An examination of the long-term CO records from MOPITT and IASI: comparison of retrieval methodology

Abstract. Carbon monoxide (CO) is a key atmospheric compound that can be remotely sensed by satellite on the global scale. Fifteen years of continuous observations are now available from the MOPITT/Terra mission (2000 to present). Another fifteen and more years of observations will be provided by the IASI/MetOp instrument series (2007–2023>). In order to study long term variability and trends, a homogeneous record is required, which is not straightforward as the retrieved products are instrument and processing dependent. The present study aims at evaluating the consistency between the CO products derived from the MOPITT and IASI missions, both for total columns and vertical profiles, during a six year overlap period (2008–2013). The analysis is performed by first comparing the available 2013 versions of the retrieval algorithms, and second using a dedicated reprocessing of MOPITT CO profiles and columns based on the IASI a priori constraints. MOPITT v5T total columns are generally slightly higher over land (bias ranging from 0 to 13%) than IASI v20100815 data. When IASI and MOPITT data are retrieved with the same a priori constraints, correlation coefficients are slightly improved. Large discrepancies (total column bias over 15%) observed in the Northern Hemisphere during the winter months are reduced by a factor of 2 to 2.5. The detailed analysis of retrieved vertical profiles compared with collocated aircraft data from the MOZAIC-IAGOS network, illustrates the advantages and disadvantages of a constant vs. a variable a priori. On one hand, MOPITT agrees better with the aircraft profiles for observations with persisting high levels of CO throughout the year due to pollution or seasonal fire activity (because the climatology-based a priori is supposed to be closer to the real atmospheric state). On the other hand, IASI performs better when unexpected events leading to high levels of CO occur, due to the less constrained variance-covariance matrix.