3

I have a dataset of 20 Landsat scenes (MSS, TM, ETM+) for 20 years (one scene per year). Scenes are in WGS 84, EPSG 32634 coordinate system and I can easily project = geometrically correct them to S_JTSK, EPSG 5514, using Project raster Tool (Data Management in ArcGIS 10.1).

As my area of interest is in mountaineous region, I would like to orthorectify scenes using DEM (from Shuttle Radar Topography Mission, res ~ 30m) but I don´t know really how to do this. I can´t find my rational polynomial coefficients (RPCs) needed for automatic ortorectification in my data and really I can´t imagine how to do this orthorectification. Any help please?

  • 2
    If your scenes are already in UTM, they are already geometrically corrected and you will not be able to do rigourous orthorectification. If you are not happy with the geometric correction 1) start from raw images or 2) try to improve your results with image coregistration methods. – radouxju Oct 16 '14 at 8:03
  • Hi @raduxju, thank you for your comment. I was thinking that geometrically correcting is "reprojecting" and subsequently I have to do orthorectification, so "reprojecting with DEM". That is not right? How can I check that my data are sufficiently correct? are Landsat data from all sensors (MSS, TM, ETM+) orthorectified using DEM? thanks again for your response. – maycca Oct 16 '14 at 8:11
  • reprojecting does not need a DEM, it is the process of changing the coordinate system assuming that the input is geometrically correct. Most Landsat are quite well orthorectified, but you should check where you got them from, and a visual check of the superposition is useful to detect potential errors in the automated Landsat processing chain (they are rare, but there are some). – radouxju Oct 16 '14 at 8:34
  • I´ve got them from GloVis.com. And what do you think with "visual check of the superposition"? I have overlapped them with my .shp dataset and with raster map. It was quite good, pixels were little bit shifted after Project raster process. When I´ve applied "snap to raster" in Environment settings, it was really good. – maycca Oct 16 '14 at 8:46
  • The reason why I am not sure with my Landsat data processing is that I have 2 images from 1 year. 1st is a mosaics (made up from several DOY in 1 year), 2nd is from one DOY=day of year. I am trying to do Maximum Likelihood classification. I was assuming that classification result from "mosaic" will be worse that results from one DOY, what is not true. That is the reason why I was thinking that mosaics was ortho corrected and my row data wasn´t. Where else could be problem? Or how differently can I check for potential errors? thanks again – maycca Oct 16 '14 at 8:53
5

Answer for others so confused people as I am:

To know how to deal with downloaded raw Landsat data - what else in pre-processing do I need?

Firstly check their processing level in_MTL.txt file (included in downloaded Landsat image: http://landsat.usgs.gov/Landsat_Processing_Details.php)

Processing level = DATA_TYPE

  • L1T - terrain corrected processing.
  • L1Gt - systematic terrain corrected processing.
  • L1G - systematic corrected processing.

in MTL file source: http://landsat.usgs.gov/processing_level_of_the_Landsat_scene_I_have_downloaded.php

If possible, Landsat scenes are processed to Standard Terrain Correction (L1T) (Level 1T -precision and terrain correction with DEM). If there are no ground-control or elevation data for L1T , the best level of correction is applied (Level 1G-systematic or Level 1Gt-systematic terrain).

Source: http://landsat.usgs.gov/Landsat_Processing_Details.php

Final geodetic accuracy of the product depends on the accuracy of the ground control points and the resolution of the DEM used.

If I am not wrong, my Landsat data are finally in L1T so I don´t have to orthorectify them - they have to be already ortorectified.

0

I suggest you refer to the following publication: C. Devaraj and C.A. Shah, 2014. Automated geometric correction of Landsat MSS L1G imagery, IEEE Geoscience and Remote Sensing Letters, 11(1): 347-351.

  • Hi, Welcome to GIS@SE. It would be of much more value if you could edit you answer ad provide links to the documents you refer to. best Regards. – Mark Cupitt Oct 25 '14 at 11:07

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.