Getting Landsat 8 imagery into Google Earth

When Landsat 8 imagery was made available on Amazon AWS our first question was: ‘how easy will it be to view it in Google Earth?’. This is our experience trying to answer that question.

The Landsat 8 satellite has a number of different cameras that capture imagery in different wavelengths. For details of the various sensors see this article. What we want to do is try to get a true colour image into Google Earth, which involves combining the three colour bands (2,3 and 4) and the panchromatic band (band 8) to provide extra detail (a technique known as pansharpening).

The first step is to decide which part of the globe you wish to look at. Landsat 8 covers almost the whole globe every 16 days. The imagery is arranged into strips and rows and you need to determine the strip and row of your point of interest. We used the USGS Global Visualization Viewer for this purpose. It requires the Java plugin, which Chrome says will soon be unsupported. An alternative is USGS Earth Explorer. Both websites also help you find what Landsat imagery is available for a given point of interest. Look for one with either minimal cloud cover or as close as possible to the date you are interested in. We chose an image of the Cape Town region captured on July 1st, 2015. Take note of the scene ID, which in our case was LC81750832015182LGN00.

Next you need to find the relevant imagery on Amazon AWS. Amazon has a well written explanation for how the imagery is organized and how to access it. In our case, we chose to download it using a web browser directly from the URL, which for our image was:
http://landsat-pds.s3.amazonaws.com/L8/175/083/LC81750832015182LGN00/index.html

Next, download the only file that is a *.txt file, as well as the files ending in B2.TIF, B3.TIF, B4.TIF and B8.TIF. Note that they are quite large files, typically about 400MB for all of them.

Next, we need to combine them and we came across a set of useful tools available for free for non-commercial use from GeoSage, which do most of the work for you. (Note that they are windows only.)

After downloading and installing the GeoSage’s Spectral Transformer for Landsat-8 (GUI), we ran it and opened the text file that was downloaded earlier. However, when we tried to run it, it could not read the TIF files as it doesn’t recognize the compression used. So we had to open and re-save each of the images to remove the compression. We managed to do this with the free Gimp image editing software. Having done that, Spectral Transformer worked without a hitch and produced a full colour pan-sharpened and stretched image.

All that remains is to put it in Google Earth as an image overlay and position it appropriately. The instructions for Spectral Transformer suggest that the output image is a GeoTIF and contains the necessary positioning information, but we were unable to get that to work, so we did it manually.

The result is an image that, apart from a little cloud cover, looks nearly identical to the Google Earth imagery – which of course makes sense because, as we mentioned yesterday, Google Earth at this zoom level shows an image that was created from Landsat imagery.


A Landsat 8 image imported into Google Earth looks almost identical to the default imagery.

Closer inspection reveals slight differences in crop patterns and water levels in a lake, as the image is of a different date than the Google Earth imagery. If you zoom in any further Google Earth switches to Digital Globe imagery.


Left: Imported Landsat 8 image. Right: Google Earth default image (also from Landsat).


The resolution of Landsat 8 imagery is only 15m for the Panchromatic band and even worse for the colour bands. You can make out larger ships and the football stadium, but individual houses or cars cannot be distinguished.

So, in conclusion, it is relatively easy to get Landsat imagery into Google Earth, but nevertheless probably not worth the effort unless you have a specific reason for doing so. If there was a particular event that was visible at 15m resolution (large wildfires or volcanic eruptions, for example) that happened to coincide with a Landsat 8 pass of that location, then it might be worth it. It could also be useful for looking at large scale vegetation cover changes over time, but if you want to compare more than one or two images, then consider using Google Earth Engine, which not only provides a wealth of tools and computing power specially designed for analysing Landsat imagery, but you might find someone has already done the analysis you are interested in. Another possible reason for getting Landsat imagery into Google Earth is that there may be interesting things to find in the non-visible colour bands that are not part of the current Google Earth imagery.

The image created by Spectral Transformer was 173MB, but we have compressed it with jpeg compression down to 23MB so that you can try it out if you wish, using this KML file.

About Timothy Whitehead

Timothy has been using Google Earth since 2004 when it was still called Keyhole before it was renamed Google Earth in 2005 and has been a huge fan ever since. He is a programmer working for Red Wing Aerobatx and lives in Cape Town, South Africa.



Comments

  1. Wguayana says:

    New Imagery July 24: CumanĂ¡, Venezuela

  2. The problem with Google earth engine is it hasn’t been updated everywhere, either that it I haven’t found SB easy way to use it

  3. Given the above review, we add more info and some clarification benefiting those who are interested in.

    1 – A specific report on this topic was released by GeoSage in February 2015: “Easily take daily fresh Landsat-8 imagery into Google Earth Pro with 2 simple button clicks”. It is still available at GeoSage http://www.geosage.com/products/spectral_transformer/landsat8/SpectralTransformerForLandsat8_GE.pdf

    2 – Amazon AWS is the latest avenue to provide Landsat-8 imagery (with individual band files for separate download), but there are other three conventional ways to obtain the RAW (uncompressed) and FULL (all bands and metadata) Landsat-8 imagery data, which provide a single zip file for easy download:

    – USGS GloVIS (http://glovis.usgs.gov/)
    – USGS Earth Explorer (http://earthexplorer.usgs.gov/)
    – libra by developmentseed (https://libra.developmentseed.org/, allowing visual selection of right Landsat-8 scenes)

    Tim mentioned the first two but the third method is really cool for the casual use by the general public.

    Once the downloaded imagery data are unzipped, users may use the Spectral Transformer for Landsat-8 (GUI) to process Landsat-8 imagery for (1) bands combination, (2) image stretching, and (3) image pansharpening, with two simple button clicks.

    3 – The processed natural- or false-color composites in GeoTIFF format from image stretching (with 30m-resolution) or image pansharpening (with 15m-resolution) can be readily displayed in Google Earth Pro (not the Google Earth general version). It is all automatic since the free Google Earth Pro can read GeoTIFF files directly. In Google Earth Pro, it may not be super fast to display the full processed scene due to its large size, and users may just crop an Area of Interest (AOI) for a quick display.

    An example with much sharper imagery result is displayed in Google Earth Pro:
    http://www.geosage.com//products/spectral_transformer/landsat8/demo/LC81750832015182LGN00_B432_Fused_AOI.jpg

    Or download its KMZ file (~6MB) at
    http://www.geosage.com//products/spectral_transformer/landsat8/demo/LC81750832015182LGN00_B432_Fused_AOI.kmz

    (Another useful tip: Tim used free Gimp image editing software to convert a compressed GeoTIFF file to uncompressed; probably a more popular tool in the geospatial imaging processing field would be free gdal_translate.exe – http://www.gdal.org/gdal_translate.html
    Its typical usage is as follows: gdal_translate.exe b2_compressed.tif b2_uncompressed.tif)

  4. Frank Taylor says:

    Hats off to Tim for an interesting experiment and sharing the process and results! And, big thanks to the commenter above from ImageAnaysis (@1GeoSage) with tips on how to better use tools to process the Landsat 8 imagery.

Leave a Reply