You are hereAttribution of change

Attribution of change


Many processes act simultaneously to change landscapes.  For the purposes of management, modeling, and understanding landscapes, we must understand not just when and where landscape change occurred, but what caused it.   We call this "change attribution."  

In our workflow, trained human interpreters make the initial calls about what caused change in several hundred individual patches on landscapes.  As humans, we are very good at labeling what caused a change.  Using time-series satellite data and historical photos, we can consider not only the color and shape of an area affected by some change, but also its context on the landscape and the the processes we know may be taking place in a given landscape.  

Here is an example of the interface we use to help interpreters determine agents of change:

To see a live example of this interface, please go to:

http://timesync.forestry.oregonstate.edu/ltattribution/public/bin/

Log in with user name:  attruser

Password:  attruser

From the "Projects" drop down, select "nmfs", and from the "Scene" drop down, select "4627".  Click on a patch ID from the list on the left to zoom to that location.  

Once we have accumulated labels for several hundred polygons, an automated algorithm can take over to scale to the tens or hundreds of thousands of polygons on the rest of the landcsape.