Advanced Sampling Technologies Research Group
The Northeast Fisheries Science Center's (NEFSC) advanced sampling technology research efforts during recent years have focused on improving fisheries-independent population estimates of Atlantic herring in the Georges Bank and Gulf of Maine regions for more cost-effective and timely fisheries management. Field experiments have been conducted to evaluate survey designs, improve variance estimators, and define species-specific individual target strength measurements. Laboratory experiments have provided development, 3-D visualization, and validation of theoretical backscattering models for individual fish. Resources have also been allocated to test, evaluate, and improve advanced technologies (e.g., broadband, multibeam acoustics, theoretical backscatter modeling, neural network classifiers) in collaboration with industry, academics, and other government agencies for future implementation on NEFSC fisheries acoustic surveys. Fisheries acoustic research efforts will be expanded to include other commercially important species of fish and squid depending on future funds and staffing.
The target strength (TS) of an individual is an important scaler for converting volume or areal backscatter into abundance. TS is a decibel (dB) measure of the backscattering cross-section from an individual target, which is related to the ability of a target to reflect sound back to a transceiver. Fish with air-filled swimbladders, like Atlantic herring (Clupea harengus), have higher individual TS measurements in comparison to fish without swimbladders. TS measurements also vary with fish length (L). TS-L regressions for fish species are often derived using their total length to the nearest cm (LTL). We are presently using the TS-LTL regression (with b = -71.2) developed by ICES for Atlantic herring in the North Sea, although we recognize that Atlantic herring in the Northwest Atlantic may require a different conversion. Seasonal and regional difference in morphology, physiology, and behavior can affect the TS-L relationships of fish. Therefore, in-situ TS measurements from Atlantic herring have been collected from field experiments and surveys during 1997 to present in an attempt to define their TS-L relationship and its variability in the Northwest Atlantic.
Acoustic data have been collected with an EK500 operating three hull-mounted transducers (12-kHz single beam or 18-kHz split-beam, and 38- and 120-kHz split-beam). Biological composition is determined from High Speed Midwater Rope Trawl catches and underwater video. In general, individual in situ target strengths are difficult to obtain from aggregating fish. One hurdle with using in situ TS data is that multiple targets from tightly aggregated organisms (like schooling herring) can be incorrectly recorded as individual targets by the EK500. Using the 38- and 120- kHz split-beam data and an algorithm derived by Demer et al. (Demer, D.A., M.A. Soule, R.P. Hewitt. 1999. A multiple-frequency method for potentially improving the accuracy and precision of in situ target strength measurements. J. Acoust. Soc. Am., 105:2359-2376), we are able to reduce the acceptance of false individual targets. From our preliminary results, this multifrequency filter removes about 98-99% of the TS measurements. For our trawl-acoustic data comparisons, we use TS data that surround (4 nmi before or after and at regular survey speed) trawl hauls where more than 95% of the catch by number and weight are Atlantic herring. Another question is how to separate TS distributions resulting from two types of organisms. It is unclear from the literature how best to define (separate or truncate with thresholds) species-specific TS distributions from field data. We plan to investigate this problem using the multifrequency visualization tools of SonarData's Echoview software. This program will provide the ability to more closely examine and separate TS data in relation to multifrequency backscatter data. Examining differences between the 12, 38, and 120 kHz data should also allow improved separation TS distributions from swimbladder and non-swimbladder organisms.
Conversion of acoustic data to biological metrics such as fish length, abundance, and biomass has been studied for quite some time. In general, empirical regressions between acoustic measurements and biological metrics have resulted from these studies and are used throughout the fisheries acoustics community (see In Situ Research). However, empirical measurements alone do not ensure accurate conversion of acoustic target strength to fish length. Modeling acoustic backscatter provides a quantitative tool to examine variability in backscatter measurements, to improve estimation of target size, and to improve discrimination among types of acoustic targets. Laboratory measurements of individuals and aggregations furnish verification of model predictions and provide direction for new measurements and models.
We are collaborating with scientists, post-docs, and graduate students at academic institutions (University of Washington and Woods Hole Oceanographic Institution), government laboratories (Alaska Fisheries Science Center) and Naval Research Laboratory-Stennis Space Center), and industry (Simrad and Scientific Fisheries Inc.) to develop new methods for experimental measurements and model predictions. These collaborations have resulted in new applications and models with the potential for improving fisheries estimates using acoustic technology.
We conducted a series of backscattering measurements using live, adult alewife (Alosa pseudoharengus) in a 7x7x7m tank. A greater-than-octave bandwidth (40-100 kHz), shaped, linearly swept, frequency modulated signal was used to insonify individual alewife. An individual alewife for each series of measurements was tethered and rotated at 1o increments in two planes of orientation (dorsal/ventral and lateral). Dominant acoustic scattering mechanisms were identified through both spectral and time-domain (pulse compression) analyses. Backscattering amplitudes for all angles of orientation were modeled by using a Kirchhoff Ray-Mode model and digital images of the fish body and swimbladder morphometry. Digital images of fish anatomy and morphometry have been quantified using multiple techniques, including dissection, traditional x-rays, Computerized Tomography (CT) scans and Phase Contrast X-rays (PCX). These data have been used in the development of a new model as well as further development of two existing models: 1) conformal mapping technique, 2) Kirchhoff Ray-Mode (KRM) model, and 3) a modal-series-based deformed cylinder model. Interactive representations of fish body and swimbladders used in the KRM model, and the resulting acoustic backscattering response curves, surfaces, and ambits are available on the web (www.acoustics.washington.edu).
As part of our annual herring acoustic surveys we allocate 2-3 weeks for collaborative research. The goals of this research are to investigate, field-test, and evaluate the efficacy of advanced technologies for improving fisheries acoustic estimates of fish and zooplankton abundance, biomass, and sizes. One goal of the fisheries acoustics group is to incorporate multiple frequency techniques into acoustic estimates. During 2000 and 2001, we were able to conduct in situ experiments with two different systems (in addition to the standard NEFSC frequencies) that greatly expanded the frequency spectrum of our measurements. We intend to compare Sv measurements and backscatter from individuals across this frequency range and also to compare model predictions for Atlantic herring.
During the 2000 and 2001 fall herring acoustic surveys, Gerald Denny of Scientific Fishery Systems, Inc. (SFS, Anchorage, Alaska) collected acoustic broadband data on Atlantic herring aggregations in the Gulf of Maine and Georges Bank regions. The Broadband Fish Identification System was recently developed and is presently being tested by Scientific Fishery Systems, Inc. to improve target classification for fisheries applications. This system is comprised of an RDI 150 kHz broadband transducer (110-190 kHz) and modified RDI sounder hardware. The system is a portable unit which includes a light-weight (100 kg) tow-body that can easily be deployed and towed at 2-5 knots about 5 m below the surface. The SFS system's power cable (carrying 110 volts between the towed-body to a portable PC based processor in the dry-lab) was tendered to the side of the vessel during deployments. Broadband signals were processed and analyzed using a neural network, then displayed as acoustically classified targets. During these cruises, we sampled herring aggregations and attempted to find single-species aggregations of other species to train the neural network. We have collaborated with SFS to analyze and compare the broadband data with single frequency (12, 38, and 120 kHz) acoustic data.
During the 2001 fall herring acoustic survey, Redwood Nero and Charles Thompson of the Naval Research Laboratory (NRL at Stennis Space Center, Mississippi) collected low-frequency acoustic data on herring aggregations. Their system, Low Frequency Fish Sonar (LFFS), was developed for low-power, multi-frequency data collection for fish detection. The NRL USRD Model G81 Ring Transducer was powered with an Instruments Inc. model L6 power amplifier and outputs approximately 185-195 dB re 1 uPa at the frequencies of 1, 2, 3, 4, and 5 kHz. Normal operations were to transmit one CW ping (10 ms duration) at each frequency at two-second intervals. Bandwidths for each frequency are approximately 100 Hz. The sonar was deployed when herring aggregations were observed in echosounder data. The source transducer and line-receive array were towed aft of the ship via a pontoon raft at 2-4 knots. The LFFS transmit frequencies measure the resonance range of fish-sized gas-bearing organisms. We are currently collaborating with NRL to compare backscattering responses among the different systems.
During October 2001 we collected EK500 acoustic data on board the FRV Delaware II while Gary Melvin (Canadian DFO) collected multibeam data on the CSS Needler using a Simrad SM2000 multibeam system operating at 95 kHz. These data were collected on the northern flank of Georges Bank in areas of high herring concentrations. We are currently analyzing these data for comparisons of spatial distribution and calculations of fish abundance and biomass. We will also be collaborating with Dr. Kenneth Foote (Woods Hole Oceanographic Institution) and Dr. Larry Mayer (University of New Hampshire) to calibrate a multibeam system.
In the northern Georges Bank region, dense concentrations of per-spawning and spawning herring have been consistently encountered during the 1998-2001 Atlantic Herring Hydroacoustic Surveys. Surveys of this region have been completed at least twice for each annual survey using various survey designs, and the acoustical estimates have been consistent. Presently, the timing and coverage of the NEFSC Atlantic Herring Acoustic Survey appears adequate for obtaining annual population estimates for the spawning population of Atlantic herring in the Georges Bank region.
The timing and coverage of the NEFSC Atlantic Herring Acoustic Surveys in the Gulf of Maine region may be inadequate for assessment. The greater area and limited coverage in this region could easily result in herring concentrations that were not surveyed, particularly if their distributions fluctuated annually due to environmental anomalies. Cooperative efforts between agencies and industry need to be increased for surveying herring in the Gulf of Maine.
The objective of collecting data using several different designs in 2000 and 2001 was to use the spatial and temporal distributions of herring biomass to plan future surveys. Our analysis of these data using spatial statistical approaches has improved our understanding of the distribution and abundance of herring and assist with future survey design. When we examined variograms from the 1999-2001 surveys we observed the degree of similarity in spatial autocorrelation among the three designs. Given that geostatistical approaches allow for the examination of the relationship between survey design, sample location, and precision of estimates, we will be able to determine which design performed the best and use this information in future. We can also take advantage of the spatial structure in the data by geostatistical modeling to obtain minimum variance estimates from our surveys.
For the future, a design that incorporates the three strata and parallel transects will probably be appropriate for estimating the abundance of herring in the Georges Bank region. We will also have the time in the future to continue to use a stratified random design. This may allow us to understand and improve our estimates of the variance even further. It also will be helpful to have data and analyses from a more classical design that may prove to be an advantage in our assessment process.
In addition to collaboration on research topics, interagency coordination and cooperation has developed and evolved through annual workshops and cooperative survey efforts in the Gulf of Maine and Georges Bank regions. Scientists from the NEFSC, Canadian Department of Fisheries and Oceanography (DFO), Maine Department of Marine Resources (MDMR), and Gulf of Maine Research Institute (GMRI/UM) have met annually to discuss their herring acoustic research and assessments. While interagency coordination among US and Canadian fisheries scientists has long been fruitful, these acoustic workshops were the first to build working relationships among groups conducting acoustic assessments in the Northwest Atlantic. The workshops provided an informal gathering of scientists and representatives of industry to exchange information and discussion regarding the implementation of fisheries acoustic technology to improve the population estimates of Atlantic herring stocks in the Northwest Atlantic. The overall objective of the workshop was to improve cooperative operational and scientific efforts to obtain more accurate, cost-effective, and timely population estimates for Atlantic herring from various regions in the Northwest Atlantic using fisheries acoustic technologies.
During the first workshop held in Woods Hole, MA, scientists provided an overview of their field operations and post-processing procedures. Fisheries acoustic research by the NEFSC included EK500, omni-directional sonar, midwater trawling, and underwater video operations in the Georges Bank and Gulf of Maine regions. Scientist from the DFO St. Andrews Biological Station described their operations with Femto echo-sounders, side-scan sonar, trawling, and seining from commercial and research vessels on the Scotian Shelf and Georges Bank. DFO had also begun collaborative efforts with the University of New Brunswick to develop calibration and post-processing procedures for the recently developed SM2000 multibeam system. MDMR presented data using a Simrad EY500 aboard charter vessels to survey herring in coastal waters and nearshore banks in the Gulf of Maine. GMA/UM initiated a project to implement automated acoustic data loggers aboard commercial herring vessels. An overview was also provided on Simrad BI500, SonarData, and Femto post-processing software. The need to coordinate field operations, compare procedures and results, and develop cooperative herring acoustic research in the Northwest Atlantic was recognized at the first workshop. The Second Annual Northwest Atlantic Herring Acoustic Workshop occurred at the Darling Marine Center in Walpole, Maine. Overviews of recent research and assessments by NEFSC, DFO, MDMR, and GMA/UM scientists were presented. The results from an inter-vessel comparison between the F/V Mary Ellen (commercial fishing vessel chartered by GMA/UM) and FR/V Delaware were also presented. Scientists discussed the importance of accurate calibrations. Approaches for deriving herring abundance and biomass estimates using acoustical data were discussed and recommended. Representatives from the herring fishing industry also attended part of the workshop to provide suggestions for improving survey operations. This workshop was concluded with the planning and coordination of the 2000 field operations. The Third Annual Northwest Atlantic Herring Acoustic Workshop focused on specific topics identified during the second workshop that had the largest affect on the estimates and variance of the acoustical measurements. Participation was increased to include additional Canadian scientists from Quebec and Newfoundland who had analytical expertise in estimating abundance of Atlantic herring using acoustical methods. The goal of this workshop was to evaluate ongoing research and identify research requirements of fisheries acoustics to improve population estimates and variance for Atlantic herring assessment.
We participated on a study group formed by the International Council for the Exploration of the Sea (ICES) Fisheries Acoustics Science and Technology (FAST) Working Group (Study Group on Target Strength Estimation in the Baltic Sea). The goals of this study group were to identify biological and physical factors influencing fish target strength estimation in the Baltic Sea. The study group used Atlantic herring and sprat (Sprattus sprattus) as specific cases, but results from this study will have direct consequences for acoustic estimates of Atlantic herring in the northwest Atlantic. A sampling and backscatter modeling program and a fish anesthesia and radiographing protocol were developed to standardize data collection for backscatter modeling.