Title: Facilitating Flexible Hydroclimate Digital Twins Through Land Surface Model Emulation
Presenting Author: Jonathan Hobbs
Organization: Jet Propulsion Laboratory
Co-Author(s): Matthias Katzfuss, John T. Reager, Paul Wiemann, Matthew Bonas, and Manju Johny

Abstract:
Uncertainty quantification (UQ) for climate impacts often involves a combination of suitable Earth-system models and statistical methods to represent the probability distribution of geophysical quantities of interest. Devising a flexible representation can be challenging for geophysical fields with complex spatio-temporal correlation patterns. Ensemble experiments can provide valuable information, but computing and data storage resources limit the ensemble size and scope. This work implements an approach known as a Bayesian transport map (BTM) to estimate the joint probability distribution for fields of one or more related quantities of interest. The BTM also enables fast simulation from the distribution to produce realistic ensembles of arbitrarily large size, enabling UQ for science and applications. The methodology is demonstrated for hydrologic variables, including snow, soil moisture, and runoff, from an ensemble of land-surface model (LSM) experiments conducted under contemporary and future climate scenarios.