E2.3 Development and verification of terrestrial biophysical multi-physics processes

Abstract

Land Surface Models (LSMs) are multi-physics simulators that compute exchanges of water, momentum, energy, and nutrients at the Earth’s surface. Current generation LSMs routinely omit several critical biophysical processes such as lateral transport of water, energy, and nutrients in the subsurface; transport of water through the soil-plant continuum; and advective transport of energy in the soil. The numerical techniques used in current generation LSMs to obtain a solution of the combined linear and nonlinear system of equations are often outdated and inaccurate. The global terrestrial model community actively participates in model intercomparison projects for model validation (i.e., evaluating how well model simulations compare with observations), but mostly ignores model verification (i.e., the correctness of the numerical implementation of the model). The goals of this study are to (1) develop a numerically robust standalone library for solving global terrestrial biophysical process with support for flexible coupling strategies; and (2) verify the multi-physics library for various problems, including coupled soil and plant hydraulics.

To achieve the objectives of our study, we developed a sequential, open source Multi-Physics Problem (MPP) library for solving global terrestrial biophysical processes, and have integrated MPP with ELMv1. The MPP library has a flexible framework for coupling multiple physics processes (e.g., conservation of mass, conservation of energy, etc.) in multiple physical domains (e.g. soil, root, xylem, etc.). The underlying numerical engine of the MPP library is PETSc, whose DMComposite subclass provides an interface for individual modules to assemble parts of a global matrix for solving a tightly coupled multi-component/process problem. We use the Method of Manufactured Solutions (MMS) to verify the MPP library for a range of problems comprising of single and multiple components/processes in one or multiple physical domains. We conclude by advocating for implementation of MMS approaches to verify all components of the next generation of ELM.