Multi-Model Machine Learning Approach Accurately Predicts Lake Dissolved Oxygen with Meteorological and Hydrological Input

Highlights• Simulate dissolved oxygen in 5 lakes via multiple machine learning models.• A 2-step machine learning model workflow combines Gradient Boost Regressor and Longshort-term-memory.• A one-dimensional process-based hydrodynamic model provides ML models with predictors, including indices of lake thermal structure and mixing.• In a polymictic lake, the 2-step mixed machine learning model workflow showed over 90% true positive rate (TPR) of hypolimnetic hypoxia detection.• Predictors related to stratified conditions (i.e., Wedderburn number and the temperature difference between surface and bottom water) are essential for simulating bottom DO