underestimate of the degree of supersaturation and an overestimate in the severity of hypoxia that occurred at a site. The Research Coordinators estimate that approximately one full time technician is required to collect the water quality data for two deployment sites using the data sonde technology. Technician responsibilities include maintaining and deploying the instruments, conducting a quality assurance program to ensure high quality data are collected, downloading the data, and transfer of data and metadata to the Centralized Data Management Office (CDMO). All of the parameters evaluated by this study (depth, conductivity/salinity, temperature, dissolved oxygen as % saturation, and dissolved oxygen concentration as mg/l) provide valuable information about estuarine system dynamics. The depth data were particularly important for evaluation and interpretation of the temporal variability in parameters associated with tides. We did not evaluate the pH data based on a recommendation from the expert panel. The panel noted that at this time they did not know how to interpret the results of these analyses and several panel members expressed concerns about the reliability of the data over the several week deployment periods that were used. We recommend that the pH data be evaluated in the future to assess its usefulness and to determine its reliability. Recent data suggest this information may be related to the growth and productivity of some bivalve species (A. Ringwood, personal communication). The procedures used to select and establish deployment sites were not standardized across NERR sites. Many of the reserve programs selected sites that their scientists and managers felt represented either: (1) typical relatively pristine; and (2) typical “disturbed” habitat. Other reserve programs, however, selected sites that represented conditions or habitats that they felt were typical of the ecological conditions in the reserve or provided information that addressed questions of interest to the reserve. In many cases, site selections were made without detailed knowledge of the factors associated with the site that affect water quality (i.e., estuarine circulation, size of drainage basin, land cover in the drainage basin). Hydrographic processes and factors affecting water quality at some of the deployment sites are too complex to characterize using the limited data collected by the SWMP. At almost all reserves, ancillary data such as bathymetry, residence time of water at the sample site, freshwater inflow, pollutant loadings, watershed size, and land cover within the drainage area were not available or only partially available to assist with interpretation and synthesis of the SWMP data. Although general protocols for instrument deployment were developed by SWMP, differences in deployment methods affected the comparability of the data across sites and regions. Depth of deployment of instruments above the bottom sediment was not uniform for all sites. At most sites (n=24), instruments were moored 0.1 to 0.6 m (mean = 0.3 m) above the bottom sediment, a variation approximately equivalent to the length of the instrument itself. Assuming that the definition of the mooring depth was consistently used by these sites (i.e., depth always refers to the depth of some point on the instrument such as the probe), this is not a troublesome difference. At three Reserves (Padilla Bay, Narragansett Bay, and Wells Bay), depth of deployment above the bottom sediment differed between sites within the same Reserve by 0.5 to 1.0 m. At these sites the mean depth of deployment of the instrument above the bottom sediment was 0.9 m (range = 0.3 to 1.5m), similar to the mean depth of deployment of instruments above the bottom sediment at Waquoit Bay (0.8 m). Mean depth of deployment of instruments at these sites is almost three times the mean depth of deployment for all other sites. Instrument depth above the bottom sediment was not provided for 10 sites. In addition to deployment depth, standardization of the actual location (i.e., mid-channel, adjacent to the shoreline) of sampling sites needs to be considered. Although logistical considerations often determine where and how instruments are deployed, subtle differences between sites can result in pronounced differences that preclude comparison between sites. For example, at both monitoring sites used by the Sapelo Island NERR and at the Great Bay Buoy, instruments were deployed from a floating platform rather than a fixed platform; thus, depth data could not be compared to other sites. These differences frequently have major influence on the values of the parameters measured as well as analysis results (e.g., comparisons within and among reserves). Most importantly, however, they affect the value of the collected information for addressing regional and national issues (e.g., status and trends in the extent and severity of hypoxia). In order to reduce discrepancies in measurements of water quality variables due to sampling design, we recommend that Reserves collectively develop a protocol, and provide the appropriate training, to standardize instrument deployment practices. Data Management A number of problems were encountered in the process of incorporating the water quality data into a central database (MS Access) to obtain a continuous time series of data for all sites. A summary of the problems encountered with the data were first presented at the March 2000 Research Coordinators Meeting in Williamsburg, VA, but will be briefly revisited here. First and foremost, the quantity of data was highly variable among sites. Data gaps were due to numerous factors including equipment failure, staffing problems, or weather (particularly freezing water temperatures that precluded winter observations at NERRs located in the Northeast and Lake Erie). Second, the quality of data was also highly variable, largely due to the utilization or under-utilization of the “anomalous data” and “missing data” sections of the metadata report. The “anomalous data” section is included to allow each Reserve to document and explain what constitutes suspect data and why these data were removed; however, this is frequently not performed. It often appeared that some Reserves only assessed the quality of data to determine if values were within the instrument range, but not whether the data were abnormally high or low. In contrast, some Reserves reported that data were retained due to a lack of justification to remove the data; however, they felt the data were suspect. Regarding the “missing data” section, many Reserves felt that this section should only reflect data that were missing due to instruments not being deployed. As a result of this perceived redundancy, the comments in the “missing data” section often only provided a reference to the “anomalous data” section. To address this issue, the CDMO decided in August 2000 to modify the wording of the protocols for both the “anomalous data” and “missing data” sections. Lastly, we suggest that Research Coordinators include a justification of sample site selection along with the metadata. Deployment and retrieval data need to be included with the metadata, and the CDMO has taken necessary action to implement this recommendation. Deployment and retrieval data were necessary to determine deployment duration and, subsequently, the effect of dissolved oxygen drift at 1, 2, 4, 7, and 14-day intervals post-deployment due to bio-fouling (See DO discussion). Deployment-retrieval information was also necessary to explain abrupt shifts in observed water quality data, which often resulted when even slight differences in deployment of meters occurred with respect to water depth, tide stage, or time of day (See Discussion section on Instruments and Approach). Inclusion of deployment and retrieval data was also particularly helpful when Reserves simply switched out YSI meters and no gap (i.e., no “missing data”) was observed. Identification of abrupt shifts in water quality parameters due to new deployments was easily accomplished by incorporating deployment data into scatter plots of raw data. As a result of the ease and benefits of these procedures, we strongly recommend that all Reserves visually examine their data using scatter plots containing deployment information before completing the “anomalous data” section. Standardization of the NERR SWMP with a readily available, relational database such as MS Access is highly recommended for managing data at the Reserve level. The major benefit of managing data in MS Access is the large storage capability of this program, which enables data from multiple sites to be stored in a single database. Because all of the data are contained within a single database, representing a semi-continuous time series, simple queries can be used to extract facets of the data that can then be exported into MS Excel for graphing. Furthermore, use of a standard database format would save a tremendous amount of time when preparing data from the NERR SWMP for synthesis reports and projects such as this one. Synthesis Approach Our approach to analysis of the water quality data collected by the SWMP worked well for this synthesis, given the large amount of data and the complexity of the analyses. A particularly useful element of the process was convening an expert panel. The panel, consisting of scientists, NERR staff and statisticians experienced in the collection, analysis and interpretation of water quality data, assisted in developing the following analytical approach for the project:
The expert panel was especially valuable in identifying key parameters on which to base analyses and key indicators and measures for defining the extent and severity of hypoxia; discussions about stratification, summarization and evaluation of the data; and providing input on alternative sampling approaches and trends assessment. It was also helpful to have the expert panel available for guidance and advice throughout the project. We recommend a group of experts to participate in all aspects of future synthesis efforts. |
|||
Ocean and Coastal Resource Management | Ocean Service | NOAA | Department of Commerce
Privacy Policy | Disclaimer http://nerrs.noaa.gov/Monitoring/Synthesis6a.html