Snow-departure vignette for the Kootenay Lake region (v0.2.1)

[NewGraphEnvironment]

Summary

New vignettes/kootenay-lake.Rmd covering the southern Kootenays — KOTL + LARL + DUNC + SLOC, ~24,200 km². The "keep + expand" resolution of #49: drop the legacy single-WSG KOTL vignette (already removed in v0.1.4), replace it with a regional vignette that mirrors the FWCP Peace pattern, and use the v0.2.0 snow vars to tell the snow story for a region with a sharp east-west precipitation gradient. Closes #56.

Headline scientific findings (regional, 2015-2025 vs 1951-1980)

The Kootenay Lake Region story is meaningfully different from the FWCP Peace just to the north:

Variable	Peace v0.2.0	Kootenay v0.2.1
Annual SWE	-10%	-23%
Annual snowfall	-6% (~stable)	-15%
DOY-50 freshet shift	~7 days earlier	12.6 days earlier
Annual precipitation	~stable, n.s.	-7%, p = 0.02
Cumulative warming	+2.0 °C	+1.8 °C

Two distinct mechanisms:

1. The Kootenays sit warmer at the relevant elevations than the Peace, so the snow-vs-rain threshold (Knowles et al. 2006: significant fraction declines occur where winter wet-day Tmin > -5 °C) is being crossed in the calendar. Total annual snowfall is dropping, not just snowmelt timing shifting. In Peace the snowpack story was "redistribution"; in the Kootenays it's "less snow and earlier melt."
2. Precipitation itself is declining (-7%, p = 0.02). Combined with rising VPD, this is a "warmer and drier" story rather than the Peace's "warmer with mostly-stable precip."

What's in this PR

• data-raw/example_aoi_kootenay_lake.R — builds the AOI from union of 4 WSGs via bcdata. Output inst/extdata/example_aoi_kootenay_lake.gpkg.
• data-raw/example_context_kootenay_lake.R — bundled context layers (12 towns including Trail/Rossland/Red Mountain via LARL, plus Nelson, Castlegar, Kaslo, Nakusp, Slocan, New Denver, Argenta, Crawford Bay, Kimberley, Cranbrook). Found and fixed a latent bug in the inherited Peace pattern: Village (1) feature_type filter was missing, so smaller communities (Kaslo, Nakusp, Slocan, New Denver) were being dropped. Output inst/extdata/context_kootenay_lake.gpkg.
• data-raw/kootenay_lake_vignette_data.R — regional + per-ecoregion + per-WSG precompute. The per-WSG list is new (peace-fwcp.Rmd doesn't have one); 4 facets is small enough to show directly. Output inst/vignette-data/kootenay_lake.rds (404 KB) + kootenay_lake_departure_tmean.tif (3 KB).
• vignettes/kootenay-lake.Rmd — full regional vignette mirroring peace-fwcp.Rmd structure with three Kootenay-specific additions: per-WSG facet plots inside Per-Ecoregion Variation, Snowpack interpretation reflecting the snow-vs-rain threshold finding, Watershed Groups Across Ecoregions section sized for 4 WSGs (vs Peace's 16). Bibliography reuses vignettes/references.bib from Snowpack-departure methodology lit review: rag-build + 11 papers + citation map #54 — Najafi et al. 2017 covers the Columbia basin so no new lit review needed.
• data-raw/qa_snow_validation_kootenay_lake.R — ASWS QA at 5 sites in the AOI. Notable findings:
  • 74 paired station-years; pooled correlation r = 0.90 (much better than Peace's 0.51 — ERA5-Land tracks year-to-year SWE variability tightly at high-elevation Selkirk/Purcell stations)
  • Mean bias -54% — uniformly negative across both usable sites (Moyie Mountain Purcells -40%, Redfish Creek Selkirks alpine -63%)
  • Bias is stable at Moyie (p = 0.15); marginal drift at Redfish (p = 0.07, ~12 mm/yr widening). Vignette methodology paragraph carries the caveat.
• README — links both vignettes side-by-side.

Test plan

• devtools::test() clean (166 PASS, 0 FAIL)
• Vignette renders in 6.8 s; 12 csl-entries in References section
• Both gpkg files build and validate; bbox 49°N-51°N / 118.5°W-115.5°W
• All 12 town names found (caught + fixed the Village (1) filter)
• Precompute produces all 15 vars × regional + 4 ecoregions + 4 WSGs
• ASWS QA scatter is sensible
• Post-merge: pkgdown CI green
• Post-merge: kootenay-lake article visible on pkgdown site

Out of scope

• Rockies / Elk Valley (BULL, ELKR) — different climatic regime, different reporting context (FWCP East Kootenay).
• Anadromous salmon framing — Lower Columbia is dam-fragmented; vignette uses resident-salmonid framing (bull trout, Gerrard rainbow, mountain whitefish, kokanee) and freshet-shapes-channel framing.
• Removing existing KOTL bundled polygons — they stay for the README quick-start (single-watershed small/fast example).
• New methodology / lit review — reuses v0.2.0 stack and Snowpack-departure methodology lit review: rag-build + 11 papers + citation map #54's 11-paper bibliography.
• Updating README quick-start to use the new 4-WSG AOI — would bloat the example. Quick-start stays on example_aoi_kotl.gpkg.

Notes

This vignette is the second of the planned three reporting climate-departure appendices in this package family (cf #47). Ports directly to a hypothetical fish_passage_kootenay_2025 reporting context, mirroring how peace-fwcp ports to fish_passage_peace_reporting_2025.

Fixes #56
Relates to NewGraphEnvironment/sred-2025-2026#23

🤖 Generated with Claude Code