Lume Nowcast (live feed)

Arm scoreboard — the prequential decision log

Every lab grab is scored exactly once, against the predictions each arm published at the grab’s sample hour, before the result was known, and written to an immutable log (first write wins). This table reads only that log — it never pairs grabs against the stored series. Cost prices a missed exceedance at 3× a false alarm; the lowest-cost arm on a site’s own record is the one the gate serves. The field arm is the customer dashboard’s own regression, joined over HTTP; it enters the record from its first live hour and is scored on the same footing.

Loading the decision log…

Concentration nowcast

Central estimate with the calibrated 50% and 90% bands (coverage verified per region on walk-forward replay; see the promotion record for the site's own numbers). Red strip along the top = advisory engaged. Blue bars hanging from the top = local hourly rain (right axis). Dots/triangles = lab grabs (green under the limit, red = exceedance). The bold green line is what the site served; the thin dotted lines are every arm's own estimate (calibration, stack, regression, field), and the strip along the bottom shows which arm the gate served hour by hour, in the same colors.

Regulatory quantity — 30-day geometric mean

10^(mean of the served log₁₀ estimate) over the trailing 30 days (green; published only with ≥50% window coverage), the 60-day window (grey dotted), and the geometric mean of the lab grabs known in the same 30-day window where ≥4 exist (blue steps). Criteria are written on this windowed quantity, not on single samples; the dotted red line is the site's action limit.

About this feed

Every number is produced by a model that, at prediction time, used only information available at that moment. Arms compete per site under a gate: the one with the lowest decision cost (a missed exceedance priced at 3× a false alarm) on the site's own scored record serves, once the site has enough decisions; otherwise the region's record decides, and below that the site serves calibration. A new site holds its alarm until its first lab result. When a sensor is dark, service continues on site calibration with honestly widened uncertainty, and a standing alarm is held rather than cleared. Every grab's scoring is written once to an immutable decisions log (first write wins) with each arm's prediction as published at the sample hour; the scoreboard above reads only that log. Model provenance rides each row as model_version, the serving arm as served_arm. This page reads the nowcast_* tables and nothing else; the raw sensor store is upstream of a read-only boundary and cannot be affected by anything on this page.