I'm working on an area in Amazonia with high cloud persistance which makes global data on land cover quite useless at small scales, so I wanted to learn to process my own Landsat images, with GRASS functions.
I downloaded one Landsat 5 image from http://earthexplorer.usgs.gov/, corresponding to 7 raster files, one for each band; - Imported them into GRASS; - Applied i.landsat.toar function to get top of atmosphere reflectance : The full code line is i.landsat.toar --o input=LT52270571991288CUB00_B output=LT52270571991288CUB00 metfile=/home/LT52270571991288CUB00_MTL.txt. - Finally, I applied i.landsat.acca function to the outputs of the previous function.
What I get is the following: "Processing first pass...
Preliminary scene analysis: * Desert index: 0.00 * Snow cover: -nan % * Cloud cover: -nan % * Temperature of clouds: ** Maximum: 0.00 K ** Mean (cold cloud): -nan K ** Minimum: 10000.00 K Result: Scene cloud free Removing ambiguous pixels..."
This suggests that the image is cloud free, which is difficult to see looking at the image, and checking the metadata I see that the cloud cover is 29%.
Is there a step that I missed somewhere? I tried with a neighbouring Landsat scene, using exactly the same code, and in that case the correct cloud cover is found (same as in the meta data file).