A Comparison of Aerosol Parameterizations in the ACOS XCO2 Retrieval Algorithm

Abstract:

An effective parameterization of clouds and aerosols in retrieval algorithms is essential for reducing measurement errors and biases in estimates of the column-averaged dry-air mole fraction of carbon dioxide (X_CO2) from space-based measurements of near-infrared reflected sunlight. The NASA Atmospheric CO₂ Observations from Space (ACOS) X_CO2 retrieval algorithm has evolved significantly over the past several years in an effort to more accurately represent the impact of clouds and aerosols on X_CO2. Recent ACOS algorithm versions up to build 3.4 used a water cloud type, ice cloud type, and two generic aerosol types for each sounding. ACOS build 3.5 uses the same cloud parameterization, but was modified to replace the “one-size-fits-all” aerosol scheme. Build 3.5 uses a monthly aerosol climatology based on the Modern-Era Retrospective Analysis for Research and Applications (MERRA) reanalysis to choose the two most likely aerosol types for a given measurement location, along with typical optical depths. The five MERRA types available for selection are sulfate, dust, sea salt, organic carbon, and black carbon. The algorithm then uses a pre-assigned Gaussian width and height and fits for the aerosol amount and peak height based on information from the 760 nm O₂ A-band and the CO₂ bands centered near 1610 and 2060 nm.

Here we compare ACOS builds 3.4 and build 3.5 to quantify the impact of the aerosol scheme update. Two types of tests were performed. Simulated Orbiting Carbon Observatory 2 (OCO-2) retrievals and their associated aerosol and cloud profiles were compared to the “true” aerosol and cloud profiles used to create the simulated environment for a given measurement. The retrieval algorithms were also run on Greenhouse gases Observing SATellite (GOSAT) observations and compared to AErosol RObotic NETwork (AERONET) aerosol optical depth measurements in order to quantify the ability of the algorithms to retrieve information about aerosol optical depths. X_CO2 errors and biases were evaluated for both builds to test the hypothesis that a more realistic aerosol parameterization leads to a reduction in X_CO2 errors.