| description abstract | For a 14-day period in late March and early April 1979, TIROS-N operational soundings have been compared with independent temperature analyses produced by the National Meteorological Center. Over the United States and southern Canada, analyzed layer-mean virtual temperatures were interpolated in space and time to the TIROS-N soundings. Differences between retrieved and analyzed layer temperatures were calculated at each sounding location. Statistical analysis of those differences reveals the following. For this data sample, the average TIROS-N sounding is too warm near the surface, too cool in the mid troposphere and too warm at the tropopause. Soundings made under cloudy conditions, which depend heavily on radiances in the microwave channels, have larger temperature biases than soundings made in clear air. Root-mean-square differences between retrieved and analyzed layer temperatures are generally less than 2°C except in the 1000?850 and 70?50 mb layers, where they are larger. The horizontal correlation of TIROS-N retrieval errors is substantial, largerthan O.5 out to 5OOkm. There is strong evidence that horizontal temperature gradients inferred from TIROS-N soundings are systematically too weak. | |
