All Climate Events

Webinar: Subseasonal to Seasonal Scale Ocean Forecasting for Chesapeake Bay and the NE United States

Download as iCal file

Wednesday, 18 September 2019, 12:00

Wednesday, September 18, 2019. 12:00PM. Webinar: Subseasonal to Seasonal Scale Ocean Forecasting for Chesapeake Bay and the NE United States. Andrew C. Ross, Princeton University. Sponsored by NOAA’s National Ocean Service. More information here

 

We present advances in subseasonal to seasonal scale forecasting of water temperature, salinity, and oxygen in Chesapeake Bay and the broader Northeast United States and discuss potential ecological applications of these forecasts. First, we use a machine learning model trained on observations from the Chesapeake Bay Program dataset to show that dissolved oxygen concentrations (DO) in Chesapeake Bay are primarily controlled by density stratification and temperature. With this model, DO can be skillfully predicted in most regions of the bay if the current stratification and temperature are known. However, forecasting DO in advance with this model, when stratification and temperature must also be forecast, is more challenging. To attempt to solve this challenge, we next produce and assess forecasts using a dynamical model of Chesapeake Bay based on the Regional Ocean Modeling System (ChesROMS). We use ChesROMS, along with an ensemble of 35-day atmospheric forecasts produced by NOAA's GEFS model as part of the Subseasonal Experiment (SubX), to run a series of 35-day reforecast simulations for Chesapeake Bay that forecast temperature, salinity, stratification, and oxygen. When compared to both an ocean model hindcast and to observations, forecasts for sea surface temperature and stratification are skillful out to about two weeks of lead time, and forecasts for surface salinity are skillful for the majority of the forecast period. Dissolved oxygen remains challenging to forecast, and we suggest that improvements to the oxygen model, which was based on a simple parameterization, may be necessary to obtain reliable forecast skill for oxygen. We also examine the performance of the forecasts for two high-impact events, a heat wave and a hurricane, that may be predictable at the subseasonal to seasonal scale. Finally, we present some early results from ongoing work to expand the scale of our forecast experiments to cover the entire Northeast U.S. marine ecosystem, and we discuss potential applications of our models to ecological forecasting.

Location  Webcast