Using Simulation-Based Inference to Determine the Parameters of an Integrated Hydrologic Model: A Case Study From the Upper Colorado River Basin
AuthorHull, Robert Bruce
MetadataShow full item record
PublisherThe University of Arizona.
RightsCopyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction, presentation (such as public display or performance) of protected items is prohibited except with permission of the author.
AbstractHigh-fidelity process-based models are currently the standard tool to explore complex watershed processes and how they may evolve under a changing climate. While these tools are powerful, calibrating them can be difficult because they are costly to run and have many unknown parameters. To solve this problem, we need a state-of-the-art, data- driven approach to model calibration that can scale to the high-compute, high-dimensional hydrologic simulators that drive innovation in our field today. Enter Simulation- Based inference (SBI), a deep learning method for finding unobserved variables and parameters that is grounded in the fundamentals of statistics. SBI as discussed here uses a neural density estimator to learn a probability distribution of parameters by comparing simulator outputs to observed data. These ‘inferred’ parameters can then be used to run calibrated model simulations. The approach has pushed boundaries in simulator-intensive research from cosmology, particle physics, and neuroscience, but it is less familiar to hydrology. The goal of this report is to introduce SBI to the field of watershed modeling by benchmarking and exploring its performance in an example case. We use SBI to infer two common physical parameters of hydrologic process-based models, Manning’s Coefficient and Hydraulic Conductivity, in a snowmelt-dominated catchment in Colorado, USA. We utilize a process-based simulator (ParFlow), streamflow observations, and several deep learning components to confront two recalcitrant issues related to calibrating models of watersheds: 1) the high cost of running enough simulations to do a calibration; 2) finding ‘correct’ parameters when our understanding of the system is uncertain or incomplete. In a series of experiments we demonstrate the power of SBI to conduct rapid and precise parameter inference for model calibration. We propose how the general purpose workflow presented here can be adapted to other hydrology-related problems, and by anyone with a simulator they just can’t seem to find good parameters for. Some challenges and recommendations to implementing SBI in more complex hydrological settings are discussed in the concluding section.
Degree ProgramGraduate College