“Ground Truthing” maxent model


I have not seen the new 32 datasets and have been meaning to ask Tom Colson if they are in IRMA yet.

"Accuracy" in the case of the vegetation overstory dataset depends on the knowledge, experience, and intuition of whomever is interpreting the color infrared aerial photos and digitizing the polygons – the 80% accuracy is a measure of how well the photo interpreter’s data model matches the actual conditions on the ground (which again is only an estimate of accuracy since you can’t realistically visit every digitized polygon to assess reality).

Tom would have to correct me where I’m wrong, but I think the new 32 datasets have higher "accuracy" in the sense that the newer data models better reflect the true conditions on the ground – either from better quality digitizing, or from additional refinement of the prior data models based on subsequent site visits.

I don’t know if the new 32 datasets have higher resolution – but I doubt it: higher resolution data is expensive to collect – and as you point out, all the datasets would need to be the same resolution to combine in a spatial analysis, and it seems unlikely anyone has money for higher resolution than 30 by 30 meters (Outside Google and the NSA).

I do think that we can realistically (and affordably) get higher resolution data for hemlock locations (lat/long, not polygons) – meaning we use OpenForis Collect Earth to pinpoint dead hemlocks with Google Earth’s 15 x 15 m aerial images (1:2,400 scale, I presume).

This is in contrast to relying on spectral characteristics to create polygons from the 1:12,000 scale color infrared aerial images collected for the vegetation overstory layer back in the 90s.

I’ll ask the Forest Service if they already did the work of gathering point locations for hemlocks in the Smokies using remote sensing data.

If they haven’t, what I’m proposing would really come down to someone very patiently sitting down with OpenForis and systematically collecting spatial coordinates for hemlock skeletons preserved in Google Earth’s 15 x 15 m satellite images.

I have no clue how long that would take, but I’d be willing to try it out and provide an estimate.



About Tanner Jessel

I am a recent M.S. in Information Science graduate from the University of Tennessee School of Information Science. I was formerly a graduate research assistant funded by DataONE (Data Observation Network for Earth). Prior, I worked for four years as a content lead and biodiversity scientist with the U.S. Geological Survey's Biodiversity Informatics Program. Building on my work experience in biodiversity and environmental informatics, my work with DataONE focused on exploring the nature of scientific collaborations necessary for scientific inquiry. I also conducted research concerning user experience and usability, and assisted in development of member nodes with an emphasis on spatial data and infrastructure. I assisted with research designed to understand sociocultural issues within collaborative research communities. Through August 1, 2014, I was based at the Center for Information and Communication Studies at the University of Tennessee School of Information Science in Knoxville, Tennessee.

Posted on December 15, 2014, in Uncategorized. Bookmark the permalink. Leave a comment.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: