Mesospheric temperature from UARS MLS: retrieval and validation

A research algorithm is developed to retrieve temperature at 20-90 km using 63 GHz O-2 emission measurements from Microwave Limb Sounder (MLS) on Upper Atmosphere Research Satellite (UARS). The algorithm is based on a previous MLS radiative transfer model but improved to produce more accurate radiance calculations in the cases where the geomagnetic Zeeman splitting is important. A fast version of the model is developed and implemented for practical uses of the temperature retrieval, which uses a single temperature and O-2 density profile as the linearization basis. The calculated radiances and linearization coefficients are fit to a set of explicit functions of the geomagnetic field and its direction at tangent heights of 0-120 km, which are pre-stored in order to speed up the computation. The new algorithm has been used to process all the data available during 1991-1997 before MLS 63 GHz radiometer was powered off. The estimated precision of MLS temperature varies from 2 K at similar to20 km to 8 K at similar to80 km and increases sharply above similar to90 km. The retrieved MLS temperature are compared against CIRA'86, satellite, lidar, and rocket observations. Comparisons to CIRA'86 seasonal climatology show that the differences are latitude-and-season dependent and generally < 5 K below 50 km and 10 K in the mesosphere. Comparisons with other satellite observations (ISAMS, HRDI, CRISTA1) show different patterns but a cold bias at 85-90 km seems common in all these comparisons. Comparisons to ground-based lidar measurements suggest that MLS temperatures are warmer by 2-4 K in the stratosphere and colder by 5-15 K at 85-90 km. The MLS-minus-lidar difference shows a 3-10 K cold bias near 70 km for most of the sites selected. The comparisons with rocket measurements are similar to those with lidars at these altitudes, giving cold biases in the MLS temperatures at 85-90 km. Most of these biases are understandable in terms of sampling and resolution differences, and some biases can be reduced with further improvements in the MLS retrieval algorithm. Despite the existing biases, the MLS temperature have been found useful in studying large-scale mesospheric phenomena such as the temperature inversion layer. (C) 2002 Elsevier Science Ltd. All rights reserved.