Comparisons of
CLAVR-x Clear Radiances
The goal of this page is to present results showing the quality of the clear radiances produced by the CLAVR-x cloud mask. These results are derived from the clear pixels with no additional filtering. It takes only one clear pixel for a clear value to be reported in a grid-cell.
Comparison with CLAVR-1 (PATMOS)
Based on the successful use of PATMOS clear radiances to derive aerosol climatologies, we are confident the PATMOS clear radiance record is of high quality at least over the ocean. PATMOS used the CLAVR-1 cloud mask which is basis of the CLAVR-x cloud mask. As shown in the cloud amount comparisons with ISCCP and HIRS, the modification to CLAVR-x cloud mask has resulted in more realistic distributions of global cloudiness. Our goal here is to demonstrate that this improvement in the global cloudiness did not come at the expense of the accuracy of the clear radiances.
To compare CLAVR-x to CLAVR-1, we produced CLAVR-x products using the one degree equal-area grid cell format used in PATMOS. In this analysis, we show the clear 11 µm radiances and 0.63 µm reflectances in addition to the cloud amounts derived from each algorithm. The NDVI values are presented because of the known sensitivity of NDVI to cloud contamination. Lastly, we present the numbers of days for the months that each grid-cell contained clear pixels as determined from CLAVR-1 and CLAVR-x because the availability of clear pixels can be as important as the radiometric accuracy for some applications. These results are shown for the monthly means from July 1991 (NOAA-11) and January 1996 (NOAA-14). The PATMOS data is the new A2 data available from the SAA.
Results of NOAA-11 July 1991.
| CLAVR-1 |
CLAVR-x |
Difference |
Zonal |
|
| Ascending
Clear Ch4 (11 µm) Radiance Ocean |
view |
view |
view |
view |
| Ascending
Clear Ch4 (11 µm) Radiance Land |
view |
view |
view |
view |
| Descending Clear
Channel 4 (11 µm) Radiance Ocean |
view |
view |
view |
view |
| Descending
Clear Channel 4 (11 µm) Radiance Land |
view |
view |
view |
view |
| Ascending Clear
Channel 1 (0.63 µm) Reflectance Ocean |
view |
view |
view |
view |
| NDVI |
view |
view |
view |
view |
| Ascending
Total Cloud Amount (ISCCP + UW/HIRS*) |
view |
view |
view |
view |
| Descending
Total Cloud Amount (ISCCP + UW/HIRS*) |
view |
view |
view |
view |
| Ascending
Number of Days per Month with Clear Pixels |
view |
view |
view |
view |
| Descending Number of
Days per Month with Clear Pixels |
view |
view |
view |
view |
Results for NOAA-14 January 1996 (coming)
| CLAVR-1 |
CLAVR-x |
Difference |
Zonal |
|
| Ascending
Clear Ch4 (11 µm) Radiance Ocean |
view |
view |
view |
view |
| Ascending
Clear Ch4 (11 µm) Radiance Land |
view |
view |
view |
view |
| Descending Clear
Channel 4 (11 µm) Radiance Ocean |
view |
view |
view |
view |
| Descending
Clear Channel 4 (11 µm) Radiance Land |
view |
view |
view |
view |
| Ascending Clear
Channel 1 (0.63 µm) Reflectance Ocean |
view |
view |
view |
view |
| NDVI |
view |
view |
view |
view |
| Ascending
Total Cloud Amount (ISCCP + UW/HIRS*) |
view |
view |
view |
view |
| Descending
Total Cloud Amount (ISCCP + UW/HIRS*) |
view |
view |
view |
view |
| Ascending
Number of Days per Month with Clear Pixels |
view |
view |
view |
view |
| Descending Number of
Days per Month with Clear Pixels |
view |
view |
view |
view |
Preliminary Conclusions from this Page:
- It is difficult to definitively prove if CLAVR-x radiances
are better than CLAVR-1. There are regions where a consistent improvement
is seen and CLAVR-x detects more cloud and its clear pixels are darker and
warmer than CLAVR-1's. In other regions, the above results show contradictory
results. For example, in January in the Tropical West Pacific, CLAVR-x
detects more cloud, it's clear reflectances are similar but it's clear
radiances are slightly colder. Another contradiction is seen in some
regions where the CLAVR-x NDVI is higher than the CLAVR-1 value but the clear
11 µm radiance is less. We have have not resolved these contradictions.
- DIfferences due to the CLAVR-x scheme using a terrain data base
are evident.
- CLAVR-x detects more cloud than CLAVR-1
except over snow-covered land.
- The zonal and global distributions of cloudiness from CLAVR-x agree better with ISCCP and UW/HIRS than CLAVR-1.
- Though CLAVR-x detects more cloud than CLAVR-1, the availability
of grid-cells with clear pixels is not significantly diminished compared
to CLAVR-1.