Project risk register · 11 July 2026
Risks tied to evidence, owners, and retirement paths
Eight current camera-payload risks drawn from committed trace, radiometry, cost, requirements, and validation findings. Ratings are qualitative design- review judgements, not numerical probabilities.
Review position
The open exposure is concentrated in measured engineering gaps and explicit human decisions.
Committed-evidence register
Current risks and the work or decision that retires each one.
| Risk | Driver evidence | Likelihood | Impact | Mitigation or retirement path | Owner |
|---|---|---|---|---|---|
| Slow survey optics cannot concentrate enough energy from the weakest subpixel targets. | The committed F-number / aperture trade gives F/5 only 26% 2x2 ensquared energy at 11 µm. At the 50 mm survey scale, the 1 cm ember proxy is below the advisory gate at SNR 3.2 and the weak warm edge is at SNR 0.1. | High if the survey branch remains near F/5. | High: target sensitivity fails even when geometric image quality appears acceptable. | Retarget the optimizer to the 25 mm F/3.2 and 50 mm F/4 cases, then report detector-plane energy and SNR in labelled-preliminary traces. | RFS-042 to RFS-046 |
| Custom-payload recurring cost can exceed the fleet unit-cost ceiling. | The committed fleet model estimates USD 22k-57k recurring per custom payload against an indicative USD 66k total payload budget. The recorded 500 ha / 120 m case requires three payloads and allows about USD 22k each. | High under the current custom-cost allowance. | High: the fleet misses either the budget or hourly coverage target. | Keep the 120 m survey path COTS-first; mature custom reflective optics for the waiver/fine-GSD or cued-spotter regimes; revisit when cost evidence or budget changes. | Greg / AAO budget decision; RFS-019 |
| Above-120 m coverage depends on an operational authorisation the research study does not possess. | The fleet model requires BVLOS / altitude authorisation for waiver figures. The recorded source note makes 120 m AGL binding and keeps the waiver as upside only. | Medium for the waiver-dependent branch. | High: without authorisation, only the lower-coverage 120 m column is available. | Judge the baseline at 120 m. Do not use waiver coverage for acceptance until AAO/RFS records the authorised altitude and operating basis. | AAO / RFS altitude-authorisation decision |
| Folded reflective candidates can fit the envelope while blocking their own beams. | The corrected multi-mirror checkpoint measures only 0-6% nominal minimum unobstructed fraction across the four- through seven-mirror seeds; several fields are fully blocked. | High for the current seed geometries. | High: image-quality results from blocked rays cannot support a viable prescription. | Keep obstruction in the merit function, require at least 90% unobscured throughput per field, and retain a measured negative result if the branch cannot clear the geometry. | RFS-043 to RFS-046; RFS-052 and RFS-053 |
| Alignment and fabrication tolerance can erase nominal performance in a higher-mirror-count branch. | The corrected checkpoint reports a relative tolerance-yield proxy of about 0.30 for the seven-mirror seed; the RFS-035 record ties the reduction to corrected surface normals and OPD evaluation. | High for the current seven-mirror seed. | High: low yield undermines fabrication readiness, cost, and repeatability. | Do not promote a seed on nominal image quality. Recalculate a tolerance budget after obstruction and image quality pass, with manufacturability ahead of cost in shortlist selection. | RFS-019 and RFS-028 |
| A detector branch may expose display video but not replayable raw or radiometric measurements. | The committed radiometry requirements reject AGC-only imagery for metrics and require source frames, detector state, calibration links, and checksums. Exact product evidence remains absent from the evidence ledger. | Medium until an exact detector interface is verified. | High: display-only data cannot support quantitative detection, calibration, or false-alarm claims. | Apply the same data contract to COTS and custom paths; withhold quantitative claims when source or calibration provenance is missing; implement replayable record types. | RFS-047; AAO detector-branch decision |
| The 5% sortie-level false-alarm tolerance is not demonstrated on representative held-out data. | The requirements checkpoint sets the tolerance. The dataset plan requires held-out burns/sites/sensors, a no-fire stress set, radiometric inputs, and local EmberScope data; local data does not yet exist. | High that the target remains unverified in the current evidence set. | High: excessive alarms can make a sensitive camera operationally unusable. | Implement replayable records and deterministic target tests, then measure held-out no-fire and burn performance per minute, hectare, and sortie. | RFS-047 to RFS-049; optional field scope requires AAO decision |
| Later stakeholder answers can supersede the advisory defaults and reopen derived targets. | The 2026-07-05 source note closed RFS-011 for research use after the requirements delay, but any dated Dani/RFS answer supersedes the matching default and reopens dependent artifacts. | Low for the current research run. | High if target area, altitude, payload branch, or airframe basis changes. | Preserve current defaults and provenance. When a dated answer arrives, update the source chain and rerun only affected requirements, coverage, and trace artifacts. | Greg / AAO requirements decision; RFS-054 complete |
How to read this register
Measured branch failures stay separate from charter decisions.
“High if…” describes a condition observed in a committed branch, not every possible EmberScope design. The next traces may retire a risk or record a useful negative result.
The 120 m baseline, advisory requirements defaults, and COTS-first survey path let the research study proceed. Vendor responses, hardware purchase, and above-cap flight authority are not prerequisites for its feasibility conclusions.
Return to the project review index to inspect the mission, optics, detector, validation, cost, and decision-history evidence around these rows.