Diagnostics Interpretation Playbook

This page defines operational checks for the three key diagnostic plot families.

Strata Diagnostics (strata-*.png)

Healthy signals:

  • Top strata capture expected cumulative area coverage.

  • Relative abundance bars are non-empty and interpretable.

  • SI violin/scatter distributions show plausible ecological structure.

Red flags:

  • Near-empty abundance in retained strata.

  • Extreme SI spikes isolated to tiny samples.

  • Missing or collapsed strata labels inconsistent with run settings.

VDYP Fit Diagnostics (vdyp_fitdiag_*.png)

Healthy signals:

  • Smoothed curve tracks binned median trend.

  • For AU-level first-growth synthesis, the default accepted direct-fit family is smoothed_bin_pchip rather than the older NLLS-first approach.

  • SI levels show coherent ordering when signal exists.

  • Tail handling avoids unrealistic late-age collapse/oscillation.

Red flags:

  • Outlier-driven overfit (for example impossible young-age volumes).

  • Inverted SI dominance where ecological signal should be monotonic.

  • Apparent fit to sparse/noisy bins without fallback/merge behavior.

  • Unexplained reversion to an older NLLS-oriented first-growth fit when the canonical AU-level default should be smoothed_bin_pchip.

TIPSY vs VDYP Overlays (tipsy_vdyp_*.png)

Healthy signals:

  • Managed and untreated curves are scenario-coherent AU-by-AU.

  • Relative magnitudes and timing match intended silviculture assumptions.

Red flags:

  • Managed trajectories nonsensically underperforming/overshooting without explicit scenario rationale.

  • Strong discontinuities that point to handoff parse or unit mismatch issues.

Escalation Path

  1. Validate input data and stratification assumptions.

  2. Validate VDYP fit settings and bin collapse behavior.

  3. Validate TIPSY DAT fixed-width alignment and species/FIZ compatibility.

  4. Re-run targeted units with deterministic run IDs and preserved diagnostics.