Atmospheric Chemistry and Physics Discussions www.atmos-chem-phys-discuss.net 1680-7367 1680-7375 5 5 2005 10.5194/acpd-5-8879-2005 http://www.atmos-chem-phys-discuss.net/5/8879/2005/ http://www.atmos-chem-phys-discuss.net/5/8879/2005/acpd-5-8879-2005.html http://www.atmos-chem-phys-discuss.net/5/8879/2005/acpd-5-8879-2005.pdf 8879 8923 2005-09-19 The direct inversion method for data assimilation using isentropic tracer advection M. N. Juckes British Atmospheric Data Centre, SSTD, Rutherford Appleton Laboratory, Chilton, Didcot, Oxfordshire, OX11 0QX, UK A new data assimilation algorithm is applied to MIPAS and SBUV measurements of stratospheric ozone. The results are validated against HALOE, POAM III, SAGE II and III, OSIRIS and ozonesonde data. The new assimilation algorithm has the accuracy of the Kalman smoother but is, for the systems studied here with up to 200 000 variables per time step and 61 million control variables in total, many orders of magnitude less computationally expensive. The analysis produced minimises a single penalty function evaluated over an analysis window of over one month. The cost of the analysis is found to increase nearly linearly with the number of control variables. Compared with 850 profiles from Electrochemical Concentration Cell sondes at 29 sites the analysis is found to be merely 0.1% high at 420 K, rising to 0.4% at 650 K (813 sonde profiles). Comparison against the other satellites imply that the bias remains small up to 1250 K (38 km) and then increases to around −10% at 1650 K (44 km). Between 20 and 35 km the root-mean-square difference relative to HALOE, SAGE II and III, and POAM is in the 5 to 10% range, with larger discrepancies relative to other instruments. Outside this height range rms differences are generally larger, though agreement with HALOE remains good up to 50 km.