Machine Learning (ML) has found tremendous success in the field of hydrology in various applications ranging from forecasting to new hypothesis generation. However, ML models require extensive training with large quantities of a priori data and their results often lack physical constraints and generalizability. Traditional physics-based models are good at improving process-level understanding and prediction at certain spatiotemporal scales. The limitations of physics-based hydrologic models are that they require extensive parameterization which is often subject to human bias and executing such models becomes computationally infeasible at higher resolution across large spatial and temporal scales even with the state-of-the-art high-performance computers. The integration of ML models with physics-based hydrologic models can leverage the strengths of both models and provide innovative solutions to solving complex water quality and quantity problems. This Research Topic focuses on the integration of ML in hydrologic modeling with topics including, but not limited to:
● Hybrid ML and physical based models or differentiable modeling;
● Use of ML as surrogate models to predict variables at higher resolution and/or across scales;
● Leverage ML or hybrid models to gain physical knowledge;
● ML-assisted parameterization and calibration of physics-based hydrologic models; and,
● Use of ML for generating inputs for hydrologic models.
Machine Learning (ML) has found tremendous success in the field of hydrology in various applications ranging from forecasting to new hypothesis generation. However, ML models require extensive training with large quantities of a priori data and their results often lack physical constraints and generalizability. Traditional physics-based models are good at improving process-level understanding and prediction at certain spatiotemporal scales. The limitations of physics-based hydrologic models are that they require extensive parameterization which is often subject to human bias and executing such models becomes computationally infeasible at higher resolution across large spatial and temporal scales even with the state-of-the-art high-performance computers. The integration of ML models with physics-based hydrologic models can leverage the strengths of both models and provide innovative solutions to solving complex water quality and quantity problems. This Research Topic focuses on the integration of ML in hydrologic modeling with topics including, but not limited to:
● Hybrid ML and physical based models or differentiable modeling;
● Use of ML as surrogate models to predict variables at higher resolution and/or across scales;
● Leverage ML or hybrid models to gain physical knowledge;
● ML-assisted parameterization and calibration of physics-based hydrologic models; and,
● Use of ML for generating inputs for hydrologic models.