Add fl_dem_aoi() + Parsnip Watershed Group floodplain showcase vignette#37
Merged
NewGraphEnvironment merged 8 commits intoMay 7, 2026
Merged
Conversation
Approved phases scaffolded into planning/active/: - Phase 1: implement fl_dem_aoi() - Phase 2: PARS floodplain showcase vignette + data-raw cache - Phase 3: polish + release prep MRDEM-30 via /vsicurl/ as the primary path per rtj/docs/dem-sources.md; LidarBC via STAC stays as example pattern, not a wrapper function. Relates to #34 Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Single entry point that takes an sf AOI and returns a SpatRaster cropped to a buffered extent. Source defaults to MRDEM-30 DTM via /vsicurl/ (the settled choice in rtj/docs/dem-sources.md for watershed-scale BC work). Local files, /vsicurl/ HTTP COGs, and /vsis3/ S3 COGs all work identically — terra::rast handles dispatch. Implementation lifts the access pattern from rtj/docs/dem-sources.md:50-65 verbatim with parameter wiring: buffer the AOI in its own CRS, transform to the raster's CRS for the lazy crop, reproject after crop (never before — the 84 GB MRDEM COG must not be reprojected as a whole). snap = 'out' on the crop ensures the result fully covers the buffered AOI. Buffer semantics: applies to whatever sf geometry is passed as aoi. Polygon AOI → region-bounded crop. Streams sf → tight, memory-efficient crop along the stream corridor (the wedzin production optimization, now first-class). Tests: 8 passing including a live MRDEM /vsicurl/ range read gated by skip_on_cran + skip_if_offline. lintr clean. Full suite 172 PASS. Phase 1 of #34 complete. Relates to #34 Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Generic data-raw script — any WSG via 'wsg <- "PARS"' at the top, output filenames namespaced (pars_aoi.gpkg, pars_streams.gpkg, etc.). Switching the WSG re-runs the whole pipeline. Pulls boundary from whse_basemapping.fwa_watershed_groups_poly and streams from bcfishpass.streams_vw via fresh::frs_stream_fetch — streams_vw carries channel_width / upstream_area_ha / map_upstream that fl_valley_confine()'s flood model needs (the bare FWA stream networks table doesn't). Runs fl_dem_aoi() against the MRDEM-30 default (streams-as-aoi for memory- efficient crop) → fl_valley_confine() → fl_valley_poly(). Two XYZM bugs surfaced during the run. fwapg streams carry route measures in M, and GEOS can't buffer XYZM: 1. fl_dem_aoi() fixed at source — wraps sf::st_zm() before the sf::st_buffer() call. Tests still PASS 18/18. 2. fl_valley_confine() also calls sf::st_buffer() internally for the channel_buffer feature; same crash. Worked around in the data-raw script with an explicit sf::st_zm(streams). Hardening fl_valley_confine the same way fl_dem_aoi was hardened is filed as a follow-up; out of scope for #34. PARS cache totals 11 MB across 5 files (under the 15 MB target): pars_aoi.gpkg 0.31 MB PARS WSG boundary (5597 km²) pars_streams.gpkg 2.7 MB 5,581 order-4+ segments from streams_vw pars_dem.tif 6.2 MB INT2S MRDEM-30 clip, 4089x4862 cells pars_valleys.tif 88 KB fl_valley_confine() binary output pars_floodplain.gpkg 1.2 MB polygonized valleys DEM stored as INT2S (Int16) — VCA at 30 m doesn't need sub-metre vertical precision, and INT2S+DEFLATE+PREDICTOR=2 is ~5x smaller than FLT4S. Phase 2a of #34 complete. Relates to #34 Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Demonstrates the fl_dem_aoi() + fl_valley_confine() workflow on the Parsnip River watershed group (PARS, 5,597 km², FWCP Peace). Loads cached outputs from inst/vignette-data/ — heavy compute lives in data-raw/wsg_vignette_data.R, the vignette renders fast and offline in CI. Designed for portability to the Peace report appendix: the YAML output is the only thing tied to the package vignette format, swap it for the report's bookdown::pdf_book and the chunks work as-is. Walks through: - Why MRDEM-30 (links to rtj/docs/dem-sources.md rationale) - AOI + streams loaded from cache - The fl_dem_aoi(streams, buffer = 2000) call (eval=FALSE, shown as docs) - The fl_valley_confine() pipeline - Hillshade + floodplain hero map (terra::plot) - Summary table: 449 km² floodplain on order 4+ streams (8% of PARS) - One-paragraph 'adapt to another WSG' note pointing at the generic data-raw script Renders cleanly to 604 KB HTML. Full test suite 172 PASS. Phase 2b of #34 complete. Relates to #34 Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- Bump DESCRIPTION version 0.2.1 → 0.3.0 (new exported function = minor) - NEWS entry under '# flooded 0.3.0' covering fl_dem_aoi(), vignettes/pars-floodplain.Rmd, and data-raw/wsg_vignette_data.R - .Rbuildignore: add ^planning$ and ^\.claude$ (cleared the 'top-level files' R CMD check NOTE) - Delete stale vignettes/figure/ knitr leftover from previous vignette renders (cleared the 'vignettes' R CMD check NOTE) - .gitignore: add *.Rcheck/ and *.tar.gz so local check artefacts don't pollute the working tree R CMD check status now: 0 errors, 0 warnings, 0 notes (with _R_CHECK_FORCE_SUGGESTS_=false to skip whitebox, which is unrelated to this change). devtools::test() 172/172 PASS. lintr clean on new files. Phase 3 of #34 complete. Relates to #34 Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Iterated heavily on the PARS showcase vignette in response to feedback: Stream source switched to bcfishpass.streams_bt_vw — every row is already classified accessible (access IN (1, 2)) by construction, so filtering 'best accessible habitat order 3+' is a single SQL via fresh::frs_db_query. Replaces an attempted wedzin-style frs_network + frs_break_find pipeline that hit an upstream fresh bug (frs_break forwarding points_table args into frs_break_find which doesn't accept them). min_order lowered 4 → 3. Waterbodies (lakes + wetlands) now picked up via waterbody_key linkage on the order-3+ streams and passed to fl_valley_confine(waterbodies=...) so the binary valley raster fills lake/wetland cells. flood_factor=4 (ff04 — functional floodplain). Resulting floodplain: 481.9 km² (8.6% of PARS), 8,436 streams, 200 lakes + 464 wetlands. Context layers added to the maps: railways, roads (filtered to FSR / resource roads on the hero map; full network on the inset), First Nations reserves (with black diamond markers + halo'd english_name labels), parks / protected, and named streams (italic labels on the inset). gq registry colours used for parks / reserves. All layers fresh::frs_clip'd to the WSG and the DEM terra::mask'd to the WSG so nothing renders outside the basin. Vignette restructure: - Methodology section moved to the top with ecological context paragraph - fl_params() and fl_scenarios() rendered as knitr::kable tables with bookdown::html_vignette2 numbered captions - Citation keys converted to inline citations via xciter (helper keys_to_pandoc() pre-formats bare semicolon-separated keys) - bibliography: vignettes/references.bib regenerated by data-raw/bib_regenerate.R (rbbt → BBT, mirrors the cd repo pattern) - ## References auto-populated by pandoc Cache consolidation: 9 sf layers now live in a single multi-layer GeoPackage (pars.gpkg). Rasters stay as standalone GeoTIFFs (gpkg's raster spec is for tile pyramids, wrong format for analytical DEM / binary valleys raster). Direct GitHub raw download links for all three files surfaced in the vignette so non-R users can grab the bundle. DESCRIPTION: DBI, fresh, xciter added to Suggests; Remotes points at NewGraphEnvironment for fresh + xciter. Clears the 'unstated dependencies in vignettes' R CMD check NOTE. Final R CMD check: 0 errors, 0 warnings, 0 notes. Tests 172/172 pass. Phase 4 of #34 complete. Relates to #34 Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Set echo = FALSE on map-floodplain, map-detail, params-stamp, and scenarios-stamp. The maps and tables are the value of those chunks; the rendering plumbing (terra::plot + halo'd text() + kable + xciter key-formatting) was dominating the page without showcasing the package. The eval = FALSE snippets that demonstrate fl_dem_aoi() and fl_valley_confine() stay visible — those are what flooded actually exposes. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
NewGraphEnvironment
deleted the
34-add-dem-source-helpers-file-stac-for-aoi
branch
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
fl_dem_aoi()— AOI-driven DEM fetch helper. Defaults to MRDEM-30 via/vsicurl/(the rtj-doc-settled choice for watershed-scale BC work) but accepts any local path,/vsicurl/HTTP COG, or/vsis3/S3 URL. Buffered crop in source CRS, reprojection after crop. Replaces hand-rolled per-project DEM plumbing.vignettes/pars-floodplain.Rmd— watershed-scale showcase runningfl_dem_aoi()+fl_valley_confine()end-to-end on the Parsnip River WSG (5,597 km²; 481.9 km² floodplain, 8.6% of WSG, atflood_factor = 4). Hero + south-east detail maps with parks, FN reserves, waterbodies, roads, railways, named streams.data-raw/wsg_vignette_data.R— changewsg <- "PARS"to any 4-letter BC WSG code; usesbcfishpass.streams_bt_vwfor accessible bull-trout habitat. Cache: 14 MB across 3 files (multi-layerpars.gpkg+ 2 GeoTIFFs). Direct GitHub raw download links surfaced in the vignette.DBI,fresh,xciteradded to Suggests withRemotes:for the org-only deps. R CMD check 0/0/0; tests 172/172 PASS; lintr clean on new files.Related Issues
fl_valley_confineXYZM hardening)Test plan
devtools::test()— 172/172 PASS (includes a live MRDEM/vsicurl/range read gated byskip_on_cran()+skip_if_offline())lintr::lint_package()— 0 lints in new exported code (some indentation lints in the data-raw script are stylistic and Rbuildignored)R CMD check— Status: OK (0 errors, 0 warnings, 0 notes)pkgdown::build_site()— pars-floodplain vignette renders cleanly to HTMLNotes
The vignette is designed to drop into a Peace report appendix — chunks read against the cached gpkg + GeoTIFFs, only the YAML output flavour is tied to the package vignette format. Bibliography regenerates via
data-raw/bib_regenerate.R(rbbt → BBT, mirrors the cd repo pattern).🤖 Generated with Claude Code