Marine Ecology and Acoustics Lab
Current Lab Activities and Projects
Taxonomical classification of reef fish with broadband backscattering models and machine learning approaches (2017-2021)
Complex environments with biologically diverse fish assemblages represent a challenge for detection and classification of fish species, specially when visual methods are hindered. Acoustic devices like echosounders present a promising alternative since they can cover larger distances than visual cameras and are not affected by turbidity. However, lack of species-specific acoustic patters has delayed the implementation of modern classification methods.
Using acoustic scattering we model the swim bladder of 130 fishes across 6 species, and examine their backscatter responses (echoes), at frequencies of 30 kHz – 200 kHz and orientation angles (± 44o). Several machine learning techniques (Bayesian, SVM, KNN, and CNN) are used to classify their backscatter responses, with overall accuracies > 80%. The modeling and classification approaches indicate that a taxonomic distinction based on morphologically dependent scattering responses is possible and may provide the capacity to acoustically discriminate among fish species.
​
Figure 1: TS(f,θ) backscattering spectra of two characteristic individuals (columns Indiv1 and Indiv2); and average TSavg(f,θ) (column TSavg); for six species examined: blue striped grunt (BSG); grey triggerfish (GTrigg); red snapper (RedS); tomtate (Tomt); vermilion snapper (VermS); and white grunt (WG). Angles limited to ± 44o. All TS are in dB re 1m2, angles in degrees with respect to horizontal.
Figure 2: (left) oil droplets and gas bubbles rising from the marine seep, detected by acoustic methods. (right) acoustic instrumentation and setup (taken from 2019 NOAA report).
Comparison of broadband measurements and acoustic models of gas bubbles and oil droplets in the water column (2019-2021)
Oil droplets and agas bubbles from an oil marine seep in the northern Gulf of Mexico were recorder in 2019 by visual and acoustic methods. This study acoustically models the bubbles and droplets and compares them with their measured echoes. A methodology to estimate the oil flux, base only on acoustic data is derived and compared with visual estimates. Machine learning methods are applied to visual and acoustic data to expedite their processing.
ASV Sediment profiling on NPS canals (2021-2022)
National Park Services (NPS) funded a test project to derive a sediment depth profile in two of their Florida canals. Phosphorus levels, due to sediment accumulation, increase dramatically when the gates that communicate the canals open, impacting the downstream ecosystem. We estimated the sediment depth with echosounders and multibeam sonars and quantify sediment bed-load in proximity to gates to improve dredging operations and reduce the impact on the aquatic ecosystem.
Figure 3: Boswell lab ASV with acoustic and sediment profiling capabilities.
Figure 5: 3D bathymetry of L29 and L67A canals
Figure 4: Bathymetry of L29 and L67A canals
Figure 6: Acoustic sediment mapping of L29 and L67A canals
Areal and aquatic motion caption tank (2021-2022)
Our lab along with other FIU colleagues are part of a larger effort for developing and implementing marine vehicles focus on monitoring aquatic environments. Part of that effort is the construction of a DoD funded tank for aquatic and aerial motion capture. The tank, that is built on BBC, has 10 underwater cameras, 12 aerial cameras, and the computing capabilities to track and model the motion of underwater, surface, and aerial vehicles. It is an ideal test bench for new vehicle’s designs and their integration with drones, obstacles, and the environment.
Figure 7: Motion capture tank design.
Figure 8: Motion capture tank facility at BBC
FDEP Water quality and Health Monitoring System for Biscayne Bay
Despite many years of concern about the health of Biscayne Bay, it is continually deteriorating due to low water quality caused by nutrient runoff, limited water circulation, and increasing water stagnation. Recurring incidents like fish kills, algal blooms, and sewage leaks have drawn the attention of the community and the state, raising concerns about the economic and ecological consequences of vital ecosystems. Researchers are now unable to forecast when or why outbreaks will occur, as well as where they will be most destructive to coastal ecosystems and communities. But real-time, synchronized data from distributed sampling platforms and sensing devices could improve environmental monitoring and serve as early warning systems within dynamic environments, as well as build mechanistic and predictive models of HAB dynamics, facilitate actionable interventions, and drive adaptive monitoring decisions. As an FEEP funded project, FIU is implementing a monitoring system of the Bay, that can rapidly detect the onset of HABs and other water quality issues capable of widely disseminating information regarding conditions, in addition to relevant levels of the algae and their toxins in coastal regions, is critical in the efforts to protect and recover it and would be a useful tool for institutions involved in its preservation.
Figure 9: ASV with water quality surveying capabilities.
Gulf of Mexico Deep Pelagics. (2011-Present)
DEEPEND (Deep-Pelagic Nekton Dynamics) is a recently funded GOMRI consortia that will investigate deep-pelagic communities on short-term (sub-generational) and long-term (evolutionary) timescales to appraise extant recovery and potential future recovery of these communities using a suite of integrated approaches. These approaches include: 1) a direct assessment of GoM deep-pelagic community structure, with simultaneous investigation of the physical and biological (including microbial) drivers of this structure, in order to document biodiversity and 'natural' variability; 2) a time-series, 'hindcast,' comparison of biophysical data from 2015-2017 (DEEPEND sampling) to 2010-2011 DWHOS data; 3) an examination of differences in genetic diversity among key species; and 4) an assessment of the extant and potential future consequences of the DWHOS on the shallow and deep-pelagic biota.
Vertical Migration of Sound Scattering Layers
Sound scattering layers (SSLs) are ubiquitous features throughout the world’s oceans and the organisms comprising these layers are important components of these vast ecosystems. The depths at which SSLs occur are dynamic and often dependent on the depth of the water column and time of day, which gives rise to well-recognized and remarkable diel vertical migration (DVM) patterns. Commonly, the main organisms that comprise SSLs are euphausiids, crustaceans, siphonophores and myctophids which serve as primary prey resources for larger predators, like marine mammals. The dramatic vertical movement between biomes offers opportunities for intense ecological interactions between SSLs and surface-oriented communities including predator-prey relationships and energy transfer. Understanding how the community structure within SSLs varies will help to better characterize the important ecological role these organisms play.
Broadband Acoustic Swimbladder Modeling
A central challenge to quantifying change in marine ecosystems and evaluating ecosystem services is the acquisition of spatially and temporally appropriate data to examine patterns in diversity and abundance of dominant scatterers, and in particular exploited target species and important prey resources. Broadband acoustic technology offers a promising tool that might allow inference at relevant taxonomic and spatiotemporal scales. We are investigating use of a boundary element method with high-resolution computed tomography data to examine numerically modeled acoustic scattering responses (12-250 kHz) of dominant fishes across a range of orientations. The modeled broadband scattering responses will be compared with in situ measurements of broadband scattering from fishes obtained using newly avaialble Simrad EK80 echosounders. Preliminary analyses suggest that broadband scattering responses in this frequency range are sensitive to the fine-scale morphological variations among common fish species. Further analysis will give insight into the appropriate frequency domain to examine the efficacy for taxonomic resolution.
Goliath Grouper Spawning Aggregation Study
The goliath grouper fishery has been closed since 1990 due to declines in their population after extended periods of overfishing. However, the population has made a significant recovery since protection was put into place, and years of research indicate that Goliath Grouper populations are on the rise. An important part of their recovery has been the occurrence of large spawning aggregations in the fall months throughout Florida. To better understand their population dynamics, characterize the effect of aggregations on surrounding reef fish communities, and to develop an accurate but rapid acoustic survey design to remotely estimate abundance, we initiated an annual goliath grouper survey with colleagues from MOTE Marine Laboratory in 2016. Since then, a combination of active and passive acoustic surveys, along with visual and camera surveys are conducted in Jupiter, Florida to capture the seasonal changes in Goliath Grouper abundance during the spawning season. Photo Credit: Liz McNamee
Pacific Herring Genomics
Pacific herring are an ecologically valuable species in the Gulf of Alaska supporting many trophic levels however, failures to recover to pre-fishery abundances make it difficult to assess ecological impacts and resolve the debate over fishing quotas between Native Alaskans and Alaska Department of Fish and Game (ADFG). Current hypotheses as to why herring have failed to recover converge on increased predation, disease, thermal intolerance and changing habitats however, very few genomic studies have been conducted. Recognition of the potential for shifts in stock organization to occur are crucial; it is imperative to examine the degree of connectivity between identified stocks in Lynn Canal. Identifying the most differentiated and poorly connected stocks could guide managers to emphasize actions of protection under the assumption that high differentiation and poor connection could indicate high vulnerability to depletion. Knowledge gained from this study will be of great interest in marine ecology and evolutionary biology because it directly addresses questions of what factors maintain and distribute genetic variability in marine populations and the importance of their influence.
Autonomous Survey Platforms
Our lab has worked closely with SeaRobotics to develop a modular unmanned 4m long vessel to perform acoustic surveys in shallow coastal waters (e.g., estuarine systems). The modular design permits the vessel to be broken down and shipped for operations in remote regions (e.g., Arctic). The vessel is capable of executing a pre- programmed transect series collecting repeatable, high-resolution, spatially-referenced physical and biological data which is telemetered in real-time back to the base station. In its current configuration it is equipped with a calibrated split-beam echosounder (Simrad EK60, 120 kHz), an imaging sonar (Soundmetrics DIDSON, 1.8MHz) and wide-swath multibeam echosounder (Mesotech M3 multimode sonar). Most recently this has been modified to participate in a dye-tracer study in a shallow water tropical lagoon and instrumented with a CTD. The ASV is capable of simultaneously collecting high-resolution data of the substrate characteristics, bathymetry and the organisms distributed above the seabed and capable of carrying additional payload components (e.g., water samplers, passive acoustic receivers, etc.).
Red Snapper Abundance Survey
We are currently working to understand absolute abundance of red snapper (Lutjanus campechanus) in the Gulf of Mexico. This research addresses one of the most urgent problems currently facing U.S. Gulf of Mexico (Gulf) fisheries management. The notable Red Snapper supports one of the most economically valuable finfish fisheries in the Gulf of Mexico. Their profound popularity led to severe stock depletion through overfishing, and the stock is still overfished. The best possible management is obstructed by the lack of robust abundance data. Thus, our overarching goal is to provide an independent estimate of Age-2 and older Red Snapper absolute abundance in the northern Gulf of Mexico.
Coupled Stressors Affecting Fish Behavior
Estuarine mangrove systems are essential fish habitats that provide important ecosystem services for a wide range of inhabitants as nurseries and foraging grounds. As climate change is projected to increase global temperature over the next 50 years, and coastal development is expected exacerbate the issue, increasing stress on these environments will result in the degradation of habitat quality and ecosystem function. A factorial design consisting of two temperature levels and two turbidity levels (high and low) was used to document changes in fish schooling behavior and predator-prey interactions using an acoustic imaging sonar (DIDSON). We are trying to quantify predation rate during high thermal stress and high turbidity periods to determine if prey fish mortality changes. Recent studies have focused on a single stressor at the species level, but by investigating the relationship between multiple stressors and their effect on behavior at the community level, we will gain a better understanding of how fish communities will respond to increases in temperature and turbidity related to climate change and coastal development.