slv.measurements.mobile.receptors#

STILT receptors for the TRAX record: one multipoint receptor per 2-km segment crossing.

The paper-2 inversion samples TRAX CH4 at the staged 2-km network points (load_trax_points()). Each staged point owns a segment: the 50-m network points nearest to it. Both point sets are generated the same way, lair.geo.points_along_line on the UTA line network (load_trax_points()), so shared track is one row of points (load_trax_network_points()); a crossing on one line releases only from that line’s points of the segment (release_points()). Every time the train crosses a segment (find_segment_crossings()) it gets one PYSTILT MultiPointReceptor that releases particles from all of the segment’s 50-m points at the median crossing time (build_trax_receptors()). PYSTILT spreads numpar evenly over the release points and writes their mean footprint, so the 50-m footprints come out already aggregated to the 2-km point. The CH4 value that pairs with a receptor is the mean over the same 50-m points during the crossing (i.e. the transect-matrix row averaged over the segment).

Segment membership is fixed by geometry alone: it does not depend on the inlet lag or on which periods pass calibration QC, so receptors can be built (and run) before those decisions are final; QC later selects which crossings enter the inversion. Crossings are found from the GPS positions of the on-track rows of obs.parquet, so periods with no LGR data get no receptors.

Dwells. Half the record is not a traverse at all: the train sits outdoors at a line terminus between runs or parked on a yard track, which is 39 % of the CH4 rows at the yards plus 31 % of the crossing time at the termini. Sitting still at a known outdoor position is a stationary measurement, so find_dwells() picks those periods out (the train stays within radius for at least min_duration) and build_dwell_receptors() turns each hour of one into a single PointReceptor at the dwell’s median position. Indoor shed air is excluded upstream by loading fixes with location="outdoor".

Both builders release when the air was sampled – a crossing at its median fix time, a dwell hour at the median fix time inside that hour – rather than on the hour, and both accept an hours window (local time) so a project can start with the afternoon and add the rest later without rebuilding what it already has.

Attributes

`LINE_LETTERS`	Letters of the TRAX lines as `load_trax_points()` tags them (bit i of a line mask).
`RECEPTOR_COLUMNS`	Column order of a PYSTILT receptor CSV (`r_idx` groups rows into one receptor).

Functions

`build_dwell_receptors`	PYSTILT receptor table (`RECEPTOR_COLUMNS`) for the dwell periods.
`build_trax_receptors`	PYSTILT receptor table (`RECEPTOR_COLUMNS`) from a crossings table.
`filter_receptor_hours`	Keep rows whose `time` falls in those local-time hours (`None` keeps all).
`find_dwells`	Periods where the train stayed put: parked on a yard track or waiting at a terminus.
`find_segment_crossings`	Split the GPS record into crossings of the 2-km segments.
`label_dwell_site`	Where each dwell is: `yard_name` (JRRSC / MRSC, from the packaged storage polygons) and `segment` (nearest 2-km segment, NaN beyond `max_dist`).
`load_trax_fixes`	`Time_UTC`, `Latitude_deg`, `Longitude_deg` of the `location` rows of the cached TRAX record (`obs.parquet`), one row per distinct second, time-sorted.
`load_trax_network_points`	The 50-m network points with their 2-km segment membership.
`release_points`	Index of the segment's points on every line in `lines` (all of them if `lines` is empty).