Aviation has one of the most remarkable safety records of any industry, and part of the reason is cultural: when something goes wrong, it gets reported, recorded, and studied. Wildlife strikes are no exception. The FAA’s National Wildlife Strike Database alone holds hundreds of thousands of reports stretching back to 1990. ICAO’s IBIS system aggregates reporting internationally. National authorities, airlines, and airports maintain records of their own.
This is a genuinely valuable body of evidence. It has driven habitat management standards, informed engine certification testing, and shaped wildlife hazard management programmes at thousands of aerodromes. Nothing in this article diminishes that.
But if you want to use this data for prediction, forecasting where and when strike risk will be elevated, you run into a problem the wildlife hazard community has known about for decades: the record is a picture of reporting behaviour as much as it is a picture of risk.
The shape of the problem
Strike reporting is voluntary in many jurisdictions and contexts. Reporting rates differ between airlines, between airports, between countries, and across time. Reporting culture in 2026 is very different from 1996, which means longitudinal comparisons need careful handling. Studies of under-reporting have produced a wide range of estimates over the years; the honest summary is that a substantial fraction of strikes never enter the record, and that fraction is not constant across operators or eras.
The gaps are not random, which is what makes them dangerous for modelling. Minor strikes are less likely to be reported than damaging ones. Strikes identified during runway sweeps may be logged differently from those reported by crews. Species identification varies with access to remains, feather identification services, and local expertise; a large share of records identify the species only partially or not at all. Altitude and phase-of-flight fields are unevenly completed.
Feed this record naively into a model, and the model will faithfully learn all of it: the risk patterns and the reporting patterns, indistinguishably blended. A model trained on raw counts will conclude that airports with diligent reporting cultures are dangerous, and that operators who under-report are safe. It will mistake silence for safety.
Absence of reports is not absence of risk
This is the discipline that has to sit underneath any credible forecasting effort. In practice it means several things.
First, treat reporting behaviour as something to be modelled, not ignored. Where reporting intensity can be estimated (from traffic volumes, programme maturity, or the ratio of minor to damaging reports), it becomes context for interpreting the counts.
Second, prefer signals that don’t share the record’s biases. A strike report is a lagging, human-filtered indicator. Meteorology is neither: weather is measured continuously, consistently, and independently of anyone’s reporting culture. The same is true of seasonal and migratory patterns. Fusing these signals with the strike record lets each source compensate for the other’s blind spots. That is precisely the approach we take at Avigilance, and the subject of the next article in this series.
Third, say what you don’t know. A forecast built on this data should carry its confidence openly, and there are questions, particularly at low-reporting aerodromes, where the honest output is a wide interval and a stated limitation, not false precision.
Why this matters now
The industry manages wildlife risk largely reactively: strike reports feed trend reviews, which feed adjustments to habitat management and patrol scheduling. That loop works, and it has made aviation safer. But it is a loop that starts after the event.
The pieces now exist to add a forward-looking layer: decades of strike history, dense meteorological data, better ecological datasets, and pattern-matching methods capable of finding structure across all of them. What has been missing is the discipline to combine them in a way this community can trust. Trust, in aviation safety, is earned by showing your evidence, your uncertainty, and your working.
That is the standard we hold ourselves to at Avigilance. On 5 August I’ll be presenting our approach, including how we handle everything described above, at the aviation wildlife management conference in Cleveland (session CS1A-P7, 09:00). If your operation lives with this risk, we would genuinely like to compare notes: https://www.avigilance.com/wildlife#register