What Role Does Hydrological Science Play in the Age of Machine Learning?

Abstract

This paper is derived from a keynote talk given at the Google's 2020 Flood Forecasting Meets Machine Learning Workshop. Recent experiments applying deep learning to rainfall-runoff simulation indicate that there is significantly more information in large-scale hydrological data sets than hydrologists have been able to translate into theory or models. While there is a growing interest in machine learning in the hydrological sciences community, in many ways, our community still holds deeply subjective and nonevidence-based preferences for models based on a certain type of "process understanding" that has historically not translated into accurate theory, models, or predictions. This commentary is a call to action for the hydrology community to focus on developing a quantitative understanding of where and when hydrological process understanding is valuable in a modeling discipline increasingly dominated by machine learning. We offer some potential perspectives and preliminary examples about how this might be accomplished.

Issues in generating stochastic observables for hydrological models

“…Indeed it has been suggested that machine learning method might be more successful in converting rainfalls into streamflows than models based on process representations (e.g., Kratzert et al, 2019; Nearing et al, 2021). Such data‐based methods are heavily dependent on the range and quality of the training data set available, though they will be able to compensate for inconsistencies in a data set if those inconsistencies show some form of consistent structure (see the discussion in Beven, 2020b).…”

AI-Driven Cross-Domain Knowledge Discoveryand Hypotheses Generation for Enhanced Earth System Predictability

“…parametric). These limitations are especially severe for hydrologic models of Earth's water cycle extremes and their impacts (Nearing et al, 2021). Data-driven knowledge acquisition could help to identify and reduce these model biases (structural errors) thus improving the model predictability.…”

“…parametric). These limitations are especially severe for hydrologic models of Earth's water cycle extremes and their impacts (Nearing et al, 2021). Data-driven knowledge acquisition could help to identify and reduce these model biases (structural errors) thus improving the model predictability.Earth scientists have abundant physics-based experimental data (both observational and simulated), but they do not yet know how to discover and extract predictive knowledge from these large heterogenous dynamic data streams.…”

Emerging technologies and radical collaboration to advance predictive understanding of watershed hydrobiogeochemistry

“…Given the complexity of hydrobiogeochemical data and processes that occur across scales and compartments of a watershed, this strategy holds significant promise for advancing predictive understanding of watershed hydrobiogeochemical behavior. Nearing, Kratzert, et al (2020) advocated the importance of integrating ML into hydrological workflows.…”