Hydrodynamic and biochemical impacts on the development of hypoxia in the Louisiana–Texas shelf – Part 2: statistical modeling and hypoxia prediction

Abstract. This study presents a novel ensemble regression model for forecasts of the hypoxic area (HA) in Louisiana–Texas (LaTex) shelf. The combines zero-inflated Poisson generalized linear (GLM) and quasi-Poisson additive (GAM) considers predictors with hydrodynamic biochemical features. Both models were trained calibrated using daily hindcast (2007–2020) by three-dimensional coupled hydrodynamic–biogeochemical embedded Regional Ocean Modeling System (ROMS). Compared to ROMS hindcasts, yields low root-mean-square error (RMSE) (3256 km2), high R2 (0.7721), mean absolute percentage biases overall (29 %) peak HA prediction (25 %). When compared shelf-wide cruise observations from 2012 2020, our provides more accurate summer forecast than any existing (0.9200); RMSE (2005 km2); scatter index (15 %); (18 %), fair-weather windy-summer predictions. To test its robustness, is further applied global produces 2012–2020 adjusted HYbrid Coordinate Model (HYCOM). In addition, sensitivity tests suggest an aggressive riverine nutrient reduction strategy (92 needed achieve goal 5000 km2.