Development of a statistical method for estimating the radiative cooling from satellite observations

Cooling rate estimation within the atmosphere is necessary and important for use in climate studies. In this study, we follow Ellingson's physical-statistical approach to develop and validate a new technique to obtain estimates cooling rates by using radiance data from the High Resolution Infrared Radiation Sounder/2 (HIRS/2) instrument flown on NOAA operational satellites.;For clear-sky conditions, nonlinear models are found to be best for the atmosphere above 500 mb, while for below 500 mb the best results are obtained by using the brightness temperatures to estimate parameters of a simplified radiation model which in turn is used to infer cooling rates. The rms errors are 0.26, 0.27, 0.19, and 0.07°C/day in the 1000 to 700, 700 to 500, 500 to 250, and 250 to 10 mb layers, respectively, which are 16%, 20%, 32%, and 58% improvements from the previously developed linear models. For cloudy-sky conditions, the rms error ranges are similar to those for the clear-sky analysis. An analytic cloud scheme is designed to handle partial cloudy conditions with multiple layers and different categories of clouds for flux or cooling rate estimation.;Sensitivity studies are performed to determine possible error sources for clear conditions. Combined with random instrument errors, the rms errors for the regression models are 0.27, 0.28, 0.22, and 0.11°C/day for the four layers, respectively. These errors are smaller than those associated with radiation model calculations that result from the propagation of observational errors in the model.;The model validation shows good agreement between regression estimates and model calculations for the uppermost three layers. For the lowest layer, significant improvements are obtained over most areas of the globe, while there are still large differences over specific continental areas where additional improvements are needed. Further validations with cloud data and simultaneous satellite data are recommended.;This study suggests that direct use of satellite data with a physical-statistical model is a viable alternative to radiation transfer model calculations for estimating the global distribution of cooling rates, and has the advantages of rapid speed and comparable or smaller errors than model calculations based on satellite retrieved sounding data.