03 · Verification Free forecast vs INEA Macaé radar — control vs radar-cycled WRF

Three trajectories on the same grids and times. After placement (§01) and the per-window 3D-Var cycling (§02), the pipeline releases a free forecast out to 21 h driven by ndown from d01 → d02 → d03 → d04. The reflectivity truth is the INEA Macaé composite — the same radar whose volumes were assimilated (not an independent truth); rain gauges (INMET / Alerta Rio / CEMADEN) provide the independent precipitation check.

Delta-CSI (radar minus control) by precipitation threshold, window and distance
ΔCSI (radar − control) by hourly precipitation threshold × time window × distance from the radar. Green = DA improved the score.
Radar blind spots. The truth field here is a single radar, and ~19% of d03 is blocked to it (W/NW behind the Serra do Mar, plus man-made dead sectors). See the beam-blockage study — found two ways, the radar's own persistently dead sectors and a terrain beam-blockage simulation — for where rain can fall unseen and why those gates are masked out of the assimilation and the verification truth field.

Open the interactive verification report

Per-domain plots (d01, d02, d03, d04) · cycled vs control vs radar truth · CSI / POD / FAR / bias heatmaps and time series

Open report

Outputs inventory — where the truth lives on disk

1) Cycled DA (WRFDA 3D-Var analyses)

Root: containers/_instances/assimilator/fcst_cycled_2024012100/

Per cycle: wrfvar_output (analysis), fg (background), wrfbdy_d01, statistics, cost_fn, radar_omb_oma_01.

2) Final forecast — d01 21 h

Same root:

3) ndown chain — d02 / d03 / d04

Root: containers/_instances/assimilator/fcst_ndown_2024012100/

Each dX/ contains wrfout_d01_2024-01-21_09:00:00 .. 2024-01-22_06:00:00 (afternoon) or the 03 Z → 00 Z + 1 analogue (morning). Note: after ndown, the child was renamed to d01 in the namelist.

All forecasts complete. Audit note (2026-06): the afternoon published control was a GFS 21/00Z cold start, which mixes spinup/lineage with the DA signal (audit finding AU3-5); a fair-lineage control_v2 was re-run from the GFS 20/12Z pre-forecast — afternoon conclusions on this page are restricted to the morning pair, which was always fair. The afternoon d04 control run finished on 2026-04-13 (final file control_afternoon/d04/wrfout_d01_2024-01-22_06:00:00 on disk). The full d01 / d02 / d03 / d04 control vs radar comparison is unblocked.

Further reading