C-DRONE GUIDE · 7 SEPTEMBER 2026
Heat islands and canopy plans: at what scale to measure, and what the drone adds to the satellite
A municipality that wants to act on summer overheating quickly runs into a question of scale. The urban heat island — higher temperatures in the city core than in the surrounding countryside, a mostly nocturnal phenomenon whose peak Météo-France places between 4 and 6 in the morning — reads very well on satellite imagery at conurbation scale. But no operational thermal satellite goes below seventy metres of resolution, and the finest of them does not revisit on demand. Yet the concrete decisions are taken at the metre: should this schoolyard be de-paved, this avenue planted, this asphalt replaced with a light-coloured surface? That is precisely the gap a drone survey fills — thermal for the surfaces, photogrammetric for the canopy. Here is how to combine the two, what indicator to derive for a canopy plan, and above all what the measurement does not tell you.
Published on 7 September 2026, reviewed on 11 September 2026 — regulations in force as of September 2026.
What the satellite sees, what the drone sees: metres and hours
Figures settle this debate better than principles. The thermal infrared channels of Sentinel-3 (SLSTR, channels S7 to S9) work at 1 km resolution at nadir: one pixel swallows a whole neighbourhood. The thermal bands of Landsat 8 and 9 (TIRS sensor, bands 10 and 11) are collected at 100 m: one pixel covers a square and the roofs around it without telling them apart. The finest civil thermal sensor available today, ECOSTRESS, carried on the International Space Station, has a nadir pixel of 38 m by 69 m, resampled to 70 m in the distributed products, and only covers the Earth between 52° north and 52° south. At those scales you rank neighbourhoods; you do not arbitrate a scheme.
The timing constraint matters just as much. Landsat revisits the same spot every 16 days per satellite (8 days combining Landsat 8 and Landsat 9), with an overpass time set around 10:12 at the equatorial crossing: mid-morning, therefore, when the heat island is still far from its maximum. The peak occurs at night, between 4 and 6 in the morning, once the countryside has released its heat while urban mineral surfaces keep releasing theirs. A daytime sun-synchronous satellite will never see that moment; a drone takes off at the hour you choose, including in the middle of the night.
The accuracy gap is not merely theoretical. A team led by Dongwoo Kim published in 2021, in the journal Remote Sensing, a direct comparison over seven periods in an urban park between surface temperature measured by a drone-mounted thermal camera and Landsat 8 data, both checked against contact thermometers on the ground: the drone reached an R² of 0.912 for a root-mean-square error of 3.502 °C, against an R² below 0.3 and an error of 7.246 °C for Landsat (see the study on Google Scholar). In other words: over a park with mixed surfacing, satellite imagery no longer discriminates between ground types — and that discrimination is exactly what you are buying. The right protocol does not pit one against the other: the satellite ranks sectors at conurbation scale, the drone decides at street scale. Our guide to urban heat island thermal mapping details how a campaign runs; what follows concerns the trade-off and the canopy side.
The canopy side: a rate per city block, not a municipal average
The same flight that carries the thermal camera produces, through photogrammetry, a digital surface model and a digital terrain model; their difference gives a canopy height model. Three things come out of it: the ground footprint of the tree cover, the height of each crown, and — crossed with the cadastral parcels or the GIS block layer — a canopy rate per city block rather than a municipal average that steers nothing. A multispectral sensor on the same flight adds vegetation indices, which flag the trees whose chlorophyll activity is dropping before the foliage visibly yellows: our guide to the municipal tree inventory covers that health dimension tree by tree.
Why the block rather than the town? Because the cooling effect is neither linear nor scale-independent. A team led by Carly D. Ziter published in 2019, in the Proceedings of the National Academy of Sciences, a campaign measuring air temperature every 5 metres along ten urban transects using a bicycle-mounted system (see the study on Google Scholar). Findings: daytime temperature decreases non-linearly with tree canopy cover, cooling becoming markedly stronger above 40% cover, and the effect is greatest when cover is analysed at the scale of a city block, that is radii of 60 to 90 metres. Daytime variability measured within the same urban landscape averaged 3.5 °C.
The same study carries a lesson greening plans would do well to absorb: at night, the effect of tree cover is limited, average nocturnal variability falling to 2.1 °C, and it is then the reduction of impervious surfaces that remains decisive. Two levers, two times of day: planting acts on daytime comfort, de-paving acts on the night — the very moment the heat island peaks and tropical nights weigh on health. A serious municipal plan therefore tracks two distinct indicators, canopy rate and impervious-surface rate, and the drone survey feeds both in a single pass. On the second, the high-resolution orthophoto from the flight offers a level of detail national land-cover databases do not reach: that is the gap we document in our guide to the triennial land-take report.
These indicators have an immediate use: when an authority sets itself a numeric target, it needs a repeatable measurement to prove it. The Canopy Plan of the Métropole de Lyon, adopted in 2017 and built around 25 actions, targets 300,000 trees planted and 30% shaded surface across the territory by 2030. A target of that kind is only worth something if it is audited with the same method year after year: same flown extents, same season, same computation rules.
What the drone does not measure, and the precautions of flying over a city
This point needs to be plain, because it is the most frequent misunderstanding in technical meetings. A thermal camera measures a surface temperature: that of asphalt, of a roof, of foliage. It measures neither air temperature nor the thermal comfort felt by a pedestrian. The latter also depends on humidity, wind speed and received radiation — which is what dedicated indices try to integrate, starting with the UTCI (Universal Thermal Climate Index), established by the International Society of Biometeorology in 2009, which combines air temperature, humidity, wind, mean radiant temperature and clothing insulation into an equivalent temperature. Note in passing that the Ziter et al. study cited above measured air temperature at ground level: the two quantities are related but are not substitutes.
Practical consequence: the drone does not replace weather stations or fixed sensors. Those give a continuous time series at a few points; the drone gives a dense snapshot over an extent. The two complement each other, and a well-run campaign in fact places a few reference probes on the ground during the flight — that is what allows the airborne measurement to be checked, exactly like the contact thermometers in the Korean study.
A second limit, this one physical: infrared measurement assumes a correct emissivity. Most urban surfaces sit in a comfortable range — concrete between 0.85 and 0.97, vegetation around 0.95 — but polished metals drop below 0.10 (as low as 0.04 for polished aluminium) and behave like mirrors in the thermal infrared: the camera reads a reflected temperature, not the object's own. Glazing raises the same problem at grazing incidence. A new steel-deck roof or a glass canopy can therefore appear aberrant on the map, and that must be flagged in the report, not smoothed away. Controlled conditions are also required: no recent rain, light wind, stable sky, a consistent time slot from one sector to the next — the Kim et al. study shows moreover that the drone's own accuracy degrades in the wrong season (mean error of 2.5 to 3.5 °C in spring and summer, 3.8 to 5.4 °C in autumn and winter). We detail these criteria in our guide to the measurement window for drone thermography.
A third point, often underestimated at ordering time: flying over dense urban fabric engages both the framework for flights over built-up areas and personal data protection. The thermal camera at this resolution identifies nobody, but the visible orthophoto produced on the same flight does capture private courtyards, terraces, vehicles and their plates. The purpose must be defined and documented, faces and plates blurred, retention bounded, and the public informed — through the municipal website or on-site notices. Our guides on flying over public space in built-up areas and on GDPR applied to professional drone work set out these obligations. A campaign scheduled early in the morning, outside busy hours, settles a good part of the issue upstream.
Who commissions this kind of mission, method and 2026 prices
The order usually comes from a parks department or a planning department, sometimes from an ecological transition directorate, sometimes from the consultancy assisting them. Also common are inter-municipal authorities running a canopy plan, urban planning agencies producing a territorial observatory, thermal engineering firms and landscape architects needing a baseline before sketching, social housing landlords weighing up greening across their estates, development-zone promoters having to justify the climate treatment of their scheme, and municipalities engaged in transforming their schoolyards — the subject we cover site by site in our guide to the cool schoolyard and de-paving audit.
The calendar is constrained: the thermal component happens in summer, on a hot day, ideally with one late-afternoon flight and one early-night flight; the canopy component is flown in leaf-on season, in stable light. Two reasons to order dominate: the baseline before works — indispensable if a gain is later to be demonstrated, notably in a funding application under the urban renaturation strand of the French Fonds vert — and the post-works survey, which objectifies that gain in degrees and in points of cover. The constraint not to overlook is comparability: a baseline flown on a September afternoon does not compare with a post-works survey flown on a July morning.
Ranges observed in France in 2026, excl. VAT:
| Service | Range (excl. VAT, 2026) |
| Single site: schoolyard, square, forecourt — day + night thermal and canopy model | €800 to €1,600 |
| Sector of 20 to 50 ha: calibrated surface temperature map and canopy rate per block | €2,000 to €4,500 |
| Multispectral component added (tree cover vitality indices) | +30 to 50% on the flight price |
| Post-works survey, same protocol as the baseline | 50 to 70% of the initial campaign price |
These amounts cover the flight, radiometric processing, GIS layer production and the report; they exclude microclimate modelling (UTCI-type comfort simulation, which belongs to a specialist consultancy), the installation of fixed sensors, and construction supervision. A whole town is best handled by prioritised sectors rather than in one block: more useful and far cheaper. Our drone thermal imaging page presents the service; for a municipality, an inter-municipal authority, a planning agency or an engineering firm, request a quote stating the candidate sectors, your plan's deadline and the GIS format expected.