1

My Landsat 5 and Landsat 8 imagery are not lining up in ArcMap (also tried QGIS). (See attached images) L8 seems to be ok and lines up with the shapefile where as with L5, the shapefile is farther inland.

The data frame as well as both sets of imagery are in WGS_1984_UTM_Zone_21N.

The shapefile has been projected to the same.

*edit: Sorry for the lack of details :) The 2017 image (and all L8 images I downloaded from EarthExplorer) line up correctly with the polygon (mangrove restoration site). All L5 images I'm using (2007-2011) place the polygon too far inland.

The polygon was projected to the same crs as the landsat imagery (WGS_1984_UTM_Zone_21N).

Any ideas why the L5 imagery is off? Could it be with the imagery itself? Now that I think about it, I uploaded the shapefile in EarthExplorer, and I do remember one scene that I looked at and the polygon was too far inland.

2007 L5 imagery

2017 L8 imagery

  • what is the source of your green polygon? what do you like to do? also, you didn't define the problem in your question. please as a new user be sure to take the Tour to learn about our focussed Q&A format – Moh Feb 18 '18 at 13:06
  • Have you turn off the "on the fly projection" to see what happens ? – PROBERT Feb 18 '18 at 14:15
  • @PROBERT I tried this answer to "turn off" on the fly projection but same resutl – Oldgreg22 Feb 18 '18 at 20:12
  • 2
    @User2009 Thanks for the reminder about the Tour. The green polygon was made in Google Earth -> exported as KML - > KML to layer (ArcMap) ->exported polygon witin the layer as a shapefile. The polygon represents the site of a mangrove restoration project. I'm using an NDVI time series to look at how the restoration is working over 10 yr period. - The problem is that the polygon is too far inland in the 2007 image (and all L5 images I downloaded). The 2017 imager (and all L8 images) show it's proper location. – Oldgreg22 Feb 18 '18 at 20:15
  • 1
    @PROBERT forgot the link – Oldgreg22 Feb 19 '18 at 9:54
1

I figured out what the issue was. The Landsat 8 imagery was Tier 1 data, the Landsat 5 imagery was Tier 2. From what I understand, Tier 2 data has not been orthorectified due to issues such as too much cloud coverage (which was the case in this location). For this particular location, there were no Tier 1 level data available for Landsat 5.

General notes in case others have similar issues:
- I had no issues with using polygons drawn in Google Earth and importing in ArcMap - worked well on several other locations with Landsat 5 and 8 Tier 1 data.
- Georeferencing the images could be a workaround, but I found it difficult to pinpoint landarks in the Landsat imagery.
- In EarthExplorer, you can choose to omit Tier 2 data.
- More information on Landsat and the Tiers I mentioned above - Landsat Processing Details and What are Landsat Collection 1 Tiers?

0

From your comment about the polygon source, we can get the answer. First of all Landsat8 and Landsat5 can have different resolutions depends if it's MSS or TM also Landsat 5 can have some issues like noise, data loss.

And of course, between them (ca. 10 years) so there are differences in the coordinates of your area (e.g tectonic movement).

Also, the data source (Aerial images) of google map is unknown, so all of these reasons lead us to get a difference.

But we still can say that Landsat8 will have more corresponds than Landsat5 when you add a feature from GoogleEarth over the required part.

Please have a look at the two images (you can see the differences).

There is (Landsat Quality Assessment (QA) ArcGIS Toolbox) maybe can be useful.

Also, many tools and information about Landsat 5,8 can be found here.

PS: If you working with time series, you can make GNSS analysis (Long procedure) to check the differences in the coordinates over the 10 years. GNSS analysis can be done by getting local GPS measurements, and then download coordinates (Rinex) for (e.g 10 years) then make some analysis for the specific area and time series to get how these coordinates differ using some free tools (TEQC and RTKLib).

  • I checked another polygon (also from google earth) and the 2007 and 2017 imagery just a bit north of this site. There also differences in the imagery but much less severe. -What do you mean by "L5 and L8 have different resolutions"? They are both 30x30. -I searched for GNSS analysis but didn't find anything I though I could use - any suggestions? - Is there anyway to use the landsat data I have? Maybe this sounds stupid but is there a way to recalibrate/adjust the imagery? – Oldgreg22 Feb 19 '18 at 9:53
  • please check the updated answer for more details – Moh Feb 19 '18 at 10:39
  • Thanks for the extra info but that's a bit out of my experience level. I may have figured out the issue, the images are Tier 2 and I believe they are not georeferenced. I'm not sure if this is correct. – Oldgreg22 Feb 19 '18 at 12:09
  • you welcome, and good luck – Moh Feb 19 '18 at 12:11
0

The georeferencing on Landsat 5 is significantly off at times.
You simply have to adjust it using any standard georeferencing tool, like the ones in QGIS or ArcGIS. The challenge is to find reliable referencing points, as urban development and general changes to the environment can be significant. However, in your case, there appear to be plenty usable points available.
You can also try using one of the tools for autogenerating points for referencing, but they usually do not work well when the changes are significant between the two images.

  • I may have figured out the issue - the L5 data is Tier2 and the L8 is Tier1. This location does have alot of cloud cover which led to the L5 data not being processed to Tier 1 standards. {"Tier 2 scenes do not meet the Tier 1 criteria due to significant cloud cover, insufficient ground control, and other factors."](landsat.usgs.gov/…) - I don't understand why Tier2 products would be listed under "Landsat Collection 1 Level-1 data" as being "suitable scenes for time-series pixel level analysis" – Oldgreg22 Feb 19 '18 at 12:08
  • so georeferencing the Landsat 5 imagery to the Landsat 8 imagery should solve the problem? – Oldgreg22 Feb 19 '18 at 12:12
  • @Oldgreg22 - yes, adjusting the L5 to match L8 should work reasonably well. – Mikkel Lydholm Rasmussen Feb 19 '18 at 13:38
  • Do you have ERDAS Imagine ? It has an excellent tool that you can do that using Georeference Wizard. I have done this before with ASTER over a Landsat... really good toolbox. – PROBERT Feb 19 '18 at 15:06
  • @MikkelLydholmRasmussen ok I'll give it a go. Thanks again. – Oldgreg22 Feb 19 '18 at 16:41

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.