Winter 2018 GEOG457/657: Lab 2 – Working with PCA
Sentinel Data Download
Sentinel data can be downloaded from the the European Space Agency Data Hub:
The web interface is similar to Earth Explorer and although it is newer in it look and feel – It is a bit clunky though. For a better interface for looking at Sentinel data – you can use a web interface Roger introduced in Geog 432 – remotepixel.ca
Look for the data we are using for this weeks lab. Use the remotepixel web interface and look for a scene for the area around Hazelton/Terrace collected 2017-08-03 (with a scene label – 09UWA). Once you click on the scene number, you need to look through the dates and then find the date we are looking for.
You are not required to download any data today – but these are the methods for obtaining sentinel data. Scott usually looks in remotepixel, then downloads from ESA (this way you get all the metadata.
Loading and clipping Sentinel Data with PCI
The data we are using for this lab is in:
We are getting quite used to using the MTL file to load the TIF files download from Earth Explorer – and we have the same benefits with Sentinel data and PCI. Open focus and navigate down through the sentinel folder (into the folder created by unzipping the sentinel data (ending in SAFE)), and open the file called manifest.safe
You will prompted to open data based on its resolution – open the 10 metre data. You will be presented with the visual band data in the “natural look” 4-3-2. Play with the data using the NIR band to see changes in the presentation. In order to make things move a bit quicker – lets subset out the data. Great a new layer is around 6000 X 6000 pixels that include Terrace in the scene.
We have similar data to that of Landsat 8 (although much higher resolution). If we were to perform a classification for with this 10 metre data – what would be missing in comparison to the OLI data? What features could be more difficult to classify?
Just for fun – because we know that is what it is all about, perform a quick unsupervised classification using the 10 metre bands to see what kind of results you get.
PCA with Sentinel-2 and Landsat data
Now lest try a Principal Component Analysis. PCA is an function in the Algorithm Toolbox. Open it up and run through it (open up the help as well to see what is happening with the algorithm). You are going to include all the channels and write out the same number of Eigenchannels. Use greyscale to the viewer to look at the data and also set the report out to be “long“.
What happened here!! What order a the PCA Eignchannels loaded into Focus? How do the results look in comparison to the lecture on PCA (remember the lecture was based on Landsat data.)?
Open up the Multi Spectral bands in Landsat 8 scene in the terrece_sentinel_oli folder for oli data. Clip the data using the extent of the dataset you created with the Sentinel data, BUT do not bother with bands 1,8 or 9. What is the resolution (number of pixels and rows)? How does this compare with the 6000×6000 Sentinel clip?
Once clipped – perform the same PCA analysis – again using all the bands and matching number of Eigenchannels
Now what happened!! Can you explain the reasons why this data set fits the lecture (other than the obvious fact the lecture was written for Landsat)?
Comparison with NIR/Red ratio
Perform a NIR/Red ratio for both clipped datasets. How do the results compare with each other? How do they compare with PCA 2 for both datasets?
We are going to perform Top Of the Atmospheric and Ground Reflectance Corrections to both the Sentinel-2 and Landsat 8 data. In the lecture we spoke about sensor adjustments in relation to where the satellite is relative the earths atmosphere as well as using sun angle to compensate for the ratio of potential irradiation on a pixel to the radiation being measure from the surface (reflection). We are going to use the metadata from each sensor to perform the analysis that PCI kept when you clipped out the data using the MTL and SAFE file. The Atmospheric correction tools can be found under Analysis –> Atmospheric Correction.
Open up the TOA wizard and select your subsetted pix file for the Sentinel data. Keep all the defaults and create an output directory(s) to put the results in (maybe atmospheric_correction/toa). Review the results of the TOA once it is loaded into the project. What did we notice about the values of the pixels of the TOA layer in comparison to the original clipped layer?
Perform the same correction to the OLI scene. How do the TOA and original pixels compare?
Use the same wizard method to perform ground reflectance corrections for both scenes. You may want to play around with calculating the sun angle this time as well as using band metadata or the “Import from Text File” option. The *.cal files PCI is expecting can be found in the /opt/geomatica_2017/atcor/cal folder.
Building a Multi Spectral Sentinel-2 Data set using both 10 metre and 20 metre resolution data
We are going to make a Sentinel dataset with visual bands, SWIR and NIR that approximates OLI data- but with Sentinel resolution(s). These steps produces a pix file that holds all the Sentinel-2 bands at a common resolution.
20 Metre Resolution dataset
- Load in the original SAFE file for Sentinel-2 with both 10 metre and 20 metre data
- Clip the data to a region that has Terraace and the large lakes below and above the town at a 3000X3000 pixel dataset using the 20 metre data (file 1)
- Clip the 10 meter data using the extend from the layer above (file 2)
- Use the RESAMP algorithm to resample the 10 metre data into a new 20 metre resolution file (file 3)
- Now use transfer layers utility to move this newly 20 metre data (file 3) into the orgianl 20 metre clip file (file 1)
10 Metre Resolution dataset
How do we create this using the same tools as above?