IMPROVED RETRIEVAL OF INTEGRATED WATER-VAPOR FROM WATER-VAPOR RADIOMETER MEASUREMENTS USING NUMERICAL WEATHER PREDICTION MODELS

Water vapor radiometer (WVR) retrieval algorithms require a priori inf ormation on atmospheric conditions along the line of sight of the radi ometer in order to derive opacities from observed brightness temperatu res. This paper's focus is the mean radiating temperature of the atmos phere (T-mr), which is utilized in these algorithms to relate WVR meas urements to integrated water vapor. Current methods for specifying T-m r rely on the climatology of the WVR site-for example, a seasonal aver age-or information from nearby soundings to specify T-mr. However, val ues of T-mr, calculated from radiosonde data, not only vary according to site and season but also exhibit large fluctuations in response to local weather conditions. By utilizing output from numerical weather p rediction (NWP) models, T-mr can be accurately prescribed for an arbit rary WVR site at a specific time. Temporal variations in local weather conditions can be resolved by NWP models on timescales shorter than s tandard radiosonde soundings. Currently used methods for obtaining T-m r are reviewed. Values of T-mr obtained from current methods and this new approach are compared to those obtained from in situ radiosonde so undings. The improvement of the T-mr calculation using available model forecast data rather than climatological values yields a correspondin g improvement of comparable magnitude in the retrieval of atmospheric opacity. Use of forecast model data relieves a WVR site of its depende ncy on local climatology or the necessity of a nearby sounding, allowi ng more accurate retrieval of observed conditions and increased flexib ility in choosing site location. Furthermore, it is found that the cal culation of precipitable water by means of atmospheric opacities does not require time-dependent tuning parameters when model data are used. These results were obtained using an archived subset of the full nest ed grid model output. The added horizontal and vertical resolution of operational data should further improve this approach.