development of swem
The System-Wide Eutrophication Model (SWEM) was developed for the New York City Department of Environmental Protection (NYCDEP) to assistin water quality management planning. SWEM is a comprehensive regional management tool which has the capability to diagnose the causative agents of dissolved oxygen deficit in New York/New Jersey Harbor and adjacent waterways. NYCDEP has utilized SWEM to predict likely improvements in dissolved oxygen resulting from specific management actions.
New York Harbor, New York Bight, and Long Island Sound have historically been modeled as independent system compartments. Historical modeling efforts of eutrophication in the individual system compartments include the Harbor Eutrophication Model (HEM) (HydroQual, 1999a), the New York Bight Model (HydroQual, 1989), and the Long Island Sound Model, version 3 (LIS3.0) (HydroQual, 1996). SWEM represents the first regional modeling effort to integrate the three major system compartments in a complex eutroph-ication framework. SWEM was constructed to be analogous in technical complexity to its predecessor models, HEM and LIS3.0. The New York Bight Model has much simpler and less sophisticated kinetics than HEM, LIS3.0, and SWEM. SWEM has several technical advantages over both HEM and LIS3.0.
swem spatial domain
The spatial domain of SWEM includes the Hudson River from Albany to the Battery, the East River, the Harlem River, Long Island Sound, Upper and Lower New York Bay, Jamaica Bay, Raritan Bay, the Raritan River, Arthur Kill, Kill van Kull, Newark Bay, Hackensack River, Passaic River, and the New York Bight extending out into the Atlantic Ocean to the shelf break and to Cape May and Nantucket Shoals. The drainage area covered by SWEM is 57,800 km2 (34,700 square miles). The drainage area of SWEM includes 11 major tributary basins, 325 municipal wastewater treatment plants, 750 combined sewer overflows, and a population of more than 26 million.
field program - calibration and validation data
The calibration and validation of eutrophication models such as SWEM requires a synoptic database including hydrodynamic and water quality measurements for both the water column and sediment. Model calibration involves the comparison of model calculations to measured values. Model validation involves the further comparison of modelcalculations to measured values, but for a set of conditions different from calibration conditions. SWEM was calibrated to data collected during 1994-5 and validated for data collected in 1988-9.
The 1994-5 monitoring program conducted in support of SWEM included three components: loadings, hydrodynamics, and water and sediment quality. Monitoring of loadings involved sampling of thirty sewage treatment plants STPs and eleven tributaries eight times over the course of the year. In addition, eighteen CSO and fifteen storm-water locations were sampled three times and atmospheric deposition samples were collected at ten locations for seven precipitation events. Hy-drodynamic monitoring included the deployment of eight moored instruments which continuously recorded temperature, salinity, currents, pressure, andmeteorological conditions over twelve months. The water and sediment quality monitoring included more than 100 stations at two depths for nine synoptic events. The variables measured include physical parameters, plankton, nutrients, dissolved oxygen, and sediment fluxes.
Other data sources include NYCDEP Harbor Survey, monitoring by the Connecticut Department of Environmental Protection, Consolidated Edisonmonitoring of the Hudson River, and special studies performed by the Interstate Environmental Commission and HydroQual. The major source of data for model and data comparison under validation conditions is the database collected for the Long Island Sound Study for the development of LIS3.0 and the NYCDEP Harbor Survey.
Calibration and validation of water quality models is essential to demonstrate the model's credibility and utility for use as a management tool. This process consists, in simplest terms, of inputting measured pollutant loads, river flows and other necessary inputs into the model calculations and computing the predicted concentration of the response variables (nutrients, algae, dissolved oxygen, etc.) in space and time. The calculated response outputs are then compared with observed data in the water column and bed sediments of the receiving waters.
SWEM contains a number of rate and distribution coefficients for various source and sink terms. Laboratory data and repeated application of these types of models in numerous water bodies has produced reasonable, first estimate values to be assigned initially in the calibration process. Selected parameter values as appropriate are then adjusted within limits to improve the model's reproduction of observed field data.
During the calibration procedure, parameter values are adjusted one at a time so that the consequences of, or the sensitivity of the model to, an assigned parameter value is well understood. For this reason, model calibration is a long and time consuming task. In the case of SWEM, more than 100 calibration simulations were performed.
An overall goal of model calibration is to avoid having a situation where small adjustments to a single unmeasured model input produce large changes in calculated model results. Such a situation rarely occurs with sophisticated models such as SWEM which essentially are closed mass balances. Less sophisticated models which are not closed mass balances (i.e., do not include linked sediment flux or hydrodynamic transport submodels) can suffer from this problem.
A control on the calibration procedure is the necessity to reproduce observed field data for many dependent water quality variables simultaneously. For example, parameter values cannot be adjusted to produce a satisfactory reproduction of measured chlorophyll unless a satisfactory reproduction of dissolved oxygen, nutrient, and light measurements is also achieved, demonstrating that an entire process is properly modeled and the parameter adjustment is not an exercise in curve fitting to a single parameter.
The starting point for calibration of SWEM for 1994-5 conditions was the parameters used in predecessor models, particularly LIS3.0 and HEM. Only minor modifications were made to the parameters applied from LIS3.0 and HEM for SWEM. The identical set of parameters for the final SWEM 1994-5 calibration was successfully carried over to the application of SWEM to 1988-9 conditions, proving that two distinctly different sets of conditions in New York/New Jersey Harbor could be modeled with a uniform set of equations and parameters. Further, essentially the same framework as SWEM has been applied to other systems such as Chesapeake Bay and mesocosms of Narragansett Bay with consistent and similar parameters (Di Toro et al., 2000). These results demonstrate that modern eutrophication models such as SWEM are valid and reproducible and can be used for predictive purposes.
The calibration and validation of the water quality model portion of SWEM included model and data comparisons both along spatial transects and over time at specific locations. Specifically, an individual SWEM calibration or validation simulation included model and data comparisons for approximately 2,600 spatial profiles and 650 temporal profiles. The calibration and validation of the SWEM benthic sediment model involved model and data comparisons for almost 500 temporal profiles per simulation. The major water quality variables of concern for SWEM calibration and validation included salinity, temperature, dissolved oxygen, chlorophyll-a, BOD-5, carbon, nitrogen, phosphorus, and silica.
Figures 11.3 and 11.4 present model and data comparisons for August 1995 for eighteen variables along a spatial transect that runs down the centerline of the Hudson River Estuary south from Poughkeepsie, New York through the Battery and the Narrows toward the apex of the New York Bight. For comparison purposes, a ten-day average of model results closest to the time period during which the data were collected is used.
salinity and temperature
On Figure 11.3, panels A and B show model and data comparisons for salinity and temperature. Comparisons are shown for both surface and bottom waters. The model results for salinity are presented by two sets of curves. The dashed lines show the salinity calculated by the hydrodynamic submodel. The solid and broken lines show the salinity used in the carbon production submodel. Temperature results as calculated by the hydrodynamic submodel are shown by solid (surface) and broken (bottom) lines. The general agreement between model results (ten-day averages) and the observed data for salinity and temperature (grab samples) is an indication that the physics and transport have been correctly calculated. Both model and data show that the Hudson is a vertically stratified system.
Total nitrogen. On Figure 11.3, panels F, G, H, and I show model and data comparisons for several organic and inorganic nitrogen forms (TON, NH4, NO2 + NO3, TN). The model results shown by the solid and broken lines reproduce the data extremely well, capturing sharp gradients in all nitrogen forms. The increase in ammonia
Was this article helpful?