Why a single bear sighting is not enough: the case for each animal’s history over time
7 min read
On 8 August 2026 — the same day we published the first version of this method — the Large Carnivores Section of the Province of Trento’s Wildlife Service issued a notice about the online maps of individual bear sightings. It states, in its own words, exactly the problem our method tries to address: “Online maps that collect individual bear sightings capture a situation tied to a specific moment and cannot be used to establish the presence or absence of animals in a given area.” And further: “already a quarter of an hour after a report the animal may be a kilometre away”; over a day, “movements of tens of kilometres.” The Province’s conclusion is blunt: “the geolocation of a single sighting cannot be used to predict where the animal is” — and, symmetrically, “the absence of reports should not be read as a guarantee that no wildlife is present.”
There is a further reason not to publish precise positions, the Province writes: publishing a specific location “can also have a boomerang effect, drawing in onlookers keen to seek a close encounter with the animal.” The only map the portal does publish covers areas — not individual points — where female bears with cubs of the year have been reported, “as part of institutional and scientific monitoring”; and even here the notice cautions: “Areas not shown on the map cannot be considered, with certainty, free of females with cubs,” because “the picture built from reports takes shape progressively over the course of the season.”
The method does not change
Last month we proposed a method for this exact problem. We restate it here, unchanged: aerial scanning with RTK drones, a thermal camera and a 4K camera, on systematic, repeatable flight grids — not ad hoc searches — with a seasonal window chosen deliberately. Our models distinguish a bear from other animals of similar size and track its movement at a distance, without close pursuit: the system does not decide, it flags. The output is an alert to the competent authority, with position, time and images. That information lets those with the authority organise, under the procedures in force, the fitting of the GPS radio collar — we do not fit the collar, and we do not decide to fit it.
The constraints set out in our previous piece still hold — authorisation for overflying protected areas, a declared stand-off distance, automatic suspension at the first sign of a reaction: we do not repeat them here, they apply the same way.
This is where the method answers the problem the Province describes. A single sighting, by its nature, expires within a quarter of an hour. Systematic, repeated scanning, by contrast, accumulates: position, routes, recurrence, times, elevation, use of territory over time. It is no longer an ageing point but an updating series, cross-referenced with terrain and prior history to derive the areas worth scanning next. Operators see a single map, with each individual’s history: not “where it was,” but where it returns, when, and how often.
The time genetic identification takes
The monthly report for July, published on 6 August, describes the case where this difference matters most. August marks the start of hyperphagia — the pre-winter drive to build up fat reserves — and with it, the Province writes, a rise in damage. The peak is in cattle losses in upper Val di Sole: “seven incidents with more than 10 head killed,” livestock “harder to manage with prevention measures than sheep and goats.” The institutional response: “an intensive monitoring effort is under way, to genetically identify the individual(s) involved,” before adopting — the report explains — the further actions set out in the Pacobace, starting with the fitting of a radio collar. Repeated damage is also recorded between Val di Breguzzo/Val d’Arnò and Val Daone, “where the wolf has preyed too,” and “enquiries are under way to establish, if possible, the identity of the animals involved”; on Monte Bondone, above Trento, repeated sightings are “probably attributable to a young individual.”
It is the right method, set out in the interregional plan: but it takes time, because it waits on the sample and the laboratory, and meanwhile the damage continues. Our method does not replace genetic identification: it arrives earlier, with positional information useful for narrowing down where to look; it delivers afterwards, over time, the continuity a single identification does not give. Certain identity remains genetic — the record of where an individual returns, and when, is a different, complementary piece of data.
Seven archives, one question
Citizen reports, damage assessments and compensation files, genetic samples in the laboratory, radio-collar data, camera-trap images, inspections by specialist dog units — like the one confirming, in late July, that a car-struck bear near Vezzano had moved off unaided — and formal orders. None of these archives, on its own, answers the question that matters to a decision-maker: which individual, where, with what history. It is the same problem, in a different shape, that we deal with elsewhere: proof of customs origin scattered across production, purchasing, logistics and customs is not an isolated case — it is the same question, put to different systems that do not talk to each other.
The platform we build inside our confidential data analysis service brings scans and GPS data together into a per-individual history. It does not replace the other archives — reports, compensation, genetics remain with those who manage them — but the history we build can be cross-referenced with that data by those entitled to see both — the same logic behind our work, in other files, for the public sector. See the service · Talk to an engineer
Not a public map
The same caution the Province asks for individual sightings applies here. The positions our scanning produces do not feed an app or a browsable site: the recipient is the competent authority, access is logged, and there is no public exposure — consistent with the 8 August notice on the boomerang effect of a published location. A system that, in trying to inform, ended up attracting onlookers seeking a close encounter would have failed its purpose just as much as a drone that disturbs the animal it is meant only to observe.
What we do not know
We have no accuracy data for our models on this species, and we do not claim any: this is an operational proposal, not a measured result. We do not know the real timeline of the Val di Sole genetic identification, nor its outcome. We do not know which flight authorisations apply case by case: that depends on the area, the period and whoever holds authority over the territory, and must be verified with the relevant body each time. Were this proposal to become an operational trial, objectives and verification criteria would need to be declared in advance — the same principle used to judge whether a system has actually worked — not afterwards, once the result is in. We hold no assignment from the Province of Trento on this subject: what is written here is a public proposal, not work in progress.
Comply and decide
Comply: the record of who saw what, when, with which instrument, and who looked at that data remains available to produce for whoever has the authority to ask for it — the same material an investigation must be able to produce, whichever way the final decision goes.
Decide: the single map with a history for each individual — not the single point that expires within a quarter of an hour — is the operating model on which an AI flags and an operator decides, within the perimeter of those with authority. We build it in two modes: on-premises, on autonomous machines needing no deep integration into the client’s network, or a dedicated cloud with a data centre in Italy — always with shared management, the same principle governing whoever administers a company AI system.
Do you manage a territory where a single sighting is not enough to decide? Talk to an engineer: the first session is free of charge.