Data assimilation and machine learning for hydrologic forecasting: review and outlook

Abstract:

We discuss recent progress, synergistic potentials and challenges in two data-centric approaches, data assimilation (DA) and machine learning (ML), for advancing hydrologic forecasting. DA is a statistical approach based on Bayesian filtering to produce optimal states and/or parameters of a dynamic model using observations. By extracting information from both model and observational data, DA improves not only the performance of numerical modeling, but also understanding of uncertainties in predictions. While DA complements information gaps in model and observational data, ML constructs a new modeling system by extracting and abstracting information solely from data without relying on the conventional knowledge of hydrologic systems. With increasing diversity and volume of data available in real-time, two data-centric approaches are getting more attention in hydrology as a means to enhance predictions in the uncertain environment with climate and land-use changes caused by humans. We summarize the commonality, differences, and recent progress of DA and ML in hydrologic sciences and discuss potential benefits of using both approaches. A focus will be given to how DA and ML can solve the issues in hydrologic forecasting including and not limited to the timing errors, the predictability of hydrologic extremes, the predictions in ungauged basins (PUB), and the theory guided data science (TGDS).