An Interpolation Method to Reduce the Computational Time in the Stochastic Lagrangian Particle Dispersion Modeling of Spatially Dense XCO 2 Retrievals.

A growing constellation of satellites is providing near-global coverage of column-averaged CO2 observations. Launched in 2019, NASA's OCO-3 instrument is set to provide XCO2 observations at a high spatial and temporal resolution for regional domains (100 × 100 km). The atmospheric column version of the Stochastic Time-Inverted Lagrangian Transport (X-STILT) model is an established method of determining the influence of upwind sources on column measurements of the atmosphere, providing a means of analysis for current OCO-3 observations and future space-based column-observing missions. However, OCO-3 is expected to provide hundreds of soundings per targeted observation, straining this already computationally intensive technique. This work proposes a novel scheme to be used with the X-STILT model to generate upwind influence footprints with less computational expense. The method uses X-STILT generated influence footprints from a key subset of OCO-3 soundings. A nonlinear weighted averaging is applied to these footprints to construct additional footprints for the remaining soundings. The effects of subset selection, meteorological data, and topography are investigated for two test sites: Los Angeles, California, and Salt Lake City, Utah. The computational time required to model the source sensitivities for OCO-3 interpretation was reduced by 62% and 78% with errors smaller than other previously acknowledged uncertainties in the modeling system (OCO-3 retrieval error, atmospheric transport error, prior emissions error, etc.). Limitations and future applications for future CO2 missions are also discussed.