Computer Modeling and the Bay
	Polutant Transport and Predicting Water Quality | Predicting Fecal Coliform | Predicting Currents

In order to understand water currents and the movements and fate of pollutants we can go out and measure them using the appropriate equipment and techniques, and we will have a good idea of what is happening in that system for the period of measurement at the locations we sampled. If we need to know what is happening at some other time, when conditions have changed, or in another part of the area of study, we need to revisit and resample. But sampling is time consuming and expensive. And what if we wish to find out what conditions would be like if we made changes that have not yet happened, such as dredging or reducing sources of pollution? What other approaches do we have for generalizing our understanding of a system so that it is not necessary to resample it for each new condition and area we wish to explore? Computer modeling of hydrodynamics, pollutant transport and water quality provides a useful answer to these needs.

A computer model creates a simplification of the natural system that mathematically defines and calculates each of the processes that we wish to simulate. The actual area that is simulated is represented within the model as a gridwork of discrete subareas (grid cells). Properties within each grid cell are uniform over its entire area and depth at any point in time. The simulation time is likewise broken up into discrete steps, so that changes over time take place in steps rather than as a continuous process. These changes in the representation of reality are required to allow the computer to calculate what is taking place in the simulated space. The modeler compensates for these misrepresentations of the real world by making the gridded representation of the space as detailed as possible (consistent with how much we actually know about it) and the time steps as short as possible. These choices, grid size and time step, will be different for each application, depending on the actual type of model being used and the computational requirements of that model, on how much we know about the system being simulated, and on the precision required for the particular application and the amount of time the modeler has available to devote to it. The choices will always be a tradeoff of these various factors.

A computer modeling exercise might proceed like this: A deeper channel is to be dredged to the port in City Harbor. How will this change in the bottom alter currents in the channel and in other parts of the harbor?

• A sampling program is set up to measure the existing conditions - tide heights are measured over time in the harbor, currents are measured in the existing channel, profiles of salinity are collected at a series of stations over a whole tidal cycle, and a survey is taken to measure the water depths (bathymetry) of the whole harbor region and the passages leading to the harbor.
• A model system is selected that is consistent with simulating the harbor system within the time constraints of the project. In selecting the model, the modeler is also selecting a series of mathematical representations (algorithms) of water movement and conservation.
• The model is set up to represent City Harbor, with the grid size selected to provide the resolution needed to answer the questions posed. The decision of model grid detail is a made based on the modeler's experience with similar systems, as well as on how realistic the simulation needs to be (or can be afforded). Bathymetry data are used to represent the water depths in the model, and the gridded area is created to represent the area within the shoreline of the real harbor.
• The model is run using the data already collected on tides, salinity and currents. Changes may be made in the model setup to improve the fit of the model output to the observed data. This process of model "calibration" may or may not be a lengthy process, depending on the data available and the accuracy required.
• With the model calibrated, conditions may be altered in the model setup (specifically, we would change the channel bathymetry in the example) and the model is rerun. The new model output is a prediction of the tides, currents and salinity that are to be expected if the proposed dredging takes place.

The cost and complexity of computer models cover a huge range. A simple model may represent the simulated area as only a few boxes and take a day or two of the modeler's time to set up and run. A large modeling project may represent a large area with many thousands of grid cells, taking years to complete and costing a million dollars or more.

This picture shows a representation of the Providence River and upper bay of Narragansett Bay, RI. A model run is to simulate conditions in the Providence River and is shown gridded as it will be run. The grid cells in this picture have been colored to represent the depth of each grid cell in the model.

Home	Contact Us	Glossary