trifecta + balance + run_provincial: filter dispatch list to bundle species (skip 15 known-empty WSGs)

[NewGraphEnvironment]

Problem

Every provincial run dispatches 15 WSGs that have zero target species for the bcfishpass bundle (BT/CH/CM/CO/PK/SK/ST/WCT). Each errors ~30-80s in with No species resolved for AOI '<wsg>'. Net waste per provincial run: ~12 min compute, 15 stub-error RDS files.

The 15 affected WSGs (carry only ct/dv/gr/rb per inst/extdata/configs/bcfishpass/overrides/wsg_species_presence.csv):

WSG	Flagged species	Note
ATLL	dv, gr, rb
BRID	ct, rb	Falls 20807
CHUK	dv	No target species observations
GRNL	rb	No target species observations
KAKC	rb	Not accessible to target species
KUSR	dv, gr	No target species observations
LEUT	rb	Kenny Dam
LFRT	gr, rb	Mackenzie basin
LKEC	dv, gr	Mackenzie basin
LNRS	ct, rb
MURT	rb
MUSK	dv, rb
PITR	dv, rb	No target species observations
UISR	dv, rb	Forrest Kerr Generating Station
UPET	gr, rb	Mackenzie basin

Root cause

trifecta_provincial.sh (line 50), balance_provincial_buckets.R (line 123), and run_provincial_parity.R (line 49) all use the same broad presence inclusion list:

spp_cols <- c("ch","cm","co","pk","sk","st","bt","wct","ct","dv","rb")
has_spp <- apply(wsg_pres[, spp_cols, drop=FALSE], 1,
                 function(r) any(r %in% c("t","TRUE",TRUE)))

The check looks for ANY of 11 species presence flags. The bcfishpass bundle only classifies the first 8 — so a WSG flagged only for ct/dv/gr/rb passes the dispatch filter but errors during lnk_pipeline_species() (which intersects with cfg$species).

Fix

Narrow the presence check to the bundle's cfg$species (8 bcfp species). Caller-passed species override still works for default-bundle runs that want broader coverage.

spp_cols <- tolower(cfg$species)  # 8 species for bcfishpass bundle
has_spp <- apply(wsg_pres[, spp_cols, drop=FALSE], 1,
                 function(r) any(r %in% c("t","TRUE",TRUE)))

Predicted impact: provincial run drops from 232 → 217 dispatched WSGs; ~12 min wall-time saved; zero stub-error RDS files.

Prior context

• Surfaced in research/provincial_parity_2026_05_01.md Errors section; "filed as a follow-up" never actually filed (this issue).
• Reproduced unchanged in 2026-05-11 provincial run (same 15 WSGs, link 0.35.0).

Acceptance

• trifecta_provincial.sh + balance_provincial_buckets.R + run_provincial_parity.R filter against bundle species.
• Dry-run prints 217 WSGs (was 232) for the bcfishpass bundle.
• research/provincial_parity_v0.35.0.md documents the 15-WSG enumeration as known-empty.

[NewGraphEnvironment]

Fixed in 8f8c7b6 — #157: dispatch filter uses bundle species only. run_provincial_parity.R + data-raw/balance_provincial_buckets.R now filter the dispatch list to WSGs that have at least one bundle species present (via apply(wsg_pres[, cfg$species], 1, any %in% c('t','TRUE',TRUE))). 15 known-empty WSGs (ATLL/BRID/CHUK/GRNL/KAKC/KUSR/LEUT/LFRT/LKEC/LNRS/MURT/MUSK/PITR/UISR/UPET) skipped pre-dispatch instead of erroring 30-80s in. ~12 min saved per provincial run.

Released as part of v0.36.0 (PR #165, merge commit 27f5f7b).