The previous tutorials used the Low-Rate ocean products. SWOT also produces a High-Rate pixel cloud — SWOT_L2_HR_PIXC — the geolocated radar returns (tens of millions of points per granule) with water/land classification, height, and backscatter per pixel. It targets inland and coastal water, and is handy near the coast, in estuaries, and wherever the 2 km ocean grid is too coarse.

Both PIXC tools live under the unified odsl-swot-swath CLI.

What you'll learn

  • Discovering and downloading PIXC granules for a time window and box.
  • Making summary maps of a granule's classification, height, and backscatter.

Prerequisites: tutorial 1; Earthdata credentials in ~/.netrc for downloading (tutorial 2).

1. Discover first, download second

PIXC granules are large (hundreds of MB to a few GB each), and a time+bbox query can match many of them. Preview what a query returns before committing the disk space:

odsl-swot-swath pixc-download \
    --start 2024-06-01 --end 2024-06-02 \
    --bbox -98 18 -80 31 \
    --out-dir ~/swot/pixc \
    --manifest-only

--manifest-only queries CMR and writes only the manifest — SWOT_L2_HR_PIXC_D_manifest.csv (and .json) in --out-dir — listing every matching granule with its size and link. Inspect it, then re-run the same command without --manifest-only to download:

odsl-swot-swath pixc-download \
    --start 2024-06-01 --end 2024-06-02 \
    --bbox -98 18 -80 31 \
    --out-dir ~/swot/pixc

Notes:

  • --start/--end take dates (2024-06-01) or full ISO timestamps; the bbox uses [-180, 180] longitudes.
  • Always set --out-dir — the built-in default is a data/ folder inside the repository, which is not where you want gigabytes of granules.
  • Downloads are idempotent (existing files are skipped; --overwrite forces).
  • --max-granules N caps the download — useful to grab one granule for a first look.

2. Summary maps of a granule

odsl-swot-swath pixc-plot ~/swot/pixc/SWOT_L2_HR_PIXC_*.nc

This reads the pixel cloud and writes a multi-panel PNG (<granule stem>_summary.png next to the input, or --output): a classification panel (land → open water categories) plus one panel each for height, water_frac, sig0, phase_noise_std, and cross_track.

Useful options:

  • --variables height sig0 — plot only those panels (classification is always first).
  • --water-only — drop land-classified pixels, so water features stretch the colour scale.
  • --max-points — granules have tens of millions of points; the plotter stride-samples down to this cap (default 300,000). Raise it for more detail, at the cost of time and memory.
  • --wrap-longitude — for granules straddling the antimeridian.
  • --show — interactive window instead of just the PNG.

3. Going further

The pixel cloud is point data, not a grid: analyses beyond quicklooks usually start by opening the pixel_cloud NetCDF group directly —

import xarray as xr
ds = xr.open_dataset("~/swot/pixc/SWOT_L2_HR_PIXC_....nc", group="pixel_cloud")
print(ds[["latitude", "longitude", "height", "classification"]])

— then filtering by classification and binning to whatever geometry your problem needs (the common.swath_regrid helpers work on any scattered lat/lon points).

Troubleshooting

  • Query matches nothing — PIXC coverage targets rivers, lakes, and coasts; a purely open-ocean box can legitimately return few granules. Widen the time window or move the box coastward.
  • Plot looks sparse — that's the --max-points stride sampling; raise the cap.
  • Out of disk — check the manifest's sizes before downloading, and use --max-granules while exploring.

Series recap: 1 — Getting started · 2 — Downloading · 3 — Finding passes · 4 — Subset cubes · 5 — Regridding · 6 — PIXC. For advanced topics — per-cycle MPI combining (swot/combine_passes.py) and geometric eddy/internal-wave separation (swot/geometric_separation_eddy_waves_ssh.py) — see the swot package README.