Experimental prototype — not an official warning system

Sorrento — 18 September 2026

Model audit. Frozen timestamp. Not a narrative engineered to make vShark look successful. Sorrento is not the centre of the WA platform.

← Sorrento live pageCity of JoondalupCACHEDModel 0.1.0-dev frozen

What this is

Two views are kept apart. The left-hand column is only information that was observable from open weather/marine APIs or standing operational facts (patrol season). The right-hand column is what became public after the incident. Post-event sightings are not fed into the pre-event model.

What the system could have known then

  • season · operational

    Surf Life Saving daily patrols at this beach were not scheduled to begin until 26 September 2026 (ABC). Pre-season: no assumed lifeguard coverage at 10:00.

  • place · known

    Sorrento is a popular swimming beach ~20 km north of the Perth CBD, immediately south of Hillarys Boat Harbour, with a groyne near Sorrento Surf Club (public geographic fact).

  • 06:00–10:00 · unknown

    No live shark-detection feed is connected to vShark. No tagged-receiver, aerial-patrol, or SharkSmart activity layer is ingested. Shark observation coverage: NONE.

  • 06:00–10:00 · known

    No official beach closure or shark warning for Sorrento is recorded in this reconstruction before the incident reports.

  • 06:00–10:00 · known

    Open-Meteo marine SST ~19.1–19.2 °C; swell ~0.9–1.1 m from the west-south-west; little wind; no rain. Regional grid, not a Sorrento thermistor.

  • 06:00–10:00 · unknown

    Tides, turbidity, chlorophyll, currents, fronts, subsurface temperature, and in-situ water clarity: not available in this build (missing).

  • 06:00–10:00 · unknown

    Prey, whale, seal, fisheries and carcass observations: none in vShark. Do not infer absence of sharks from absence of these feeds.

  • week · known

    A surfer was attacked at Glenfield Beach, north of Geraldton, on Monday 14 September 2026 (Reuters / SBS). That is Mid West, ~400 km north — background WA context, not a Sorrento detection.

What we know after the event

  • ~10:00–10:15 · after

    Witnesses described a male swimmer attacked about 50 m off Sorrento Beach near the groyne by Sorrento Surf Club, in company of another swimmer. Descriptions of shark length in early reports vary and are unverified.

    WA Today; ABC witness accounts

  • 10:15 · after

    St John WA received reports of the attack at 10:15 local time. DPIRD SharkSmart states DPIRD received a report of a possible shark incident at approximately 10:15 am.

    ABC / RNZ; SharkSmart notice

  • ~10:30 · after

    WA Police said emergency services were called to a reported shark-related incident off Sorrento Beach at around 10:30 am. A male swimmer had been bitten and later died as a result of the attack.

    WA Police via Reuters / Guardian / SBS

  • after 10:30 · after

    Air and sea search (water police, police air wing, SLS helicopter, Fisheries vessel). Beach closed. Shark siren activated by the surf club (WA Today).

    ABC; WA Today; Guardian

  • after event · after

    City of Joondalup closed beaches in its jurisdiction. Reports describe the closed stretch as Marmion to Iluka / Burns Beach, until at least noon Saturday 19 September 2026 (~12 km).

    ABC; SBS; City of Joondalup via Reuters

  • timing unclear · after

    ABC: a member of the public reported seeing a 2.5-metre shark off the beach. Time relative to 10:15 is not established in the sources used here — treated as AFTER / uncertain, not as a pre-event detection.

    ABC via RNZ

  • ~12:00–13:11 · after

    A shark estimated at about 5 m was reported to SharkSmart at Boyinaboat Reef just after noon / 1.11 pm. Reuters: no confirmation it was the animal involved. WA Today: sighted ~1.6 km west of Hillarys Boat Harbour just after 1 pm.

    Reuters; WA Today

  • species · after

    Public reports refer to a white shark at Boyinaboat after noon. Species of the animal involved in the 10:15 incident is not treated as confirmed in this audit.

    Reuters; SharkSmart

Frozen environmental reconstruction (Open-Meteo)

Hourly forecast/marine grid near Sorrento, timezone Australia/Perth. 10:15 and 10:30 are interpolated. This is not a beach thermistor.

ClockSST °CSwell mPeriod sWind m/sAir °CRain
06:0019.10.9411.82.2614.60
08:0019.11.0412.13.8617.30
09:0019.11.0612.253.9420.30
10:0019.21.0812.43.4421.60
10:1519.21.0912.383.5622.40
10:3019.21.0912.353.6923.20
12:0019.21.1811.84.926.70

Frozen model output — THEN inputs only

The frozen model does not produce VERY HIGH risk at 10:00. Detection signal forced to zero. Perth-metro species-weight hypothesis (white 0.50 / tiger 0.30 / bull 0.20). Same logistic seeds as 0.1.0-dev. Not retuned after the fatality.

10:00 AWSTMODERATE34.3/100Confidence MODERATE
ClockBandIndexConf.EHazard
06:00MODERATE34.5MODERATE0.550.5223
08:00MODERATE34.5MODERATE0.550.5223
09:00MODERATE34.5MODERATE0.550.5223
10:00MODERATE34.3MODERATE0.550.5202
10:15MODERATE34.3MODERATE0.550.5202
10:30MODERATE34.3MODERATE0.550.5202
12:00MODERATE34.3MODERATE0.550.5202
  • No live shark detections — this is not an operational alert
  • Water temperature (~19.2°C) favours great white shark thermal suitability
  • Elevated expected human water use

Unknowns at the time

  • Whether any SharkSmart or Water Police report existed before 10:15 that vShark could have ingested (no API connected).
  • The 2.5 m public sighting time (ABC “earlier”) — not placed in the pre-event column.
  • Exact swimmer count, activity mix, and distance from the groyne (post-event witness estimates only).
  • Tide height, rip/channel state, visibility, and local bathymetry at 10:00.
  • Whether a tagged animal was nearby (acoustic network not connected).
  • Aerial patrol coverage that morning.

Lessons (not marketing)

  • Against information that was available from Open-Meteo and the absence of any connected shark-observation feed, 0.1.0-dev assesses Sorrento on the morning of 18 September 2026 as MODERATE public index with MODERATE combined confidence.
  • White-shark thermal suitability at ~19.2 °C is the strongest environmental term. Detection signal is zero because no live detections were connected or asserted.
  • A fatality occurred. That outcome is not used to raise the pre-event band. The scientifically useful result is the gap: no detection layer, no patrol, low ecological coverage, and an uncalibrated development model.
  • LOW CONFIDENCE on shark presence would be the honest operational reading if this were a live warning product. It is not. Official advice after 10:15 is the safety instruction.

Sources

Language on this page is ecological and operational. It does not describe sharks as monsters, infestations, or hunters of humans. A comparable non-event control (Sorrento, Friday 11 September 2026) is registered so this case is not analysed in isolation.