Recorded presentationJ3.1
How Good is Good? Establishing Performance Goals for Marine Forecast Systems
Alan Blumberg, Stevens Institute of Technology, Hoboken, NJ; and N. Georgas, T. O. Herrington, and M. S. Bruno
Coastal ocean observing and forecast systems are used for a variety of purposes. For example, they are often used in a retrospective mode to provide inputs to water quality and sediment transport models that are used for environmental planning. These inputs govern the advection, diffusion, chemical transformation, and eventual deposition of pollutants within regional water quality models such as HydroQual's RCA. The coastal ocean inputs to the environment quality models contain uncertainty that could significantly affect the results of the analysis. It becomes important to identify the biases and errors associated with the coastal ocean modeling.
This analysis looks at the performance of the latest NYHOPS (New York Harbor Observation and Prediction System). That system now provides medium term (48hr), water level, wave, 3D currents, temperature, and salinity forecasts beginning in the 1990s. Its evolving network of operational coastal ocean and estuary sensors provides real-time observations of weather and ocean conditions throughout the New York/New Jersey (NY/NJ) Harbor Estuary, Hudson River, and NJ coastal waters.
To establish confidence in this forecast system, a detailed and very comprehensive validation exercise has been conducted. Observations from 55 water level stations, 56 ADCP bins at 6 stations, 30 water temperature stations, 18 salinity stations, and 12 wave height stations formed the basis of the validation. The evaluation includes statistical comparisons of model/observed pairs (e.g., bias, root mean square, central frequency, positive/negative outlier frequency, maximum duration of positive/negative outliers)for multiple ocean parameters. The errors that remain are likely due to environmental variability that we know about but cannot yet simulate (e.g., breaking internal waves) and environmental variability that we don't even know about. These environmental uncertainties will be placed in a perspective useful for other forecast systems.
Joint Session 3, Advances in Modeling and Forecasting
Tuesday, 28 September 2010, 1:30 PM-3:15 PM, Capitol AB
Next paper