Glossary

Each entry gives the meaning used in these pages. Where a term is defined in code, the file is cited.

Authority. Scientific authority for the method is the Python implementation. This is recorded in src/lunartrace/standing.py as scientific_authority = "python".

BLOCKED. A standing level in the projection model’s Standing literal, listed with the levels METHOD_EXECUTED, NUMERIC_REPRODUCED and SCIENCE_VERTICAL_QUALIFIED (src/lunartrace/projection.py). It is a proof ceiling that means the result is blocked. No Site01 result in these pages carries it.

Cell centre. The point at the middle of a grid cell. For column $c$ and row $r$ on the production grid it is $x = x_0 + 5(c + \tfrac12)$ and $y = y_0 - 5(r + \tfrac12)$ (see Method, equation 1). ROI membership is decided by whether the cell centre is inside or on the boundary of a polygon (see the ROI entry below). The raster is pixel-registered, so stored coordinates identify these centres.

CAS (content-addressed store). A store that names each file by the SHA-256 digest of its bytes. The layout here is cas/sha256/<2>/<64>, where the first two hex digits form a subdirectory (research/sources/pgda78-site01.manifest.json, custody.cas_layout). Two copies of the same bytes share one name. The store for this witness is local and outside Git. The STUDY plane re-verifies each blob’s digest before use (src/lunartrace/science/study.py, resolve_inputs).

EvidenceStanding. A model in src/lunartrace/standing.py that records which evidence is available in the bootstrap state of the package. Its fields are basis = "frozen-aggregate", observation_rows = "unavailable", temporal_pair_rows = "unavailable" and image_alignment = "unverified". It is a statement of availability and admits nothing.

Included fraction ( $f_\tau$ ). The number of included cells divided by the number of jointly valid cells in the same ROI at threshold $\tau$ . It is null when the joint count is zero and zero when joint cells exist but none is included (see Method, equation 9). It is a valid-cell fraction, not an area fraction.

Inverted-CDF quantile. The quantile definition used for included slopes. The value at probability $p$ is the smallest observed sample whose empirical CDF is at least $p$ . This is Hyndman and Fan type 1, and the code calls numpy.quantile(..., method="inverted_cdf"). Each quantile is therefore one of the observed values (see Method, equation 10).

Joint validity. A cell is jointly valid when its slope value is valid and its slope RMS value is valid. Validity means finite and within $[0, 90]$ degrees for slope, and finite and strictly greater than zero for slope RMS (see Method, equation 4).

Manifest. A JSON file that records a source witness. The schema is lunartrace.source_witness/v0. It holds the source identifiers, the status (SOURCE_FROZEN), the citations, the native semantics (grid, CRS and nodata), the rights status, the custody information and the entries. Each entry has a role, a file name, a size, a SHA-256 digest and its location in the CAS (research/sources/pgda78-site01.manifest.json).

METHOD_EXECUTED. A standing level and also the execution status recorded in a receipt. It means the method ran on the recorded inputs and produced the recorded result. It is an execution record and not a scientific standing. The receipt status is always METHOD_EXECUTED (src/lunartrace/science/study.py).

NOT_ADMITTED. An admission label used in the workbench for ROI-level outputs. The workbench pairs it with METHOD_EXECUTED (src/lunartrace/workbench/roi.py; src/lunartrace/workbench/lab.py). It means the output has not been admitted as scientific evidence.

NUMERIC_REPRODUCED. A standing level. It means a primary result agreed with an independent implementation under the stated comparison criteria (exact counts, histograms and quantiles, and a relative tolerance of $10^{-12}$ for means and fractions). Its docstring states that it “records reproduced numbers and is not a scientific admission” (src/lunartrace/projection.py). It is the standing of the frozen Site01 numeric result.

Oracle. The independent implementation in research/oracles/slope_quality_oracle.py. It uses NumPy, tifffile, pyshp and the standard library, and does not use GDAL, pyogrio or shapely. It checks the same frozen inputs and writes a canonical result. It is compared with the primary result by tools/compare_slope_quality.py.

Receipt. The execution record written by the STUDY plane as receipt.json. It records the schema lunartrace.execution_receipt/v0, the method and implementation identifiers, the status, the source witness and manifest digest, the input digests, the taus, the SHA-256 of the canonical result, timestamps, git state and environment versions. It is not deterministic, and it is unsigned (src/lunartrace/science/study.py, run_study).

ROI (region of interest). A polygon in metres, in the grid’s polar stereographic CRS. Each ROI is a single polygon with no holes. A cell belongs to an ROI if its centre is inside or on the boundary of the polygon (see Method, equation 6). The Site01 ROI archive has seven polygons, numbered 0 to 6 in shapefile order.

Slope. The terrain slope, in degrees, from the slope raster (role slope). Valid slope values are finite and lie in $[0, 90]$ degrees, with both endpoints included (see Method, equation 2).

Slope RMS. The root-mean-square uncertainty of the slope, in degrees, from the slope RMS raster (role slope_rms). It is the quantity compared with $\tau$ . Valid values are finite and strictly greater than zero (see Method, equation 3).

Standing. A label for the proof ceiling of a result. The levels are METHOD_EXECUTED, NUMERIC_REPRODUCED, SCIENCE_VERTICAL_QUALIFIED and BLOCKED. The handoff document orders the first three as METHOD_EXECUTED < NUMERIC_REPRODUCED < SCIENCE_VERTICAL_QUALIFIED (docs/r2/SCIENCE-TO-OPUS-UI-HANDOFF.md), and BLOCKED is a separate state. A client may not display a claim stronger than the standing (src/lunartrace/projection.py).

Tau ( $\tau$ ). The inclusion threshold, in degrees, on the slope RMS raster. A joint cell is included when its slope RMS value is at or below $\tau$ . The default values are $0.5$ , $2.0$ and $25.0$ degrees. Tau is a data-quality cutoff, not an uncertainty interval (see Limitations).

Valid-cell fraction. Another name for the included fraction. The name states that the denominator counts valid grid cells, so the value is not an area.