Latitudinal and Seasonal Dependent Zenith Angle Corrections for Geostationary Satellite IR Brightness Temperatures

Since late October 1998, the National Center for Environmental Prediction (NCEP) Climate Prediction Center (CPC) has archived half hourly, geostationary satellite, full disk, full resolution, "window" (10.7-11.5 microns) InFrared (IR) brightness temperature data area files from the GMS-5, GOES-10, GOES-8, METEOSAT-7, and the METEOSAT-5 satellites using the Man-computer Interactive Data Access System (McIDAS) environment provided by National Environmental Satellite Data Interactive Service (NESDIS). Beginning February 1999, from McIDAS area files, CPC began computing and archiving among other products; global, 60S-60N, half-hourly, merged geostationary satellite mean 0.5 degree IR temperature. Mislocation of cold cloud due to parallax is removed. However, limb IR temperatures from GOES-10 and METEOSAT-5 IR are generally 5-30 K colder compared to locations where GMS-5 retrived nadir view IR only one half hour earlier. By collocating IR from one GOES satellite for which zenith angles are restricted to less than 26.5 degrees and ranging the domain of zenith angles from the other GOES from 44-78 degrees, limb-darkening errors are determined to increase with zenith angle and toward limb temperatures of 230 K (indicating cloud effects) with maximum values greater than 20 K. A temperature dependent zenith angle correction derived from several studies is universally applied to IR from each satellite and removes bias over the tropics, however, over-correction increases from about 25S and 25N, toward the poles. To determine latitude dependence of the errors, the ratio of accumulated differences in observed IR (obtained from overlapping satellites) over the differences in their IR corrections, were binned latitudinally 60N to 60S for Feb 17-27, 1999. The ratio was near 1 in the tropics and smoothly drops off to about 0.4 at 60N and 60S. The latitudinal dependency was repeated for April 12-20, and May 24-June 2, 1999, and indicated a seasonal dependence. After applying this latitudinal and seasonal adjustment to zenith angle correction, the resulting IR bias is smaller with smooth bias patterns resembling inter-satellite calibration more than satellite geometry dependency. Root mean square error of the limb to nadir view IR is reduced by approximately 50% for most satellite pairs at the larger angles. Histograms (5 K) of limb corrected IR obtained from zenith angles of 76-78 degrees nearly match corresponding nadir satellite for 6 of the 8 satellite cases. The process for constructing the error profiles was repeated but by only using cases where the standard deviation of the IR pixels of the nadir satellite were less than 5 and from 5-12.5 K to isolate cases of horizontally smooth cloud. The resultin error pattern was very similar to using all observations which the errors increased with zenith angle but began to decrease for limb temperatures colder than 230 K. For the cases when the standard deviation of the nadir satellite was 12.5-20 K and greater than 20 K (more cumulus convection-type clouds), the resulting error pattern showed an increase in error not only with angle but continuously toward colder temperatures.