Friday, March 19, 2010

Good News

Good News! After many emails back and forth with Kristina from Portland we were able to figure out the Temporal Outline issue for Hood. She eventually found another file in a different projection and was easily able to re-project it to the Data Frame Projection I’m using and it fits perfectly! Now I have access to all the years I mentioned in a previous post.

I spent most of today going through and documenting the steps taken to determine the threshold value I’m going to use for analysis. I tried several different classification schemes, all with varying results but found the best to be a “defined interval” classification scheme within ArcMap where I specified the interval to be .05. I then looked at the CCannon outlines and the Temporal Outlines from Hood to delineate the upper boundary and edges of both the Reid and the Elliot glaciers and then hand outlined the areas of “debris-free” ice at the bottom based off the 2005 NAIP .5m resolution imagery. This gave me outlines of the ice for both well defined glaciers. I then calculated the square meter area of the outlines I created. Then I went through and tried different classification thresholds based on the scheme I described above. Previously I had been using a “Natural Breaks” classification where .3 seemed to work ok, but I found after looking through the ArcMap help files and online that the defined interval would work better. With this I was able to get a more precise threshold value. I eventually got it down to three values. 0, .05, and 1. I then went though a series of steps (not going to describe them here but have them written down) to look at the area those threshold values created within a 60m buffer of the debris-free ice outline I had created for both glaciers. The goal was to find the best threshold where the area would be closest to that of the outline I drew based on the NAIP as well as visual “fit” within the outline I drew. This has been a little tricky as the 30x30 resolution of the Landsat has trade-offs. When I calculated the figures out .05 seemed to be the best threshold. The Reid outline I created was 542,294m2 and the .05 threshold created an area of 538,200m2 which fit pretty good and the difference is less than 1%. The Elliot came in under 1% difference as well with the .05 threshold. So, I think I’ve done pretty well justifying how I’ve picked that threshold and I’ll apply it to all the band ratio’s for both Hood and Rainier.

The first image below shows how the Temporal outlines now fit great and the one below is how I determined the threshold. I figured using two glaciers instead of one, and ones on opposite sides of the mountain would be even better justification for the threshold value I picked. I’m hoping to be clearer than many of the papers I’ve read on how the very important value was chosen.

Have a great weekend!

1 comment:

Jon Ellinger said...

This looks great!
Tuesday, March 30, 2010 - 05:08 AM