Large Eddy Simulations for Methane Flux Monitoring
Motivation and Objectives
The quantification of ecosystem-scale methane emissions is important for reducing the uncertainty of greenhouse gas inventories. Turbulent trace gas fluxes can be difficult to measure directly at the desired spatial and temporal scales, but classical similarity theory can be used to relate them to measured time-averaged vertical concentration gradients. However, these simple flux-gradient methods have several limiting assumptions, the violation of which will contribute uncertainty to the estimated fluxes. This project aims to inform the ideal measurement locations in the field and to quantify the uncertainties of the field measurements based on analysis of large eddy simulation results.
Methods
Flux-gradient methods are based on Monin-Obukhov similarity theory. The turbulent vertical flux of a given variable is expressed as a function of its vertical mean gradient and an empirical atmospheric stability function. Since Monin-Obukhov theory assumes horizontal homogeneity and neglects advection effects, the dynamic wind flow over potentially complex terrain will not behave ideally and the estimated fluxes will possess advection-related errors. Also, since the validity of the "universal" nature of the empirical stability functions is questionable, the characterization of the effects of non-neutral atmospheric stability will be uncertain.
We are developing a workflow for mesoscale-microscale coupled CFD simulations at an arbitrary agricultural site of interest. The mesoscale simulations use the industry-standard Weather Research and Forecasting (WRF) model [1], and the microscale simulations are performed with the GPU-accelerated FastEddy (FE) large eddy simulation code (LES) [2]. There are five main steps to this workflow:
- Pre-process the WRF simulations using the WRF pre-processing system (WPS) [3], including the definition of the mesoscale domain and the definition of the boundary conditions based on real-world observed meteorological data.
- Run the WRF model with the real-data initialization.
- Obtain geospatial data for the microscale domain, including surface elevation and land cover, and build the computational mesh for the LES.
- Compute the initial and boundary conditions of the FE simulation by post-processing the WRF results and interpolating to the spatial and temporal resolutions of the LES.
- Run the FE simulation.
[1] Weather Research and Forecasting model
Acknowledgements
This research is part of a Sustainability Accelerator project at the Doerr School of Sustainability. Computational resources have been provided by both the Stanford Research Computing Center and by the National Center for Atmospheric Research.