ISCCP Known and Fixed Data Errors ----------------------------------------------------------------------------------- * Spurious satellite zenith angle dependence (artifact in Indian sector) * Change in land surface pressures from TOVS * Incorrect snow cover (has been corrected) * Incorrect precipitable water amounts and surface temperatures (has been corrected) * Spurious sea ice reports * Flipped array indices in North Polar DX (has been corrected) * Spurious land pixels in METEOSAT DX * D2 versions 0,1,2 contain un-corrected METEOSAT-3 reflectances (has been corrected) ----------------------------------------------------------------------------------- Spurious satellite zenith angle dependence (artifact in Indian sector) Global maps of cloud properties retrieved in the ISCCP analysis often show an artifact between longitudes of 50E and 100E. This feature is the most noticeable manifestation of a spurious, but weak dependence of the retrieved cloud properties on satellite zenith angle: as cosine of the satellite zenith angle (MUE) decreases from 1.0 to 0.3, total cloud amount increases by about 15-20%, cloud top temperature decreases by about 4-6K, and cloud optical thickness decreases by about 2-3. Most results are obtained for MUE values > 0.4 which reduces the overall effect somewhat. The dependence of total cloud amount is discussed in detail in Rossow and Garder (1993: J. Climate, 6, 2370-2393). This effect is thought to be caused in part by geometric effects on broken cloudiness and by variation in detection sensitivity to optically thin clouds; however, the total effect is not completely understood. The dependences of cloud top temperature and optical thickness are thought to be caused in part by the use of an inaccurate model of cloud microphysics in the retrieval model; the introduction of an ice crystal model in the new D-series data in place of the liquid droplet model for cold clouds in the C-series reduced but did not completely eliminate this dependence. Research is on-going to understand and develop a correction for this effect: D1 data report the value of cosine of satellite zenith angle that could be used to make corrections. - March 22, 1996 Change in land surface pressures from TOVS Between May and June 1986, NOAA increased the global mean surface pressure used to calculate topographic surface pressures (functions of topographic height and surface temperature only) over land from 1000 mb to 1013 mb. This change increases surface pressures over land by about 1%, but the maximum value is still 1000 mb. Surface pressure over ocean remains 1000 mb. Since the TOVS atmospheric temperature and humidity profiles are used in the ISCCP analysis, this change may also produce a very small change in the cloud top pressures of lower level clouds over land. The change is partly mitigated by retaining 1000 mb as the maximum surface pressure in processing the sounder data. No correction of the data is planned. - March 22, 1996 Incorrect snow cover In the first release of the new D-series datasets, there were two errors involving snow cover data. In the D1 and D2 datasets for January - April 1990 (Version Number 0), the correct snow cover is reported, but the snow cover was erroneously eliminated in the cloud analysis. This error will produce small increases in cloud amount (because of false visible radiance detections), decreases in surface reflectance and increases in cloud optical thickness in areas where snow cover varies during the month. In the D1 and D2 datasets for 1992 (Version Number 0), the correct snow cover was included in the cloud analysis, but incorrect values are reported in the datasets. The affected D1 and D2 datasets have been corrected and replaced (Jan90-Apr90 and Jan92-Dec92 Version 1). - April 12, 1996 Incorrect precipitable water amounts and surface temperatures A processing code error caused monthly mean values of the atmospheric water vapor column amount (precipitable water) to be excessively high at pressures < 560 mb. Since the monthly mean values are sometimes used to fill in when daily observations are missing, a small fraction of the daily precipitable water amounts are erroneous. These excessive values also cause retrieved surface and cloud top temperatures to be overestimated. The effect is largest for surface temperatures. The DX/D1/D2 datasets for April, May, August and October 1992 (Version 0 or 1) are affected, the first two months more so than the other two. The TOVS dataset for 1992, and the DX/D1/D2 datasets for April, May, August, and October 1992 have been corrected and replaced. - September 9, 1996 Spurious sea ice reports Beginning in 1992, a new sea ice dataset is employed in the ISCCP cloud analysis that comes from an analysis of satellite microwave observations. One problem with this sea ice product is the presence of spurious sea ice along coastlines. A procedure was applied to attempt removal of these spurious sea ice reports, which was not entirely successful. The procedure does eliminate isolated sea ice at the scale of the microwave map if it is surrounded by open water; however, it does not eliminate such sea ice if it is adjacent to the coast or if there is a cluster of sea ice values. Both of these conditions may occur occasionally along complicated coastlines. Therefore, some spurious values of snow-ice fraction in the D1/D2 datasets may occur at coastlines -- the occurrence frequency appears to be very low. This problem affects ONLY the snow-ice fraction since the cloud analysis is essentially insensitive to the presence of sea ice in near-coastal water: in other words, the cloud analysis is already so conservative in such a mixed area that the presence of sea ice makes no difference. Therefore, the cloud and surface properties reported are unaffected by this error. The error occurs only in 1992 and 1993 data. The sea ice and snow product will be revised later to reduce this problem and the information concerning the frequency and locations of its occurrence will be documented. - December 31, 1996 Flipped array indices in North Polar DX In the DX dataset the location of each pixel is given by latitude/longitude and by array indices (i,j). The latter are used primarily when displaying the DX data as pictures. The NOAA polar orbiter dataset is divided into three lower latitude sectors and two polar sectors. The array indices given in the North Polar DX datasets were incorrectly reversed so that the data, when displayed, appeared to be flipped from West to East (as if viewed from inside the Earth instead of from outside the Earth). The associated latitudes/longitudes for these data were correct, however. This error affected North Polar DX data for 1986, 1990-1992, but did not affect data in other years. The affected DX data has been corrected and replaced in the archives. - June 24, 1997 Spurious land pixels in METEOSAT DX In the METEOSAT DX data there are 4 pixels just south of the African coast at 0 degrees longitude that are mislabeled as land pixels. This error arises when the calculated longitude is exactly 360 degrees in floating point representation, causing an overflow in a look-up table. Although the cloud analysis is somewhat different over land and ocean, these differences are reduced for such isolated pixels, so that the effects on the reported cloud and surface properties should be minimal. This error affects METEOSAT DX data for 1986, 1990-1993, but does not affect data in other years. No correction of this data is planned. - December 31, 1996 D2 versions 0,1,2 contain un-corrected METEOSAT-3 reflectances In D2 1992 versions 0,1,2, the special METEOSAT day/land reflectance correction (see cloud data document page 53) was not applied to METEOSAT-3 while it was in the GOES-EAST position (July-December). D2 1992 Version 3 is now available which includes this correction. - March 20, 1997 Last Modified June 24, 1997 by Ralph Karow (rkarow@giss.nasa.gov)