Landsat 8 example using pan-sharpening in Orfeo Toolbox

February this year USGS has sent another satellite up into orbit to continue observing the earth. This satellite, named Landsat 8, provides the same spatial resolution in all bands as Landsat 7. That is one pixel equals 15m pan-chromatic and 30m for all other bands (excluding thermal). In addition, it has a couple extra bands for water, cloud and surface temperature (this link explains the other similarities and differences in more detail).

Here is a sample that I have pan-sharpened using the open source Orfeo Toolbox. Bands displayed are 6, 5 & 2, which is comparable to 5, 4 & 1 in Landsat 5 & 7. This image was capture 27th of May 2013. Most of the bright green paddocks are canola and you can see some of the earlier cereals coming through.

Bands 6, 5 & 2. Extract from scene Path 91, Row 80 on  27-May-2013
Landsat 8: Bands 6, 5 & 2. Extract from scene Path 91, Row 80 on 27-May-2013

More info: As suggested in comments, Spectral Transformer for Landsat-8 imagery by GeoSage is another free option for pan-sharpening Landsat 8 imagery. I have tried it and it works well if you don’t mind following some simple command line instructions. Check it out at this link.

Even more info: To get an idea of the best Landsat 8 band combinations and comparison with Landsat 7 bands check out this blog article from ESRI.

Using MultiSpec to composite Landsat images

As most GIS users will know the Landsat archive of satellite images is available for free download from USGS. This is a valuable resourse of historic and recent course resolution images covering most of the earth. I find the USGS Global Visualization Viewer the best way to sort through and download images. You need to add images to the basket. Often you can download them directly from the basket, but sometimes they need processing time which can take a day or two. Note that images from L7 2003 onwards have lines of no data caused by a faulty sensor. Up until recently L5 was still working producing quality images but is now out of service. Shortly we will have Landsat 8 – it has been launched and more free imagery is not far off!

Once the image file is downloaded you will have a compressed .tar.gz file which can extracted using built in Windows functions or 7Zip. Once extracted you have a TIFF (.tif) file for each of the bands of the image. For L7, that is 9 TIFF files about 60mb each except band 8 which is higher resolution band for pan-sharpening usually 200mb+.

The next step is to composite these bands to make a coloured image to load into your GIS. If you are a Windows or Mac user I suggest you try out MultiSpec. Download and extract to a folder of you choice. I keep it at C:MultiSpec for simplicity.

A comprehensive guide on how to composite images has been written up by the creators of MultiSpec and is available here. The process I give below is a simplified explanation based on the MultiSpec tutorial.

1. Launch MultiSpec.

2. Go to File > Open and select all bands you want to composite into one image.

Open multiple files
Open multiple files

3. For the ‘Set Display Specifications’ dialogue press OK (Try setting Bands to Red: 5, Green: 4 and Blue: 1. This is not critical for this exercise but is useful to visually inspect the image inside MultiSpec).

4. Press Ctrl + R to open ‘Set Image File Format Change Specifications’

Set Image File Format Change Sepecifications
Set Image File Format Change Specifications

5. You can select Channels and reorder by choosing ‘Subset…’ (I usually include bands 1-5) and also select file format in the ‘Header’ drop box (I always stick with GeoTIFF format). Press OK and choose where you want your composite image to reside and give it a useful name (do not forget to add the .tif extension).

6. Load up your favorite GIS desktop software, QGIS for me, and add your new TIFF file.

7. Edit layer properties and try a few different band combinations and contrast enhancements.
For good visibility of vegetation try Red: Band 5, Green: Band 4 and Blue: Band 1.
For a more natural look try Red: Band 3, Green: Band 2 and Blue: Band 1.

Try different band combinations and contrast enhancement
Try different band combinations and contrast enhancement

I hope this helps make Landsat imagery more accessible to everyone.

Flood water satellite image NW NSW & SW QLD

If you were wondering where the agmapsonline homepage image comes from – this is it. A  Landsat5 image captured on 23-03-2010. We have had two floods that have exceeded this flood level since.

Notice that the cotton paddocks remain completely protected. In fact you will find many cotton farmers still irrigating during large flood events. Other areas where the water spreads out is useful for filling moisture profiles and dropping valuable nutrients.

This is not a ‘true colour’ image. Of the three layered bands used to create the image I have only used one colour band and that is blue. The other bands used to produce the image are near infrared (NIR) and short wave infrared (SWIR). The SWIR, NIR and Blue bands are applied to the red and green and blue scales respectively to produce the image that you see.

This combination shows up healthy vegetation as bright green and water as blue (even though inland water is generally brown). Learn more about this here.

Click on image to view it full size.