AntarcticaData

Processing pipeline for a multi-band, multi-constellation GNSS dataset recorded continuously aboard R/V Laura Bassi on a 216-day closed-loop voyage Trieste → Panama Canal → Ross Sea (Antarctica) → Cape Horn → Trieste (26 September 2025 to 29 April 2026). A u-blox ZED-F9P-15B receiver (firmware HPG L1L5 1.40) with a Tersus AX4E02 antenna logged raw observables, broadcast navigation, the RF spectrum of both bands (MON-SPAN) and receiver health telemetry to one .ubx file per day (216 files, 176.9 GB).

• Data paper: Bertalanič, B., Dimc, F., and Blatnik, A.: A multi-band, multi-constellation shipborne GNSS dataset from a 216-day closed-loop voyage between Europe and Antarctica, Earth System Science Data (accepted). Discussion paper: https://doi.org/10.5194/essd-2026-432
• Dataset (Zenodo, CC-BY 4.0): https://doi.org/10.5281/zenodo.20344100 (concept DOI, resolves to the latest version)
• Code: MIT licence. Version v1.1 is the version used in the paper.

Layout

.
├── ubx/          Binary UBX parser (mmap + structured NumPy dtypes, vectorised; no pyubx2)
├── analysis/     Trajectory, daily QC, multipath, RF spectrum, TEC, scintillation proxy,
│                 anomaly and spoofing catalogues, jamming confound
├── ppp/          PRIDE PPP-AR wrapper, ANTEX preparation, real-time vs PPP comparison
├── rinex/        convbin wrapper (RINEX 3.04 conversion)
├── figures/      Figure scripts of the paper
├── tables/       Table scripts of the paper (LaTeX + CSV)
├── verify/       Validation analyses of the paper, pytest tests and parser fixtures
├── config/       pipeline.yaml, receiver.yaml, ports.yaml
├── Snakefile     Per-day workflow over the 216 days (idempotent)
├── run_ppp.sh        RINEX conversion and kinematic PPP for all days
├── run_strengthen.sh PPP checks, real-time vs PPP comparison, PPP berth repeatability
└── run_revision.sh   Cruise-wide tables, statistics and figures of the paper

Install

Python ≥ 3.11. Cartopy needs the GEOS and PROJ system libraries (macOS: brew install geos proj; Debian/Ubuntu: apt install libgeos-dev libproj-dev).

git clone https://github.com/sensorlab/AntarcticaData.git
cd AntarcticaData
python -m venv .venv && source .venv/bin/activate
pip install -e .          # add [dev] for pytest, ruff and mypy

External programs (RINEX and PPP)

Only the RINEX archive and the PPP comparison need two programs that cannot be installed with pip. Parsing, the per-day analysis and most figures run with Python alone. Both programs build with a standard Linux toolchain (gcc, gfortran, make, bc) and must be on PATH.

Program	Used for	Source	Notes
convbin	UBX → RINEX 3.04 (rinex/convbin_runner.py)	RTKLIB demo5 fork (rtklibexplorer)	The deposited RINEX files were written by CONVBIN EX 2.5.0. Mainline RTKLIB maps some F9P signals incorrectly (e.g. B2a as B2I).
pdp3	Kinematic PPP-AR (ppp/pride_runner.py)	PRIDE PPP-AR (Geng et al., 2019)	The paper used version 3.2.8. Build with the upstream install.sh.

# convbin (RTKLIB demo5, v2.5.0)
curl -sL https://github.com/rtklibexplorer/RTKLIB/archive/refs/tags/v2.5.0.tar.gz | tar xz
make -C RTKLIB-2.5.0/app/consapp/convbin/gcc CC="gcc -Wno-error=implicit-function-declaration"
# put RTKLIB-2.5.0/app/consapp/convbin/gcc/convbin on PATH

# pdp3 (PRIDE PPP-AR)
git clone https://github.com/PrideLab/PRIDE-PPPAR.git
cd PRIDE-PPPAR && ./install.sh

ppp/pride_runner.py passes the frequencies explicitly (-frq G15 E15 C25 J15), because PRIDE's default needs GPS L2 and BeiDou B3I, which this receiver does not track. It also passes the first epoch as the session start (-s). Otherwise pdp3 searches for the first epoch on a whole second; the receiver's epochs lie 9–20 ms before the second, so the search takes tens of minutes and on some days starts the session minutes before the end of the file.

Receiver antenna calibration. PRIDE takes the IGS20 antenna file named in each day's clock product (for this cruise igs20_2353 to igs20_2415) and downloads it when it is missing. The downloaded file has no entry for the receiver antenna, and PRIDE then applies zero phase-centre corrections without a warning. Run this once before processing. It downloads those versions into PRIDE's table folder and appends the TRSAX4E02 NONE calibration from the NGS file ngs20.atx (also in the dataset):

bash ppp/prepare_pride_tables.sh /path/to/ngs20.atx

ppp/pride_runner.py stops on any day whose antenna file still lacks the entry.

Orbit, clock and ionosphere products

pdp3 downloads the orbit, clock and bias products itself (Wuhan University rapid multi-GNSS products, WUM0MGXRAP, by default) and caches them in ~/.PRIDE_PPPAR/data. verify/tec_vs_gim.py downloads the CODE global ionosphere maps (IONEX) from the anonymous AIUB server. Neither needs an account.

ppp/igs_products.py is an optional helper that downloads IGS final products from NASA CDDIS. It is not needed for the results of the paper. CDDIS requires a free NASA Earthdata account, configured once in ~/.netrc (chmod 600 ~/.netrc):

machine urs.earthdata.nasa.gov login YOUR_USER password YOUR_PASSWORD

Configure

The default config/pipeline.yaml expects the repository next to two folders, and the run scripts write to the same work/ folder:

parent/
├── AntarcticaData/          this repository
├── antarctica2026/data/     the daily YYYYMMDD.ubx files
└── work/                    all derived output (staging, derived, figures, tables, logs)

For another layout, edit the paths in config/pipeline.yaml (relative to the repository folder unless absolute) and UBX_DIR at the top of run_ppp.sh. config/receiver.yaml describes the receiver and antenna setup, and config/ports.yaml lists the port stops used for the berth-repeatability analysis.

Data access

The 216 daily .ubx files are on Zenodo (https://doi.org/10.5281/zenodo.20344100) in eight monthly 7z archives (ubx_YYYY_MM.7z, 63.3 GB in total). Unpack them into one folder and point paths.ubx_dir at it. The RINEX 3.04 files are in rinex3_all.tar, one folder per day with obs.crx.gz (Hatanaka-compressed and gzipped) and nav.rnx.gz.

Quick start (one day)

# Parse one .ubx → 8 Parquet files + 1 Zarr block in work/staging/<yyyymmdd>/
python -m ubx.cli parse /path/to/20250930.ubx

# Per-day analysis
python -m analysis.qc_summary  --day 20250930
python -m analysis.trajectory  --day 20250930
python -m analysis.multipath   --day 20250930
python -m analysis.spectrum    --day 20250930
python -m figures.make_figures --day 20250930 --quick

Full dataset

snakemake --cores 32 -p                  # default target: parse + analysis + figures
snakemake --cores 32 -p analysis_all     # only per-day analysis
snakemake --cores 32 -p figures          # only figure rendering
snakemake --cores 32 -p rinex_all        # RINEX 3.04 conversion (opt-in)
snakemake --cores 32 -p ppp_all          # kinematic PPP-AR (opt-in, heavy)

The workflow is idempotent (based on file modification times). Days are found by scanning paths.ubx_dir for YYYYMMDD.ubx. On a 32-core workstation, parsing and the per-day analysis of all files take about one hour.

Products of the paper

After the per-day stages, three scripts reproduce the remaining products of the paper. Run them from the repository folder with the environment active.

bash run_ppp.sh --jobs 8      # RINEX conversion and kinematic PPP for all days
bash run_strengthen.sh        # PPP checks, real-time vs PPP comparison (ppp.compare,
                              #   ppp.percentiles), PPP berth repeatability
bash run_revision.sh          # file inventory and gaps, framing integrity, observation
                              #   accuracy, multipath summaries, autocorrelation-aware
                              #   statistics, OGS cross-check and ship meteorology

PRIDE PPP-AR needs up to about 3 h per day of data, so the PPP of the whole cruise takes two to three days with eight days processed in parallel. python -m verify.ppp_run_summary lists, per day, the systems used, the antenna file and any product fallbacks.

Stages

1. parse: .ubx → per-message Parquet (zstd-3, 128 MB row groups) + Zarr for the MON-SPAN spectra.
2. analysis: trajectory, QC, code multipath (M1/M2 after Estey and Meertens), RF spectrum and RFI, TEC (validation), scintillation σ_φ proxy.
3. figures and tables: the figures and tables of the paper.
4. PPP: kinematic PRIDE PPP-AR per day, compared with the receiver's NAV-HPPOSLLH solution.
5. RINEX: convbin (RTKLIB demo5) → obs.rnx + nav.rnx (RINEX 3.04).

Recorded UBX messages

Class-ID	Name	Purpose	Rate
0x02-0x15	RXM-RAWX	Raw observables: pseudorange, carrier phase, Doppler, C/N₀, lock time, tracking status per signal	2 Hz
0x02-0x13	RXM-SFRBX	Broadcast navigation subframes (ephemeris and almanac source)	per subframe
0x02-0x14	RXM-MEASX	Pre-PVT multi-band measurements	2 Hz
0x01-0x07	NAV-PVT	Position, velocity and time fix	2 Hz
0x01-0x14	NAV-HPPOSLLH	High-precision latitude, longitude and height	2 Hz
0x01-0x35	NAV-SAT	Per-satellite C/N₀, elevation, azimuth, used flags	2 Hz
0x0A-0x38	MON-RF	AGC, noise floor, jamming indicator, antenna status per RF block	1 Hz
0x0A-0x31	MON-SPAN	RF spectrum, 2 RF blocks × 256 bins, 128 MHz span, 500 kHz per bin (RF0 centred at 1583.5 MHz, L1; RF1 at 1191.5 MHz, L2/L5)	1 Hz
0x0A-0x39	MON-SYS	CPU load, run time, receiver temperature	1 Hz

The receiver tracks seven satellite systems (GPS, GLONASS, Galileo, BeiDou, QZSS, NavIC and SBAS). GPS Block IIF/III, Galileo, BeiDou-3 and QZSS are recorded on L1 and L5; GLONASS, SBAS, BeiDou-2 and the older GPS satellites on L1 only; NavIC on L5 only.

Tests

pytest verify/                 # parser and schema tests
RUN_BIG=1 pytest verify/       # adds an end-to-end test on a full day of staged data

verify/fixtures/sample_60s.ubx is a 60-second clip for the parser round-trip and schema tests. The validation analyses in verify/ (e.g. tec_vs_gim.py, ogs_cross_validation.py, daily_inference.py) need the full Parquet and Zarr outputs and are not part of the default pytest run.

References

Methods and external programs used by the pipeline:

• RTKLIB demo5: Everett, T. (rtklibexplorer fork), RINEX conversion (convbin). https://github.com/rtklibexplorer/RTKLIB
• PRIDE PPP-AR: Geng, J., Chen, X., Pan, Y., Mao, S., Li, C., Zhou, J., and Zhang, K.: PRIDE PPP-AR: an open-source software for GNSS PPP ambiguity resolution, GPS Solut., 23, 91, 2019. https://doi.org/10.1007/s10291-019-0888-1
• u-blox documentation: u-blox AG, ZED-F9P-15B High Precision GNSS Module Data Sheet (UBX-23009090) and u-blox F9 HPG L1L5 1.40 Interface Description (UBX-23006991).
• Code multipath M1/M2: Estey, L. H. and Meertens, C. M.: TEQC: The multi-purpose toolkit for GPS/GLONASS data, GPS Solut., 3, 42–49, 1999.
• IGS products: Johnston, G., Riddell, A., and Hausler, G.: The International GNSS Service, in: Springer Handbook of Global Navigation Satellite Systems, 967–982, 2017.

Citation

If you use the code, please cite the paper and the dataset:

Bertalanič, B., Dimc, F., and Blatnik, A.: A multi-band, multi-constellation shipborne GNSS dataset from a 216-day closed-loop voyage between Europe and Antarctica, Earth Syst. Sci. Data (accepted), discussion paper: https://doi.org/10.5194/essd-2026-432, 2026.

Blatnik, A., Bertalanič, B., and Dimc, F.: A Comprehensive GNSS Dataset from the 41st Laura Bassi Expedition to Antarctica, Zenodo [data set], https://doi.org/10.5281/zenodo.20344100, 2026.

The DOI of the final paper will be added on publication.

License

MIT, see LICENSE.