C‑DRONE
Night aerial view of an illuminated urban area

C-DRONE GUIDE · 7 SEPTEMBER 2026

Monitoring an active volcano by drone: Piton de la Fournaise, La Soufrière, Mount Pelée

On 19 September 2022, Piton de la Fournaise erupted. Sixteen days later, when the flow stopped, volcanologists had more than twenty satellite images, four series of aerial images and eighteen successive flow outlines. The erupted volume was estimated at 8 to 11 million cubic metres by satellite thermal sensors — and at 11.6 ± 0.5 million cubic metres from digital surface models built from the air. That gap in precision sums up what drones bring to volcanology: they will not tell you when the volcano will wake up, but they measure to the centimetre what it left behind. For an observatory, a prefecture, a road authority or a power network operator, that difference drives the return to service. Here is what a flight brings on France's three active volcanoes, what it does not replace, and the very tight framework it has to fit into.

Published on 7 September 2026, reviewed on 11 September 2026 — regulations in force as of September 2026.

Three active volcanoes, three observatories, a locked-down access framework

France has three active volcanoes, all in its overseas territories, and each has its own permanent observatory attached to the Institut de physique du globe de Paris (IPGP):

The OVPF was born of a disaster: after the 1977 eruption that partly destroyed the village of Piton Sainte-Rose, the observatory became operational in late 1979. It now runs about a hundred instruments across five networks — seismology, deformation (tiltmeters, extensometers, GPS receivers), geochemistry (SO2, H2S, CO2), cameras and weather — and since 2020 also covers Mayotte through REVOSIMA. Piton de la Fournaise is one of the most active volcanoes on Earth: at least 188 documented eruptions since 1640, and a new cycle since 2014 with 24 eruptions between 2014 and June 2023, each lasting 0.5 to 50 days and producing 0.3 to 35 million cubic metres of lava.

The two Caribbean volcanoes are in a different phase: no eruption, but heightened watch. La Soufrière is held at the yellow alert level by the OVSG-IPGP; since 1992 its seismic, fumarolic, thermal and deformation activity has followed a fluctuating but broadly increasing regime. Prefectural order no. 2024/010/CAB/SIDPC of 2 May 2024 bans access to the Dupuy and Tarissan sinkholes and to the area south of the South crater, and restricts crossing of the summit safety perimeters to hiking guides trained in volcanic risk by the OVSG and carrying gas masks: some fumaroles there exceed 200°C. Mount Pelée entered a reactivation phase in 2019; on the OVSM-IPGP's recommendation of 4 December 2020, the prefecture placed it at the yellow watch level of the ORSEC-Volcano plan, still in force in 2026. The site "Volcanoes and forests of Mount Pelée and the pitons of northern Martinique", some 14,000 hectares or 12% of the island, was also inscribed on the UNESCO World Heritage list on 16 September 2023.

Drone rules there are the European ones, identical to mainland France — it is the terrain and the zoning that change, as set out in our guide to drone services in France's overseas departments. On a volcano, two further layers stack on top: the ORSEC volcano plan and national park regulations.

What the drone actually measures: extent, thickness, volume, morphology

The 19 September to 5 October 2022 eruption at Piton de la Fournaise is the best-documented textbook case. A team led by Magdalena Oryaëlle Chevrel, with the OVPF and the IPGP, published in 2023 in the journal Volcanica the full daily monitoring of that flow (see the study on Google Scholar). The set-up combined three scales: more than twenty satellite images, four series of aerial images, and eighteen flow outlines defined at a rate of one to three per day across the sixteen days of the eruption.

The drone's exact role there is instructive. Two drone surveys were carried out, not over the whole flow but over the vent area: a DJI Mini with a 12 MP sensor on 28 September, then a DJI Mavic 3 with a 20 MP sensor on 3 October, in oblique and vertical views over a 90-metre-radius circle centred on the main vent, at flight heights of 78 m and 191 m. The result: digital surface models at 5.8 and 8.8 cm/pixel. In parallel, two microlight surveys covered the entire flow field, producing models at 8.51 and 12.9 cm/pixel. The drone therefore did not replace the crewed aircraft: it delivered very high resolution where the cone was being built, while the microlight delivered broad coverage.

The processing chain is exactly that of an ordinary topographic survey: Structure-from-Motion photogrammetry to generate point clouds, then cloud differencing between the earlier reference model and the models acquired during and after the eruption, to obtain outline, thickness and volume. The choice of ground resolution is no detail — it is the same trade-off set out in our guide to ground sampling distance (GSD) in drone photogrammetry, and the volume calculation rests on the same mechanics as drone stockpile and volume measurement, applied here to cooled lava rather than to a pile of aggregate. The precision gap is striking: the volume computed from the local digital surface model comes out at 11.6 ± 0.5 million cubic metres, against 8 ± 4 and 11 ± 3 million estimated by integrating satellite thermal flux. One order of magnitude of uncertainty removed.

Beyond volume, a flight delivers the exact flow footprint for hazard maps, the morphology of the eruptive cone and the trace of the fissures, a thermal survey of zones still hot weeks after the flow stopped, and detailed observation of sectors closed to ground access. For a road, power or water network operator, those deliverables are what allow a return-to-service decision: where lava covered the carriageway, how thick, and which zones are still too hot to send crews into.

What the drone does not do — and the constraints people underestimate

A drone does not predict an eruption. Forecasting rests on continuous time series: seismology, deformation measured permanently by GNSS, tiltmeters and satellite radar interferometry, and soil gas geochemistry. That is the job of the roughly one hundred instruments in the OVPF network, recording day and night for decades. A drone flight is a snapshot: excellent for characterising a state, useless for detecting an underlying trend. No one-off aerial campaign replaces a permanent observatory network, and a company suggesting otherwise is selling hot air.

Gas measurement by drone deserves the same honesty. It exists, it is spectacular, and it matters for hazard assessment: a team led by E. J. Liu published in 2020 in the journal Science Advances the fluxes measured at Manam volcano, Papua New Guinea, using instrumented unoccupied aerial systems — 3,760 tonnes per day of CO2 and 5,150 tonnes per day of SO2, with stated uncertainties (see the study on Google Scholar). But read the paper to the end: thirty-one authors, a dozen laboratories, multi-gas sensors, retrievable bag samples for carbon isotope analysis, and long-range beyond-visual-line-of-sight flights. This is an instrumented research set-up, built for a scientific campaign, not a service a commercial drone operator can offer off the shelf. An operator who needs SO2 fluxes goes to a geochemistry laboratory, not to a remote pilot.

Then come the physical constraints, and they are brutal. The radiant heat of a fresh flow forces a safety distance and shortens battery life. Corrosive gases attack electronics and connectors. Ash is abrasive: flying into a plume wears out motor bearings and fogs optics within minutes — there is no good reason to do it. Altitude thins the air and cuts lift: above 2,000 metres, endurance and payload drop sharply, a constraint detailed in our guide to flying drones in mountains and nature parks. Add high-altitude wind, no mobile coverage across vast sectors, and long walk-in approaches.

The administrative framework is just as restrictive. On Réunion, the ORSEC volcano plan defines four phases: vigilance (eruption possible), alert 1 (eruption probable or imminent), alert 2 (eruption under way, split into 2-1 inside the Dolomieu crater with no threat, 2-2 inside the Enclos with no direct threat, 2-3 outside the Enclos with a threat) and safeguard (eruption ended or stabilised). From the alert stage, public access to the whole Enclos is banned and aircraft movements in the volcano area require prefectural authorisation. The volcano also sits within the core of the Réunion National Park, whose director's order on motorised overflight bans drones outright over five sectors (La Roche Écrite massif, Piton des Neiges massif, the summit and ramparts of Grand Bénare, the ramparts around Grand Bassin, and the ramparts of Rivière des Remparts) and bans them within a 200-metre radius of some twenty-five viewpoints, including the summit of Piton de la Fournaise. Seven grounds for exemption exist — public-service missions, forestry work, scientific and conservation activities, technical operations, audiovisual productions, servicing of isolated sites, public events — but since 16 March 2026 applications go exclusively through the Démarche Numérique platform, with a complete file filed at least 15 days before the activity. A volcano mission cannot be improvised the day before.

Who commissions this kind of mission, method and 2026 prices

The order rarely comes from a private individual. It comes from volcanological observatories and research laboratories (subcontracted photogrammetric acquisition, when the science team would rather focus on interpretation), from prefectures and civil protection services (post-alert debriefs, updating hazard maps), from exposed local authorities and their natural-risk consultancies, from network operators — roads, power, water supply — needing to document a buried section before reopening it, and from national parks tracking vegetation recolonisation on recent flows.

The method is settled: reconnaissance and filing of the exemption application at least fifteen days ahead, definition of a reference digital model (ideally pre-eruption), a flight plan combining vertical and cross-oblique views with GNSS control points, Structure-from-Motion processing, then point-cloud differencing to extract footprint, thickness and volume. Tracking deformation between two campaigns follows the same logic as our guide to landslide monitoring by drone: it is the repeatability of the protocol, more than the accuracy of any single campaign, that makes two dates comparable.

Ranges observed in France in 2026, excl. VAT, for overseas missions including mobilisation and walk-in approach:

ServiceRange (excl. VAT, 2026)
Post-eruption photogrammetric flight over a restricted area (under 50 ha): orthophoto + digital surface model€2,000 to €4,500
Radiometric thermal survey of a flow field with a hot-zone report€1,500 to €3,500
Repeated morphological monitoring campaign (2 to 4 passes over an eruptive cycle, model differencing)€6,000 to €15,000
Instrumented gas measurement mission (SO2, CO2)off catalogue: research framework, laboratory partnership

These amounts cover acquisition, photogrammetric processing and delivery of the models; they exclude volcanological interpretation, hazard modelling and gas measurement, which belong to an observatory or a laboratory. The price of a volcano mission depends above all on three factors: how long the overflight exemption takes to obtain, the available weather window (often a few hours in the early morning) and the distance equipment has to be carried. For an observatory, a local authority, a network operator or a natural-risk consultancy, request a quote stating the volcano concerned, the area to cover, the current ORSEC phase and whether an earlier reference digital model exists.

Request a free quote

Put it into practice

Also worth reading