A. Relative Azimuth Representation Change
When processing commenced with the 1987 data, the relative azimuth value at nadir (solar azimuth minus sensor azimuth) was set to a value equal to the solar azimuth because the relative azimuth is undefined at nadir (sensor zenith is 0). On average about .016% of the land pixels (±3700 pixels/day) are affected in a daily image and less in a 10-day composite. The incorrect nadir relative azimuths have been run through a radiative transfer model and it was determined that the effect on the reflectance value is less than 0.08% in the reflectances and even smaller in the NDVI. The processing software was modified to set the relative azimuth angle at nadir to missing. As a result, the atmospheric correction for these pixels is not calculated and no NDVI is calculated or binned.
Secondly, the relative azimuth is no longer calculated as the absolute value of the solar azimuth angle minus the sensor azimuth angle, but rather is represented by a full 360 degrees. During processing of the 1987 data, if the difference of solar azimuth and relative azimuth is negative (e.g., -20 degrees) the absolute value was taken rather than setting the value to 340 degrees. About one-quarter of the land pixels are affected by this, but the exact locations vary by season. Only those pixels east of nadir and north of the latitude of solar declination are affected. This change affects only the relative azimuth angle data layer in the Daily and Composite Data Set and was corrected beginning with 1988 data processing. Since in the atmospheric correction software the cosine of the realtive azimuth was taken, this change did not affect the reflectances or the NDVI.
B. CLAVR Algorithm Error Correction
A few errors were found in the CLAVR algorithm as implemented in the Pathfinder system. The most significant change is that the Reflectance Gross Cloud Test rarely flagged data as cloudy in the 1987 data set. This was due to a coding error which has been corrected beginning in 1988 and as such interannual comparisons of this layer SHOULD NOT USE DATA FROM 1987. Furthermore, please be reminded that the CLAVR data layer is an unvalidated, experimental research data layer and the Pathfinder Land Science Working Group recommends that users be cautious about using this layer for data applications and analysis.
C. Thermal Calibration Error Correction
A correction to the 'negative radiance from space' was applied incorrectly. These corrections, which change occasionally over the life of the instrument, come from a table provided by NOAA. Only data from January 1, 1987 through October 16, 1987 is affected.
The processing software was corrected and several days from 1987 were tested to determine the magnitude of the correction. On average, the change in brightness temperature in Channel 4 was 0.71 degrees K and the change in Channel 5 was 0.41 degrees K. Most of the changes occur in cold cloud tops with warm brightness temperatures being essentially unchanged. This change does not affect the NDVI, visible or geometry layers. Cold areas may have a great enough change in brightness temperature that a CLAVR threshold may flag pixels that would have not been flagged otherwise, but this would be a very rare case.
In addition, it should be noted that during the period of mid-March 1987 and mid-October 1987 there were several days in which there was no thermal calibration information. This is not a processing error but rather a result of communications or level 0 processing errors when the 1b orbital data were generated. The quality control comments indicate those images in which this occurs. The visible channels are fine in these time periods, however caution should be used when using thermal data from composites or daily images from these time periods.
The ozone data used in calculating the ozone absorption term of the atmospheric correction comes from the Total Ozone Mapping Spectrometer (TOMS) data set. The TOMS data files used in the Pathfinder processing are monthly files containing all daily data. During processing of the 1987 data, data were produced using the ozone from the first day of the each month. Beginning with 1988 data processing, the atmospheric correction for each day uses the ozone data for that day.
Data were processed from several seasons to determine the effect on the data set. On average, Channel 1 changed by 0.02% reflectance and Channel 2 changed by 0.13% and NDVI changed by 0.004. The change to using daily ozone is an improvement in the data set, but the overall effect is small enough that interannual comparisons can still be made.
The quality control (QC) flag layer of the Pathfinder data set contains useful and important information about a variety of data conditions. Most of the conditions flagged are simply for information (e.g., the interpolated data flag) and some indicate a condition with minimal impact on the data (e.g., the Channel 1 and 2 non-standard flag which indicates that a default ozone value was used).
One exception to this is the QC flag that indicates that a NOAA QC flag was set in the input data (Pathfinder QC flag 16). This can be caused by a number of conditions which are described in the NOAA Polar Orbiter User's Guide (Kidwell, 1991). Some of these conditions, such as bad or unavailable calibration data, result in data values which are out of range or are obviously incorrect. However, in some cases NOAA QC flags on the input data are simply for information and do not indicate error conditions in the data.
When bad data (extremely high or extremely low values) are present in the daily data, processing the 8-11 days may include the bad data in the composite. This was changed so that during compositing all data with a Pathfinder QC flag indicating that a NOAA QC flag was set are filtered out. This means that those composites from October 1, 1986 through October 31, 1988 may contain data which was flagged with a NOAA QC flag which may be incorrect. Composites from time periods outside this range (10/1/86 - 10/31/88) will not include any data with a NOAA QC flag, however, there are many cases of extremely high or low data values which result from bad digital counts or calibration in the GAC orbits and for which there is no readily available detection method. These data may be present in composites even when all data with NOAA QC flags have been removed.
The Pathfinder processing team plans to regenerate the composites from the 10/1/86 - 10/31/88 time period and update the archive. Users will be informed when this update has occurred. the timing for this archive updates is uncertain, users are advised, in the interim, to carefully check the 10-day composite data and verify that this problem has a minimal effect. Note that this change involves only the Composite Data Set, the daily data are not effected.