Over the past decades, the Louisiana–Texas shelf has been suffering recurring hypoxia (dissolved oxygen < 2 mg L⁻¹). We developed a novel prediction model using state-of-the-art statistical techniques based on physical and biogeochemical data provided by a numerical model. The model can capture both the magnitude and onset of the annual hypoxia events. This study also demonstrates that it is possible to use a global model forecast to predict regional ocean water quality.