Observations that resolve no CalCOFI grid cell — reported, never dropped

Until v2026.08.11 every ingest's core projection filtered WHERE grid_key IS NOT NULL, so an observation whose event did not land on a station grid cell never reached obs at all — while the sample arm kept the event. That asymmetry is what let four calcofi_mets cruises reach a release as 11,762 underway samples with zero observations, their 1.7M measurements reachable only through the supplemental table.

Usage

check_ungridded_obs(con, obs_tbl = "obs", verbose = TRUE)

Arguments

con

a DBI connection holding obs

obs_tbl

name of the observation table (default "obs")

verbose

logical; message the headline

Value

a data.frame, one row per dataset_key: n_obs, n_ungridded, pct_ungridded, n_no_position (ungridded AND no lat/lon at all), and finding, a sentence ready to paste into a questions.csv context cell

Details

Excluding them was also inconsistent with the pipeline's own reasoning: obs_mets_full had already been deliberately gated on a position rather than on grid_key, because "a ship on transit is legitimately outside the CalCOFI station grid" — and calcofi_phytoplankton is region-pooled and has emitted ungridded obs from the start. The headline table now agrees with both: no grid cell is not a reason to delete an observation.

It IS a reason to ask. An ungridded observation is one of three things and the pipeline cannot tell them apart: a genuinely off-grid position (transit, an historical station outside the modern pattern), a coarser spatial notion (a region-pooled sample with no point at all), or a coordinate error — the sign-flipped Longitude_W that put five CalCOFI cruises in the Taiwan Strait was invisible precisely because being off-grid silently removed the rows. So this reports every dataset's share and is meant to drive a questions.csv entry per dataset, not to be quietly tolerated.