This analysis demonstrates machine learning effectively predicts salinity changes in a tidal estuary under climate change, indicating potential to reduce simulation uncertainty.
Key Points
Predictions of salinity improved by reducing root mean squared error by up to 20% with machine learning models.
The supportive ensemble methods include generalized linear model to address local simulation biases effectively.
Artificial neural networks showed superior performance in simulating daily salinity compared to random forest and support vector regression.
Implications reveal the potential of machine learning and hydrodynamic modeling in future estuarine studies.