Progress for Boise

I've been doing a project for the Rocky Mountain Research Station (RMRS) in Boise, ID, recently. Parts of it have been extremely frustrating, but the overall point of the project is somewhat interesting. The RMRS is attempting to model stream surface temperature based on the radiation each stream receives from the sun. The logic is that a dense canopy (forested) over a stream or river will block a lot of the direct sunlight and therefore not influence the stream temperature very much. In areas that have no canopy (like a recently burned area, for example), it is anticipated that the water there will be more influenced by the direct sunlight.

We're trying to see if this idea will work with moderate-resolution satellite imagery like Landsat. I'm personally a little skeptical because image registration is so important when studying riparian areas. If the image alignment is off even one pixel, you may completely miss the boat (hah, that was punny, wasn't it?).

I completed a NDVI change product between two dates (1989 and 2002) and a vegetation classification of the 2002 image today. The NDVI change was a little fun because I had to radiometrically correct the 1989 NDVI image to match the 2002 image. I'd never tried that derived image layers, such as a NDVI. I created it by computing a Model II regression line capturing the relationship between temporally invariate clusters in both dates. My r-squared wasn't very good (probably due to image mis-registration), but it was fun to try this.