A new approach used to analyse annual variations, may be important in systems where ecological forecasts are needed in the face of long-term changes resulting from climate change, biological invasion or other phenomena.
Annual hypoxia forecasts for the Gulf of Mexico and Chesapeake Bay have been performed to raise awareness of hypoxia and how it is connected to nutrient loads, to manage it better and to test our understanding of the phenomenon. This last aspect is a particularly important way to test the validity of model-derived scenarios for decision makers.
The overall suitability of a model is a balance between how accurate it is and how quickly it can respond to ecological changes. This is important because lakes, rivers and other water bodies have experienced changes in their sensitivity to nutrient loading over the past couple of decades.
“These changes in a system state are difficult to recognize in real time because there is substantial year-to-year variability”, says researcher Mary Anne Evans. “The forecasting ability of a hypoxia model that incorporates Bayesian parameter uncertainty and variability for the Gulf of Mexico and Chesapeake was, surprisingly, optimized by calibration to short, recent datasets.”
Using a “moving-window” calibration approach, with windows of different sizes, she found that the model calibrated with a three-year moving window had a higher accuracy than models using longer windows in both systems. “The approach, which can also be used to analyse annual variations, may be important in other systems where ecological forecasts are needed in the face of long-term changes resulting from climate change, biological invasion or other phenomena.”
“Future work will explore the interaction of upper- and lower-water layers. In the Gulf of Mexico the upper layer, in which phytoplankton biomass is produced, flows in the same direction as the lower layer, in which biomass is decomposed and oxygen is used up.”