Members of the FAR Lab work on a variety of topics. Often, we are working on more than one project at a time. Here are just a few of the topics that we are currently conducting research on.
Kresimir Williams
Acoustic surveys are an important component of resource management effort for walleye pollock (Theragra chalcogramma), the largest fishery in the US. During the course of the survey, conducted by NOAA research ships, fish aggregations seen by the echosounders are sampled using survey trawls. The trawl catches are expected to correctly represent the species composition and size distribution of the fish in the aggregation sampled. However, trawls usually do not catch all sizes and species of fish with equal efficiency, resulting in potential errors in survey results. My research focuses on determining how fish of different sizes interact with the trawl, and what the effect of size–dependent escapement from the trawl has on the estimates of pollock abundance.
One way to sample escaping fish is to re-capture them using small nets attached to the outside of the survey trawl. From these samples, the total amount of escapement can be estimated and compared to the fish retained in the codend, and the accuracy of the catch data can be evaluated.
b) Attachment of the pocket nets onto the outside trawl
Fish interaction with the trawl can also be observed using instruments placed in the trawl. We have used a dual frequency identification sonar (DIDSON) and cameras to look at fish in the trawl and describe how they react during the capture process. These observations provide insight into size specific behaviors of pollock, and will aid in making modifications to the trawl to reduce fish escapement.
b) Individual fish tracks from the DIDSON data.
This research has been sponsored by the Midwater Assessment and Conservation Engineering program, part of the NOAA Fisheries Alaska Fisheries Science Center.
Julian Burgos
My project consists on describing and simulating walleye pollock spatial and temporal patterns. Walleye pollock is a fish from the Gadidae family. It has a large role in food webs in the Bering Sea and Gulf of Alaska, and sustains a very large commercial fishery. Although large scale (>100km) distribution patterns are relatively well known from bottom trawl and acoustic surveys, patterns at smaller spatial (and temporal) have not received much attention. Quantifying mesoscale (0.1 - 10 km) patterns is important because many ecological processes, including predator-prey interactions, occur at these scales.
Acoustic surveys are conducted regularly by the Alaska Fishery Science Center (NOAA) to estimate the biomass of walleye pollock. The data collected during these surveys provide a synoptic view of the distribution of pelagic and semi-demersal species at a fairly high resolution (i.e. 20 m horizontal and 0.2 m vertical). If these data are displayed in echograms, different distributions patterns are evident, including schools, large pelagic shoals and benthic shoals or 'carpets'. I use echo-trace classification and landscape ecology indices to obtain quantitative metrics of spatial pattern and to develop a classification typology for spatial patterns. I also use statistical methods like geostatistics and generalized additive models to explore the relationship among spatial patterns and between pattern and environmental variables.
I am also developing an individual based simulation of walleye pollock. In this simulation individual fish behave socially, gathering with other individuals to form aggregations, and reacting to environmental cues (temperature, light, bottom topography and food “patchiness”). The objective of this simulation is to generate realistic walleye pollock spatial and temporal patterns. This will help us understand how these patterns are generated. In addition, the simulation will generate prey fields to be used as input in marine mammals predation models.
Simulated horizontal distributions of walleye pollock obtained with conditional gaussian simulation, using two levels of patchiness.
Each pixel represents 0.85 km2
This research has been sponsored by the Alaska Fisheries Science Center.
Sandy Parker Stetter
Hypoxic conditions can influence the horizontal or vertical distributions of pelagic organisms. In Hood Canal (Washington), low oxygen has resulted in fish kills during late summer. Our studies examine how changing dissolved oxygen levels affect the distribution of nekton (i.e., fish and large invertebrates) and the ecology of the pelagic zone.
Nekton response to dissolved oxygen in Hood Canal
This 2006 study examined the responses of pelagic nekton to a midwater oxygen minimum (MWOM) in Hood Canal. MWOM occur in freshwater and marine systems, but marine nekton response to MWOM has not been investigated. Our results suggest that the MWOM did not affect invertebrate nighttime vertical distribution, but did limit the extent of fish vertical migration, providing a seasonal, nighttime prey refuge.
On left: Dissolved oxygen profile in September, with low DO between 10-35 m.
On right: Acoustic echogram nighttime image of same site, showing lack of nekton in low DO zone.
Pelagic nekton distribution relative to dissolved oxygen in Hood Canal
Building on work from 2006, this project will map fish and invertebrates distributions throughout Hood Canal before and during low-oxygen conditions. Our June and September 2007 nighttime surveys will have trawling support from Washington Department of Fish and Wildlife (WDFW). We will use this data to examine patterns in nekton distribution related to oxygen gradients.
High-resolution survey of low DO refugia in Hood Canal
We will expand upon our 2006 study of MWOM in Hood Canal to investigate potential use of the low oxygen layer as a prey refuge by invertebrates. We will examine fish and invertebrate density distributions at a low- and moderate-oxygen site between June (pre-hypoxia) and September (potential MWOM). A portion of this study will also compare among Hood Canal and Puget Sound sites. A UW capstone year student will be involved in an analysis of invertebrate and fish migration rates under different oxygen scenarios.
Patrick Adam
I am interested in the time and scale dependent patterns of nekton and plankton abundance. Current methods for evaluation of abundance and dispersion include trawl and acoustic surveys. Due to the slow speed with which an acoustic survey is conducted, typically less than 10 knots, poor estimates of the temporal variation of these animals may result. Ship avoidance behavior of nekton species may also affect abundance estimates. Acoustic surveys typically miss the upper 10 meters of the water column, or less if using a towed-body transducer, which potentially eliminates a significant portion of the biomass depending on species of interest. Incorporating the use of an aircraft mounted Light Distance And Ranging (LIDAR) sampling platform may minimize these limitations and provide key insight into movements of fish schools and plankton over hourly time series. Laser light is used much like sound from a sonar device and can detect the presence and distance of objects in the water. Unlike acoustics, LIDAR is capable of detecting fish and plankton in the surface waters to approximately 20m depth in nearshore (<30 nautical miles) waters.
Because the LIDAR instrument is mounted in an aircraft it can survey considerably more area in the same amount of time as ship based surveys, or the same area in a fraction of the time. My research will compare both sampling techniques to characterize the effects of survey speed and transect layout on backscatter (energy reflected from fish and plankton targets) estimates using coincident spatial and temporal surveys.
The two figures below are display the change in difference between LIDAR and acoustic backscatter measurements taken at increasingly distant spatial locations (denoted by the colorbar) and temporal lag times, from a coincident (ship and plane measuring the same parcel of water at the same time) location. Data is from August 22 – 23, 2005 off the coast of Washington. Note the increasing difference between LIDAR and acoustic backscatter measurements as lag time between coincident location increases from +0.5 - +3.5 hours for August 22. The trend for August 23 is less clear.
This research has been sponsored by the National Oceanographic Partnership Program (NOPP).
Steve Barbeaux
Commercial fishing vessels have long been used as sampling platforms for scientific studies and many national, state, and provincial agencies contract fishing vessels to conduct scientific research to support fisheries management objectives.
Echosounders capable of collecting scientific quality acoustic data have recently become available to the commercial fishing industry and researchers around the world have begun to use acoustic data collected from commercial fishing vessels in a broad range of fisheries research applications (Stanley 2000; Dorn et al 2002; Barbeaux et al 2005; O’Driscoll and Macaulay 2005). In the North Pacific the use of scientific quality echosounders by the Alaska pollock (Theragra chalcogramma) fishing fleet has provided an opportunity to collect acoustic data opportunistically during normal fishing operations. Collecting opportunistic acoustic data from commercial fishing vessels allows researchers to inexpensively obtain data from multiple platforms during a single time period.
The costs of collecting acoustic data opportunistically is restricted to the price of the digital media used to record the data (~$220 US for 120GB), and the staff time to install and collect media from the vessels. A multiple platform approach potentially offers a more synoptic view of a population than a survey conducted on a single vessel. Although opportunistic data are not collected on a systematic grid and may not be used to obtain population abundance estimates directly, the broad temporal extent and high spatial resolution of the opportunistic data may facilitate investigations on the distribution and behavior of fished aggregations. Due to the unstructured “sampling design” and the multiple platforms from which these data are collected there are significant challenges to its use as a quantitative tool for fisheries managers. The project I am working on will investigate the appropriate use of these data and develop methods for over-coming obstacles intrinsic to this potentially valuable data source.
This research has been sponsored by the Alaska Fisheries Science Center.
This research has been sponsored by the Alaska Department of Fish and Game
Cairistiona "Kirsty" Anderson
The data produced by acoustic instruments, such as echosounders, only tell us that sound is being reflected (backscattered) to a greater or lesser extent at particular depths in the water column. The challenge is take this data and extract biologically useful information from it. Traditionally, scientists have relied on a combination of recognizing acoustic features in echograms and then fishing nets through those features to find out what animals are producing them (e.g. in the echograms below we can say with some confidence that the very high intensity patches near the bottom are produced by dense schools of adult walleye pollock). However, this approach only works to distinguish organisms that form distinct features and relies on individual users being able to consistently recognize those features.
Gulf of Alaska, N. Pacific: Two echograms showing backscatter at different frequencies and four virtual “echograms” showing probability of membership of different classes (warmer colors = higher probability). The upper virtual echograms appear to identify the core of the dense schools of adult walleye pollock and the less tightly packed fish around them respectively, whilst the lower panels separate out other components of the ecosystem.
More recently, scientists have exploited the simultaneous collection of acoustic data at more than one frequency to use the most apparent pair-wise differences in the amount of backscatter at the different frequencies to identify specific organisms. My research is focused on taking an unsupervised probabilistic classification technique developed for the analysis of gene expression data (www.seqexpress.com) and adapting it to provide a rigorous method for identifying the different constituents producing the backscatter in multi-frequency echosounder data. This will have applications a) in known systems to provide a rigorous method for extracting the backscatter produced by the species of interest (see above) and b) in unknown systems to provide insights into community structure and composition (see below).
Over Mid-Atlantic Ridge, N. Atlantic: Two echograms showing backscatter at different frequencies and three virtual “echograms” showing probability of membership of different classes (warmer colors = higher probability). In this relatively unknown system, the different classes provide clues as to the spatial organization of the community, and so indicate where fishing efforts should be targeted to identify the different constituents.