Regarding Atmospheric correction and Elevation DEM Projection system

Question

I am trying to conduct atmospheric adjustment of Landsat 8 image bands using 6s algorithm (i. atcorr grass gis tool). It seems, in the input section, an elevation raster map should be provided. In my literature review, I could not find any reference discussing the DEM's relevance in the process. It is not a required option either. Any explanation how and where it fits in the overall process? In my control file, I am already providing average height above msl (of surface I want to correct). My earlier attempts at correction led to max values above 1 (which is unusal, I was told). Want to explore if the DEM makes any difference or not. Anyways, I may need the DEM for topographic corrections later.

Correction: Earlier, While generating Radiance values from DN number using i.landsat.toar; Atmospheric correction method at Metadata section was set to one of the options. It should be left blank while generating Radiance. This lapse may have contribute to later 6s Algorithm adjusted bands to have max value higher than 1.

I am trying to be consistent with the data. The Elevation source in line 12 of metadata references GLS2000. Online query led me to GLSDEM. I am assuming this is the DEM used in the L1T correction.

The original DEM is in WGS1984 geographic coordinate system. However, the image bands are in WGS1984 UTM projected coordinate system. Do I need to convert the DEM from geographic to projected coordinate system at all? I am asking because conversion will require bilinear interpolation and this will introduce distortion in the elevation values.

Update: Original Geometric projection GLSDEM do not even coincide with TOAR processed Radiance Bands once imported in Grass GUI. Most likely, have to switch to projected coordinate system.

Answer 1

What I have done in the past is as follows. Get the Hydrosheds DEM (or AsterDEM, whatever you choose) and import into a WGS84 Long/Lat GRASSS Location. Switch to the UTM location where you have your Landsat8 imagery. Run r.proj to reproject the DEM to the UTM projection. Now you can use the reprojected DEM in i.atcorr

As for your question regarding the relevance of a DEM layer to the atmospheric correction process, I think that's a bit beyond the scope here, but in any uneven terrain the elevation does indeed impact the surface radiance. The better atmospheric correction tools (ERDAS, GRASS, etc) that go beyond 6S use a DEM.