Date of Award

Spring 2025

Project Type

Dissertation

Program or Major

Biological Sciences

Degree Name

Doctor of Philosophy

First Advisor

Jennifer Miksis-Olds

Second Advisor

Jennifer Dijkstra

Third Advisor

Richard Smith

Abstract

ABSTRACT By identifying soundscape properties and taxa detected with metabarcoding that accurately predict substrate composition, assumptions about propagative properties of the environment can be rapidly formed without the use of active acoustic sensing or invasive environmental sampling techniques. “Sentinel indicators,” are variables, processes, or components within ecosystems that can be measured or quantified and have sensitivity to environmental pressures. Passive acoustic monitoring (PAM) and metabarcoding of seawater samples (MSS) are two minimally invasive potential methods of identifying sentinel indicators of soundscape and propagation properties of marine habitats by detecting the presence of marine taxa and extracting information from soundscapes. To better understand how soundscape features are related to environmental components that impact sound propagation, an observational experiment was conducted comparing the soundscapes of three different habitat types in coastal Gulf of Maine (GoM) waters: eelgrass, macroalgae, and sand. Using the Soundscape Code (Wilford et al., 2021) to rapidly and quantitatively characterize soundscapes of different habitats revealed soundscape properties are indicators capable of discriminating among coastal habitats at fine spatial scales using discriminant analysis. Impulsiveness and Uniformity metrics are important sentinel indicators for predicting habitat composition from acoustic recordings, demonstrating the value of quantifying soundscape properties beyond amplitude. This ability to efficiently obtain information about the acoustic environment is valuable for informing propagation models associated with marine biota as well as for quantifying uncertainty in the use of acoustic sensing instruments by humans such as echosounders. Both PAM and MSS are viable, minimally invasive methods for long-term detection of marine biota. A comparative study of these observational techniques was conducted to determine whether taxonomic detections from MSS could be predicted by Soundscape Code metrics (SCMs), and to identify which method serves as a more reliable detector of sound producing taxa. The present investigation revealed that there was little overlap in sound producing taxa detected by either method. Snapping shrimp were the dominant biological sound source in acoustic recordings across all habitats and locales, but were not detected with MSS. Fish were the second most prevalent source of biological sound, though these detections were not identified to species level in the acoustic recordings. Using MSS, 28 taxonomic groups of fish were detected with the MiFish-U primer set, many of which include sound producing species. Phocidae was the only family of marine mammal detected in metagenomic samples, while there were no confirmed detections of marine mammals in acoustic data that could be positively taxonomically identified. Vessel traffic and construction were dominant, high amplitude anthropogenic sound sources in nearly every habitat due to the close proximity of recorders to areas with high anthropogenic activity. These high levels of anthropogenic sound likely masked many biological sound sources and may have also influenced the behavior of sound producing taxa in a way that further reduced their detectability. Overall, MSS proved the stronger identifier of sound producing taxa, though this was a product of the broad taxonomic scope of taxa detectable with metabarcoding and was not a reliable indicator of which taxa make the greatest contribution to soundscapes. To determine whether Soundscape Code metrics were able to serve as reliable predictors of the taxonomic profiles of GoM habitats, partial least squares discriminant analysis (PLS-DA) models were developed at the Order and Family levels. The model that explained the greatest amount of variance in taxonomic profiles used 35 Orders that were identified as indicators of geographic locales using Indicator Species Analysis (ISA). However, of the four best performing models, none possessed any predictive power, demonstrating that Soundscape Code metrics did not serve as reliable predictors of taxonomic profiles for the sampled habitats. The likely explanation is that the soundscapes recorded were dominated by anthropogenic and geophysical inputs, as well as snapping shrimp, all of which were absent in the taxonomic profiles detected with MSS. Reliable predictive models still may be possible for soundscapes dominated by biological sound sources, but for the habitats investigated, a strong relationship did not exist between acoustic predictors and genetically detected biological responses.

Download

Share

COinS