Designing Eddy Covariance Networks to Characterize and Quantify Surface CO2 Leakage

Recent studies suggest that micrometorological methods, when combined with geophysical inversion techniques, can accurately locate, quantify, and define the spatial distribution of CO2 efflux arising from natural sources or simulated leaks from carbon storage reservoirs. In this contribution, we build on this work by presenting a method that optimizes an eddy covariance monitoring network to resolve the spatial distribution and magnitude of a hypothetical leakage scenario. Specifically, we convolve source area footprints that represent atmospheric conditions typical of a particular site with a set of different leakage scenarios to determine those locations at which eddy covariance station measurements yield the most information about the surface CO2 flux distribution. For each selected location, a synthetic time-series is produced, noise is added commensurate with anticipated site conditions, and the spatial distribution of surface CO2 fluxes is estimated using a linear least-squares inversion of the time series and footprint functions. The difference between this inverted flux and the hypothetical flux distribution is used to determine the location an additional station using convolution of this difference with the footprint functions. The new station's time series, with noise added, is then included in the linear least-squares inversion, and the process is repeated for an arbitrary number of eddy covariance station locations. By assessing the goodness of fit between the hypothetical and inferred flux distribution, as well as the total CO2 leakage discharge, we can determine the optimal number of stations, as well as their locations, that is required to accurately quantify a particular leakage scenario. We applied this method to design an experiment at the ZERT CO2 release facility in Bozeman, MT, where a CO2 release from a horizontal well in the shallow subsurface produced a series of patches of high CO2 flux aligned along the surface trace of the well. Our analysis suggests that 2-3 eddy covariance stations should be able to accurately characterize the spatial distribution of this CO2 leakage signal, and quantify the total CO2 discharge from the well. Our results provide a rational means of designing experiments that strive to quantify the total amount, spatial distribution, and temporal variability of both anthropogenic and natural surface CO2 leakage in areas where geologic or anthropogenic information constrains a set of likely leakage geometries, such as along geologic faults or wells.