Comparison of Two Soundscapes: An Opportunity to Assess the Dominance of Biophony Versus Anthropophony

Sound travels further through water than light and is one reason why many marine animals use sound to communicate and gain information about their surroundings. Scientists collect recordings of these underwater sounds to gain information on species’ habitat use, abundance, distribution, density, and behavior. In waters where visibility is severely limited or access is difficult or cost-intensive, passive acoustic monitoring is a particularly important technique for obtaining such biological information over space and time. The “soundscape” of an ecosystem is defined as the characterization of all the acoustic sources present in a certain place (Wilford et al., 2021). A soundscape includes three fundamental sound source types (Figure 1): (1) anthropophony, or sounds associated with human activity; (2) biophony, or sounds produced by animals; and (3) geophony, or sounds generated by physical events such as waves, earthquakes, or rain (Pijanowski et al., 2011). Studying soundscapes can provide biological information for a specific habitat, which could then be linked to ecosystem health status and other bioindicators. This information can be used to monitor the habitat over time, allowing for rapid detection of habitat degradation, such as in response to human-driven events. Comparison of Two Soundscapes: An Opportunity to Assess the Dominance of Biophony Versus Anthropophony


SOUNDSCAPE DATA ACQUISITION
Acoustic recordings can be collected using devices that are either fixed to the ocean floor or floating/navigating in the water column; by stations cabled to a land-based laboratory; or by instruments towed from boats (i.e., hydrophones) or attached to animals (i.e., bio-loggers). New technologies permit long deployments (months) that generate large amounts of acoustic data. Analyses of these data are very labor and time intensive, so automation is highly desirable.

SOUNDSCAPE ANALYSES
Because the study of underwater soundscapes is relatively new, there is not yet a standardized way of processing acoustic data (Wilford et al., 2021). Thus, given the variety of instruments, mooring types, and deployment settings available, it can be challenging to compare results between different data sets. However, some initiatives, like the International Quiet Ocean Experiment (IQOE), are creating standards for underwater sound processing.
When analyzing acoustic habitats, different approaches can be considered. Common examples include the detection and quantification of specific events or the calculation of acoustic indices, which are summary statistics that describe the distribution of acoustic energy and can sometimes be correlated with certain biological or ecological habitat properties. Apart from classical acoustic indices, sound ecological indices could reveal the status of marine ecosystems, but they require previous knowledge about each sound type and its characteristics. One common approach to visualizing the soundscape is to use a spectrogram, a visual representation of a sound's intensity and frequency over time. A spectrogram allows identification of interesting acoustic events and their timing, even for sounds outside the human hearing range.

NOISE POLLUTION
Over the last many decades, human activities at sea such as pile driving, dredging, or shipping have increased, contributing to and sometimes dominating underwater sound levels. When anthropophony masks biophony, marine animals that rely on sound to detect predators or prey, to find or communicate with mates or offspring, and/or to navigate can be harmed (Duarte et al., 2021). Thus, it is important to describe and record the soundscapes of places that are currently less and more disturbed to quantify current noise levels.
Knowledge of these "baselines" will enable us to measure additional human-driven degradation to the oceanic soundscape and the resulting impact on marine life. Jacques Cousteau's first impression of the ocean was that it was silent. We now know it has always been filled with natural sounds. 3 of decree 3573 that licenses for megaprojects, such as port construction, must be approved by ANLA, which is also responsible for monitoring environmental implications.

CASE STUDIES
Here, we describe two study regions with vastly different soundscapes, characterized by extremely different shipping densities ( Figure 2). The first study region, the Gulf of Tribugá, Colombia, is "less disturbed" by undersea noise (closest to pristine). It serves as a general marine soundscape baseline for comparison with possible future disturbances from port construction and operation. By contrast, the second study region, the Belgian part of the North Sea (BPNS) is located in a "more disturbed" area of very exploited shallow waters. Its baseline is being used to monitor the effects of noise reduction policies. We chose October 16, 2020, at 12:00 until October 17, 2020, at 07:30 (local time) as the day for our soundscape comparison.
Our hypothesis is that biophony dominates the Gulf of Tribugá while anthropophony dominates the BPNS.

Gulf of Tribugá
The main goals of the PHySIColombia Project were to identify which sound sources exist in the Gulf of Tribugá

SPECTROGRAM VISUALIZATION
We identified different sounds using spectrograms generated by Raven Pro 1.6 software ( Figure 5). Marking the spectrograms manually when each sound type occurred allowed us to determine the schedules on which animals, natural events, and human-made noises operated. The largest contribution to the Gulf Tribugá soundscape was from singing humpback whales, then shrimp, and finally fish. Anthropophony was primarily from small boats, but occasionally from one or two larger shrimping boats. The loudest geophony sounds came from rain and wind, while the sloshing of the tide and crashing of waves onshore commonly existed in the background.
In contrast, anthropogenic noise dominates the BPNS soundscape. The identified sounds were generated by large ships, probably commercial or fishing. Another identified sound is possibly dredging or trawling, which is concentrated at about 1 kHz or below and is constant and prolonged.

SPECTRAL PROBABILITY DENSITY COMPARISON
To compare the soundscapes of both locations, we computed the spectral probability density (SPD) of each location using pypam (https://github.com/lifewatch/pypam). SPD is useful for computing the statistical distribution of underwater noise levels across the frequency spectrum (Merchant et al., 2013). To compute the SPDs, the audio files were divided into one-minute samples. Frequency distribution and the probability of each frequency appearing at a certain sound level (from 20 to 140 dB re 1 μPa) were computed, and both sites were processed to remove the direct current (DC) electrical noise generated by the instruments. The data from the BPNS location were downsampled to match the sampling rate used in Tribugá so that the frequency and time resolution of both SPD computations would match.
The 1 st , 10 th , 50 th , 90 th , and 99 th percentiles of the SPD represent the intensities and contributions of sounds in the soundscape. The 1 st percentile represents sounds that occur 99% of the time but are low intensity, and the

CONCLUSIONS AND PERSPECTIVES
By first establishing acoustic baselines in less and more noisy ocean regions, monitoring soundscapes over time can be a cost-effective method for assessing the health of marine ecosystems. Some scientists are developing acoustic indices that would link acoustic features to biodiversity or other biological indicators (Wilford et al., 2021).
Few standards exist for sensor deployment configuration, making ecosystem comparisons challenging or not feasible, and no global acoustic indicator yet exists. However, various groups are working to standardize marine acoustic FIGURE 6. Spectral probability density of the two locations. Oneminute window of one day, no overlap, NFFT 4096, histogram bin size of 1 dB re 1 μPa. Boxes with dashed lines show possible boat sounds, continuous lines indicate humpback whales sounds, and dash-dot lines show shrimp sounds. Overlap in frequency with the sound of the humpback is an example of masking biophony.