How many homes does a single delivery drone flight affect?
Pick a depot, enter a destination within 3 km — see the noise corridor modelled on the published TCD acoustic data.
Where is the delivery going?
Departing from Marina Market, Cork
Calculating noise corridor…
Result
—
building exposures from a single delivery (drone noise above ambient)
Modelled estimate calibrated to published TCD acoustic research on the Manna delivery drone (Nash & Kennedy 2024; Kennedy 2025). Levels shown are per-overflight at ground level.
Couldn't complete
Flight settings
These apply to the next calculation
Enter two Eircodes (or pin two locations on the map), then click Estimate.
Basemap & view
Terrain exaggeration1.0×
What does this do?Visual only — stretches terrain relief in 3D view (about sea level). In MSL the flight tube stays flat at true altitude and the ground rises toward it; in AGL the tube rides the exaggerated ground, so both rise together. Does not change the model or the cross-section (always true metres). 1.0× = true scale.
⚠️ At this vertical scale the drone looks higher above buildings than it really is. Building heights are not exaggerated, only the terrain and the flight tube — so the drone-to-rooftop gap is stretched by the same factor. Judge clearance over houses at 1.0×.
⚠️ Your browser has disabled 3D terrain (Canvas2D fingerprinting protection — common in private/incognito windows or strict privacy settings). The ground can't be raised here, so terrain exaggeration has no visible effect. To see terrain relief, allow fingerprinting/canvas for this site, or open it in a normal window. The flight path and cross-section still work.
Noise corridor
Soundscape context (EPA)
Show on map — calc always uses all sources combined
Beyond-Visual-Line-of-Sight operating areas from AirNav Ireland's UAS Geographical Zones. Tap a zone for its conditions and authority. (Cork BVLOS Area 1 covers the Marina delivery area.)
Wildlife & protected areas
Bat roost locations are deliberately withheld by the NBDC to protect them. Data: NPWS (CC BY 4.0), OpenStreetMap (ODbL).
About this tool
DronePath estimates the noise corridor of a delivery drone flight using a three-stage modelling chain established in the Trinity College Dublin acoustics research group:
Source measurementNash & Kennedy (2024), In-Flight Evaluation of UAV Noise Emission, QuietDrones 2024, Manchester. Acoustic test campaign of the Manna delivery drone at Manna's Moorock test facility, March 2024. Established source noise characteristics, directivity, and spectral content for the specific 8-motor contra-rotating aircraft used in commercial operations. 10.17866/rd.salford.27930075
Propagation modelling validationCussen, Garruccio & Kennedy (2022), UAV Noise Emission — A Combined Experimental and Numerical Assessment, Acoustics 4(2). Validated ISO 9613 propagation modelling for UAVs against field measurements, achieving ±2.1 dB agreement at distances up to 30 m. 10.3390/acoustics4020018
Application to delivery operationsKennedy (V1 November 2024, V2 May 2025), UAV Noise Footprint, Manna Aero Noise Report, Trinity College Dublin. Consultancy report applying the validated methodology with Nash & Kennedy source data to the Blanchardstown delivery operation. Provides DronePath's anchor values: 59 dBA flyover at 65 m cruise, 66–68 dBA hover at 14 m above garden, and a 52 dBA urban background reference for Blanchardstown.
Independent measurement corroboration is provided by Amplitude Acoustics report D251213RP1 R3 (April 2026), prepared in support of Fingal planning application FW25A/0428 and publicly available through the Council's planning portal. The report includes direct measurements at residential receivers in the Coolmine area and broadly aligns with the modelled predictions above.
DronePath uses a simplified point-source approximation calibrated to the anchor values above, rather than re-implementing the full ISO 9613 chain. Trade-offs are documented in the caveats below.
How the model works (and what it isn't)
DronePath does not measure anything and is not a standards-compliant acoustic assessment. It is an interpolation tool that spreads two published numbers across a map. Specifically:
The source level is back-calculated, not measured. Starting from the two anchor values in Kennedy 2025 — 59 dBA at 65 m (flyover) and 67 dBA at 14 m (hover) — the tool works backwards to an equivalent A-weighted source level of roughly 95.6 dBA at 1 m.
Propagation uses simple spherical spreading, not ISO 9613. The level at any distance is estimated as source level minus 20·log₁₀(distance) minus a flat atmospheric absorption term (0.005 dB/m). There is no ground-effect modelling, no barrier or building screening, and no frequency-band treatment — it is a single A-weighted figure throughout.
Hover uses a hand-applied correction. The back-calculated source level over-predicts the reported hover figure by about 5 dB, so a flat −5 dB correction is applied during hover to reconcile the two anchor points. This stands in for proper directivity modelling, which the tool does not do.
A standards-compliant assessment (ISO 9613, as used in the underlying TCD research via iNoise software) would measure the source level directly, model ground effect, barriers, directivity and atmospheric absorption per frequency band, and apply the result over real terrain. DronePath does none of that. It rides entirely on the fact that the underlying TCD work — Nash & Kennedy 2024 for source measurement, Cussen et al. 2022 for ISO 9613 validation — was done properly. Treat DronePath's output as a way to explore published findings across a map, not as an independent measurement or a compliant noise assessment.
What the noise bands mean
Bands are coloured by the dB delta above the ambient baseline you set (default 42 dBA, an estimate for quiet suburban Cork streets like Blackrock, Ballintemple, and Ballinlough during the day). The TCD report uses 52 dBA, an urban Dublin baseline.
Each band is named for what a listener at ground level would experience during a single overflight — the loudness ratio versus the ambient baseline and the absolute dBA range. They are descriptive, not regulatory.
Why no regulatory limit is shown
An earlier version of this tool framed results in terms of Ireland's EPA NG4 noise guidance ("exceeds NG4 daytime limit of 55 dBA"). That framing was wrong and has been removed. Per feedback from Dr John Kennedy of TCD (27 May 2026):
NG4 uses a rating level based on a daytime from 7 am to 7 pm or evening time from 7 pm to 11 pm. A single drone flight needs to be averaged over this time window to see if it exceeds the NG4 guidance. A single event above 55 dB(A) does not breach the NG4 limit; the level must exceed 55 dB(A) after being averaged over the day and having any penalties applied. […] All of the existing legislation was developed for other noise sources which are much louder. A train or a bus is much louder than a drone but they are also confined to specific tracks and roads. There is need for research and legislation across Europe to set limits which are tailored specifically to drones.
BS 4142 has the same problem — it averages over hours. DronePath measures something neither of those standards is built for: the per-event level a homeowner experiences when a drone passes overhead, repeated many times per day. We've removed the regulatory framing so the numbers aren't misread either way (as "this corridor breaches a limit" or as "this corridor is fine").
How to read the result
The numbers are about per-event lived experience, not regulatory compliance. A band labelled "Loud (57-67 dBA at ground)" means a listener inside that band hears the drone at that level for roughly 30 seconds during each pass. Each delivery flies the corridor twice. Multiply by the expected daily flight count to estimate cumulative exposure.
"Buildings" vs "houses"
Counts are labelled as buildings, not houses, because OpenStreetMap's tagging is uneven. OSM building polygons used here are classified as:
Unknown use — tagged just building=yes, the most common tag in Irish OSM. In residential street grids these are almost always houses, but the OSM data does not confirm this.
The summary panel and band-detail modal show the confirmed-residential vs unknown split so you can judge how heavily the counts depend on the untyped buildings.
What this tool does not account for
Building shielding (real back gardens are often 5-10 dB quieter than open-line-of-sight predictions)
Frequency-resolved directivity — the along-track/abeam amplitude lobes and the angle-dependent character are now modelled (see below), but not the full per-third-octave-band directivity (which varies 10–20 dB across emission angles in Nash & Kennedy 2024)
Real flight-path routing — outbound and return are modelled as the same straight line
Tonal / impulsive character of drone sound — the rotor whine and modulation are part of what makes drones disproportionately annoying, but the model treats only A-weighted overall level
Time of day or duration penalties
Atmospheric conditions (temperature, humidity, wind direction — these are inputs to full ISO 9613 modelling, not modelled here)
Multiple simultaneous aircraft — the cumulative-impact issue cited by Fingal County Council in its 19 May 2026 refusal of the Coolmine planning application is not addressed by this tool, which models single flights
A note on where the measurements come from
No academic source provides direct acoustic measurements of the Manna delivery drone in residential settings. The Nash & Kennedy 2024 measurements were conducted at the Manna test facility in Moorock, County Offaly — a flat open peatland chosen specifically because it is free of contaminating noise sources and reflective surfaces. Application to residential settings (Blanchardstown in Kennedy 2025, and the Cork and Dublin streets DronePath models) relies on ISO 9613 propagation modelling of the Moorock source data, validated to ±2.1 dB in Cussen et al. 2022. The Amplitude Acoustics 2026 planning report does include direct noise measurements at residential receivers in the Coolmine area — these provide the only published residential-setting measurements of this aircraft and broadly align with the modelled predictions.
On directivity modelling
Earlier versions treated the drone as an omnidirectional point source. As of v0.6.13 the corridor models the flyover directivity measured by Nash & Kennedy 2024: the airframe shields side (abeam) emission, so for a given slant distance the level is higher ahead of and behind the flight direction than to the sides, and the rear radiates more than the front. The corridor is therefore elongated along the flight axis — reaching further behind the depot than ahead of the customer — while the perpendicular (abeam) band widths are kept as the calibration reference, so the validated levels and building counts are unchanged. The fore/aft lobes are modelled at +2 dB (front) and +3 dB (rear) relative to abeam, within the ~5 dB overall range reported. The sound-character labels additionally encode the frequency-dependent angular emission: low-frequency blade-pass tones dominate directly below (small angle from nadir), while higher harmonics carry further toward wide side angles. What is not yet modelled is the full per-third-octave-band directivity (a 10–20 dB swing across angles) — that needs the frequency-resolved propagation the underlying TCD work performs in iNoise.
On distance and atmospheric absorption
Levels fall with distance from two effects. The first is spreading — sound spreads over a larger area as it travels (−6 dB per doubling of distance). The second is atmospheric absorption, where air progressively soaks up sound, and this is strongly frequency-dependent: high frequencies are absorbed far faster than low ones. As of this version DronePath models this per octave band rather than with a single average figure. The drone's measured sound spectrum (Cussen/Garruccio/Kennedy 2022) is dominated by the 1–4 kHz range, and the ISO 9613-1 absorption there is steep — roughly 3.7 dB/km at 1 kHz, 9.7 at 2 kHz and 32.8 at 4 kHz, versus about 1 dB/km at 250 Hz. Each band is propagated with its own coefficient and the result re-summed, so the audible footprint rolls off faster at long range than a flat coefficient would predict (the far edge of the faint band shrinks by roughly 10–20%), and the character dulls with distance as the high-frequency whine is absorbed first, leaving the lower hum — which is why the sound-character labels shift from "whine" near the line to "thump/hum" further out.
Propagation assumptions: 10 °C, 70 % relative humidity, 101.325 kPa — ISO 9613-1's reference condition, and a close match for Irish daytime weather; the coefficients carry ±20% accuracy. Favourable (downwind, low temperature-gradient) propagation is assumed, which is the louder case — real upwind or daytime-lapse conditions roll off faster and produce shadow zones. The drone is an elevated source (~65 m), so unlike road traffic there is little ground attenuation and a clear line of sight is assumed; building screening is not modelled (real back-garden levels may be several dB lower).
A note on conservatism: where a modelling choice is genuinely close — a level sitting right on the boundary between two impact bands — DronePath rounds toward the smaller impact (the drone level rounded down, the background up). The intent is that the tool should never be open to the charge of overstating drone noise: where it is uncertain, it under-claims rather than over-claims.
How buildings are counted
A building is counted once, by its centroid, in the loudness band that contains it. When EPA noise is used as the ambient, each building is then re-graded against its own local background rather than a single flat baseline: the band reflects how far the drone's peak rises above the Lden mapped at that spot, with road, rail and industrial layers energy-summed where each is present. Where no EPA contour is mapped (quiet streets below the 55 dB threshold) the flat baseline is used instead. A building whose drone peak falls below its local background is reported separately as masked and left out of the band totals — the background is already louder there, so the drone is unlikely to be picked out. The drone level used in this comparison is taken at the quieter edge of each band, keeping the count on the conservative side. The practical effect is that homes on busy roads (or near rail and industry) drop into lower bands or out, so the totals are smaller and more honest near loud infrastructure; anywhere a source isn't mapped its masking is missing, so counts there may be overstated.
Why the centroid — and what it misses. The EPA background for each building is read at its centroid, the single point at the middle of its footprint. This is a deliberate, disclosed simplification, because the "correct" point depends on which receptor you mean and the data can't resolve it. EPA's Lden is a facade value (the most-exposed wall, at 4 m), so a centroid sitting set back behind that wall reads quieter than the facade EPA would assess — a house whose street-facing front is inside a road contour but whose centre falls just behind it is treated here as quiet. That makes the masking from road, rail and industry modest, so counts near loud infrastructure lean slightly high rather than low. We deliberately do not correct this by pushing the test point out toward the road, because the receptor that matters most for an overhead drone is the opposite one: the open back garden, which the building itself screens from road noise and which is genuinely several dB quieter than the facade — exactly where a facade-matched background would wrongly mask the drone away. Pinpointing that garden would need per-building orientation we don't hold, so rather than invent either a facade buffer or a garden-shielding allowance, we test one honest, reproducible point and state plainly that it captures neither the loudest (road-facing) nor the quietest (sheltered-garden) part of a dwelling exactly.
On delivery time, descent and ascent
A tethered delivery is not a flyover: the drone descends from cruise to a low hover, lowers the package on a tether and climbs back, spending tens of seconds near-overhead. DronePath estimates each phase from the cruise altitude, the hover height and the flight distance. Manna's published figures anchor the cruise speed (50–80 km/h; the model uses the lower, conservative end), the cruise altitude (50–80 m) and the low tethered hover, and report an average door-to-door time around 2 min 40 s. The climb rate, descent rate and the dwell while the package is lowered are not published, so they are engineering estimates exposed as adjustable inputs — the intended use is to time a real flight with a stopwatch and tune them until the modelled total matches. The descent/dwell/ascent total is the figure that matters for a neighbour, because it is a sustained near-overhead exposure rather than a brief pass.
Because exposure depends on time, the SEL shown for the customer, the depot and a nearby location is now integrated over those phases rather than read off a single peak: the level is sampled second-by-second as the drone descends, hovers and climbs (or, at the depot, takes off and lands) and energy-summed. The result is markedly higher than a single flyover — a lingering hover can be 15+ dB more exposure than one pass — which is the whole reason dwell matters. The depot is scored on every flight's takeoff and landing, the most repeated exposure in the system; its very-near-field levels are floored at a 10 m slant, where the point-source model stops being reliable, so the depot figure represents a close neighbour rather than someone on the launch pad.
On A-weighted decibels and drone noise
DronePath reports A-weighted decibel figures because this is the standard metric in environmental noise research, including the underlying TCD work. Acoustic research on drone noise specifically (Torija et al. 2020, cited in Cussen et al. 2022; Nash & Kennedy 2024) notes that A-weighting may understate the perceived annoyance of drone noise — the contra-rotating motor configuration produces low-frequency tones and high-frequency motor whine that A-weighting attenuates. Two drone events at the same dBA level as a passing car may be perceived as substantially more disruptive than the car (Christian & Cabell, cited via Torija, found drones as annoying as road vehicles ~5.6 dB louder). This is a known limitation of using dBA for drone noise; no consensus alternative metric exists yet.
The numbers should be treated as order-of-magnitude estimates for relative comparison, not as absolute predictions. For planning-grade noise assessment, consult an acoustic engineer.
What's new
A short summary of recent feature changes — not a full version log. The current build number is shown in the header badge.
Works outside Ireland now — with an honest caveat. You can place a depot or address anywhere: buildings (OpenStreetMap) and terrain (global elevation) load as normal. But EPA background-noise mapping only exists for the Republic of Ireland, so beyond it the model measures loudness against a flat baseline — which can overstate how audible a drone is — and a persistent notice says so, with the impact-zones wash drawn hatched. Out-of-Ireland runs fetch everything live, so give them a little longer.
40 main towns load instantly — now with background noise baked in. For the larger towns of the Republic, the homes, terrain and EPA road/rail/industry/airport background are pre-computed, so picking one paints a properly masked footprint in a second or two instead of fetching dozens of map tiles and contour queries live. Under the depot box, a "… towns load instantly" expander lists them as tap-to-load chips — or just type any town name as before.
When a result is degraded, it now says so — and keeps saying so. If the EPA background data doesn't load, the model falls back to a flat baseline that overstates loudness, so the impact-zones wash is drawn hatched and a warning now stays up (rather than briefly flashing) until you re-run successfully — including on the pre-baked town caches, whose live EPA fallback can still time out. If terrain elevation hasn't loaded (home popups show Ground "?"), a persistent notice flags that hills and valleys aren't modelled, with the fix (open 3D once, then re-paint).
Impact zones cover the full reachable area, shaped by where flights can actually go. The wash is the potential-exposure envelope across the whole service area — not just the sampled delivery spokes — but a spot is only painted where a feasible flight can reach it: it stops at the approved BVLOS zone boundary and carves a shadow behind no-fly gaps and (in held-altitude mode) behind ridges the flight can't clear. Colour is how far a pass directly overhead clears the local background; opacity is how often that spot is overflown, densest near the depot.
You're now told when live data doesn't load. When the EPA background-noise service or the OpenStreetMap building data is busy or unreachable, a small dismissable notice now appears (instead of failing silently), so you know the estimate may be incomplete — patchy masking or missing homes — and can re-run. Repeated failures within one run collapse into a single notice.
Calculating…
Modelling the corridor and counting buildings — the OpenStreetMap step can take 10–20s.