Marmion
Perth Metro, WA · City of Joondalup · Relative human–shark interaction conditions
Model audit · 18 September 2026
What vShark could have known at 10:00 AWST is kept separate from what became public after the incident. The frozen reconstruction does not raise the live 0–100 because a fatality occurred.
Standing local context (true at 10:00)
- This location is inside the City of Joondalup coastal strip covered by the 18 September 2026 official closure.
- Surf Life Saving daily patrols on this metropolitan coast were not scheduled to begin until 26 September 2026.
Barriers are used at Cottesloe, Coogee, Quinns Beach and Bathers Beach. Some are seasonal and not installed until October. That is not a swim clearance at this beach.
Experimental · uncalibrated
Evidence freshness: 10 Nov 2026, 3:02 pm
Dominant evidence class for the index: Environmental habitat suitability
Public record on file (CACHED) — not used as live detection in the 0–100.
Why?
- •No live shark detections — this is not an operational alert
Shark interaction risk is assessed as MODERATE (public index 25.1/100) with moderate combined confidence. Primary drivers: No live shark detections — this is not an operational alert. No live shark-detection feed is connected. This is an experimental development assessment, not an operational warning. Always follow lifeguard instructions, beach closures, and official WA SharkSmart / DPIRD advice. vShark is not a government or SLSA product.
Species conditions
Mixed assemblage / transition hypothesis for the Perth metropolitan coast. Do not treat any single species as dominant without evidence.
Recent detections / sightings
- Great white sharkCACHEDpublic sighting · unverified · 18 Sept 2026, 1:11 pmPublic source
- Species not confirmedCACHEDofficial incident · verified · 18 Sept 2026, 10:15 amPublic source
CACHED rows are curated public notices, not a live tagged-shark feed, and do not enter the 0–100 as live detections.
Environmental state
No environmental observations stored yet. Open-Meteo ingest writes to a regional centroid.
Human exposure
- Class
- estimated
- Popularity proxy
- moderate
- Model exposure
- 0.45
Activity classes: swimming, ocean swimming, surfing, bodyboarding
Estimated, not observed counts. No false precision.
Persistent spatial context
- beach
- reef
Geomorphology is persistent context, not a transient observation.
Data gaps
- turbidity proxy
- live detections
- recent detections
Missing data reduces confidence; it does not mean low risk.
Nearby locations
- Sorrento1.5 km
- Hillarys2.1 km
- North Beach2.7 km
- Trigg3.3 km
- Whitfords5.2 km
- Scarborough6.1 km
Data coverage
Where vShark knows little is as important as where it knows more. Missing coverage does not mean low risk.
- Shark observation coverageLOW
Curated public notices for 18 September 2026 (SharkSmart incident, Boyinaboat report). Not a live tagged-receiver feed. Records are CACHED and do not enter the 0–100 as live detections.
- Ocean coverageMODERATE
Open-Meteo marine SST and wave height at a regional centroid — not a beach sensor.
- Weather coverageMODERATE
Open-Meteo forecast weather at a regional centroid.
- Ecology coverageLOW
No live prey, whale, seal or productivity feed. Seasonal notes are hypotheses or climatology.
- Human exposure coverageLOW
Estimated from popularity class, weekday/weekend and swell. Not a head-count.
- Sensor coverageNONE
No acoustic, camera, drone or eDNA adapters connected.
- Historical coverageMODERATE
Frozen model audit for 18 September 2026 is on file. Not a complete incident archive.
NOW / HISTORY — recent assessments
Append-only model outputs. New model versions do not rewrite these rows.
| When | Band | Index | Confidence | Model |
|---|---|---|---|---|
| 10 Nov 2026, 3:02 pm | MODERATE | 25.1 | MODERATE | 0.1.0-dev |
| 10 Nov 2026, 11:42 am | MODERATE | 25.1 | MODERATE | 0.1.0-dev |
| 10 Nov 2026, 11:06 am | MODERATE | 25.1 | MODERATE | 0.1.0-dev |
| 9 Nov 2026, 6:49 pm | MODERATE | 25.1 | MODERATE | 0.1.0-dev |
| 9 Nov 2026, 2:02 pm | MODERATE | 25.1 | MODERATE | 0.1.0-dev |
| 9 Nov 2026, 7:53 am | MODERATE | 33.4 | MODERATE | 0.1.0-dev |
| 9 Nov 2026, 4:54 am | MODERATE | 33.4 | MODERATE | 0.1.0-dev |
| 9 Nov 2026, 2:23 am | MODERATE | 33.2 | MODERATE | 0.1.0-dev |
| 8 Nov 2026, 9:29 pm | MODERATE | 33.2 | MODERATE | 0.1.0-dev |
| 8 Nov 2026, 6:54 pm | MODERATE | 33.2 | MODERATE | 0.1.0-dev |
| 8 Nov 2026, 4:57 pm | MODERATE | 33.5 | MODERATE | 0.1.0-dev |
| 8 Nov 2026, 12:24 pm | MODERATE | 33.5 | MODERATE | 0.1.0-dev |
| 8 Nov 2026, 7:15 am | MODERATE | 33.3 | MODERATE | 0.1.0-dev |
| 7 Nov 2026, 5:43 pm | MODERATE | 33.3 | MODERATE | 0.1.0-dev |
Evidence hierarchy
- Direct recent detection — Tagged receiver, verified instrument detection, or equivalent in the relevant window.
- Verified sighting — Sighting confirmed by a responsible authority or trained observer.
- Multiple corroborated observations — Independent reports that agree in space and time.
- Strong ecological proxy — Prey aggregation, carcass, or similar with documented association — still a proxy.
- Environmental habitat suitability — Temperature, depth, and similar suitability — not presence.
- Historical seasonal association — Climatology / phenology prior. Not a live observation.
- Long-term background prevalence — Baseline occurrence in the region. Weakest for now-casting.
Methodology
Relative interaction risk, not attack probability. Official advice outranks this page. Methodology · Data sources