A222-0004
Regional bias exploration in OCO-2/3 retrieved XCO2

Wednesday, 16 December 2020
Poster
Steffen Mauceri, Jet Propulsion Laboratory, Pasadena, CO, United States, Peyman Tavallali, Jet Propulsion Laboratory, Pasadena, United States, Lukas Mandrake, NASA Jet Propulsion Laboratory, Pasadena, CA, United States, Annmarie Eldering, NASA Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA, United States and Michael R Gunson, Jet Propulsion Lab, Pasadena, CA, United States
Abstract:
The Orbiting Carbon Observatory-2 and 3 (OCO-2 and OCO-3) makes high resolution spectral radiance measurements in the oxygen A band at 765 nm and carbon dioxide bands at 1.61 µm and 2.06 µm. An optimal estimation algorithm is used to invert these measurements to column-averaged dry air mole fractions of carbon dioxide (XCO2). The optimal estimation retrieval requires many assumptions that are challenging to validate. This can lead to systematic biases and unphysical variance over space and time in the retrieved XCO2.

To investigate the magnitude and possible causation of these biases we apply a machine learning technique to identify non-physical XCO2 variances over small areas of OCO coverage. Since the features we use are a subset of the variables in the atmospheric state vector (e.g. water vapor, aerosols, surface pressure, surface albedo, ...) retrieved using the optimal estimation, our approach allows us to extract the most important state variables and their contribution to unphysical variabilities in XCO2. This information can then be used to guide future model improvements and allow high resolution analysis of the behavior of XCO2 over different geophysical locations.

<< Previous Abstract | Next Abstract >>