I wanted to perform atmospheric correction using MODIS surface reflectance products for validating my atmospherically corrected images (from Landsat 8). As Landsat 8 only became available in 2013, is it safe to use MODIS data of 2013 and beyond?
-
what do you mean by 2013 and beyond ? The important thing is to use Landsat-8 and MODIS from the same date. In any case, this is useful to detect artefacts but you cannot really "validate" with another image (spectral bands are not exactly the same, BRDF is not exactly the same etc)– radouxjuCommented Jun 9, 2016 at 4:54
-
i mean 2013 and any date after that 2014, 2015..– user18058Commented Jun 9, 2016 at 4:58
-
@radouxju: I mean 2013 and any date after that 2014, 2015.. Yes, I am using images from same date. My (little) understanding is MODIS and Landsat bands are almost overlapping. Also, I assume I can use MODIS surface reflectance products in lieu of spectra data from the field to check how my atmospheric method works with me Landsat/how good the atmospheric correction is?– user18058Commented Jun 9, 2016 at 5:28
Add a comment
|
1 Answer
Yes - you can use MODIS imagery. The MODIS sensors have been degrading a bit and as such, there are some problems there. Also, given that you don't have exactly the same bands, view angles are a bit different, vastly different spatial scale and so on, the comparison will most likely be quite noisy and as such, it may indicate that a correction that actually went well, is wrong, or the other way around.
Another option would be to compare your atmospheric correction, with the atmospheric correction that the USGS have applied to Landsat 8, which is available through EarthExplorer etc. Just look for the Landsat 8 Surface Reflectance Product. This product is generally good, but there are some spatial anomalies and such, but for comparison purposes, it should be fine.
answered Jun 9, 2016 at 6:00
-
I see. Informative! I was of the opinion that MODIS SR products are generally better than Landsat SR product. Regarding differences in spatial resolution, cant I take care of that by aggregating (using the aggregate tool in ArcGIS, for example) Landsat (30m) to MODIS (500m) spatial resolution, and then perform regression? Commented Jun 9, 2016 at 6:53
-
The relationship between "high" resolution / Landsat reflectance and coarse resolution / MODIS reflectance could be a study in an of it self. Plenty of issues to investigate relating to aggregation, spatial correlation and decorrelation in in-homogeneous area. All in all, lots of assumptions to consider, and you want to simply say "it all works perfectly from the start". You should at least note in your report / paper / thesis that you know that it is a rough assumption. Commented Jun 9, 2016 at 7:11