aggregate L2 data using dask

The new version offers a versatile method to grid the point elevation estimates using parallelism based on the dask package. By default the l3.build_dataset calculates the median elevation difference to the reference DEM. You can pass a custom aggregation function in case you are interested in other statistics.

For my testing region, I could run it on my local notebook, it took only 10 minutes for a 2-by-2 km grid on 16 CPUs, and less than 30 minutes for the "full-resolution" 500-by-500 m grid.

There are some limitations, however. So far, the aggregation period has to be a number of months and the aggregation function always works on the elevation difference to the reference DEM. The latter is fairly easy to adapt to your needs by changing the build_dataset function.

Parallel processing added

Processing L1b data to L2 is now parallelized.

Also, downloading L1b data now uses multiple threads.

However, it is, yet, not possible to have a backgroud thread to download the data while processing L1b files.

Patch bug finding swath start

Previous version occasionally discarded waveforms if POCA couldn't be found.

Tested to mostly work

This is the second pre-release. Never the less, this version is better tested and there were substantial improvements.

For the test cases, swath processing including group recognition and POCA retrieval behave as expected.

Full Changelog: v0.0.0...v0.1.0

Pre-beta release

This release is a preview of the cryoswath package.

While it can already build datasets of gridded surface elevations, it takes some adaptions to run it on another machine. On trying, you will need to obtain a reference DEM, provide glacier outlines, and set the file paths. Currently, there may be hardcoded preferences that, e.g., limits the use of cryoswath to the Arctic. However, if you patiently replace those parts you will be rewarded being the first user.