version: "1.0.0" name: drought-analysis description: "Streamflow drought analysis: day-of-year percentile index at near-natural reference gauges per the drought recipe; regulated gauges refused for drought claims."

drought-analysis

Drought characterization from streamflow, done at gauges that measure hydrology.

The regulated-gauge rule (hard refusal, non-negotiable)

A drought index requested at a REGULATED gauge is REFUSED for drought claims: flow statistics there measure operations, not hydrology. The refusal cites the gotcha (nwis-regulated-gauge) and offers a near-natural reference gauge. This rule fires before any computation, whatever the framing of the request; the gauge's regulation status and the measured evidence behind the refusal are read from the gotcha, not carried here.

Behavior, in order

1. Parse and show back: basin/gauge, target period, climatology

window (default the recipe's fixed climatology window, stated).

1. Consult the bundle for THIS analysis first. Consult installed

knowledge concepts first, as the core consult-knowledge skill sets out, by the gauge, the parameter, and the drought method; the nwis skill lists the concepts this plugin resolves to, and this analysis reads the regulated-gauge and provisional-data gotchas, the NWIS streamflow dataset, and the drought-index recipe (knowledge/recipes/drought-index.md) among them. Restate what each changes about the plan and cite it by path. A concept added or corrected since you last ran is found this way; the regulation evidence, the provisional-window rule, the fixed climatology window, and the drought class thresholds all live in concepts and are read from them, never from this skill.

1. Load through load-nwis (its gate and qualifier discipline).
2. Compute per the recipe: its day-of-year percentile method against

the recipe's climatology, with the qualifier discipline the provisional-data gotcha requires; report the summaries the recipe defines (the median percentile and the below-threshold fractions), naming the recipe's class thresholds as read from it.

1. Validate where the recipe anchors apply (recipe sites/years

reproduce within expected_uncertainty; anything else is a finding).

1. Report with uncertainty framing: the recipe's method-sensitivity

band, the climatology window used, qualifier accounting, and the concepts consulted.

Must NOT

• Never present a drought index from a regulated gauge as hydrology;

refuse and offer a reference gauge. (Hard refusal: invariant, refusal-shaped, universal.)

• Never hardcode expected values; the recipe is the authority.
• Never compare indices computed against different climatology windows

without saying so. (Method discipline: percentiles read against different baselines are not comparable, regardless of dataset.)

Dataset-specific rules (which gauges are regulated and the Lees Ferry evidence, the provisional-versus-approved split, the fixed climatology window, the USDM-style class thresholds, the recipe's expected values and method-sensitivity band) are NOT restated here: they live in the nwis-regulated-gauge and nwis-provisional-data gotchas, the nwis-streamflow dataset, and the drought-index recipe, and are consulted per step 2. That is what lets a corrected threshold, a new gotcha, or a re-measured anchor change this skill's behavior without editing it.