Dataset Viewer

The dataset viewer is not available because its heuristics could not detect any supported data files. You can try uploading some data files, or configuring the data files location manually.

canVODpy reference stores

The reference stores of canVODpy, which computes vegetation optical depth (VOD) from GNSS signals (GNSS Transmissometry). They are GNSS stores and VOD stores (Icechunk) built from the canVODpy test data with a fixed set of settings, called setting profiles. New canVODpy code must reproduce them: a change that alters the stored observations, the satellite geometry or the VOD shows up as a difference.

The stores were built by canVODpy at commit bd2e6c7b from the test data at commit f0764c4. The releases of the test data are archived on Zenodo (10.5281/zenodo.19708759, all versions).

Setting profiles

Profile Sites Settings
default Rosalia (RINEX v3.04, SBF, NMEA), MOFLUX (RINEX v2.11), ExampleSite (stripped RINEX v3.05), Hainich (NMEA); one day each final SP3 orbits (CODE), up to 15 signals (SIDs) per system, no preprocessing, SNR only, VOD of each receiver pair
sbf-broadcast Rosalia (SBF) as default, but the satellite geometry from the broadcast ephemerides the SBF files carry, instead of the final SP3 orbits
gridded as default as default, plus observations aggregated to 1-minute means and assigned to the cells of a 2° equal-area grid

The profiles themselves are in the canVODpy repository, under reference/profiles/, and the reference stores guide describes them in full.

Files

One zip per store: <profile>/<site>-gnss.zip and <profile>/<site>-vod.zip, 26 zips and 165 MB in all. SHA256SUMS lists the sha256 of every zip. Each zip unpacks into stores/<site>/<gnss|vod>/, so the zips of one profile can be unpacked into the same folder.

The zips are deterministic: packing the same stores again gives the same sha256. They hold only the stores: no settings files, no logs and no local paths. The stores name no real person or computer: their metadata is the made-up identity of canVODpy's tests, and their commits don't record the computer that built them. Each store records the receiver position its test data carries, links to this dataset and names the other store of its site (GNSS or VOD).

Use

canVODpy downloads, checks and unpacks the zips itself. From a checkout of the canVODpy repository with the test data:

just reference-check

This runs every profile again and compares each new store with its reference store. It reads the zips of one commit of this repository, named in canvodpy/src/canvodpy/reference-stores/published.yaml, and checks each zip's sha256, so a later change here never changes what a given canVODpy version compares with.

Citation

Cite canVODpy (10.5281/zenodo.18496233) and the test data the stores come from (10.5281/zenodo.19708759).

License

Apache License 2.0, as the test data.

Downloads last month
52