Global observations from the satellites of the TIROS-N/NOAA series over polar regions, where thein situmeasurements are scarce, constitute a valuable help for the study of mesoscale meteorological phenomena after interpretation of these observations into atmospheric and surface parameters. The 31 (Improved Initialization Inversion) system, which relies on a pattern-recognition-type approach, has been applied to NOAA-9 and NO A A-10 passes of both hemispheres and at different periods of the year. Results show that the 31 method is able to identify cyclonic conditions. Comparisons between retrieved products and conventional products show a good agreement. The first assimilation experiments of 31 temperature profiles into a fine mesh model seem promising.