Freshwater soundscapes: a cacophony of undescribed biological sounds now threatened by anthropogenic noise

The soundscape composition of freshwater habitats is poorly understood. Our goal was to document the occurrence of biological sounds in a large variety of freshwater habitats over a large geographic area. The underwater soundscape was sampled in freshwater habitat categorized as brook/creek, pond/lake, or river, from five major river systems in North America (Connecticut, Kennebec, Merrimack, Presumpscot, and Saco) over a five-week period in the spring of 2008. Over 7,000 sounds were measured from 2,750 minutes of recording in 173 locations, and classified into major anthropophony (airplane, boat, traffic, train and other noise) and biophony (fish air movement, also known as air passage, other fish, insect-like, bird, and other biological) sound categories. Anthropogenic noise dominated the soundscape of all habitats averaging 15 % of time per recording compared to less than 2 % for biological sounds. Anthropophony occurred in 79 % of recordings and was mainly due to traffic and boat sounds, which exhibited significant differences among habitats and between non-tidal and tidal river regions. Most biophonic sounds were from unidentified insect-like, air movement fish, and other fish sound sources that occurred in 57 % of recordings. Mean frequencies of anthropogenic noises overlapped strongly with the biophony, and comparisons of spectra suggest that insect- like and air movement sounds may be more susceptible to masking than other fish sounds. There was a significant decline in biodiversity and biophony with increasing ambient sound levels. Our poor understanding of the biophony of freshwater ecosystems, together with an apparent high temporal exposure to anthropogenic noise across all habitats, suggest a critical need for studies aimed at identification of biophonic sound sources and assessment of potential threats from anthropogenic noises.


125
Biological sounds were largely unidentified and subjectively classified into broad 126 categories based on our previous experience [7, 9- 183 Data analyses 184 Many environmental factors that potentially influence soundscape composition such as 185 habitat category, diel period, river position and river system were statistically confounded, but 186 were compared based on various subsamples of the data in an exploratory examination of their 187 potential to influence the soundscape composition. The exploratory nature of these comparisons 188 is emphasized given the "snap-shot" nature of the data collection and lack of control of 189 environmental variables such as time-of-day, temperature, season (although all data were 190 collected over a five-week period), amount of human development (ranging from remote 191 wilderness to heavily populated urban areas), turbidity, water depth, and propagation distance.    Table 1).

226
Biophony occurred in 57 % of the recordings, while anthropophony occurred in 63 % 227 (  Percent of the stations where sound types were observed. P = probability of a significant difference in the frequency of sounds between day and night based on a chisq test (ns = not significant, * = < 0.05, ** = < 0.01, *** = <0.001). Traffic sounds were the most common component of the anthropophony occurring in 37 238 % of the recordings. Biophony was observed in significantly more recordings at night than 239 during the day (84 % vs 51 %, Table 2). All components of the biophony except insect-like and 240 bird sounds occurred in significantly more night recordings than day recordings. No significant 241 diel differences in occurrence among recordings were observed for the anthropophony.  Table 3). Unclassified sound accounted for just 9 % of the total sounds by number and less than 254 5% of the sounds by percent recording time (S7 Fig, Table 3).  Table 3). The composition of the biophony is 258 shown in the expanded plots.
259 260 Biological sounds tended to be more diverse at night averaging 7.9 types per recording 262 (standard error = 1.7, maximum = 32) compared to 2.5 types per recording (standard error = 0.3, 263 maximum = 16) types during the day. Although the relative contribution of biophony and 264 anthropophony by both number and percent recording time were similar between day and night, 265 the composition of the sounds changed (Fig 3, S7 Fig, Table 3) There were no significant differences in the percent of recordings among habitat 275 categories within the biophony during the day (S3 Table). At night, insect sounds occurred more 276 frequently in the brook/creek habitat (however, the brook/creek sample size was low), while 277 other biological and bird sounds were absent from the pond/lake habitat. Traffic sounds were the 278 most widespread noise and were significantly more frequently occurring in brook/creek habitat 279 during the day. In contrast, boat sounds were absent from brook/creek (S3 Table). No significant 280 differences in the frequency of occurrence of anthropophonic sounds were observed among the 281 habitat categories at night (S3 Table).
282 Insect and other fish sounds dominated the biophony by percent time in all habitats during the 283 day, but air movement sounds dominated pond/lake and river habitats at night (S8 Fig, S4 284 Table). Other fish dominated brook/creek habitat at night but the sample size was low (n = 2).
285 Traffic sounds dominated soundscape percent time during both day and night in brook/creek 286 habitat, while boat sounds dominated other habitats (S8 Fig, S4 Table).
287 River gradient pattern 288 There were no consistent trends among rivers in biological or noise sounds in main-stem 289 river habitats moving along the river gradient from headwaters to mouth, although the highest 290 elevation locations tended to have little or no biological sounds (S1 Database S1 online). When 291 day data from all rivers were pooled after grouping locations into non-tidal (N = 46) and tidal 292 regions (N = 20), all boat noise categories were significantly more frequent in tidal regions (S5 293 Table). Similar trends were observed for sound rate and percent time (S9 Fig, S5 Table). Boat

311
Background ambient sound levels (RMS category) had a strong influence on biological 312 sound occurrence (Fig 4, S7 and S8 Tables). Air movement sounds significantly declined from a 313 high of 72 % of recordings to a low of 6 % of recordings from the lowest to highest ambient 314 sound level categories. Similar, but non-significant, trends in occurrence were observed for FRT 315 and other fish sounds (Fig 4). Rate and percent time of biological sounds followed similar trends 316 with significant declines in air movement, fish and total biophony with increasing RMS level (S7 317 Table). Biodiversity also declined significantly from 4.2 to 0.7 sounds/recording with increasing 318 ambient sound level (S7 Table).
319  (Fig 2, S6 Fig, Table 1) also suggests the potential for masking (an example 382 of which can be seen in S1 Fig and the corresponding S1 Audio). In contrast, the lack of overlap 383 between peak frequencies of biological sounds and peak ambient frequency (Fig 2, S6 Fig)

410
Our observations suggest that boats running idle while docked, anchored, or drifting are a 411 major component of freshwater soundscapes (Fig 3), and have the potential to mask some 412 biophony such as insect sounds, but we are not aware of studies that examine its potential 26 437 river systems of different lengths and elevation gradients requires a gradient approach. It is 438 interesting that while the biotic community changes considerably from high elevation reaches to 439 estuarine reaches, the changes in the biophony type contribution to the soundscape are minimal, 440 suggesting that although soniferous species may change, the broad sound categories are more 441 consistent. Sampling along the river coenocline at a higher spatial resolution would likely reveal 442 more subtle shifts in the biophony, due to species assemblage changes. A gradient in impacts 443 from different types of anthropogenic noise is expected as we observed a striking transition from 444 remote wilderness to increasingly developed urban areas while traveling from the river 445 headwaters to the sea. Some of this transition was captured in the comparison between non-tidal 446 and tidal reaches of the rivers where there was a shift between dominance of the soundscape by 447 traffic noise in non-tidal reaches to a dominance by boat noise in tidal reaches (S9 Fig, S5 448 Table), highlighting different potential for impacts in different habitats.

449
Why has the freshwater soundscape been neglected by scientists, resource managers, 450 conservationists and environmentalists for so long? We believe that in large part it is due to a 451 general lack of appreciation of the importance of inland fisheries on regional, national and 452 international scales which can lead to a lack of scientific inquiry [21]. A lack of appreciation of 453 the importance of the soundscape is exacerbated by a pervasive lack of awareness that fish and 454 other aquatic organism produce sounds that can be monitored remotely and how that can be a