Abstract:
Geophysical methods are increasingly used to monitor subsurface processes, providing valuable information to qualitatively constrain or quantitatively calibrate predictive models for flow, transport, and other physical or chemical processes. In groundwater hydrology, the dominant paradigm for model calibration entails solution of deterministic (or stochastic) inverse problems to identify optimal estimates (or realizations) of model parameters; this approach is limited in several respects. First, sources of structural model error (e.g., incomplete physics, errors in boundary conditions, etc.) are commonly discounted, the consequences of which are rarely quantified but possibly important for joint inversion. Second, the conventional approach is poorly suited to real-time (`online') estimation or forecasting, as data are analyzed in batch rather than recursively assimilated; i.e., the models calibration is not amenable to incorporation of new data as it becomes available. In this presentation, some recent examples of Kalman-based filtering and smoothing are reviewed, with applications to recursive estimation for groundwater/surface-water exchange and groundwater recharge. The algorithms discussed are amenable to real-time application for estimation and potentially forecasting. Filtering and smoothing frameworks (e.g., Kalman-based) allow for calibration of mechanistic process-based (`white-box') models, or data-driven time-series or transfer function (`black-box') models, or hybrid (`grey-box') models between these end members. In other fields, such as real-time autonomous navigation, recursive estimation and forecasting are common and used effectively within control systems. We see enormous potential for expanded use of filtering and smoothing in hydrology and hydrogeophysics; however, this will require (1) a reframing of how hydrologists and geophysicists conceive of model calibration, and (2) an understanding that models exist on a continuum between white- and black-box models.
Bio:
Fred Day-Lewis joined PNNL in 2021 as a Chief Geophysicist in the Environmental Subsurface Science Group within the Earth Systems Science Division. Prior to starting at PNNL, Fred worked for the U.S. Geological Survey for 18 years as a Research Hydrologist. Fred has worked on a variety of applied-research projects related to subsurface characterization and monitoring, groundwater remediation, groundwater/surface-water exchange, geophysical inverse problems, thermal methods, and hydrologic parameter estimation. Fred currently serves as an associate editor for the journal Groundwater. He previously served as an associate editor for Water Resources Research, Geosphere, and Hydrogeology Journal. Fred is a past president of the American Geophysical Union Near Surface Geophysics Section, and a current Vice President of the Environmental and Engineering Geophysics Society. He was elected Fellow of the Geological Society of America in 2015 for seminal contributions to hydrogeophysics, and he was elected a Laboratory Fellow at PNNL in 2021. He is a 2023 recipient of the Harold Mooney Award from the Society of Exploration Geophysicists for contributions to the near-surface geophysics community. He received a PhD from Stanford University and BA and BS from the University of New Hampshire.
