An Efficient Algorithm for Retrieving CO2 in the Atmosphere From Hyperspectral Measurements of Satellites: Application of NLS-4DVar Data Assimilation Method

Abstract

A novel and efficient inverse method, named Nonlinear least squares four-dimensional variational data Assimilation (NLS-4DVar)-based CO2 Retrieval Algorithm (NARA), is proposed for retrieving atmospheric CO2 from the satellite hyperspectral measurements, in which the NLS-4DVar method is used as the optimization method. As the NLS-4DVar method works independently of the tangent linear model and adjoint model, the time-consuming calculation of the weighting function matrix is unnecessary, and the computation complexity is tremendously reduced while maintaining the retrieval accuracy. This is extremely important for space-based CO2 retrievals with large data volumes. Observing system simulation experiments (OSSEs) over four different sites around the world showed that the NARA algorithm could retrieve XCO2 and CO2 profiles effectively. To further evaluate the NARA algorithm, it was used for real CO2 retrievals from target-mode observations of Orbiting Carbon Observatory-2 (OCO-2) over Lamont, Oklahoma, and Darwin, Australia. The results were compared with that of ground measurements of Total Carbon Column Observing Network (TCCON). The mean difference of XCO2 between NARA and TCCON over Lamont, from 180 observations, was −0.15 ppmv with a standard deviation (SD) of 0.76 ppmv. Over Darwin, the mean difference, from 180 observations (90 points over land and 90 points over the ocean), is −0.17 ppmv (SD: 1.26 ppmv). The preliminary results showed that the efficient NLS-4DVar-based algorithm could provide great help for satellite remote sensing of CO2, and it may be used as an operational procedure after further and extensive evaluations.