Analysis of currently-available results from the collection of OCO-2 inverse models of CO2 suggests that they suffer from significant biases in atmospheric transport. This is revealed through…

This federal contract award to Colorado State University (CSU) is for research and analysis services related to inverse modeling of carbon dioxide (CO2) data from the Orbiting Carbon Observatory-2 (OCO-2) satellite. The objectives are to assess biases in atmospheric transport within current OCO-2 inversion models, develop methods to tune these models to maximize information extraction while avoiding overfitting, and evaluate the ability to detect interannual flux variability, particularly the 2015-2016 El Niño. The research aims to improve inversions of satellite XCO2 data, refine bias corrections, and cross-validate against new in situ data in the Amazon. This award is funded by the National Aeronautics and Space Administration (NASA) and has a potential value of $171,604.68. No set-aside designation is mentioned. As a public land-grant university, Colorado State University has extensive experience providing research and technical services to U.S. federal agencies on topics such as climate science, environmental management, and agricultural technology.

Analysis of currently-available results from the collection of OCO-2 inverse models of CO2 suggests that they suffer from significant biases in atmospheric transport. This is revealed through forward simulations of SF6 and CO2, and supported by evidence from optimized fluxes considered as an ensemble. We propose continued meta-analysis of the inversion ensemble to target and remove some of this transport error by comparing the meridional transport of CO2 in models. We will do this by decomposing that transport into components associated with the mean meridional circulation, stationary eddies, and transient eddies. We will develop a method for tuning inverse models to maximize information extraction from observations while avoiding overfitting, by introducing the novel "l-curve" diagnostic. Expected tropical variability will be assessed using the multi-scale terrestrial model intercomparison project (mstmip) ensemble. An empirical orthogonal function decomposition of that variability will allow US to summarize flux anomalies and to rank them by their importance and spatial scale. These modes will then be propagated as pulsed emissions through atmospheric transport models to assess the observability of flux anomalies as a function of observing system and of atmospheric transit time between flux and observation. Interannually-varying flux priors from a near-real time terrestrial biosphere model will be developed and applied to atmospheric inversions, with the specific goal of better representing the 2015-16 EL nino. Atmospheric CO2 simulations will be evaluated against emerging in situ data constraints from sites in the amazon basin, including new light aircraft profile time series. Online bias correction will be implemented within carbontracker, both in order to refine standard bias correction parameterizations, and also to look for systematic variability in residuals from OCO-2 retrievals. Such patterns in the residuals may be due to heretofore-unidentified retrieval biases. The proposed research will address three stated goals of the solicitation. First, IT involves performing and substantially improving inverse model analyses of satellite XCO2 data. Second, IT includes work to improve existing bias corrections and to identify additional systematic features of the retrievals that cannot be explained by inverse models. These features may be caused by previously unidentified biases. Third, IT concentrates on cross-validation in the undersampled tropics of south america, where refined flux uncertainties and the introduction of new datasets will help to constrain CO2 distributions and highlight unexplained features in OCO datasets.