snowreach: virga classification and blind-zone simulation for vertically pointing radars

Two methods for deciding whether the snow a vertically pointing radar detects aloft reaches the ground, released with Roversi et al., Virga hidden within the blind zone of spaceborne radars: implications for surface snowfall estimates in coastal East Antarctica (Atmospheric Chemistry and Physics, submitted).

• A classification of radar profiles into precipitation, virga and noise, according to whether the echo survives to the lowest gate free of ground clutter. Two continuity criteria reject the near-surface spikes that ground clutter produces and the single-gate returns that carry no vertical structure.
• A simulation of the lowest observable height of a spaceborne radar, which hides the lower gates of a ground-based record and reports what the blind zone costs in occurrence and in accumulated water equivalent.

This is the method rather than the analysis pipeline of the paper: it carries no figure scripts and no intermediate files. Over the four years of the Mario Zucchelli record it reproduces the published accumulation table to within 0.3 %, and the excess at the simulated heights to the rounding of the paper.

The classification needs only equivalent reflectivity in dBZ over (time, range), so it applies to any vertically pointing profiler. A reader of the archived files also gets a reader for them, because the published NetCDF stores absolute time as a data variable and orders its dimensions the other way round.

One month of the record ships as a worked example, so both command-line scripts run as soon as the package is unpacked.

The reflectivity to snowfall rate relations collected in snowreach.zesr are published by others and are included only so that the simulation can return an accumulation. The papers they belong to are named in the code and in the README, and they are what should be cited for the coefficients.

Acknowledgement. The package was assembled from the analysis code of the paper with Claude Opus 5.5 (Anthropic), used through Claude Code: it extracted the two methods from a larger working repository, wrote the command-line scripts and the documentation, and checked the result against the published table. The methods are those of the paper, and the decisions about what to release and the verification of the outcome are the author's.