How the two experiments work

An IMU (inertial measurement unit) measures turning and acceleration. Local Level Lab uses those measurements in two ways: professional airborne analysis, which is primary, and a simpler stationary crowdsourced experiment under development.

Primary: NASA airborne analysis

An aircraft following local level over a globe gradually changes its orientation as it moves across the curved surface. The candidate called “rotating globe” adds a standard background angular-rate term; “still globe” retains curvature without that term. These are model names, not determinations of the field's physical cause. The specified still flat disc uses a different geometry. All three are fitted against the same raw IMU increments and receiver positions, with the same allowed instrument and motion errors.

The archive selector requires at least four continuous minutes with every supported ground-speed observation at least 700 km/h, plus frozen completeness, timing and motion rules. It saves file hashes, exact intervals, durations, speeds and exclusions before model fitting. The public inventory is reused rather than selecting attractive gyro readings.

Fit each complete selected stretch jointly. Fixed shorter sections check residuals at the same fitted parameters; they are not separate flights or fresh calibration fits. Gaps and configuration changes split intervals. Joining files needs demonstrated timing, sample and installation continuity. Whole-stretch fits require credible error assumptions for their entire duration; residual sections can expose violations but cannot prove stationarity.

Interpret globe versus the specified flat disc first, using the better of the two globe fits. Report the background-field comparison only after all required fits converge and both processing cases favor globe. The two primary cases allow constant gyro offsets of ±1 degree/hour and either no Earth-rate removal or an unknown shared amount removed upstream. Those limits are analysis assumptions, not verified per-unit drift specifications.

Keep three questions separate: which surface geometry fits, whether a retained background angular-rate field is needed, and what physically explains that field. Gyros measure angular motion relative to an inertial reference. A successful field fit does not itself explain that reference or establish heliocentrism. The zero-field globe does not represent every theory of a stationary Earth: a theory proposing a nonzero inertial field needs its own quantitative predictions. Explanations predicting identical measurements cannot be distinguished with those measurements. The interpretation policy applies to current model comparisons and the proposed freely estimated field alike.

The fit observations come from two separately recorded streams: Group 4 IMU increments and positions reconstructed from Group 10001 primary-receiver output. Applanix's V6 interface specification, section 3.3.2.1 describes Group 10001 as data emitted by the receiver. Group 1 is the combined IMU/GPS navigation answer; it supplies decoder checks and motion screening, never fitted gyro or position observations. This establishes a separation of recorded sources, not statistical independence of every error or proof of every historical receiver's processing.

Receiver coordinates and height conventions still carry geodetic assumptions. The globe candidate uses WGS84 geometry; the disc uses its declared alternative metric. Separate streams therefore do not make the comparison independent of all models. The planned channel and coordinate controls will test whether the separation requires the observed gyro increments. The provenance record states the current evidence and limits. Exact device identification remains useful parallel work rather than a prerequisite for modeling with survey-grade data.

The contribution is a reproducible test of moving-platform measurements under competing geometries. A first-of-its-kind claim has not been established. Earth's shape is separate from geocentric or heliocentric explanations; a globe result alone decides neither.

The original first study found twelve complete globe preferences. Completed numerical recovery adds twenty-nine, giving 41 globe preferences, none for flat and 12 unresolved stretches. Each recovered comparison passed all six primary fits, its scheduled repeat and fresh controls. The same intervals, measurements and physical assumptions were retained. Completed comparisons now include both recorded IMU8 and IMU21 configurations; their conversion scales differ, but the selection and physical comparison rules are the same. Unfinished calculations do not give the models equal support. The consistent completed comparisons are a substantial empirical finding. Establishing a statistical significance level additionally requires validated uncertainty and accounting for shared flights or instruments; the stretches are not automatically independent replications. Those checks quantify the finding's strength without discarding the observed model preferences. A better relative fit must still be checked for absolute adequacy, numerical stability and statistical uncertainty. See the archive guide, current findings and analysis roadmap. The method development record organizes the completed timing, calibration, uncertainty and solver checks and explains what each establishes.

Secondary: stationary crowdsourced measurements

Place an IMU on a solid, level surface. Point a marked horizontal axis approximately toward true north, then reverse it by 180° with the same face up between still blocks. A geographic horizontal rate changes sign in sensor coordinates; a constant sensor offset does not. Repeating A–B–B–A helps separate a linear offset change.

The analyzer estimates the signed reversal signal first, without latitude or a predicted Earth rate. A later population study would compare sealed estimates with cos(latitude), fitting the amplitude freely and accounting for orientation uncertainty, instrument, day and site dependence. Horizontal northward rotation has the same sign in both hemispheres; this is not a primary vertical-rate test.

This measures a rotation field rather than deciding Earth's shape by itself or explaining the field's ultimate cause. It does not require an aircraft, flight identity, continuous GPS, turbulence corrections or in-flight reversals.

Consumer slow-rate sensitivity and calibrated uncertainty still need physical controls. Reversal removes constant offsets, not arbitrary changing drift. Longer averaging cannot restore a signal erased by firmware or guarantee resolution beneath a fixed digital step. The implemented output is a preliminary diagnostic and always abstains from a scientific decision. See the stationary method, protocol and bench checks.

Separate evidence, shared principles

Preserve originals and failures; select by recording quality rather than model agreement; measure uncertainty; replicate across instruments; publish assumptions and exclusions. NASA observations do not certify consumer sensors. Ground recordings do not inherit airborne thresholds or imitate permanent aircraft installations.

Earlier passenger-flight methods remain in the technical methodology history and aircraft validation history. Frozen studies remain unchanged. The rotation-field roadmap describes the separate future airborne field measurement and its interpretation.