Estuarine forecasts at weather to subseasonal scales

We evaluate the skill of 35-day forecasts of water temperature, salinity, and oxygen in Chesapeake Bay using a numerical model and an extensive suite of reforecast simulations. By comparing the forecasts with both observations and the results from a realistic hindcast simulation, we find that surface temperature can be predicted with skill exceeding climatology and persistence for up to two weeks of lead time, and bottom temperature and surface and bottom salinity can be skillfully predicted for even longer. The oxygen forecasts are skillful when evaluated against the hindcast simulation but not when evaluated against observations, suggesting that improvements to the oxygen model (which is based on a simple parameterization) are necessary to accurately forecast the observed variation of oxygen. We also closely examine forecasts of the impacts of two extreme events, a hurricane and a heatwave, that previous studies have shown to be potentially forecastable with long lead times. Finally, we discuss possible improvements to the model system that could further improve the forecast skill. By proving that skillful forecasts of estuarine conditions are for long lead times, this study is an important step towards incorporating forecasts in water quality and fisheries management decisions and in preparation for extreme events.