lnk_score_severity: classify crossings by biological impact #11

lnk_score_severity() — classify crossings by biological impact

The core scoring function. Classify crossings into severity levels (high/moderate/low) based on actual crossing measurements rather than the binary BARRIER/PASSABLE provincial classification. Thresholds are configurable per project, species, and life stage.

Signature

lnk_score_severity(
  conn,
  crossings,
  thresholds = lnk_thresholds(),
  col_drop = "outlet_drop",
  col_slope = "culvert_slope",
  col_length = "culvert_length_m",
  col_width = "downstream_channel_width",
  col_severity = "severity",
  to = NULL,
  verbose = TRUE
)

Parameters

• conn — DBI connection
• crossings — schema-qualified crossings table (after overrides applied)
• thresholds — output of lnk_thresholds(). Named list with high, moderate, low severity specs.
• col_drop, col_slope, col_length, col_width — column name parameters. Defaults match PSCIS field names. Users in other systems swap these without touching scoring logic.
• col_severity — name of the output column written to the crossings table
• to — if NULL, updates crossings in-place. If specified, writes scored copy to new table.
• verbose — report severity distribution

Returns

Updates or creates a table with the col_severity column populated. Returns the table name (invisibly) for piping.

Design intent

• Beyond binary: provincial barrier_result treats very different crossings identically. A 1.2m outlet drop and a 0.3m drop with steep slope are both "BARRIER" but have very different biological impact.
• Measurement-based: uses actual culvert dimensions, not just the assessment checkbox. This is the technological uncertainty from SRED — translating measurements into biological impact scores.
• Column-agnostic: col_drop = "outlet_drop" is the PSCIS default. A New Zealand user might pass col_drop = "perch_height". The scoring logic is identical.
• Threshold-driven: all cutoffs come from lnk_thresholds() — nothing is hardcoded in the function body.

Scoring logic (default thresholds)

Examples must show

1. Score crossings with default thresholds — show the severity distribution
2. Custom thresholds for a project focused on bull trout (higher drop tolerance)
3. Column remapping for non-PSCIS data: col_drop = "perch_height", col_slope = "pipe_gradient"
4. Full pipeline: lnk_override_apply() → lnk_score_severity() → lnk_break_source() — the three-step workflow that produces fresh input
5. Verbose output showing "High: 234, Moderate: 891, Low: 2,103" — users need to see the distribution to validate thresholds
6. WHY this matters: show two crossings with same barrier_result = "BARRIER" but different severity scores

File layout

• R/lnk_score_severity.R
• tests/testthat/test-lnk_score_severity.R

Relates to #1
Relates to NewGraphEnvironment/sred-2025-2026#24