1. Introduction
NORAH2 (NOise of Rotorcraft Assessed by a Hemisphere approach, version 2), released by EASA in December 2024, is the first dedicated European tool for helicopter noise modelling. It provides decision-grade outputs for the measurement and reporting of rotorcraft noise, supporting the development of more effective noise management policies. NORAH2 computes sound levels by sampling pre-computed acoustic hemispheres along each trajectory point. For every rotorcraft type and each combination of speed and vertical angle, a dedicated hemisphere encodes the full 3-D directivity—i.e., how the aircraft radiates sound in all directions for that specific flight condition.
1.1. Policy Context and Motivation
Environmental noise is a major public health concern in Europe, with well-established links to annoyance, sleep disturbance, and cardiovascular effects. The Environmental Noise Directive (2002/49/EC, END) provides the policy framework for the strategic assessment and management of noise exposure from major transport sources, including roads, railways, airports, and agglomerations [
1]. Under this directive, Member States must produce strategic maps and action plans every five years using harmonised metrics, most notably for the day–evening–night level (Lden) and for the night-time level (Lnight), as well as consistent assessment methods defined by CNOSSOS-EU [
2,
3].
While END explicitly includes helicopters within the definition of “aircraft,” the obligation to map noise only applies to “major airports” with more than 50,000 movements per year. Many heliports and corridors fall below this threshold, even though community annoyance depends strongly on event characteristics such as maximum levels, tonality, and intermittency, rather than on total annual counts. In practice, helicopter noise is often under-represented in official assessments, leaving affected communities without the same transparency as those exposed to road or rail noise.
ICAO Certification standards administered by EASA provide another regulatory dimension. Helicopters are certified under specific chapters (e.g., Chapter 8, for helicopters over 3175 kg maximum certificated take-off mass, and Chapter 11, for helicopters with or less than 3175 kg MTOW—ICAO Annex 16, Volume I—for defined flight phases (take-off, flyover, approach), but these conditions do not fully capture real-world operations such as hovering, low-altitude manoeuvres, or complex terrain interactions [
4,
5]. Moreover, fixed-wing-oriented modelling tools used in standard aircraft noise mapping often fail to represent rotorcraft directivity or mode-specific emissions.
The gap is therefore clear: robust, transparent tools are needed to model helicopter operations with sufficient fidelity to support regulatory metrics, planning, and community engagement. NORAH2 addresses this by segmenting trajectories into operational modes, coupling them to rotorcraft-specific noise hemispheres, and producing gridded exposure metrics (e.g., LAeq, Lden, Lnight, and number of events above certain levels—N60, N65, N70) that are compatible with EU reporting. In parallel, the rapid emergence of Urban Air Mobility (UAM) and eVTOL concepts reinforces the urgency of consistent, rotorcraft-aware noise prediction for urban settings.
Against this background, the present study sets out specific objectives and scope, described in
Section 1.2.
1.2. Research Scope
The main objective of this study is to demonstrate the capability of NORAH2 to model helicopter noise in an urban airspace, using the case of Cannes, France. A secondary objective is to assess the suitability of open flight radar data, specifically ADS-B trajectories from the OpenSky Network [
6], as an input source for noise assessments. Together, these elements provide a reproducible pipeline that links open traffic data to mode-specific rotorcraft noise modelling.
The scope of the work includes the following:
reconstructing trajectories from ADS-B archives and converting those to form NORAH2 input files;
generating noise contours for standard regulatory metrics over a georeferenced basemap of the study area;
illustrating the potential of these outputs to inform discussions on helicopter noise in the context of urban noise management.
Certain aspects are purposely excluded. For instance, population exposure calculations (number of people or households per noise band) are not presented here, although this process is straightforward once contours are available. Likewise, validation using dedicated microphone measurements is outside the present scope, as it depends more on site-specific circumstances than on NORAH2’s capabilities.
By focusing on the methodological pipeline and the resulting exposure maps, the study aims to generate example noise contours from reconstructed helicopter trajectories. These contours are not presented as official exposure maps, but as a demonstration of how rotorcraft operations can be represented within established regulatory metrics. In this way, the work illustrates the feasibility of integrating helicopter noise into wider urban noise analyses, using reproducible methods that combine open traffic data with NORAH2 modelling.
3. Methods
This section describes the procedures applied to process ADS-B trajectories, convert coordinates, segment operations, optimise positional, speed, and altitude profiles, and prepare model inputs for NORAH2. The sequence of steps was designed to ensure reproducibility and transparency in helicopter noise modelling.
3.1. Study Area: Cannes Coastal Corridor
Cannes is a coastal city on the French Riviera (Côte d’Azur) in the Alpes-Maritimes department of southeastern France. Its municipal population is about 74,000 inhabitants [
29]. Cannes is internationally recognised for the annual Film Festival, which attracts thousands of visitors and major figures in the film industry.
Helicopter transfers along the Riviera and especially during the film festival period generate recurring community concerns due to low-altitude operations near coastal neighbourhoods. A Le Monde feature (May 2023) described the “incessant buzz” of rotorcraft as an entrenched part of mobility in the region [
30]. The article underscored a striking contrast: while helicopters enable rapid transfers between Nice, Monaco, Cannes, and inland estates, they also generate persistent annoyance among coastal residents. This juxtaposition of glamour, accessibility, and disturbance positions Cannes as a relevant case where environmental questions converge with social perception. Such reports highlight the need to test whether modern noise-modelling tools can provide evidence beyond anecdotal accounts.
Helicopters often operate at lower altitudes and closer to populated areas compared to fixed-wing aircraft, creating distinct exposure patterns. Yet systematic quantification of helicopter noise exposures remains limited. This absence of noise exposure information provided the motivation for a proof-of-concept study: translating helicopter operations into mapped metrics using the NORAH2 model.
The study period coincided with the Cannes Film Festival 2025 (12–25 May), when helicopter transfers typically play a prominent role in the noise landscape. The analysis was deliberately scoped to a representative coastal corridor of Cannes, an 8.3 km × 2.5 km grid extending from the Cannes–Mandelieu helicopter base along the shoreline to Pointe Croisette, rather than to the entire commune. This subset was selected because most movements during the study period occurred over this area, with the most affected receptors located in coastal neighbourhoods. The intention was to demonstrate NORAH2’s end-to-end workflow (from trajectory sourcing to mapped metrics), not to produce a comprehensive city-wide noise map. Results should therefore be interpreted as illustrative of method and capability, with the approach readily extensible to a wider study.
Figure 1 presents an overview of the study grid (in red) superimposed on the Cannes coastline, with helicopter trajectories (in yellow) recorded during the two-week study period, highlighting the concentration of operations between the Cannes–Mandelieu heliport and the Port de Cannes helipad. Operations extended beyond the study area to avoid boundary-related noise level artifacts (end of yellow lines).
The two main facilities within the study area are the Cannes–Mandelieu Heliport, adjacent to Cannes–Mandelieu Airport, and a helipad at the end of the pier at Port de Cannes towards the right of the figure.
Table 1 summarises arrivals and departures at these two sites during the study period, broken down by period of day. The majority of operations were concentrated at Cannes–Mandelieu, with activity peaking during daytime hours, although transfers at the Port Helipad also contributed to localised exposure.
The number of operations per rotorcraft type is presented in
Table 2.
3.2. Data Sources: ADS-B, Weather, Terrain, Land Cover
3.2.1. ADS-B Signals: Background and Accuracy
Automatic Dependent Surveillance-Broadcast (ADS-B) is an aircraft positioning system that complements existing primary and secondary radars. Unlike secondary surveillance radar (SSR) it provides conspicuity by broadcasting (as the “B” indicates) a signal at regular intervals, without interrogation. These broadcasts can include a wide range of aircraft state parameters: registration, squawk, 3-D position, barometric and geometric altitude, speed (indicated airspeed and ground speed), heading, vertical rate (climb or descent), QNH setting, autopilot engagement, and others.
ADS-B differs fundamentally from the other two radar systems in current use. Primary radar relies on reflected signals, yielding only basic range and bearing and requiring no action by the aircraft. SSR operates via interrogation and reply, with different modes (A, C, S) providing increasing levels of detail; even Mode S, however, is considered less informative than ADS-B. Introduced operationally in 2003, ADS-B has since proven its value and is now mandatory for many commercial aircraft in specific classes of controlled airspace and under IFR conditions.
By receiving ADS-B broadcasts, a ground station can reconstruct an aircraft’s trajectory at 1 Hz intervals. Aircraft can also receive each other’s signals, which improves situational awareness and enables ADS-B to function as a collision avoidance tool. The signals are unencrypted and free to receive, enabling inexpensive receivers or software-defined radios (SDRs) to capture them. Platforms such as the OpenSky Network, FlightRadar24, and FlightAware aggregate data from large numbers of distributed receivers and make flight trajectories widely accessible.
ADS-B operates at 1090 MHz, whereas the global VHF airband spans 108–137 MHz (with 118–137 MHz used for verbal communications). While higher frequencies generally reduce range at equal transmit power, ADS-B’s digital modulation with error correction ensures high reliability. In practice, inexpensive ground-level receivers can often capture ADS-B transmissions from aircraft 150–200 NM away, with Earth’s curvature being the principal constraint.
The horizontal position reported by ADS-B is highly accurate because it derives from GNSS constellations (GPS, GLONASS, BeiDou, Galileo), often within a metre. Vertical positioning is less precise: GNSS geometry limits access to satellites below the horizon, and signals are refracted at low angles through the atmosphere [
31]. Errors of around ±50 ft are common, though temporal continuity and filtering can reduce noise.
Finally, since ADS-B signals are unencrypted, they could in theory be subject to interference or malicious use. At present, such risks are not considered critical for research applications.
3.2.2. OpenSky: Flight Radar Data
The present study relies on ADS-B trajectory data collected and distributed by the OpenSky Network [
3], which provides access to historical, crowdsourced ADS-B signals.
Data extraction, for the period 12–25 May 2025, was constrained to a fixed bounding box covering the Cannes shoreline study area and limited to altitudes ≤2000 m above mean sea level (AMSL). The dataset provided 1 Hz messages containing icao24 (Mode-S), callsign, UTC timestamp (later converted to local time), latitude, longitude, geo-altitude, baro-altitude, velocity, heading, vertical rate, and on-ground status (all in metric units).
Each message was linked to aircraft metadata using the icao24 identifier. Rotorcraft were retained while fixed-wing traffic was excluded. For each icao24, independent operations (opnum) were defined as continuous sequences of points with no long-time gaps and no large position jumps. The native sampling frequency of 1 Hz was preserved throughout the analysis, with no down-sampling applied to the source data.
3.2.3. Weather Data
Atmospheric parameters were obtained from the Open-Meteo Historical Weather API [
32]. This service provides hourly values for key meteorological variables, including temperature, relative humidity, mean sea-level pressure, surface pressure, and wind vectors. Data were downloaded in CSV format for the coordinates and date range of the study area.
Table 3 presents mean values of selected parameters for day, evening, night, and aggregated 16 h and 24 h periods. For each time partition, a single set of representative meteorological values was applied to all operations within that partition, and frequency-dependent atmospheric absorption was computed using those values.
3.2.4. Topographic Data
Terrain elevation data were sourced from France’s national mapping agency (IGN), which distributes digital elevation models via the Géoplateforme. The BD ALTI
® 25 m resolution digital terrain model (DTM) [
33] was downloaded in Cloud-Optimized GeoTIFF (COG) format, referenced to Lambert-93/RGF93. All IGN datasets are released under Licence Ouverte/Open Licence v2.0 (Etalab).
For this study, terrain elevations covering the analysis grid plus a 1.5 km buffer were extracted. Heights were sampled at grid-node centres using nearest-neighbour interpolation and assembled into a NORAH2-formatted ESRI ASCII raster (*.asc).
3.2.5. Land Cover Category Data
NORAH2 accounts for ground-surface impedance, which influences how sound travels and attenuates. Unlike terrain elevation, no authoritative spatial dataset exists that directly provides acoustic ground impedance. A practical approach is to infer impedance values from land-use/land-cover classifications.
The Copernicus Land Monitoring Service (CLMS) provides harmonised pan-European datasets suitable for environmental assessments. The CORINE Land Cover (CLC) 2018 raster (100 m resolution; version 2020_20u1) [
34] was used to classify terrestrial surfaces into thematic categories (e.g., artificial surfaces, agricultural areas, forests, wetlands, water bodies).
For acoustic ground modelling, CLC classes were reclassified to NORAH2 ground categories and associated with representative acoustic parameters (e.g., effective flow resistivity, ground factor) based on Annex D of the NORAH2 User Manual and the established acoustics literature. The resulting impedance field was exported as a NORAH2-compatible ESRI ASCII raster (*.asc) with grid geometry consistent with the terrain raster.
3.3. Trajectory Conditioning (Ground Track, Altitude, Speed)
Preprocessing of ADS-B/Mode-S trajectories is warranted because raw crowdsourced observations contain artefacts that can yield implausible kinematics if used as-is. Documented issues include delayed data and duplicate receptions in multi-receiver networks; occasional erroneous positions that appear as short target “jumps”; misassigned transponder identifiers; short gaps and dropouts that complicate interpolation; and variability in reported position integrity together with the widespread use of barometric (rather than geometric) altitude in position messages. Untreated, these effects can produce mismatches between reported position and velocity and propagate into unrealistic climb/descent rates and headings in reconstructed profiles [
35,
36,
37].
Although ADS-B messages may include both geometric (GNSS-derived) altitude and barometric altitude, the latter is used in this study as the primary vertical reference. This choice reflects aviation surveillance and operational standards: ICAO Annex 10 (Volume IV) [
38] defines pressure-based altitude as the standard vertical quantity for ADS-B reporting and surveillance processing. Barometric altitude therefore represents the altitude reference used by pilots, procedures, and controllers, and provides a temporally stable basis for reconstructing low-altitude flight profiles. By contrast, GNSS-derived geometric altitude is more sensitive to satellite geometry and short-scale variability as mentioned in
Section 3.2.1 3.3.1. Correction of Barometric Altitude
Barometric altimeters assume International Standard Atmosphere conditions (15 °C, 1013.25 hPa). Any departure from these reference values introduces a bias. To account for this, hourly meteorological observations (surface pressure and temperature) were obtained and corrected for each altitude sample using the standard hypsometric relation (Equation (1)) [
39]:
where
hcorrected is the corrected altitude (m),
hbaro is the barometric altitude as supplied in the source data and P is the surface pressure in hPa. Likewise, this altitude was corrected for ISA temperature deviation using the standard 0.4% per °C rule. These adjustments align the trajectories to actual atmospheric conditions at the time of each operation.
After ISA correction, residual negative samples persisted due to pad elevations close to sea level. Subsequently, trajectories were compared against a 25 m-resolution digital elevation model (DEM). Known helipad elevations (e.g., Heliport Mandelieu, Port Helipad) were used as anchors, shifting the altitude of the entire operation consistently up or down, ensuring that ground and taxi segments matched surveyed elevations. This also eliminated negative altitude artefacts.
Figure 2 presents an example comparing the barometric altitude (after ISA correction) versus the optimised altitude, where the shifting upwards made to align pad and ground elevation is evident.
3.3.2. Ground Track Optimisation
The original ground track positions describe rather smoothly the overall trajectory of the operation (see
Figure 3); however, when zoomed in (see
Figure 4), changes of direction are noticeable.
The procedure of ground track optimisation consisted of the following:
Detection of implausible movements. Successive positions were compared to identify step lengths and heading changes beyond defined physical limits. Where these occurred, the affected point was shifted to a position interpolated between its neighbours, bringing the movement within a realistic bound.
Stabilisation of low-speed manoeuvres near heliports. Operations within the heliport or helipad boundaries were treated conservatively, with only small corrections applied. This ensured that hover and taxi movements, which naturally involve low speeds and small displacements, were preserved.
Smoothing of residual artefacts. Runs of repeated positions were spread slightly to restore continuous progression, and stretches of consistent turning were refitted with arcs constrained to plausible radii (300–5000 m). Local heading discontinuities were straightened by aligning the point with the chord between its neighbours. These adjustments were restricted to metre-scale displacements.
3.3.3. Speed Optimisation
In OpenSky the “velocity” parameter represents ground speed. As a procedural step, true airspeed (TAS) is obtained by correcting ground speed for wind: the wind vector is vertically interpolated from the available levels (10, 80, 120, 180 m) to the aircraft’s barometric altitude and projected onto the ground track direction; the along-track wind component is then removed from ground speed (tailwind raises the ground speed relative to TAS; headwind lowers it).
Figure 5 illustrates the resulting separation between ground speed and TAS across headwind and tailwind segments.
The speed values provided by ADS-B showed short-term spikes and rapid fluctuations that are inconsistent with helicopter motion (see
Figure 5). These fluctuations are inherited by the resulting TAS profile. If untreated, these artefacts would result in unrealistic acceleration profiles.
To address this, speed time series were processed in three steps:
Filtering and smoothing. A moving median followed by a moving average was applied to reduce short-term fluctuations while preserving overall trends.
Acceleration limits. Step-to-step changes were checked against fixed numerical thresholds based on rotorcraft capabilities; where exceeded, values were adjusted by interpolation to remain within plausible ranges.
Distance consistency. A scale factor was applied so that the integrated distance from the smoothed speed profile matched the length of the ground track, ensuring consistency in total duration and distance.
Figure 6 contrasts the wind-corrected true airspeed (TAS; red) with the optimised profile after smoothing and distance-matching factorization (blue); the scale factor of the operation shown is 0.98.
3.4. Operation and Flight-Mode Assignment
Each operation was categorised as arrival (ARR) or departure (DEP) by comparing the first and last altitudes (first > last ⇒ ARR; otherwise, DEP). Diurnal classes were assigned from local time for Lden: Day (07:00–19:00), Evening (19:00–23:00), Night (23:00–07:00).
Per-second flight “modes” (APP, DEP, FOV, IGH, OGH, TAX, RIDL) were labelled via rule-based logic combining altitude bands, speed regimes, and proximity to pads/corridors, providing NORAH2 with context-aware segments for emission and propagation modelling.
The classification assigns one mode per second by combining altitude above ground level (AGL), ground speed, vertical trend, and proximity to heliports and published helicopter corridors. Proximity is evaluated against geofenced pad polygons assigned to each helicopter pad station (Mandelieu and Port de Cannes).
The criteria applied to assign mode is presented in
Table 4.
3.5. NORAH2 Configuration (Inputs, Rotorcraft Representation, Metrics)
The NORAH2 model requires a structured set of inputs describing project geometry, trajectories, aircraft types, environmental conditions, and output metrics. Configuration followed the consultation of the NORAH2 User Manual (D1.2) [
40] and the Rotorcraft Noise Modelling Guidance (SC01.D1.5d) [
25] for inputs, hemisphere selection, and computation workflow. The following subsections describe the configuration adopted for this study.
3.5.1. Project Coordinate Frame and Origin
All calculations were performed in a local Cartesian frame defined in Lambert-93, consistent with national terrain and land-cover datasets. The origin was set at the Cannes Heliport (Base hélicoptère de la Sécurité Civile, adjacent to Cannes–Mandelieu Airport), with coordinates X = 1,019,635 m, Y = 6,279,647 m, and Z = 3 m AMSL.
3.5.2. Trajectory Input
The Radar_Trajectories.csv file was used as an input. This file contained quality-controlled trajectories from ADS-B processing (
Section 3.2), including position, altitude, speed, and operation mode at one-second resolution. ADS-B trajectories do not provide sideslip information, and it cannot be reliably inferred from the available state vectors. The Slip column in Radar_Trajectories.csv was therefore set to 0° for all samples. This choice is recorded here as a modelling assumption; according to NORAH modelling guidance, slip angle primarily affects source directivity during near-ground manoeuvres such as hover, take-off, and landing, while its influence on single-event metrics (LAmax, SEL) for trajectory-based en-route or steadily climbing/descending segments is second-order compared to altitude, speed, and lateral distance. Its influence can be evaluated in a future validation study.
3.5.3. Rotorcraft Type and Representation
Rotorcraft were represented in NORAH2 by directional source hemispheres grouped by class. For types not explicitly catalogued, a proxy was selected on the basis of airframe configuration, number of rotors, number of blades, etc. Certification adjustments were applied using the EASA Rotorcraft Noise Database [
41]: the median certification value for each type was compared to that of the selected proxy, and the difference was applied as a regime-specific offset (take-off, overflight, approach).
Table 5 presents the rotorcraft types used in this study with their corresponding certification levels, their NORAH proxy, if not catalogued, and the derived adjustment applied. It summarises the regime-specific level adjustments applied where certification data provided sufficient information to support such corrections. In accordance with NORAH modelling guidance, offsets are only introduced when regime-resolved certification information (approach, flyover, departure) is available (Chapter 8) and can be meaningfully associated with the corresponding operational modes.
For helicopter classes for which certification data did not provide regime-specific differentiation suitable for this purpose (Chapter 11), no offsetting was applied. Where regime-resolved certification information was available, median certification levels were calculated for each operational regime and helicopter class, and relative level differences were derived. These relative differences were then applied as constant additive offsets to the corresponding NORAH hemispherical source levels. This treatment is intended as a starting point for proxy selection and level adjustment; in application-oriented studies, validation against field measurements should be used to refine the actual level differences between a modelled aircraft and its chosen proxy.
3.5.4. Metrics
The Metrics CSV defines how each noise indicator is assembled from day/evening/night energy and what time span it is normalised to (see
Table 6). Each row names a metric (e.g., LAeq24h, Lden, Ldn, Lday, Leve, Lnght, LAeq16h, SELA), specifies weighting factors for the Day, Evening, and Night partitions, and provides T0 [s], the normalisation duration in seconds for the analysis window. Weightings of 1, 3.162, 10 correspond to the standard 0 dB, +5 dB, +10 dB penalties used for evening and night in European indicators (e.g., Lden), while binary entries (1 or 0) select which partitions contribute to indicators like Lday, Leve, or Lnght. The Day/Evening/Night partitions follow the conventional 12 h/4 h/8 h split used in EU practice, and the penalties align with the Environmental Noise Directive definitions for Lden.
The T0 [s] field must equal the total covered seconds for the partitions included in that metric over the entire study period (e.g., 14 days of Day + Evening for LAeq16h; all 24 h periods for Lden). Correct T0 is critical: it is the divisor that converts accumulated acoustic energy into the period-average level; setting T0 shorter or longer than the data coverage will, respectively, inflate or dilute results. In practice, the list of metric names at the end of Case_Options.csv determines which lines from this Metrics file are applied; the engine then aggregates the time-series energy within each partition, applies the stated weightings (penalties), and normalises by T0 [s] to produce the requested indicators. This structure ensures that outputs align with established European definitions while remaining explicitly tied to the study’s actual temporal coverage.
3.5.5. Case Specific Files
Case_Radar_Operations defines the set of rotorcraft movements to be modelled. Each record specifies the ICAO type (“Heli”), the operation category (“OP”, e.g., ARR or DEP), and an operation identifier (“RTID”) that links to the corresponding 4-D trajectory. The Day/Evening/Night flags (D, E, N) indicate which temporal partitions the operation contributes to in the period metrics. When an ICAO type does not have a dedicated NORAH hemisphere, the type is cross-referenced via Substitutes.csv to a NORAH proxy before source modelling proceeds. In effect, Case_Radar_Operations is the run manifest: it declares exactly which operations are in scope, when they count, and which source descriptor (direct or proxy) applies to each.
Table 7 shows an extract of a Case_Radar_Operations.csv file.
Case_Run_Options provides the environmental and computational context for the run (
Figure 7 shows an example). It supplies the atmospheric inputs used for absorption and propagation (e.g., reference temperature, pressure and humidity), identifies the ground model through the DEM and associated sigma references (flow resistivity for ground impedance classes), and specifies the evaluation grid if not read from a separate grid file. Spatial limits are controlled with parameters such as GRDEXT (grid extension beyond the domain of interest) and DMAX (maximum propagation distance). Receiver height was specified via the HMIC parameter in the Case_Run_Options file and set to 1.2 m above local ground level, with ground elevation defined by the digital elevation model (DEM). Explicit building geometry was not included in the terrain representation; while building heights could in principle be incorporated, meaningful façade-level modelling would require grid resolutions incompatible with town-scale noise mapping and standard computational practice.
3.5.6. Outputs
NORAH2 produced two types of output:
Single-event footprints, listing calculated levels such as SEL and LAmax at receptor points (single event *.onl).
Multi-event outputs (Metrics) are delivered as both tabular point listings (.onl) and raster surfaces (.grd).
These outputs were subsequently converted to contours and exposure layers. From the grids, Lden and LAeq, 16 h contours were derived, and Number Above metrics (N60, N65, N70) were calculated. These post-processing steps produced the maps and indicators presented in
Section 4.
5. Discussion
5.1. Principal Findings and Contribution
This study demonstrates a real-case application of NORAH2 for helicopter noise mapping in an urban coastal corridor (Cannes), producing Lden and LAeq, 16 h contours and N60/N65/N70 event-count maps over a uniform receptor grid (
Figure 7,
Figure 8,
Figure 9,
Figure 10 and
Figure 11). Event-level capability was shown through a representative LAmax footprint (RTID186), illustrating how individual operations can be analysed in addition to period averages (
Figure 13). The pipeline, from ADS-B trajectories to NORAH2 inputs and mapped outputs, was executed with sample-preserving conditioning of position, altitude, and speed (
Section 3.2), yielding coherent maps and footprints suitable for scenario testing and stakeholder communication.
5.2. Temporal Scope and Interpretation of Indicators
Lden is defined for long-term exposure with evening/night penalties; LAeq, 16 h represents daytime average exposure. The present analysis covers two weeks, which is insufficient to characterise long-term exposure or to support health-impact appraisal, especially given the chosen period of data, but rather acts as a proof of concept. The contours should therefore be interpreted as demonstrative outputs of the modelling workflow for a defined window, not as statutory exposure maps. Statements on health relevance require indicators computed over standard averaging periods (annual Lden; seasonal/annual practice for LAeq, 16 h) and are out of scope for this dataset.
5.3. Suitability of OpenSky ADS-B (Context and This Dataset)
A secondary objective in this study was to determine whether OpenSky ADS-B data are suitable for producing kinematically plausible helicopter operations for NORAH2. For the two-week Cannes dataset, one-second state vectors provided sufficient signal quality to reconstruct operations once the conditioning workflow in
Section 3.2 was applied (sample-preserving adjustments to ground track, hypsometric altitude correction with terrain-clearance enforcement, and speed smoothing with distance consistency). The resulting trajectories were consistent with pad elevations, terrain, and expected rotorcraft behaviour, and yielded coherent single-event footprints and period maps.
Data noise typical of ADS-B, such as occasional position jumps, short runs of repeated positions, and small step-to-step speed fluctuations, was present but was mitigated by the stated conditioning without removing samples or altering the operation duration. Within this dataset, the processed outputs indicate fitness for purpose as inputs to NORAH2.
5.4. Methodological Choices and Uncertainties
Ground track conditioning. Horizontal positions can include plateaus, long steps, and sharp heading changes that inflate curvature and bias kinematics. Sample-preserving, metre-scale adjustments were applied in a fixed order (bounds on step length; protection of heliport manoeuvres; de-plateauing; turn smoothing; local heading repair). These constraints limited geometric drift while restoring physical plausibility; timing and sample count were unchanged. Residual uncertainty mainly concerns short manoeuvres that may be partially smoothed; however, the metre-scale limit was chosen to keep this small.
Altitude optimisation. Barometric altitudes deviate from true heights under non-ISA conditions; near sea level, small pressure/temperature biases can leave samples slightly below zero even after ISA correction. In this dataset, ISA adjustment alone produced little change in the count of negative altitudes. Anchoring each operation to surveyed pad elevation and applying a constant vertical shift per operation, followed by a ≥1 m terrain-clearance margin, removed ground-penetration artefacts. Remaining uncertainty is a small bias over complex terrain.
Speed smoothing and distance consistency. Speeds were first smoothed with short moving median and mean windows and a per-second change cap to suppress spikes while retaining trend. Thereafter, a single multiplicative factor per operation was applied so the distance derived from the optimised speed profile matches the original distance and duration of the operation.
Uncertainty arises primarily from the chosen median/mean window lengths and the per-second change cap. Windows set too long can dampen genuine acceleration/deceleration ramps; windows set too short leave residual small-scale variation that propagates to vertical flight-path angle and mode boundaries. A cap set too low clips real ramps; a cap set too high admits isolated spikes, inflating instantaneous acceleration. The distance-matching factor corrects only a small global mismatch introduced by smoothing and does not contribute to local distortions. Under these controls, residual uncertainty is confined to brief manoeuvres, and it should be small in magnitude.
Mode assignment. Noise hemispheres depend on the assigned flight mode; misclassification can therefore affect level estimates. Per-second labels (RIDL, TAX, IGH, OGH, DEP, APP, FOV) were derived from altitude above ground, ground speed, climb/descent trend, and proximity to pads/corridors. Residual uncertainty is concentrated at these boundaries: small changes in thresholds or timing can move a few seconds between modes (e.g., DEP↔FOV, APP↔IGH), with corresponding changes in hemisphere selection and local levels. The criteria used may be refined, and alternative or tighter ranges, potentially adding cues such as vertical-rate windows, could be adopted to further refine labels if validation indicates benefit. With one-second sampling, any mislabelling is short in duration.
Types without dedicated hemispheres were mapped to NORAH classes and, in this study, adjusted using regime-specific uniform level shifts derived from EASA certification medians. This represents a pragmatic starting point but reflects a central tendency rather than the specific noise characteristics of the local operating fleet; if the actual fleet is systematically quieter or noisier than the median assumed for a given class, corresponding bias may arise in predicted levels and event counts. Further refinement is possible by applying adjustments to different speed and angle ranges and, where available, by replacing class medians with aircraft- or type-specific certification data derived from tail-number information. In all cases, validation against measured noise levels remains necessary to refine the effective level differences between a modelled aircraft and its chosen proxy.
5.5. Verification and Validation
In aircraft noise studies, credibility rests on validation, demonstrating agreement between model predictions and independent measurements for the intended use. This is standard practice in the principal frameworks (ECAC Doc 29; FAA AEDT technical guidance), which emphasise precautions to ensure results are valid and comparable to established methods [
11,
12].
Verification and validation are not unique to rotorcraft. In established frameworks (ECAC Doc 29; FAA AEDT [
42]; UK CAA ANCON [
43]; EUROCONTROL IMPACT [
17]; EASA ANP support [
44]), validation against independent measurements is a standard requirement for credible aircraft-noise modelling. Its purpose is to show that model configuration and assumptions reproduce observations within stated tolerances and, when they do not, to support revised assumptions and re-test cycles.
Procedurally, NORAH2 requires the same validation discipline as fixed-wing models: compare predictions against measurements for representative operations and periods; if differences exceed agreed tolerances, documented assumptions must be revised (e.g., flight-procedure parameters, proxy level offsets) and re-tested until alignment is achieved. Guidance under Doc 29 explicitly points to comparing calculated and measured levels and adjusting procedure assumptions where warranted, an approach directly transferable here.
However, validation alone does not confer exposure relevance when the time period is not aligned with reporting norms. ECAC/European practice treats Lden as a long-term indicator (annual basis), and UK practice typically reports LAeq, 16 h for a defined summer period; by contrast, the present analysis covers two weeks. Consequently, even if validated, the average outputs presented in this study (Lden and LAeq, 16 h) remain demonstrative by-products of the NORAH2 workflow, useful to show feasibility, mapping behaviour, and event-level capability, not statutory exposure findings or a basis for health appraisal. Future, decision-oriented studies would need both validation and time period consistent with reporting conventions before results are interpreted for policy or health contexts.
In practical terms, verification of helicopter noise modelling in an urban context would typically rely on a combination of short- and long-term noise measurements at representative locations beneath flight corridors and near heliports. Long-term monitoring provides context on background sound levels and temporal patterns, while targeted short-term measurements during known helicopter operations enable direct comparison of predicted single-event metrics (e.g., LAmax, SEL) with observed values.
Such measurement campaigns are routinely used in fixed-wing aviation noise studies to validate both event-level predictions and aggregated indicators, and the same principles apply to rotorcraft. In this framework, discrepancies between measured and modelled levels are used to refine operational assumptions (e.g., flight profiles, mode assignment, proxy selection, or level offsets) rather than to invalidate the modelling approach itself. While no field measurements were available for the present demonstration, the modelling workflow presented here is explicitly designed to support this type of verification in future, decision-oriented applications.
5.6. Policy and Health Context
Environmental noise is recognised as a risk factor for adverse health outcomes, beyond auditory effects such as hearing loss. The epidemiological and public health literature shows that chronic exposure to environmental noise can contribute to annoyance, sleep disturbance, cardiovascular diseases (e.g., hypertension, ischemic heart disease), metabolic stress, and cognitive impacts in exposed populations [
45], and is therefore treated as a public-health concern in European policy frameworks [
46].
Aviation noise, as a specific component of transportation noise, has been associated with similar health effects. Large reviews and meta-analyses report relationships between aircraft noise exposure and increased risk of sleep disturbance, community annoyance, and cardiovascular morbidity, including elevations in blood pressure and ischaemic heart disease risk with long-term exposure above guideline levels [
47]. These associations underline why organisations such as the World Health Organisation include aircraft noise in environmental noise guidelines aimed at protecting public health [
46,
47].
Helicopter noise shares many of the acoustic and exposure characteristics of other aviation noise sources (e.g., high maximum levels, variable temporal patterns) but often occurs in dense urban or coastal environments at lower overall traffic volumes. This can result in frequent, high-amplitude events that are not well-captured by traditional annual average indicators alone yet may contribute disproportionately to community annoyance and sleep disturbance. By producing both energy-averaged and event-based indicators, modelling tools can thus better characterise the aspects of helicopter noise that are most relevant to community response and potential health impacts, supporting more informed policy and management decisions.
5.7. Application to Urban Noise Management (Policy and Stakeholder Relevance)
Regulatory alignment and gap-filling. NORAH2 produces standard indicators (e.g., Lden; LAeq, 16 h, Lday, Levening, Lnight, …) consistent with European practice, enabling integration with strategic mapping carried out under the Environmental Noise Directive and CNOSSOS-EU methods. Helicopter activity often sits below statutory mapping thresholds; NORAH2 enables route-level (corridor) mapping of helicopter operations and provides event-count indicators (N60/N65/N70: number of events exceeding 60/65/70 dB(A) at each location), capturing the frequency and intermittency of rotorcraft noise that are not well-represented by annual, airport-centric indicators.
Transparency and communication. Metric outputs and single-event footprints provide plain-language artefacts for public portals and consultation: maps of typical daytime exposure (LAeq, 16 h), long-term indicators (Lden), and frequency maps (NAxx). Publishing inputs (data sources, time windows, processing rules) alongside outputs improves auditability and trust, supporting work by metropolitan observatories and associations and municipal “noise watch” programmes.
Operational scenario testing. The segment/mode framework allows forecast studies prior to policy changes: route alignment or altitude profiles, use of specific pads/heliports, time-of-day restrictions, holding locations, and fleet-mix evolution (including eVTOL). Outputs quantify relative changes in exposure bands and event frequency, supporting proportionate mitigation design and stakeholder negotiation.
Integration with other sources. NORAH2 outputs can be assembled with road/rail/industry layers used in strategic mapping. Event-count layers remain complementary and should be reported alongside energy-average layers. This supports area-wide planning where multiple sources interact (e.g., coastal roads plus rotorcraft corridors).
Complaint handling and monitoring. Single-event footprints and Nxx layers provide a structured way to relate reported disturbances to operations, identify hotspots, and prioritise responses.
Governance and reproducibility. The pipeline is implementable with open data (traffic, terrain, land cover) and documented processing, enabling reproducible analyses that can be re-run for new periods or scenarios. This reduces barriers to adoption by local authorities and environmental noise stakeholders while keeping methods consistent with European reporting practice.
6. Conclusions
This study demonstrates a practical, reproducible workflow for mapping helicopter noise in an urban corridor with NORAH2, driven by OpenSky ADS-B trajectories and public geospatial layers. Outputs comprise Lden, LAeq, 16 h, and event-count metrics (N60/N65/N70) on a uniform grid, complemented by single-event footprints that resolve individual operations.
A process has been described whereby ADS-B data for relevant helicopter operations can be transformed into representative trajectory data for noise modelling purposes.
For local authorities and urban noise stakeholders, the approach enables corridor-level mapping where rotorcraft activity concentrates, provides event-count indicators that reflect frequency and intermittency alongside energy averages, and supports scenario exploration (routes, pad use, time-of-day, fleet mix) with transparent inputs and settings suitable for communication with residents and stakeholders. When configured over standard periods, results can be integrated with strategic mapping under European practice.
Extending coverage to seasonal/annual periods and undertaking measurement-based validation would be the natural next steps for policy applications.