Development of a statistical method for estimating the downward longwave radiation at the surface from satellite observations

A new technique was developed statistically for estimating the DLR using the HIRS/2 clear-column radiances from NOAA polar orbiting satellites. Simulated radiation data were used in the model development and validation. Two categories of regression equations were studied. They are the linear models and the emissivity/transmissivity-approach models. It was found that the emissivity/transmissivity models are superior than the linear ones in the aspects of the RMS errors of the DLR estimates as well as in the residual spread patterns. For both clear and cloudy conditions, when instrument noise is included, the regression RMS errors of the emissivity and transmissivity models range from about 4 to 9 Wm;Sensitivity studies show that DLR is the most sensitive to the error in the cloud amount on average. Overall, the combined DLR errors, excluding the contribution of the error in the surface pressure estimation, range from about 7 to 12 Wm;For cloudy skies, cloud base height and the effective cloud amount are needed. The DLR estimated with different cloud data sources, however, has not been examined. Since the sensitivity study shows that the DLR error due to the uncertainty of the cloud amount usually dominates the other error components, it is believed that an accurate estimate of the cloud amount is more important for cloudy-sky DLR estimation.;Despite these limitations noted above, our analysis indicates that the physical statistical method developed in this study should provide estimates of the DLR to within about 10 Wm...