Calculation of annual aircraft noise exposure for Geneva and Zurich airports with the next-generation program sonAIR – first results Stefan Schalcher 1 Empa, Swiss Federal Laboratories for Materials Science and Technology Überlandstrasse 129, 8600 Dübendorf, Switzerland Christoph Zellmann 2 n-Sphere AG Räffelstrasse 29, 8045 Zürich, Switzerland Jonas Meister Empa, Swiss Federal Laboratories for Materials Science and Technology Überlandstrasse 129, 8600 Dübendorf, Switzerland Jean Marc Wunderli Empa, Swiss Federal Laboratories for Materials Science and Technology Überlandstrasse 129, 8600 Dübendorf, Switzerland Beat Schäffer Empa, Swiss Federal Laboratories for Materials Science and Technology Überlandstrasse 129, 8600 Dübendorf, Switzerland

ABSTRACT The next-generation aircraft noise simulation program sonAIR was designed to precisely predict single flights with the scope of investigating and optimizing noise abatement flight procedures. Recently, sonAIR was implemented in a geographic information system (GIS) to make it suitable for noise mapping calculations of entire airport scenarios. In this study, we conducted first calculations of annual aircraft noise exposure with sonAIR for Geneva and Zurich airports, Switzerland, for the year 2017, simulating several 10'000 single flights and accounting for real conditions in great detail (e.g., detailed emission modelling and considering ground cover in the sound propagation calculation). Our calculations proved the ability of sonAIR to do annual aircraft noise calculations with many simulated flights. Comparing the results to those obtained with the best-practice program FLULA2, which is currently used for official aircraft noise calculations in Switzerland, we found that the noise contours differ on average little in areas with legally relevant noise exposure. However, locally larger differences may occur, primary due to varying ground cover and more detailed sound emission modelling. In this contribution, we present the process of sonAIR for annual aircraft noise calculations, show the most important results and discuss implications for future aircraft noise calculations around airports. 1 stefan.schalcher@empa.ch 2 christoph.zellmann@n-sphere.ch

a Shea mar ce 21-24 AUGUST SCOTTISH BENT caso

1. INTRODUCTION

The next-generation aircraft noise simulation model sonAIR has been designed to precisely predict single flights with the scope of investigating and optimizing noise abatement flight procedures. A rigorous validation exercise with noise exposure measurements of roughly 20'000 flights around Geneva (GVA) and Zurich (ZRH) airports, Switzerland [1] demonstrated that sonAIR is highly accurate. This result was confirmed in a follow-up study [2], where calculation results obtained with sonAIR were compared to those of the best-practice programs FLULA2 [3] and AEDT [4] regarding their potential to reproduce the noise exposure of single flights close to and far from the airports.

Recently, sonAIR was implemented in a geographic information system (GIS), namely in the Esri ArcGIS environment, to make it suitable for noise mapping of entire airport scenarios (yearly air traffic scenarios). However, the feasibility and applicability of sonAIR for yearly air traffic calculations still needs to be verified.

In this study, we conducted first calculations of annual aircraft noise exposure with sonAIR for ZRH and GVA for the year 2017, simulating several 10'000 single flights and accounting for real conditions in great detail. The results of the sonAIR calculations are compared with the results obtained with FLULA2, which is one of three programs in Switzerland currently recommended by the Federal Office for the Environment (FOEN) for official aircraft noise calculations [5]. Besides proving the ability of sonAIR to do yearly aircraft noise calculations, this study thus provides insights on the comparability of sonAIR and FLULA2 for the calculation of annual aircraft noise exposure. 2. METHOD

2.1. Main characteristics of and differences between sonAIR and FLULA2

A detailed overview on the most important characteristics and differences of the models sonAIR and FLULA2 can be found in [2]. Hereinafter, the major differences are discussed.

sonAIR is a spectral (third-octave bands) time step noise calculation program with three- dimensional sound emission [6, 7]. It uses separately emission models for engine and airframe noise and can take flight configuration, i.e., airplane configuration plus flight parameters such as power setting, into account if the respective information is available. In this study, however, since the flight trajectories are derived from radar data for annual aircraft noise calculations, only N1 (relative rotational speed of the low pressure compressor) was used as an input besides the 3D position and speed of the aircraft. N1 was estimated from the radar data using a parameter estimation method developed at Empa [8]. Sound propagation is calculated separately from emission with the physically based model sonX, accounting for the following effects: geometric divergence, atmospheric absorption, ground reflections, shielding by terrain, shielding and reflections of buildings (not applied here), and forest and rock reflections (not applied here) [9]. sonX uses terrain information (elevation) and land-use cover for the sound propagation calculation. FLULA2 is a best-practice model [3], where sound emission and propagation are combined in an empirical model describing the noise exposure at receiver positions on the ground as a function of the emission angle and distance between source and receiver. Yielding A-weighted noise metrics such as the L AE , L AS,max and L Aeq , FLULA2 only considers the flight operation (departure with maximum or reduced power plus cutback, and approach), besides 3D position and speed. Sound propagation accounts for geometric divergence, atmospheric absorption, shielding by terrain and lateral attenuation, i.e., excess attenuation in situations with grazing sound incidence comprising ground and meteorological effects [10]. FLULA2 also uses terrain information, but models the entire area around the airport as soft ground (grassland). Table 1 gives an overview of the main differences between the models sonAIR and FLULA2.

Table 1: Overview with the main differences between sonAIR and FLULA2.

Feature sonAIR FLULA2

Model Semi-empirical time-step model,

Empirical, time-step model, combined sound emission and propagation, A-weighted sound level

separate sound emission and propagation, spectral calculation

Emission model directivity 3D (longitudinal and lateral) 2D (longitudinal)

Sound propagation geometric divergence, atmospheric absorption, ground reflections, shielding by terrain; (shielding and reflections of buildings,

geometric divergence, atmospheric absorption, shielding by terrain, lateral attenuation

forest and rock reflections) 1

Flight trajectories 3D position, speed, N1 (+ airplane configuration if information

3D position, speed

is available) 1

Terrain data elevation and ground cover elevation; soft ground

1 not applied here 2.2. Calculation of annual aircraft noise exposure with sonAIR - workflow

This section summarizes the main steps to calculate annual aircraft noise exposure with sonAIR. Calculations for the two airports are done for the year 2017. Note that the calculations are done for large commercial aircraft (maximum take-off weight MTOW > 8'618 kg), while general aviation is not accounted for. However, the contribution of the latter to the total aircraft noise exposure is small.

1) Preparation of input data The inputs provided by the airports need to be processed. They consist of radar data from air traffic control of all the flights of one year, as well as lists of the movements. First, the list of movements is processed by (i) identifying each aircraft by its tail number, and (ii) generating the statistics of movements (number of movements per aircraft type/group, air route, and period of day). Second, radar data are inspected for physically plausible aircraft movements, smoothed, and extended to the runway. These two steps are the same for sonAIR and FLULA2. However, as sonAIR additionally needs N1 as a mandatory input, these data are estimated from the geometries of the processed flight trajectories using a method developed by Empa [8]. In the same step, a statistical selection of maximum 200 flight trajectories per aircraft type, air route and time periods was taken. This results in some 43'000 and 54'000 flight trajectories for GVA and ZRH, respectively. The time periods are defined according to Swiss legislation (Swiss Noise Abatement Ordinance (NAO) [11]): daytime (06–22 h), first (22–23 h), second (23–24 h) and last (05–06 h) hour of the night. (There is a flight ban from 24–05 h at Swiss airports.)

2) Project set-up in the GIS environment First, the calculation project is set up in the Esri ArcGIS environment. For the calculation of the annual aircraft noise exposure, the relevant area of the digital terrain (source: Federal Office of Topography (swisstopo), swissALTI3D LV95) with 25 m × 25 m resolution, and ground cover data (source: swisstopo, swissTLM3D) with 25 m × 25 m resolution is imported. Second, the runway system is set up using coordinates of the runway endpoints and designators. Third, as the calculations are complex and time consuming, the area is divided into calculation tiles to enable parallel computing. Here, tiles were set to 4 km × 4 km for GVA and 6 km × 6 km for ZRH. Forth, receiver points are automatically created with a resolution of 150 m × 150 m (a detail prescribed by

the Swiss guidelines for aircraft noise calculations [5]). Finally, the processed flight trajectories are imported. Table 2 gives an overview of key figures for the calculations performed here. Table 2: Key figures of the setup of the calculation projects for GVA 2017 and ZRH 2017 in the GIS environment.

Aspect GVA 2017 ZRH 2017

Calculation area (km 2 ) 3'078 7'812

Resolution digital terrain/ground cover (m) 25 25

Number of calculation tiles 210 240

Tiles dimension (km) 4 x 4 6 x 6

Number of receiver points 136'800 347'200

Number of flight trajectories 43'037 54'153

3) Generating cell grid points and pre-calculation of attenuation After importing the flight trajectories, the airspace is subdivided into cells. As aircraft usually fly along predefined air routes, they mostly pass similar spatial locations throughout the entire year. Consequently, it is neither necessary to calculate the same point-to-point sound propagation paths over and over again, nor to calculate all possible propagation paths within the calculation area. Therefore, for performance reasons, the attenuation is pre-calculated for the cell corners of all cells that are being crossed by aircraft flights. The concept of this attenuation database is described in detail in [7]. In the subsequent single flight simulation (see below), the attenuations of the corners of the cell that is currently flown through are looked-up in the database, and a weighted average is calculated depending on the aircraft position within the cell. To further increase calculation performance, the size of the cells is defined dynamically depending on the distance to the ground and horizontal distance to the receiver positions [7].

As the sound propagation calculation is demanding and time consuming, the calculation is executed with a hybrid modelling approach. In situations with aircraft high above ground and fully unshielded propagation paths, a simplified "on the fly" propagation calculation is applied during the subsequent single flight simulation, and no cell grid points are created. For all other situations, the attenuation database is calculated. The simplified model is applied if the aircraft is observed from the receiver in a viewing angle greater than 15°above the horizon or obstacles [7].

The resulting attenuation database contains nearly 4 million cell grid points for GVA and over 6 million for ZRH. To make the large computational effort possible, the calculations are parallelized using the predefined calculation tiles (see above). Note that the pre-calculation of the attenuation database is an initial task which has to be performed only once. It can be used for future calculations and be incrementally extended with new cell grid points which were additionally flown through by aircraft in the corresponding year.

4) Single flight simulation After the attenuation calculation, the noise footprints are defined. Footprints are grid files containing the mean sound exposure ( L AE or L AS,max ) of many flights of a certain aircraft/engine type, flight route and operating time according to NAO, at all receiver points arranged in a grid, energetically averaged over all flights, separately for each receiver location, scaled to one aircraft movement of a specific aircraft/engine type along a specific flight path during a specific period of day, and referred to 1 sec (for the L AE ).

To calculate the footprints, single flight simulations are carried out for all imported flight trajectories (grouped per aircraft/engine type, air route and period of day as defined above). A time- step approach is applied as in FLULA2, where the sound source model is moved along the flight trajectory in discrete time steps.

5) Substitution For the annual aircraft noise calculations, sonAIR requires emission models for all aircraft types given by the movement list. However, as the number of emission models available in sonAIR is limited (as in most calculation programs, including FLULA2), not all aircraft types can be simulated with the corresponding emission model. To account also for aircraft types with missing emission model, a substitution method is applied. The substitution is based on the approach proposed by ECAC.CEAC Doc 29 Vol. 1 [12]. Using this approach, aircraft types without emission model, referred in the following as "missing types", are assigned to a similar aircraft type with available emission model, referred as "proxy type". When grouping aircraft types, the differences in noise emission levels (based on official certification noise levels) between missing and proxy types are taken into account and corrected in the following superposition.

6) Superposition In a last step, the aircraft type, flight route, and time-of-day specific footprints are weighted with the number of aircraft movements, taken from the statistics of movements (see above Step 1), energetically summed up, and averaged over the length of the time period of interest. As a result, one obtains the total annual aircraft noise exposure (A-weighted equivalent continuous sound pressure level, L Aeq ) for all operating times of interest. Here the periods according to NAO were calculated, namely, daytime from 06–22 h as well as first (22–23 h), second (23–24 h) and last hour of the night (05–06 h; usually negligible). (Note that the NAO defines rating sound levels, Lr , but in the case of the noise of large aircraft, these correspond to the L Aeq [11].)

3. RESULTS

In the following, the calculations of sonAIR are compared with existing calculations of FLULA2. To that aim, a large-scale and close-range comparison are presented.

3.1. Large-scale comparison

Figure 1 shows the noise contours of the L Aeq ( L Aeq value of 53 dB, L Aeq values of 55-70 dB in steps of 5 dB) from 06–22 h (yearly average) obtained by sonAIR in comparison with FLULA2.

The noise contours obtained with sonAIR and FLULA2 generally agree quite well. However, one can observe that in the close range to the airports, sonAIR mostly yields lower noise exposure, while far away from the airports, sonAIR mostly yields higher noise exposure.

A more detailed analysis (not shown) of the noise maps along with radar data of the major departure and approach routes revealed that the approaches seem to be the reason for the higher noise exposure values obtained with sonAIR. Higher exposures with sonAIR appear in areas far away from the airports where approaches dominate the total noise exposure (e.g., approaches from south on runway 34 of ZRH). In contrast, in areas (mostly closer to the airports) where departures dominate the noise exposure (e.g. departures from runways 16 and 28 of ZRH), sonAIR tends to yield lower noise exposure. These observations are supported by a previous comparison of sonAIR with FLULA2 [2]. The study revealed that FLULA2 tends to underestimate the noise exposure ( L AE and L AS,max ) of single flights for approaches far away from touch down, particularly for the wide- body aircraft types. The reason behind this is that FLULA2 models the final approach configuration in calculations for the entire approach, and thus underestimates the airframe noise in these regions due to higher airspeed far away from touchdown. sonAIR, in contrast, by modelling airframe and engine noise separately and dynamically over flight trajectories, can account for different flight configurations in detail throughout the whole flight.

Figure 1 further shows that the noise contours of FLULA2 are much smoother than those of sonAIR, which is explained (i) by dynamic sound emission prediction in sonAIR and (ii) in particular also by the ground cover taken into account. This is further elaborated on in the next section.

Figure 1: Aircraft noise exposure 2017 obtained with sonAIR and FLULA2 for large aircraft (MTOW > 8'618 kg) for the day (06–22h) at ZRH (left) and GVA (right). Table 2 summarizes the differences in the noise exposure presented in Figure 1, for areas within, above and below the legally relevant noise exposure range according to NAO. This range corresponds to L Aeq of 53–70 dB during the day, L Aeq of 43–65 dB during the first hour of the night, L Aeq of 43–60 dB during the second hour of the night, with the lower values corresponding to the strictest noise exposure limits according to NAO, and the higher values to exposures which mostly lie within the airports' premises where no residents live.

In the legally relevant noise exposure range, the mean difference across both airports and all time periods varies from –0.51 to +0.01 dB, with sonAIR tending to yield lower exposure values (Table 2). In contrast, in areas far away from the airports, with exposures below 53 dB and 43 dB during the day and the night, respectively, sonAIR tends to yield higher exposure values (Table 2).

Table 2: Mean differences (in dB) of the L Aeq per airport (GVA and ZRH), sound exposure range (above, within and below legally relevant noise exposure) and time periods (day: 06–22h, first hour of the night: 22–23 h, second hour of the night: 23–24 h) according to NAO [11].

Exposure range GVA L Aeq day (06–22 h)

GVA L Aeq night (22–23 h)

GVA L Aeq night (23–24 h)

ZRH L Aeq day (06–22 h)

ZRH L Aeq night (22–23 h)

ZRH L Aeq night (23–24 h)

Legally relevant range

–0.36 -0.11 +0.01 –0.37 -0.51 -0.34

(NAO [11]) 1

Above Legally relevant range

–0.88 -0.92 -2.28 –0.87 -0.43 -0.25

(NAO [11]) 2

Below Legally relevant range

0.00 +0.40 +1.07 +0.06 +0.53 -0.97

(NAO [11]) 3

1 Day (06–22 h): L Aeq = 53–70 dB; night (22–23 h): L Aeq = 43–65 dB; night (23–24 h): L Aeq = 43–60 dB 2 Day (06–22 h): L Aeq > 70 dB; night (22–23 h): L Aeq > 65 dB; night (23–24 h): L Aeq > 60 dB 3 Day (06–22 h): L Aeq < 53 dB; night (22–23 h): L Aeq < 43 dB; night (23–24 h): L Aeq < 43 dB

—— > FLULA2 sonAlR 0 25 Skm Lois

In summary, results (here, noise contours of yearly air operations) obtained with sonAIR and FLULA2 generally agree well, particularly in the legally relevant noise exposure range. sonAIR yields higher noise exposure values in areas far from the airports than FLULA2 (below the legally relevant exposure range), but the results obtained with sonAIR seem more plausible.

3.2. Close-range comparison

Figure 2 shows an exemplary footprint ( L AE ) for one departure route from runway 16 to the west of ZRH for the aircraft typ Airbus A343. The noise contours are displayed in steps of 5 dB, for L AE values of 50–70 dB. Additionally, Figure 2 displays an exemplary flight trajectory (red line) and the ground cover which sonAIR uses in the sound propagation calculation (ground effects).

Figure 2 reveals the major differences between both simulations and the influence of the ground cover as additional input in the calculation on the noise contours, as already observed in Figure 1. First, sonAIR tends to yield lower L AE values closer to the airport. sonAIR takes into account the power setting (namely, N1) in its three-dimensional emission models, while FLULA2 uses fixed two-dimensional directivity patterns. Second, while the footprints obtained with FLULA2 are smooth, those of sonAIR are quite uneven. sonAIR uses the detailed ground cover data to model sound propagation effects. While areas with forests lead to strong sound attenuation, hard ground such as settlements (sealed land) and water bodies favor propagation (smaller sound propagation attenuation). In contrast, FLULA2 assumes uniform grassland as ground cover. Besides these effects, also the methodology to obtain noise contours from grid files might play a role, with the algorithm of FLULA2 yielding quite smooth noise contours.

Figure 2: Footprint ( L AE ) of the departure route E16 to the west from runway 16 of ZRH for the Airbus A343 for the day (06–22 h) and ground cover data (source: Federal Office of Topography, swisstopo).

Ground cover

Overall, with the more detailed emission and propagation modelling including ground cover, sonAIR is able to calculate the noise exposure in greater detail and more accurately than FLULA2. This is a particular strength, as it allows accounting also for single flights and noise abatement flight procedures in great detail. Nevertheless, for yearly noise exposure calculations, also FLULA2 yields sufficiently accurate results, as previous validations (comparisons with noise measurements) have shown, either based on single flights [2] or on yearly noise exposures [13].

4. CONCLUSIONS

In this contribution, we described the process of the next-generation program sonAIR for annual aircraft noise calculations. We presented the major results of the first annual aircraft noise calculations for GVA and ZRH for the year 2017 and compared them with the best-practice program FLULA2 currently used in Switzerland for official aircraft noise calculations. Thus, the study confirmed the feasibility and applicability of sonAIR for yearly air traffic calculations.

A workflow was established for sonAIR. It is similar to the one of FLULA2, which has been proven in numerous projects to be an efficient and accurate calculation approach. However, sonAIR requires additional working steps due to the greater level of detail of the calculations compared to FLULA2. The sonAIR work process includes the time-step simulation of single flights as well as the "footprint" concept. Weighting the footprints with the corresponding number of movements and energetically summing them up, period-of-day-specific noise contours (yearly average L Aeq ) are obtained. One major difference between FLULA2 (as well as other best-practice programs) and sonAIR is that FLULA2 combines the sound emission and propagation calculation and repeats this calculation for every considered flight, while sonAIR models sound emission and propagation separately and pre-calculates an attenuation database to be used as a lookup table in the subsequent single flight simulations.

In summary, the results obtained with sonAIR and FLULA2 agree well, particularly in the legally relevant noise exposure range. Nevertheless, sonAIR tends to yield somewhat lower noise exposure values than FLULA2 in legally relevant areas according to NAO (–0.4 dB on average during the day), but higher values far from the airports in areas with low noise exposures. Further, the contours obtained with sonAIR are less smooth than those of FLULA2. Detailed analyses of the results suggest that sonAIR yields more plausible results. Taking the power setting into account, separating engine and airframe noise, and considering ground cover types, sonAIR reproduces the real conditions and resulting noise exposure more accurately. Nevertheless, for yearly noise exposure calculations, also FLULA2 yields sufficiently accurate results.

In conclusion, sonAIR is a suitable candidate to substitute FLULA2 in future and become the new official calculation program in Switzerland. Prior to this, however, further calculations and sensitivity analyses are necessary to gain experience and to gain insights on the effects of, e.g., buildings, meteorological conditions and/or ground cover on calculation results. Work on these aspects are currently being performed. 5. ACKNOWLEDGEMENTS

The authors would like to thank Geneva and Zurich airports for permission to use and present the data for this study. Further, the funding by Aéroport International de Genève, Flughafen Zürich AG and Amt für Mobilität Kanton Zürich is gratefully acknowledged (assignment numbers 5214.020920/5214.025794). 6. REFERENCES

1. D. Jäger, C. Zellmann, F. Schlatter, J. M. Wunderli: Validation of the sonAIR aircraft noise simulation model . Noise Mapping 2020). 2. J. Meister, S. Schalcher, J.-M. Wunderli, D. Jäger, C. Zellmann, B. Schäffer: Comparison of the Aircraft Noise Calculation Programs sonAIR, FLULA2 and AEDT with Noise Measurements of Single Flights . Aerospace 8 (2021) 388.

3. S. Pietrzko, R. F. Hofmann: Prediction of A-weighted aircraft noise based on measured directivity patterns . Applied Acoustics 23 (1988) 29-44. 4. U. S. Departement of Transportation Volpe National Transportation Systems: Aviation Environmental Design Tool (AEDT) Technical Manual, Version 3d, 2021. 5. BAFU, BAZL, GS-VBS: Leitfaden zur Fluglärmermittlung. Vorgaben für die Lärmermittlung. Umwelt-Vollzug Nr. 1625. Bundesamt für Umwelt (BAFU), Bundesamt für Zivilluftfahrt (BAZL), Generalsekretariat des Eidg. Departementes für Verteidigung, Bevölkerungsschutz und Sport VBS (GS VBS), Bern, 2016, available online at http://www.bafu.admin.ch/bafu/de/home/themen/laerm/publikationen- studien/publikationen/leitfaden-fluglaerm.html . 6. C. Zellmann, B. Schäffer, J. M. Wunderli, U. Isermann, C. O. Paschereit: Aircraft Noise Emission Model Accounting for Aircraft Flight Parameters . Journal of Aircraft 55 (2017) 682-695. 7. J. M. Wunderli, C. Zellmann, M. Köpfli, M. Habermacher, O. Schwab, F. Schlatter, B. Schäffer: sonAIR - a GIS-Integrated Spectral Aircraft Noise Simulation Tool for Single Flight Prediction and Noise Mapping . Acta Acustica united with Acustica 104 (2018) 440- 451. 8. O. Schwab, C. Zellmann: Estimation of Flight-Phase-Specific Jet Aircraft Parameters for Noise Simulations . Journal of Aircraft 57 (2020). 9. Empa: Dokumentation des sonX Ausbreitungsmodells. Programmversion: sonRAIL V6.0.0 bzw. sonARMS V5.0.0 bzw. sonAIR V3.0.0 (vom 21. März 2019) . Bericht Nr. 5211.00687.100.14. Empa, Abteilung Akustik / Lärmminderung, Dübendorf, 2019. 10. Empa: FLULA2, Ein Verfahren zur Berechnung und Darstellung der Fluglärmbelastung. Technische Programm-Dokumentation (FLULA2, a Method for the Calculation and Illustration of Aircraft Noise Exposure. Technical Program Documentation). Version 2.1. Empa, Swiss Federal Laboratories for Materials Science and Technology, Laboratory for Acoustics / Noise Control, Dübendorf, 2005. 11. NAO: Lärmschutz-Verordnung (LSV) vom 15. Dezember 1986 (Stand am 7. Mai 2019) (Noise Abatement Ordinance (NAO) from 15 December 1986 (Effective 7 May 201')). SR 814.41, 1986. Available from: http://www.admin.ch/ch/d/sr/8/814.41.de.pdf (last viewed on 04/29/2022). 12. ECAC: ECAC.CEAC Doc 29: Report on Standard Method of Computing Noise Contours around Civil Airports, Volume 1: Applications Guide. 4th ed. European Civil Aviation Conference (ECAC), Neuilly-sur-Seine, France, 2016. Available from: https://www.ecac- ceac.org/documents/ecac-documents-and-international-agreements (last viewed on 01/19/2022). 13. B. Schäffer, R. Bütikofer, S. Plüss, G. Thomann: Aircraft noise: accounting for changes in air traffic with time of day . Journal of the Acoustical Society of America 129 (2011) 185- 199.