C‑DRONE
Aerial view of a green mountain valley and its stream

C-DRONE GUIDE · 5 SEPTEMBER 2026

Coastal flooding and the French PPRL risk plan: what a drone survey brings to a municipality or levee manager (method, limits, price)

A municipality receiving the State's information package on a coastal flood risk plan (PPRL) under preparation, an inter-municipal body that has just taken over a coastal levee system, a dune-ridge manager the morning after a westerly storm: all three face the same very concrete question. The reference level set by the State is a number — an elevation in metres in France's legal IGN 1969 datum. The ground, meanwhile, is only known through a national model whose vintage may be several years old, and the crest of the protective structure is often documented only by widely spaced profiles. In between, a drone photogrammetric survey produces, in a single day, a fine elevation model of the low-lying sector and a continuous long profile of the crest, in the same datum as the reference level. That is useful, it is measurable, and it is far from sufficient: a drone models no hydrodynamics, sees nothing underwater, and renders the ground poorly beneath a vegetated dune. Here is what it genuinely documents, what it does not replace, and what it costs.

Published on 5 September 2026, reviewed on 11 September 2026 — regulations in force as of September 2026.

What the PPRL sets, and why everything hinges on comparing a level to the ground

The coastal risk prevention plan (PPRL) is a natural risk prevention plan within the meaning of article L. 562-1 of the French environmental code: the State draws it up and enforces it, and it is prescribed by prefectural order. Since decree n° 2019-715 of 5 July 2019, the rules governing how it is built are codified in articles R. 562-11-1 to R. 562-11-9. The starting point is the reference hazard: it is determined from the largest known and documented event, or from a theoretical event of centennial frequency, whichever is stronger. For marine submersion, the order of 5 July 2019 adds to that hazard an extra height of twenty centimetres to account for short-term mean sea level rise (article 1), and defines a second hazard, described as the 100-year-horizon hazard, by adding a further margin of at least forty centimetres (article 4).

That hazard is then translated into levels — low, moderate, strong, very strong — through a grid crossing water depth and flow dynamics. The 5 July 2019 order sets the water-depth thresholds at 0.5 m, 1 m and 2 m, with dynamics qualified in at least two classes, "slow" and "rapid", and an optional intermediate "medium" class. Water depth, however, is arithmetically the level reached minus the ground elevation. A low-lying sector classed as strong rather than moderate hazard can come down to thirty centimetres of difference in the topographic data — with direct consequences for building rights, for safety requirements imposed on existing buildings, and for property values.

On the sea-level side, the data is public and referenced: the study "Estimation of extreme water level values — mainland coastline, 2022 edition", produced jointly by the Cerema and the Shom (France's hydrographic service), statistically estimates extreme sea levels by return period along the mainland coast, expressed in the legal IGN 1969 vertical datum. That study is based on tide-gauge observations and does not account for wave set-up: the PPRL hazard study adds that component. On the ground side, the national reference is Litto3D (Shom / IGN), a continuous land-and-sea elevation model distributed free of charge — but produced sector by sector and vintage by vintage, so not necessarily representative of a dune ridge's condition after the latest storm, nor of a levee's profile after it has been raised. That gap is precisely what a drone survey fills.

What a drone survey actually produces on a low-lying coastal sector

The typical mission combines an automated flight in parallel strips over the low-lying sector and the protective structure, with careful georeferencing — which is where everything is decided. A model not tied to the same vertical datum as the reference level is worthless: it needs ground control points surveyed with differential GNSS, or an RTK or PPK drone, and a deliverable explicitly supplied in RGF93 / IGN 1969. On a coastline, placing control points takes some craft: overexposed pale sand, no durable landmarks on a beach, a tidal range that closes off certain areas at certain hours.

Three deliverables bear directly on a PPRL file. The first is the fine elevation model of the low-lying sector, at a planimetric resolution below the decimetre where the national model works at the metre: it reveals the micro-relief that actually drives how a sheet of water spreads behind a structure — a sunken lane, a road embankment, a culverted ditch, a raised car-park platform. The second is the long profile of crest levels along the protective structure: masonry levee, seafront wall, dune ridge, groyne. Where the manager often holds only a few widely spaced levelling points, the drone returns a continuous line, and that line reveals the low points — a culvert crossing, a beach access ramp, a local settlement, a pedestrian gap — which are, nine times out of ten, where the water will come through first. The third is the comparison of two successive surveys, which puts a cubic-metre figure on what a storm has taken from a dune ridge or deposited at the foot of a structure.

Achievable accuracy has been measured in French conditions: a study by N. Long, B. Millescamps, B. Guillot, F. Pouget and X. Bertin published in 2016 in Remote Sensing ran three campaigns in three months over the lagoon-inlet system of Bonne-Anse Bay at La Palmyre, using a fixed-wing drone and ground control points surveyed with GNSS then differentially corrected: root-mean-square discrepancies for the surface model came out at around 10 centimetres against a GNSS profile and 17 centimetres against independent control points, with faithful rendering of bedforms of about one metre in wavelength and ten centimetres in height (see the study on Google Scholar). Keep the order of magnitude in mind: this is a few centimetres to a few decimetres, not the accuracy of direct levelling. When a debate turns on twenty centimetres of elevation, that uncertainty must be written plainly in the report rather than quietly omitted.

Two uses where the survey genuinely counts: the precautionary band and the dune ridge

The first use is the least known, and possibly the most valuable for a local authority. Article R. 562-11-4 of the environmental code classes the precautionary bands behind levee systems as very strong reference hazard — the most restrictive level of the zoning. Their width equals one hundred times the difference between the maximum water height reached upstream of the structure under the reference hazard and the natural ground immediately behind it, with a floor set by the 5 July 2019 order at fifty metres (a controlled reduction remaining possible). Two things follow. First, that width depends directly on a topographic figure: the elevation of the natural ground immediately behind the structure. Second, and this is the decisive point, the same article provides that the width may be adjusted in the light of technical evidence supplied by the structure's owner or manager. A fine, dated elevation model, georeferenced in the correct datum and accompanied by its uncertainty statement, is precisely the kind of technical evidence a municipality or inter-municipal body can put on the file — where, absent local data, the assessment falls back on the national model and on default assumptions.

The second use is post-storm. A dune ridge is not a classified structure — article R. 562-13 indeed states that natural elements between two levee sections are not part of the levee system — but it does in practice protect a low-lying hinterland, and its lowering after a run of heavy seas is what turns a moderate hazard into a risk of widespread overtopping. A flight the day after a storm, compared with the reference survey from the previous campaign, quantifies crest lowering in centimetres and lost sand volume in cubic metres, section by section. It is the piece most often missing from an emergency works or beach-recharge application, and it is also the only way to demonstrate, two years later, that the recharge held — or did not. On the other face of the same coastline, that of chronic erosion and shoreline retreat, the logic and the deliverables differ: see our guide to drone coastal erosion monitoring.

The limits, stated plainly: what a drone does not do

A drone models no hydrodynamics. It delivers topography, not a water level, not a storm surge, not an overtopping discharge, not a flood propagation. The chain of tide + atmospheric surge + waves + overtopping + filling of the low-lying basin behind belongs to hydraulic modelling carried out by a specialist consultancy, and the review by E. Chaumillon, X. Bertin, A. B. Fortunato and co-authors, published in 2017 in Earth-Science Reviews, is explicit on this: understanding a storm-induced marine flood requires crossing oceanography, geosciences, hydraulics and social sciences, no single discipline being sufficient (see the study on Google Scholar). A drone survey feeds that model; it never replaces it.

A drone does no bathymetry. Photogrammetry stops at the water surface. The nearshore, the channel, the scour hole at the foot of a riprap revetment, the submerged profile that governs wave dissipation: all of it stays invisible. That is the domain of airborne topo-bathymetric LiDAR and echo sounders, not of a multirotor. A drone survey is therefore planned for spring low tide to expose as much foreshore as possible, in the knowledge that part of the system will remain out of reach.

Photogrammetry yields a surface model, not a terrain model under vegetation. On a marram-covered dune ridge, on a grassed or scrubby levee slope, the algorithm returns the top of the vegetation. Filtering brings most of it back to the ground but leaves a positive bias of a few centimetres to a few tens of centimetres depending on cover density — precisely the order of magnitude that matters in a debate over an elevation. LiDAR, by contrast, partly penetrates the canopy: that is why Litto3D and institutional LiDAR surveys keep their place, and why good practice is to combine the two rather than set one against the other. See our guide to drone photogrammetry deliverables for the DSM / DTM distinction.

A drone does not write a hazard study and does not revise a PPRL. Only the State draws up and revises the plan, under a procedure set by the environmental code. A survey can feed a public-inquiry submission, an information package under article L. 132-2 of the planning code, or a works application; it does not on its own produce a re-zoning. Finally, it says nothing about the inside of a structure — core condition, internal seepage paths, undermining — which remains the province of in-depth technical inspection and geotechnical investigation; on that side, see our guide to drone inspection of dams and levees. A useful reminder: the principle laid down after storm Xynthia and carried into PPRL doctrine is that no structure can be considered infallible. A structure surveyed to the centimetre is still a structure that can fail.

Who commissions it, which deliverables to require, and 2026 prices

Two commissioner profiles dominate. The municipality or inter-municipal body, when a PPRL is prescribed, under revision, or when the prefectural information package arrives and the council discovers the extent of the zoning being considered. And the levee system manager — most often the authority holding the GEMAPI mandate for flood prevention — whose regulatory obligations depend on the structure's class: decree n° 2015-526 of 12 May 2015 and article R. 214-113 of the environmental code set class A above 30,000 protected people, class B from 3,000 to 30,000 and class C from 30 to 3,000, a levee under 1.5 m high being unclassified unless the competent authority requests otherwise. The higher the class, the more demanding the documentation of the structure's condition, and the more a dated, georeferenced survey earns its place on the file.

Four requirements to write into the tender, without which the deliverable will be unusable: the datum (RGF93 horizontally, IGN 1969 vertically, stated explicitly); the accuracy report (number and layout of ground control points, residuals on independent check points, stated uncertainty); the tidal window used and the day's tidal coefficient; and the file formats, with a classified point cloud and a raster model usable in the authority's GIS, not just a PDF. Add the crest-level long profile as a table of values, not only as a chart.

Orders of magnitude observed in France in 2026, in euros excluding VAT:

Weigh that against the only alternative of comparable accuracy — airborne topo-bathymetric LiDAR, far superior over a long stretch and under vegetation, but mobilised at the scale of a regional programme rather than for one municipality on a February Monday after a storm. To scope a need on your low-lying sector, your levee or your dune ridge, request a quote stating the length concerned, the levee system's class if any, whether an earlier survey exists, and the deadline driving the file (public inquiry, works application, information package).

Frequently asked questions

Can a drone survey get a PPRL zoning changed?

Not on its own. The PPRL is drawn up and revised by the State under the procedure set by the environmental code. A georeferenced, dated survey accompanied by its uncertainty report is, however, an admissible technical document: it can be submitted during consultation or the public inquiry, and article R. 562-11-4 explicitly provides for a precautionary band's width to be adjusted in the light of technical evidence supplied by the structure's owner or manager.

What vertical accuracy should be expected, and is it enough against a reference level?

With ground control points surveyed by differential GNSS or an RTK/PPK drone, the order of magnitude documented on the French coast is a few centimetres up to around two decimetres of root-mean-square discrepancy. That is enough to spot a low point in a crest or to quantify a lost sand volume; it is not enough to settle a dispute turning on ten centimetres of elevation. Always require the residuals report on independent check points.

Does a drone replace coastal topo-bathymetric LiDAR?

No, and on two specific counts. It cannot see underwater, so the nearshore and the submerged profile escape it entirely. And photogrammetry returns a surface model: under dense marram grass or on a scrubby slope, the true ground is poorly reached where LiDAR partly penetrates. The drone complements those campaigns over a shorter stretch and at a much higher frequency, particularly right after a storm.

Should the survey be done before or after a storm?

Both, and the first one matters most. Without a prior baseline campaign, a post-storm flight gives a state, not a change: crest lowering and eroded volume cannot be quantified. The useful practice for a coastal authority is to establish a baseline campaign outside the storm season, then trigger a flight after every significant event.

Request a free quote

Also worth reading