This Weeks Lab
Two DEMs are provided for a portion of Yukon territory, NTS map sheet 115B14 west.
You can find this area in google maps or earth by heading to 60 50N / 139 20W;
the main feature is the Kluane Glacier, SW from Kluane Lake.
All data are in /home/labs/geog457/lab6-dem-topo-effect
DEM 1977: dem-ntdb.pix
DEM 2007: dem-spot.pix
DEM reliability: dem-rely.tif
ETM 2001: landsat321.tif
Topo map: nts115b14.tif
The earlier DEM from geobase.ca corresponds to the NTS map sheet from aerial photography taken in 1977 and has been resampled to 40 metres resolution, and projected to UTM zone 7. [This is required to make them comparable and a tougher instructor would have made you do it as well – hold it Scott made you do it! I guess we are getting tougher.]
The second DEM is from SPOT 5m Pan imagery from 2007, with a resolution of 40m.
The PAN image is also available
Note the DEM resolution is 40 metres for all operations
1. You will need to have the two DEMs in one file for comparison.
Copy the 1977 DEM file to your own folder, and then use file-> utility-> transfer to import the 2007 DEM into your copy of the 1977 pix file. You should now have a single file with the 1977 DEM in channel 1 and the 2007 DEM in channel 2. Both are 16 bit signed channels.
Question: Why is it that we can just transfer the channels into one file and not warp them (using orthoengine) into one dataset?
2. View the two DEMs, along with the SPOT PAN image and also the 1977 vectors.
3. create shaded relief (REL), directing output only to the screen. Compare their histograms looking for any major difference between mean and standard deviations.
Note that the 2007 DEM was complicated by clouds especially in the NW corner…“we looked at DEMs from both dates now, but clouds got in our way” (a famous Canadian song writer)
Load the reliability image for the 2007 DEM (dem-rely.tif) using the pseudocolor mode, and inspect the DNs for this channel (histogram) – I’m not sure what a value > 100 is!
4. Use simultaneous display learned in GEOG432 to see the two DEMs superimposed:
Layer -> add -> RGB and load channel 1 (1977) in red and channel 2 (2007) in both green and blue. Red areas will show where the 2007 DEM is falsely low, and blue-green areas where it is falsely high – the one cloiud covered area stands out along with some smaller areas towards the south.
5. Generate the difference channel 1977 (2007 as below) and display in pseudocolour
You can do this in two ways: using either operation ARI or tools-> easi modeling. For the former, you would direct the result to your file and it will generate the new channel, or if you use easi modeling, you need to first create an empty 16 bit signed channel:
(go to maps tree on left, right click on file name, add -> rasters -> 16 bit signed)
[personally I prefer easi-modelling]
It will likely display as fairly flat in contrast, as the range is large as a result of ‘errors’, due to clouds or steep slopes. Clouds give false elevations; steep slopes invite errors.
Note that 1977 minus 2007 yields positive DNs for elevation decrease, and we might assume all negative values are errors of some kind, along with extreme positive values.
6. View the histogram for the created difference channel to understand the values created.
It should centre near zero, but what is the mean …. perhaps slightly negative ?
True changes are likely near the centre.
Edit the colours, by:
Right-click on the layer in the files tree. Select edit pseudocolours
Select range-based instead of single value (tabs); Click custom instead of standard
You want to exclude values that might be considered errors. Under the histogram, reduce the range to exclude values beyond -100 and click compress, then also reduce the positive range (compress again). You should settle on a value that seems to exclude errors but not large real glacier changes. Don’t spend too long on playing with colours – that’s all they are… and the numbers shown do not correspond to the DNs of the difference channel.. View the ntdb ice polygon vectors to see where the glaciers are (or were!). We could also create a bitmap to exclude errors either by DN or location.
7. Calculating change – Questions
Excluding errors, what is the largest change in elevation over 30 years, and hence what is the annual change for that location?
Comparing the PAN image and NTDB vectors, what is the general horizontal retreat in metres (simply counting or measuring the number of pixels)?
Load also the landsat image – unfortunately in visible bands only and compare the snout with the SPOT image – how much has it retreated in 6 years? Has the rate of retreat accelerated from 1977-2001?
Load also the scanned mapsheet – this is the map that was digitised to create the NTDB ice vectors. How much has the glacier retreated in the 30 years between NTS and SPOT?
8. Bitmap creation for volume calculation
We will use the reliability layer to create a bitmap which identifies all pixels with ‘0’ reliability (= clouds or other effects), and thus to help isolate areas of real change. We assume that non-zero values are meaningful. We might also exclude the values of 104 and 108.
First use file-> utility -> transfer to import this layer into your ‘masterfile’ (1977 DEM)
Then add a new empty bitmap – in the files tree on left, right click and add one bitmap, note its number, then in easi-modelling:
If %x > 0 and %x < 104 then
%%y = 1
(where x is the reliability channel number, and y is the new bitmap)
Display the result visually to check -
7. Volume (task VLM)
We can calculate the volume of ice that has melted, but would need to exclude the erroneous DEM values from clouds… let’s hope our bitmap did it. We could also have created a mask from the 1977 NTDB ice extent vectors using task POLY2BIT.
Use the bitmap created in 6 above to use as a mask in VLM. Now use POLY2BIT on the ntdb vectors and run VLM again using your new mask.
The result should be reported in cubic kms
Question – explain how you decided upon your base value?