Skip to content
Go to file

Latest commit


Git stats


Failed to load latest commit information.
Latest commit message
Commit time


Picking manual melt pond depths from ICESat-2 photon data plots

(a.k.a. "Science Humans vs. Science Machines")


We are working on comparing different algorithms that extract the depth of melt ponds on Amery ice shelf from ATL03 photon data (see images below). Because humans tend to be pretty talented at recognizing patterns in such unstructured data, we want to investigate how the algorithm estimates compare to where humans would draw the melt pond surface and bed in these data.

Melt pond data Sentinel-2 and ICESat-2

Fig.1 - ICESat-2 ATL03 photon elevation data over four melt ponds on Amery Ice Shelf. (Track 0081, 2 Jan 2019, ATL03 Release 003, granule ATL03_20190102184312_00810210_003_01.h5, beam gt2l)

How YOU can help:

If you'd be willing to use about 10 minutes of your time to help us by connecting some dots on four plots below, we'd greatly appreciated that. It's a good short distraction, and the more estimates we get, the better the data!

All you need to do is run the cells in this notebook in sequence, and trace out the surface and lake bed with line segments by clicking on the plots that are generated. The results should look something along the lines of this:

an example of picked surfaces an example of all picked surfaces when checking the data

To start running the notebook, launch it in Binder by clicking the logo, or go to the link below.

NOTE: Sometimes there seems to be a second, flat return surface just about half a metre below the pond surface (see below). This is due to oversaturation and dead-time of the sensor. Such second returns are not an actual surface, and should be IGNORED! examples of a second return due to sensor oversaturation and dead-time


A repo for manually picking surfaces in ICESat-2 photon data from melt ponds on Amery Ice Shelf




No releases published


No packages published