Additional inertial-data sources

Two archives offer raw observations for separate development cohorts. Neither has been added to the existing 53-stretch comparison. No new Earth-model fits have been run on these sources.

An IMU (inertial measurement unit) records angular motion and acceleration. The relevant files contain its observations, rather than only a finished navigation solution.

Longer, slower airborne surveys

The ESA report lists survey-leg speeds around 120–130 knots, or 222–241 km/h. Longer duration can compensate for slower travel in accumulated surface curvature: 240 km/h for 30 minutes covers about the same distance as 700 km/h for ten minutes. It does not automatically compensate for sensor bias or drift.

The iMAR Level-0 files are documented as ASCII records of GPS time, three accelerations, three angular rates in radians/second, and temperature. The report states that their axes have already been rotated to aircraft forward, starboard and down. It also describes separate raw GNSS observations in RINEX format. The inspected recording confirms approximately 300 Hz cadence. Header units and the report agree; the inspection does not independently calibrate the instrument.

The inventory distinguishes nine files from twelve flight labels: some flights share a file. The 50 reported line intervals total 20.52 hours; that is observing time, not fifty independent flights.

• Reported IMU files
• Verified live IMU file catalog
• Reported line intervals
• ESA catalog and access instructions
• Dataset access request

Downloaded ESA recording

187_rqh01.txt covers 8.49 hours on July 6, 2023. A streaming pass checked every row, preserved the original and reproduced its acquisition hash. Timestamp steps range from approximately 3.333 to 3.334 milliseconds. One startup row has near-zero acceleration and is recorded as a quality flag; it has not been silently removed. Eleven deterministic decoder tests check units, signed values, malformed records, nonfinite values, timestamp ordering and gap handling.

The matching GNSS position export contains 64,080 epochs over 8.90 hours. The survey report identifies it as GNSS-only precise point positioning, using satellite orbit and clock products. The export uses the software's generic “GNSS/INS Epochs” label; the report, rather than that label, establishes the stated processing provenance. Raw RINEX observations are available but have not been independently reprocessed here.

Both reported lines for this recording overlap the IMU and GNSS time coverage:

The longer line changes altitude substantially. These are descriptive survey intervals, not approved steady-flight analysis windows. Admission will use the independent trajectory and a published policy before any model-fit residuals are examined. No Earth-model fits have been run on ESA data.

IMU inspection · GNSS inspection · Reported-line trajectory details.

Downloaded Wuhan sample

The SmartPNT-POS archive includes fourteen UAV datasets among thirty numbered collections. The inspected sample is Data06_20210115_HG4930_UAV_Opensky. Its IMR header identifies a NovAtel SPAN HG4930, 100 Hz angle/velocity increments, GPS time tags and explicit conversion scales. Its mounting XML supplies a 90° rotation and antenna lever arm. The binary original and signed integer counts are preserved.

The matching raw RINEX 3.02 GNSS file contains 5,796 observation epochs over 19.32 minutes. Receiver positions have not been computed. A separately downloaded reference trajectory is explicitly labeled smoothed, combined GNSS/INS. Its reported maximum speed is only 29.2 km/h; this is motion context, not independent position evidence or a basis for admission.

The streaming inspector follows the manufacturer’s IMR specification, honors increment-versus-rate flags, preserves original time tags and applies no mounting, timing-bias or Earth-rate correction. Twelve deterministic format tests pass. IMR has no record checksum; structural checks and file hashes do not prove which corrections preceded conversion into IMR.

Complete file inventory · Airborne sample inspection · Long-record range inspection.

Admission to future analyses

Each dataset will have its own published eligibility policy, fixed before examining model-fit residuals. Source integrity, timing continuity, independent receiver trajectory, usable duration, traveled distance and aircraft motion will determine candidate intervals. Instrument performance assumptions and processing hypotheses must also be qualified separately. The ESA selection proposal specifies the intended longer, slower cohort and required software controls. It has not yet selected intervals or admitted them to model fitting.

Longer, slower survey lines can form a separate shape-analysis cohort. Short, slow UAV tracks and stationary recordings may instead support noise, drift, timing or rotation-field development. Raw availability alone admits no recording to a shape conclusion. Existing intervals, results and thresholds stay unchanged.

Downloaded originals, reference documents, metadata pages and physical-unit CSV are retained locally under data/external-imu-survey-20261018/. The build_inventory.py command regenerates inspection reports from those explicit inputs without network access, numerical integration or model fitting; hashes are recorded in inventory-receipt.json.

The ESA additions have separate acquisition and inspection receipts under esa-sample/. inventory_esa.py rebuilds the live catalog from saved directory listings; inspect_esa_gnss.py describes the GNSS export and reported lines. inspect_esa.py preserves its completed full-file inspection and refuses an accidental repeat. It performs no model fitting. The initial bounded byte-range request timed out on transfer closure and remains an incomplete diagnostic fragment; all reported full-file checks use the independently completed original.

Next analyses · Study index.