No need to shout? Harbor porpoises (Phocoena phocoena) echolocate quietly in confined murky waters of the Wadden Sea

Porpoises rely on echolocation to find prey, navigate, and communicate (Clausen et al., 2010; Verfuß et al., 2005; Verfuß et al., 2009). They strongly depend on utilizing echolocation for catching numerous very small prey items using their biosonar as the main source of target information (Wisniewska et al., 2016). The species occurs in a wide range of coastal habitats ranging from fresh water influenced areas of the Baltic Sea [e.g., Benke et al. (2014)] to deeper areas off of the continental shelf (Sveegaard et al., 2011). These areas differ to a large degree with respect to salinity, temperature, and therefore sound speed profiles. Furthermore, those areas have a highly variable sea floor ranging from sandy to silty sediment or even rocky sea floor and have therefore different reflectivity. It is highly likely that porpoises adapt their echolocation behavior to these different acoustic habitats (Kyhn et al., 2013; Villadsgaard et al., 2007). One extreme habitat is the Wadden Sea as an acoustically and geomorphologically very complex habitat. It is characterized by canyon-like underwater structures, that constantly change, within flat and shallow waters with high tidally driven particle resuspension—a habitat that is unique in European waters. The murky waters of the Wadden Sea therefore represent a worst case scenario in terms of visual conditions during daylight hours.

The effect of tidal rapids on sound propagation can mainly be determined by resonant scattering, thermal and viscous absorption by micro bubbles, and visco-interial absorption and scattering of suspended solid particles (Richards et al., 2004). All of these variables affect volume reverberation and therefore the speed of sound in the water. Richards et al. (2004) conclude that signal-to-noise ratio (SNR) for a typical technical sonar system in a detection task in shallow waters would be affected by microbubbles to a larger degree than by sediment transport. One variable that echolocating animals, like bats (Brinkløv et al., 2010) or odontocetes (Kyhn et al., 2013), can adapt is the intensity or source level of their echolocation clicks to receive an optimized acoustic image. Hence changes in echolocation behavior of harbor porpoises should be directly observable in tidal rapids with high sediment transport, like the Wadden Sea.

Source levels of harbor porpoises may furthermore be of crucial importance to stationary acoustic monitoring using click detectors to estimate abundance of porpoises (Nuuttila et al., 2018). The source level affects the area in which a porpoise can be recorded and large differences between areas of interest will lead to a wrong determination of habitat use and also to wrong assumptions for an abundance estimate using effective area of detection or detection ranges (Marques et al., 2013).

To test whether porpoises adapt their sonar system to their surroundings, we recorded echolocation clicks of free-ranging harbor porpoises in the German Wadden Sea with a vertical linear six-element hydrophone array to estimate apparent source levels (ASL) and relevant sound parameters of porpoise clicks in this challenging environment.

We used a six-channel vertical line hydrophone array (spacing 1 m, 5 m total aperture) to record porpoises on five different days at Meldorf Bay (Germany, 8°52′–8°54′E, 54°06′–54°08′ N) in the time of 22–31 Mar 2014 and 21–25 June 2016. All porpoises recorded were visually confirmed sightings. The best strategy for recording the clicks was to drift with a small boat (6 m length) across the tidal stream with the array deployed. Water depth varied between 7 (minimal depth to use the array) and 20 m (maximum depth of the tidal inlet) of sandy sediments covered by a varying layer of mud with little vegetation coverage. The drifts took approximately 10–15 min.

The array was constructed of 32 mm PVC pipes with a 20 cm diameter buoy at the surface and a weight of 10 kg below the array. All channels were amplified by a B2008P amplifier (70 dB gain, ETEC, Frederiksvaerk, Denmark) and digitized using a USB 6366 NiDaqMX device (National Instruments, Austin, TX) attached to a regular laptop (DELL Latitude E6330, Round Rock, TX) using sample rates of 2 MS/s (in 2014) and 750 kS/s (in 2016) in SASLab recorder (Avisoft, Glienicke/Nordbahn, Germany). Signals of the six TC4013 hydrophones (Reson, Slangerup, Denmark) were band-pass filtered at 100 Hz to 500 kHz (300 kHz in 2016) using hardware fourth order low and high pass filters. The array was calibrated using a series of artificial porpoise clicks [described in Dähne et al. (2013)] of decreasing amplitude at different distances from 0 to 200 m using a USB 6251 DAQ Device (National Instruments, Austin, TX) for sound production. Amplification of the signal by 26 dB was realized using an A-301-HS (A.A. Labs System, Ramat-Gan, Israel) amplifier and transmitted via a TC4033 (Reson, Slangerup, Denmark) hydrophone in a >6 m depth area of Stralsund harbour (13°6.6943′E, 54°18.1445′N). The loudest signal had a peak-to-peak source level of 168 dB re 1 μPa m. The calibration showed good agreement between distance calculated using time of arrival differences (TOAD) and calculated for GPS positions up to 80 m from the array (see supplementary material¹).

Echolocation clicks were extracted from the raw audio files using the methods described in Kyhn et al. (2009), Kyhn et al. (2013), and Villadsgaard et al. (2007) to ensure that only the loudest clicks without interfering echoes were used to estimate the distances and therefore ASL and unbiased parameters of the clicks.

For localization, the custom-made software “toadsuite” was used, which calculates the least-error position of a sound source based on the time-of-arrival differences at the hydrophone array [cf. Koblitz et al. (2016) and Lewanzik and Goerlitz (2018)] for the distance between the porpoise and the array (r). Source level (SL) was then calculated assuming spherical spreading for transmission loss (TL) using the equation

where the received level (RL) was calculated from the maximum of all channels. The absorption coefficient α was assumed to be 0.04 dB/m (Ainslie and McColm, 1998). All comparisons are based on these back-calculated ASL, since all other publications available about array recordings in shallow and deeper waters (Jensen et al., 2013; Kyhn et al., 2009; Kyhn et al., 2013; Villadsgaard et al., 2007) do not account for shallow water sound propagation, which is rather between cylindrical and spherical spreading (Marsh and Schulkin, 1962). To take account of that uncertainty we calculated the difference between spherical spreading and two other transmission loss models (18 log r and 16 log r) for the ASL.

Temperature and salinity measurements to calculate sound speed were carried out using a calibrated CTD48 probe (Sea and Sun Technology, Kiel, Germany). Calculated sound velocities suggested homogenous depth distribution of salinity and temperature in the area in the years 2014 and 2016 with a mean sound speed of 1480 m/s.

Calibration results indicated a decent localization accuracy of the array up to 80 m with a range jitter in source level of up to +/− 1.5 dB (see supplementary material¹). All detected clicks above that distance were excluded from further analysis. Applying the criteria from Kyhn et al. (2009), Kyhn et al. (2013), and Villadsgaard et al. (2007), we were able to extract 47 (23 clicks in 2014, 24 in 2016) out of over 3000 recorded clicks for further analysis.

A significant correlation between SL and distance to the recording was found (Fig. 1) with low R² of 0.369. Data of 2014 and 2016 were found to have much lower bandwidth in 2014 (3 ± 2 kHz) recorded with 2 MS/s sample rate than in 2016 (7 ± 3 kHz) with 750 kS/s (Table 1, Fig. 2). Furthermore, the −10 dB duration of the recorded clicks in 2014 (81 ± 15 s) was higher than in 2016 (57 ± 17 s). ASL_PP was similar in both years with 171 ± 3 dB re 1 μPa m in 2014 compared to 167 ± 6 dB re 1 μPa m in 2016. Combined an ASL_PP of 169 ± 5 dB re 1 μPa m was calculated.

The difference between spherical spreading and 18 log r and 16 log r transmission loss models was found to be 3.2 and 6.3 dB resulting in ASL_PP of 172 and 175 dB re 1 μPa m.

This study confirms that harbor porpoises adapt their echolocation source parameters to different acoustic habitats with large implications for acoustic monitoring. The calculated mean ASL_pp of 169 ± 5 dB re 1 μPa for both years is much lower than the so far reported mean source levels of free-ranging harbor porpoises from Inner Danish Waters (−20 dB) and British Columbia (−9 dB). When considering that sound transmission in shallow waters does not follow spherical spreading laws (Marsh and Schulkin, 1962), this difference still does not disappear, even though spherical spreading is most likely not the case in any of the studies conducted so far, but is nevertheless assumed in all cases (Jensen et al., 2013; Kyhn et al., 2009; Kyhn et al., 2013; Villadsgaard et al., 2007). Therefore, the found effect is most probably a combination of limited water depths and other factors influencing sound propagation (e.g., sediment resuspension and microbubbles) and in turn may alter porpoise echolocation behavior.

In a study very similar to ours in terms of water depths and habitat, but on Irrawaddy and Ganges River dolphins in riverine systems, a species dependent adaptation of source levels to confined and acoustically complex habitats was found (Jensen et al., 2013). Our study indicates that even within a species, but depending on habitat conditions, source levels and other echolocation parameters may vary. The reduced source level may be due to an adaptation to the confined conditions where loud signals are simply not necessary, but may also evolve as an adaptation to predators' presence as suggested by Kyhn et al., 2013. In the Wadden Sea, no major predator for the harbor porpoises exists, except for grey seals [e.g., Haelters et al. (2015)] that likely do not hear porpoise echolocation clicks due to a limited hearing range [Ruser et al. (2014), in-air]. Furthermore, porpoises may adept their echolocation parameters to different prey species. The findings of this study in conjunction with Kyhn et al. (2013) therefore indicate a certain plasticity in the behavior of vocalizing harbor porpoises.

Static acoustic monitoring, utilizing porpoise echolocation signals, represents a good method to receive long-term information about migratory behavior, abundance, and population density by recording clicks in defined radii around click-loggers. Since these loggers have only one hydrophone their detection range depends directly on the sensitivity of the hydrophone, but also on the source level of the clicks recorded. Kyhn et al. (2008) calculated that small differences of approximately 3 dB in detection threshold of single hydrophone click-loggers can result in large differences regarding effective detection area and radius. This is also reflected in playback experiments from Wales, where detection thresholds of click-loggers, but especially source level of the used sound source, had a huge impact on the estimated effective detection distances (Nuuttila et al., 2018). An SL_PP of 176 dB re 1 μPa led to an effective detection distance of 187 m, while the lowest emitted SL_PP of 153 dB re 1 μPa had only 20% probability of being detected at 50 m distance. Source level is therefore one of the variables that need to be taken into account for future abundance estimates using stationary acoustic monitoring. The hereby reported reduction in ASL has large implications for stationary acoustic monitoring, especially for large scale surveys throughout the Baltic Proper (Carlén et al., 2018), but also for making comparisons across ocean basins. A decrease in source level by 20 dB may in the worst case result in a 90% reduction in effective detection radius. Further studies on population density of harbor porpoises and comparative studies should take account of the habitat-dependent source levels to reliably estimate abundance and derive the correct comparisons on porpoise occurrence rates between areas. However, source level is only one of many variables that have an effect on the detection range of static acoustic data loggers. Porpoises may use specific echolocation behaviors predominantly in different areas (far field orientation vs fish catch) which could drastically alter the detection range. In addition, elevated background noise levels can affect the detection thresholds of the loggers used (Clausen et al., 2018).

Porpoises use acoustic cues to inspect their environment, but have to adapt their acoustic output to changing conditions and different habitats. While a number of studies have concluded that porpoises emit highly stereotyped clicks [e.g., Au (1993) and Verboom and Kastelein (1997)] some have documented variability in different characteristics like source level and bandwidth (Kyhn et al., 2013; Villadsgaard et al., 2007). Our study shows that source levels and bandwidths depend on the acoustic habitat and are lower in tidally driven areas, most probably due to high particle load caused by resuspension as well as elevated levels of microbubbles in the water. Reflections will, in turn, cause cluttered echoes that will affect the ability of harbor porpoises to orientate in the water column, but especially to find and hunt potential prey. Such a phenomenon has only been documented so far in the neotropical trawling insectivorous bat Macrophyllum macrophyllum in different vegetation causing different levels of cluttering (Brinkløv et al., 2010).

We thank Jens C. Koblitz and Holger R. Goerlitz for co-developing the toadsuite and funding its development. We are grateful to Ansgar Diederichs (Bioconsult SH) for lending us the research vessel “Alte Sorge.” Furthermore, thanks are due for support at the Institute for Terrestrial and Aquatic Wildlife Research in Büsum, especially to Andreas Ruser, Patrick Stührk, Tobias Schaffeld, and Beate Zein for support during field trials and discussions about the results. A.G. was funded by the project “TopMarine” under Grant No. Z1.2-53202/AWZ/2017/7/DMM by the German Federal Agency for Nature Conservation.