Advanced users,

may I seek for some recommendation on filtering Landsat reflectance outliers?

I have many Landsat scenes pre-processed (DN to Radiance/Reflectance, Cloud-
masked, Topo-Corrected, band-wise patched in large maps over Greece) and ready 
for further explorations.

Applying "color=grey -e  OR color=grey1.0 -e" doesn't work well for all larger 
maps (after patching -- call it mosaicking if you prefer).  That is, some maps 
appear too dark (i.e. band 3).

I guess that this may be due to abnormally (?) high reflectance values.  I 
guess those are artefacts, or not?


The univariate stats of 1+6 bands look fine to me, e.g.:

mean: 298.591
mean of absolute values: 298.591  ### this is Temperature in K

mean: 0.0453416
mean of absolute values: 0.0453416

mean: 0.0490654
mean of absolute values: 0.0490654

mean: 0.0785879
mean of absolute values: 0.0785879

mean: 0.139015
mean of absolute values: 0.139015

mean: 0.121867
mean of absolute values: 0.121867

mean: 0.0845493
mean of absolute values: 0.0845493


Yet, the max Top-of-Canopy Reflectances:

maximum: 326.271 # This is Temperature in K
maximum: 172.05
maximum: 117.96
maximum: 775.934
maximum: 1.66005
maximum: 120.506
maximum: 477.744


Is my understanding correct?
Is there a safe criterion to filter high reflectance values?  Could they be 
attributed to other sources, e.g. fires?

Can I use some different color rules/scheme which will "ignore" too high 
reflectances?  Simply "color=grey1.0" or other based on stddev, quantiles?

Thank you in advance for your invaluable time,
Nikos

Attachment: signature.asc
Description: This is a digitally signed message part.

_______________________________________________
grass-user mailing list
[email protected]
http://lists.osgeo.org/mailman/listinfo/grass-user

Reply via email to