There's a single sentence in that paper about open water percentage change - which hides *months/years* of work developing a novel approach to object based image segmentation and classification over sea ice.

It was first conceived for the 2008 satellite calval paper linked below - using a home made quadtree split/merge algorithm based on image texture, then classification using texture + radiometry. Written in IDL, it sucked, but it worked:

https://doi.org/10.1029/2007JC004181

#seaice #remotesensing

2/5

...the 2008 IDL version was trashed, and we invested in commercial software (eCognition) for the 2012 iteration. A different approach entirely - mosaick up images, split the mosaic by blocks, analyse blocks, merge results.

Also in 2012, there were some folks aboard SIPEX-II with a fancy radiometer - so I asked if they could very kindly measure the reflectance in RGB bands over open water, and deep ridge shadows. I'd had a hunch that there's a tiny difference...

#seaice #remotesensing

3/5

...and it was true! So going back to imagery work, we could segment sea ice and use a tiny difference in RGB reflectance to discriminate between open water and shadow. In the 2008 work I'd used texture + broadband brightness to do this - so having an extra physical/spectral characteristic was a real win!

...unfortunately, it all stopped when funding stopped. No more papers because I can't keep working for free and my non-work time is busy with family and .. well .. life.

4/5

...I managed to do a little work on object based image analysis tooling in Python to segment melt ponds in 2021-23, which made it into a poster and ESA Living Planet Symposium talk.

The idea was to relate melt ponds and their biota to other features on sea ice using drone-derived + EM observed topography. What kinds of things live where, and maybe why?

All still open questions... just apply money! Hopefully to me :D

/fin