Abstract
JASCO Applied Sciences undertook a sound source characterisation study of the Hywind Tampen floating offshore wind farm, approximately 140 km northwest of Bergen, Norway. Four recording instruments were deployed from the DOF Skandi Iceman to the seabed by remotely operated vehicle in February 2024 at various positions both within and around the wind farm site. Two instruments were set up in a single-hydrophone omnidirectional configuration and two with arrays of four hydrophones to provide directional noise discrimination. Three of the four recorders (one directional system and the two omnidirectional systems) were retrieved in June 2024. The remaining directional recording system remains in situ on a 12-month recording schedule and is expected to be retrieved in early 2025. Recording was conducted at a 64 kHz sample rate with 24-bit resolution, and the total volume of data collected from the first three stations was 7.3 TB.
Analysis of the recorded data was undertaken to determine the characteristics of the sound produced by the turbines at Hywind Tampen and compare this to similar sound source characterisation study of the Hywind Scotland floating system. The dominant sound emissions from the Hywind Tampen turbines are narrowband tones, principally below 200 Hz, with two notable tones at around 25 and 75 Hz being the primary contributors to the recorded sound spectra. The sources of these tones are directly related to the rotational rate of the rotor and the number of magnetic pole pairs in the generator, one directly and the other by a factor of three. Consequently, the actual frequencies generated by the rotating components of each turbine depend on the rotor RPM at any given time, where the strong tones at around 25 and 75 Hz are the frequency limits of the rotor related tones associated with the system maximum RPM. Other tones which were variously stable and unstable, continuous and intermittent were found to contribute to the spectra to a lesser extent.
Positive correlations were determined between sound levels in the frequency bands containing the 25 and 75 Hz tones and wind speed, the strength of this correlation reducing with distance from the turbines. The correlation analysis also displayed an approximate plateau of the sound levels in these bands at wind speeds above 20 kn (10.3 m/s). Similar positive correlation was observed between levels in these bands and rotor RPM, with the relationship reflecting the increase in frequency of the fundamental and triplet tones with increasing rotor RPM. Other low-level tones were also observable in the spectra, though at considerably lower intensity than those from the generator. Some of these tones were stable and intermittent, characteristic of pumps or motors under irregular operation, while others were continuous and unstable, possibly indicative of blade control systems responding to fluctuations in wind speed.
The acoustic data were manually analysed for impulses and transients caused by tension on the mooring system, which were a frequent component of the Hywind Scotland recordings. Only a handful of possible mooring transients were identified at higher wind speeds and further directional assessment found these to originate to the East of the Tampen site and they were subsequently dismissed as mooring noise. This confirmed a fundamental acoustic difference in noise signature between Hywind Scotland and Hywind Tampen, the reasons for which may be related to differences in the buoyancy of the spar structure, the mooring system, or both, rather than difference swell and other environmental factors. The exact mechanism generating the transient noises at Hywind Scotland is unknown, but this mechanism appears to no longer exist in the new substructure and mooring system utilised at Hywind Tampen. A quantitative analysis of the impulsiveness of the data from all three recorders was undertaken by assessing empirical distribution function of the one-minute kurtosis, which also confirmed that kurtosis at all recorder locations was very low, indicating a nonimpulsive soundscape. Based on this, daily cumulative SELs recorded at each station, ranging 717 m to 9.35 km to the nearest turbine, were compared to non-impulsive impact criteria from Southall et al. (2019). All daily cumulative SELs recorded during this study, at all stations, were found to lie below the thresholds for both temporary and permanent hearing threshold shifts (i.e., hearing loss) for nonimpulsive sounds for all functional hearing groups.
The recorder with directional capabilities was positioned within the wind farm in such a position to isolate the most southeasterly turbine in the array (HY06). Received levels from the direction of HY06 collected during relatively stable periods of wind were analysed for bins in 5 kn (2.6 m/s) increments from 5 to 40 kn (2.6 to 20.6 m/s) and then backpropagated to obtain statistical decidecade source levels for a single turbine operating under multiple wind speed conditions. Median broadband source levels ranged between 156.5–163.8 dB re 1 µPa²m², while 95th percentile broadband source levels ranged between 159.1–168.7 dB re 1 µPa²m². Source levels were noticeably lower for wind speeds below 20 kn (10.3 m/s), corresponding to the turbine operating at less than its maximum rotor RPM. Using the backpropagated source levels, a simple point source model was used to model the footprint of all eleven wind turbines. The largest modelled distance was 60 m for very high-frequency cetaceans (20 kn, 95th percentile) assuming an animal remains within this radius for a full 24-hour period at the depth of the greatest sound level.
The recorded spectra from the Tampen WTGs displayed very similar tonal features to the turbines at Hywind Scotland. Estimated source levels for Hywind Tampen are somewhat lower than those determined for Hywind Scotland, and this may be a consequence of the concrete Tampen substructure being less able to transfer vibrational energy combined with the lack of mooring transients elevating the overall noise signature.