On systematic errors in satellite sounding products and their climatological mean values

Abstract Estimates of atmospheric parameters derived from passive measurements of thermal radiation made by satellite‐borne radiometers have rather subtle error characteristics. Moreover, they are sensitive to the ‘first guess’ of the atmospheric state which is used, either explicitly or implicitly, in the inversion process. A proper understanding of these characteristics is desirable when using such data in climate studies or in numerical weather prediction. Relations describing the error structure of individual retrievals are derived and developed to describe the error characteristics of mean values of retrieved products as typically required in climate studies. It is shown that systematic retrieval errors will occur if the first guess used in the inversion is systematically different from the true atmospheric state. the errors are quantified for one particular inversion method‐a scheme designed to retrieve atmospheric temperature and humidity profiles and surface temperature from TOVS (TIROS‐N Operational Vertical Sounder) data. In this case the biases in retrieved surface temperature are small (but not necessarily negligible). However, for less directly observable quantities such as surface air temperature and surface relative humidity, the problem is acute. Potential methods for correcting retrieved parameters for such biases, in retrospect, using an improved first guess are discussed. They rely on information concerning the first guess applied in the inversion being available to the user of the retrieval products. This has important implications for the design of operational systems to process, disseminate and archive satellite sounding data.