Retrieval of ozone profiles from OMPS limb scattering observations.

This study describes a retrieval algorithm developed at the University of Bremen to retrieve vertical profiles of ozone from limb observations performed by the Ozone Mapper and Profiler Suite (OMPS). This algorithm was originally developed for use with data from the SCIAMACHY instrument. As both instruments make limb measurements of the scattered solar radiation in the ultraviolet and visible spectral range, an overarching objective of the study is to facilitate the provision of consolidated and consistent ozone profiles from the two satellites and to produce a combined data set. The optimization of the retrieval algorithm for OMPS takes into account the instrument-specific spectral coverage by exploiting information from spectral windows in the Hartley, Huggins and Chappuis ozone absorption bands. Thereby, ozone concentrations in the 12-60 km altitude range can be retrieved. Observations at altitudes where the measurements are contaminated by clouds are rejected by applying a cloud filter. An independent aerosol retrieval is performed beforehand and its results are used to account for the aerosol load in the stratosphere during the ozone retrieval. Results for seven months of data (July 2016-January 2017) are compared and validated against independent data sets from both satellite-based and balloon-borne measurements, indicating a good agreement. Between 20 and 50 km, the OMPS ozone profiles typically agree with the MLS v4.2 results within 5-10 %, with the exception of high northern latitudes (> 70° N above 40 km) and the tropical lower stratosphere. The comparison of OMPS profiles with those from ozonesondes shows an agreement within ±5 % between 14 and 30 km at northern mid-latitudes. At southern mid-latitudes, an agreement within 5-10 % is achieved, although these results are less reliable because of a limited number of available coincidences. An unexpected bias of approximately 10 % is detected in the tropical region at all altitudes. The processing of the 2013 data set using the same retrieval settings and its validation against ozonesondes reveals a much smaller bias; possible reasons are under investigation. [ABSTRACT FROM AUTHOR]