Radar DA case study — Macaé INEA, 16 December 2024 Nocturnal convective line · DA 14–17 Z · forecast 17 Z → 14 Z+1 · draft for WRF User Meeting 2026

CONVECT / Project Sharktopus · run 2026-06-11 · cross-case synthesis

The event. A strong nocturnal convective line on 16 December 2024, most active 18–20 Z near Arraial do Cabo (peak reflectivity 64 dBZ). The DA window ran 14–17 Z — 4 hourly 3D-Var cycles assimilating INEA Macaé radial velocity and reflectivity. The free forecast ran from 17 Z for 21 h (to 14 Z+1) on d01 → d02 → d03 via ndown.

Both branches share the same GFS lineage (fair same-lineage baseline). The picker placed d03 almost directly over the radar (centre −22.40°, −41.75°), climo <z> +0.64σ, with 16 gauge stations in the box.

d03 POD heatmap: control = 0 hits; radar DA scores one hit (POD 0→0.077 at ≥1 mm/h).

case date

2024-12-16

nocturnal event, 18–20 Z; Arraial do Cabo, 64 dBZ

climo <z>

+0.64σ

favourable — DJF climatology

DA cycles

14, 15, 16, 17 Z — hourly 3D-Var, rv + rf

headline — d03

POD 0→0.077

control missed the event entirely; radar DA hits 1 of 13 gauge-hours (≥1 mm/h)

The model was blind, and DA barely opened one eye. The control forecast produced zero hits at every precipitation threshold on d03 (POD=0, CSI=0 at ≥1/2.5/5 mm/h). Radar DA: POD ≥1 mm/h 0.00→0.077, that is 1 hit out of 13 observed gauge-hours; ≥2.5 mm/h 0.00→0.100 (1 of 10); ≥5 mm/h stays at 0.00. BIAS rises to ≈ 1.9 at ≥1 mm/h (over-forecast area). Spatial correlation of the 21-h accumulation degrades on d03 (0.17→−0.06) and MAE rises (8.1→22.0 mm). With a sample this small the case is indicative, not statistical.

Honest accounting — d02. At d02 the story is negative: POD ≥2.5 mm/h drops (0.101→0.013) and CSI with it (0.030→0.006), though the frequency bias moves towards 1 (2.52→0.98). MODE centroid is flat at d02 (109.3→113.3 km, 6 matched hours for the control against 5 for the radar run). Across both domains this case does not support a claim of benefit.

Navigate the story

01 · placement

Where did the nest go?

Interactive picker — composite anomaly, per-variable convective index frames, placed d02/d03 boxes. Climo <z> +0.64σ.

Open the picker

02 · innovation

Did the DA move the state?

4 × 3D-Var cycles, cost function, OmB→OmA fit, and analysis increments of U, V, T, Q_vapor around the radar.

Open the innovation report

03 · verification

Did the free forecast benefit?

Control vs radar-cycled WRF at d01/d02/d03 — station POD/CSI/BIAS, 21-h accum, MODE, HiRA, and CREF viewer (obs × control × radar).

Open verification