Groundfish Forum’s Exempted Fishery Permit (EFP) Tests of a Trawl Modification for Reducing Roundfish Bycatch in Bering Sea Sole Fisheries

By John Gauvin and Craig S. Rose


In April of 1997, the North Pacific Fishery Management Council recommended approval of Groundfish Forum’s request to receive an exempted fishing permit (EFP).  Groundfish Forum’s EFP application requested the Council and NMFS agree to set aside approximately 4,500 MT of yellowfin sole and associated bycatch species to be harvested in a test of the effectiveness of a trawl net innovation to reduce the capture of pollock and other round fish in flatfish trawls.  The experiment was designed to deliver a statistically reliable assessment of the effectiveness of a pollock exclusion device under a number of relevant types of fishing conditions and strategies employed by vessels of different sizes targeting flatfish in the Bering Sea.

The “open top intermediate” (OTI) trawl innovation tested in the EFP was designed by Dr. Craig Rose of the Resource Assessment and Conservation Engineering Division, Alaska Fisheries Science Center, National Marine Fisheries Service (Seattle).  Dr. Rose’s presentation at a bycatch reduction workshop describing his early tests of the device indicated significant potential to reduce pollock bycatch in flatfish trawls.  Rose had tested the OTI during a research fishing cruise in the Gulf of Alaska and another in the Bering Sea.  The scale and other characteristics of the research tows were somewhat different from flatfish fishing by “head and gut” (H&G) vessels in the Bering Sea.  After reviewing the paper, the relevant question for Groundfish Forum members was whether the gear would work when applied to their flatfish trawls.

Groundfish Forum’s interest in developing methods of avoiding pollock catches is fueled by the economics of flatfish vessels.  H&G boats generally cannot produce a head and gut product from pollock that sells for more than the costs of production.  The problem is the low relative value of pollock which is compounded by limited freezer hold space on H&G boats.  To be profitable, hold space must be used for higher-valued flatfish products.  This explains why a significant portion of the pollock caught by H&G vessels has been discarded in the past.  According to a NMFS report on utilization in groundfish fisheries, discards of pollock in the flatfish fisheries were approximately 29,000 MT in 1995.  Although H&G vessels are not the only vessels fishing flatfish that discard pollock in flatfish fisheries, more than half of these discards are probably attributable to H&G vessels.

The impetus to avoid catching pollock has lead to considerable individual company efforts to fashion pollock exclusion devices and other net modifications.  Although individual efforts to develop solutions are valuable, the ability to thoroughly and reliably test the actual effectiveness of such devices is limited during competitive fisheries.  The H&G sector’s experience has been that ad hoc tests are commonly abandoned when initial results fail to meet expectations.  In competitive fisheries, to continue to refine a device might jeopardize the economic position of the company.  Furthermore, lacking an adequate experimental design, results can be less than reliable.  At times, substantial gear modification investments have been made when more testing under an adequate experimental design would have demonstrated beforehand the limitations of a device.

There have always been incentives to avoid catching pollock rather than discarding them.  This is because discarding pollock involves sorting and handling costs to the vessel as well as the fishing costs associated with a portion of the net filled with species other than the flatfish target.  The problem has been that the techniques and fishing strategies available for avoiding pollock have been less feasible than the cost of discarding unwanted pollock.  Pollock discards are counted against the pollock quota, but until recently, the only individual cost associated with unwanted pollock catches was the handling costs and the fishing effort squandered on non-target catch.

The need for better methods to avoid pollock bycatch became acute recently with the approval and impending implementation of Amendment 49.  Starting in 1998, every groundfish vessel will have to retain all pollock catches regardless of target fishery.  These new regulations pose a great deal of potential economic impact on the H&G fleet because freezer hold space is very limited and H&G vessels cannot make fishmeal out of unwanted pollock catches.  While many trawl pollock vessels can turn pollock into fishmeal or lower recovery products such as fillets, H&G vessels cannot practically install fishmeal plants or are not designed or permitted for filleting machines.  The only options for H&G boats is to avoid catching the pollock or make products with relatively high recovery rates, thus constraining product hold capacity and hence economic returns.

Because they generally agree with the goal of reducing discards, Groundfish Forum members did not oppose the requirement to retain all pollock and cod starting in 1998.  Instead of trying to modify the objective of the plan, members have endeavored to develop excluder devices and lower profile net designs as well as increased product and market development for the pollock catches that cannot be avoided.  Groundfish Forum’s EFP was one part of its memberships’ efforts to improve fishing techniques and reduce unwanted catches.  As can be seen from the report that follows, the OTI as currently conceived will probably not be the salvation of the H&G fleet.  Modifications to the basic approach, however, have already been spawned.  Furthermore, the fleet has seen first hand the need for a well-designed test to demonstrate the effectiveness of potential innovations.  Perhaps the most important result is that the H&G fleet has seen the value of working together to improve fishing methods.

The preliminary report below details the manner in which the experiment was conducted and the most general statistical results of the field test.  This report is intended to convey the principle results of the gear test to the industry and interested public in a timely manner.  More detailed analysis of the data generated from the study will be performed by NMFS scientists and other researchers.


Dr. John Skalski provided the statistical design for this project under a consulting contract with Groundfish Forum.  The Groundfish Forum would like to thank Dr. Craig Rose for helping to design our study, providing underwater camera coverage during the experiment, and for assisting us with the statistical analysis.  Without Craig, we would not have been able to accomplish our objective.  We would also like to thank Dr. Gary Stauffer for making some of Craig’s time available for this work and for providing guidance throughout every step of the EFP.  Dr. William Karp, NMFS Observer Program, also helped us determine the best way to collect our species composition samples for the experiment.  We would also like to thank the North Pacific Council and the Alaska Region of NMFS, (Steve Pennoyer and Ron Berg in particular), for making sure the application received timely review and for trusting us with conducting the experiment and staying within the limits for catch and bycatch.  Andy Smoker and Kent Lind deserve our thanks as well for serving on the application oversight committee and writing up the Environmental Assessment and other paperwork tasks critical to the experiment.

We also want to thank the 12 NMFS-trained observers who dutifully followed an expanded sampling regime during the experiment to ensure our data were of sufficient quality for the experiment.  Alaskan Observers and Northwest Observers also deserve our gratitude for providing us with 12 experienced observers willing to conduct the expanded sampling and for accommodating our timing for the experiment.

Dr. Jack Tagart, Washingon Department of Fisheries, and Dr. Keith Criddle, University of Alaska, provided direction for the statistical analysis.  The more in-depth analysis that will be performed later by Dr. Rose and others will likely be able to take full advantage of some of the more refined analytical techniques outlined by Drs. Tagart and Criddle.

Lastly, the companies that participated in the experiment deserve our thanks.  Five of the six vessels that participated in the experiment were Groundfish Forum member vessels.  All participants incurred the expense of the gear purchases and fishing time, as well as the extra expense of having two observers and other non-employee persons aboard during the experiment.  As compensation, participants were allowed to retain their catches for sale provided they stayed within the directed fishing standards for the yellowfin sole fishery.  Our feeling however, given the results of the experiment, was that the participation was a break-even scenario at best for all participants.  Groundfish Forum appreciated the willingness of all participant companies to commit up front to the completion of the experiment regardless of sub-optimal economic returns.


To test the ability of the OTI to exclude pollock while retaining sole species under commercial conditions, a permit was obtained for an exempted fishery in the early August 1997.  This fishery included five head and gut factory trawlers and one vessel that has primarily targeted pollock for fillets.  Vessels participating in this fishery were selected based on the following information:

  1. Diagrams of the vessels fishing gear, including both experimental and control nets, if two trawls were to be used;
  2. Description of an implementation of the OTI concept that would be installed in the vessel’s trawl;
  3. A description of the vessel itself, with emphasis on facilities and configuration to allow extended observer sampling of the catch; and
  4. Written and signed agreement to abide by experimental protocols and other permit requirements.

Preliminary examination of observer data from the yellowfin sole fishery indicated a somewhat different species composition of catches by smaller flatfish vessels, with the most rapid shift in the range of 165 feet.  To assure representation of the range of vessel types and fishing strategies occurring in the fishery, three vessels greater than 165 feet in length and three less than 165 feet were selected for the exempted fishery.  Applicants were provided information on previous configurations of the OTI and developed their own designs, under the constraints that the openings in the trawl intermediate should be at least 16 feet in length and occupy at least 40% of the circumference of that length of the intermediate section.

The six vessels selected are described in Table 1.  They ranged from 133 feet in length to 215 feet.  The smaller three vessels all used two panel trawls, which had lower vertical openings than the four panel trawls used by the longer vessels.  The OTI designs for five of the vessels were quite similar, installed in an untapered section of the intermediate with openings approximating the length and circumference criteria.  The Cape Horn chose to install their OTI in a tapered section farther forward in their intermediate. Because the tapered section has a larger diameter than the untapered section, this resulted in a larger opening, installed in a section that got smaller from front to back.

Three methods were used to alternate between control and experimental configurations.  The Arica used two matched trawls, one with the OTI installed and one in a unmodified configuration (control).  The Brown’s Point and the Rebecca Irene each used the forward sections of one trawl throughout the experiment, installing either intermediate and codend sections with an OTI or a matched intermediate and codend set without the OTI.  The Cape Horn, Legacy and Ocean Peace each used one full trawl with an OTI installed throughout the experiment and installed a mesh cover over the OTI opening during the control tows.  All of these methods achieved the goal of experimental and control tows only differing by the availability of the escape opening in the intermediate.

To increase the likelihood of optimum OTI configuration, NMFS scientists used underwater video systems to observe two of the OTI sections during an unrelated bycatch research charter immediately prior to the OTI experiment.  Underwater video cameras were used to assess the fishing configuration and the reactions of pollock and sole.  In addition, a NMFS scientist with an underwater video system moved between the vessels during the experiment, documenting OTI function on all vessels except for the Cape Horn.

So that the experiment would be representative of commercial fishing efforts, vessels were allowed to fish at any location that would be open for the August 1997 yellowfin sole season, which was to open after the experiment, and were required to follow the directed fishing standards and all other rules that pertain to that fishery.  Each vessel was allocated a portion of the fisheries quotas of target species and of important bycatch species, particularly halibut, tanner crab and snow crab.  These limits were monitored throughout the fishery to assure that the vessel captains could adjust their tactics in such a way that their parts of the experiment would be completed without exceeding the catch and bycatch limits set by the exempted fishing permit.

To improve statistical measurements of the OTI’s effects on target and bycatch catch rates, vessels alternated experimental and control gears to create pairs of tows (blocks) conducted under similar conditions.  Blocking of the data helped to eliminate variations in catch between areas, days, times of day, and between vessels from the analysis.  The gear used for the first tow of each block was randomly determined and the vessel captain was not informed of the selection until after the decision of where and when to tow had been made.  The second tow of each block was made as close in time and space as practical to the first, matching speed and other towing parameters.  Tow duration was allowed to vary within blocks with the goal that similar catch sizes were obtained from experimental and control tows.

A preliminary analysis (reported in Groundfish Forum’s EFP application’s technical appendix by Skalski Statistical Services), using observer data from yellowfin sole fisheries, indicated that a sample size of 150 pairs would be sufficient to have an 60% chance of detecting catch differences of 10%.  To accomplish this, each vessel was assigned a goal of 25 complete blocks (pairs of one control and one experimental tow) to achieve during the experiment.  A small number of short test tows were used to allow a vessel moving into a new area to determine if the catch composition warranted a pair of full tows.  If the initial tow of a pair was unsuccessful (i.e. gear damage or crab pot in the net) or the catch composition made further towing at that site inadvisable (excessive crab or halibut bycatch) vessels were allowed to abandon an incomplete block in a limited number of cases.  These test tows and incomplete blocks were not included in the experimental analysis, though they were accounted for in tracking the target and bycatch limits for each vessel and for the experiment as a whole.

Catch sampling followed NMFS fisheries observer protocols for North Pacific H&G trawlers with a few modifications.  Each vessel contracted two NMFS certified observers who alternated 12 hour shifts to assure that all catches were sampled.  Catch volumes were estimated by determining the shape of the filled codends and taking necessary codend measurements or by taking measurements from tanks into which the catch was placed.

To assure a reliable tracking of each vessel’s use of their halibut bycatch cap and to improve survival of discarded halibut, as many halibut as possible were sorted out of the catch on deck as the catch was dumped into the live tank.  The observer worked with the deck crew to achieve a census of these halibut and to measure them before they were returned to the sea.  To assure that the rest of the catch was available for sampling, no fish were moved out of the tanks into the factory until the deck sampling was completed and the observer went down to the factory.  While this procedure cannot currently be used in commercial fisheries, due to the constraints of sampling for the vessel incentive program (VIP), a specific exemption from these requirements was made for this project

Samples of the catch, to determine density and species composition, were taken by filling baskets from conveyor belts as the catch passed from the holding tank to the factory. Basket samples were taken at several points throughout the emptying of the tank.  The procedures of the NMFS observer manual were followed, except that larger sample sizes were encouraged to improve catch estimates, with a goal of at least 300 kg from each catch. Any remaining halibut coming out of the live tanks were taken from the conveyor and measured.

Subsamples of some species were selected to determine their size composition.  First priority for these length samples was the principal target species, yellowfin sole. Sample of other species were measured as time allowed.

The position and time of the start and end of each tow were recorded by bridge personnel.  They also recorded the depth, towing speed, their own estimate of catch weight and whether the experimental or control net was used. An underwater datalogger was attached to each net to record depth, light level and temperature every 10 seconds while the trawl was fished.  Because these sensors measured ambient conditions at the trawl throughout each tow, their data were used to measure towing durations and to categorize the trawl tows into day or night conditions.

The goal of this experiment was to measure the proportion by which the OTI changed the catch rates of target and bycatch species, particularly yellowfin sole and pollock.  To adjust for varying tow durations, all analyses were done on catch per unit of time rates, not raw catches.  Tow duration was the length of time between when the trawl depth, as indicated by the datalogger, stabilized at the beginning of the tow until the depth began to decrease during retrieval.  Where datalogger information was not available, due to sensor failure or other problems, the difference between the bridge recordings of start and end times were used. This was adjusted by the average difference between bridge and datalogger durations from tows when both were available.

To allow tests for proportional differences with additive models, a (natural) logarithmic transformation was applied to all catch rates.  This also improved the normality of the catch rate distributions.  The parameter which was used as a measure of the effect of the OTI was the difference between the transformed catch rate from each tow with the OTI and the comparable rate from the control tow in the same block (pair).

This parameter was calculated for each block for each major species in the catch.  Means and confidence intervals of LnOTI’s were untransformed, using the exponential function, and decreased by 1 to provide estimates of the proportional change in catch rates due to the OTI.  To test whether there was a significant difference between the effects of the OTI on catch rates of two species (or size groups), the difference between their LnOTI parameters was used as the test parameter.One concern in designing and analyzing this experiment was that light levels and diurnal behavior patterns could change the availability of some species to the trawl or the effectiveness of the OTI.  Mean light level during the tow, as measured by the underwater dataloggers were used to partition tows into day or night categories.  For tows where light data were not available, time cutoffs were used to make this classification, based on the light observations were compared to the mean time of day for the tow.  Comparisons between transformed catch rates for control tows between day and night were used to test whether availability of each species varied diurnally.  Where this was the case, all mixed blocks, those where a daytime control tow was paired with a night experimental tow, or visa versa, were excluded from the analysis, because the effect of the OTI could have been confounded with the diurnal difference in availability.The LnOTI values for each species and for the combined flatfish catches were analyzed with t-tests and analysis of variance (ANOVA) to test the following questions:

  1. Does the OTI have a significant effect on the catch rates of pollock, flatfish and cod?
  2. Are the effects on flatfish and pollock significantly different?
  3. Do these effects vary between vessels?

In addition to the significance tests, estimates and confidence intervals for each significant effect were generated.  The data were also examined along with information on the vessels, their operations and video observations of the gear and the fish to try to identify the reasons for any differences.

Any effect of the OTI on the size selectivity was of interest because of significant price differentials between large and small flatfish.  To test whether the effect of the OTI varied for different sizes of the same species, catch rates by size class were generated and analyzed in the same manner as the species catch rates.  Yellowfin sole were partitioned into four size classes, less than 25 cm, 25 to 29 cm, 30 to 34 cm and greater than 35 cm.  Length-weight relationships from the NMFS Bering Sea groundfish survey indicate that these length classes correspond to weight classes of: less than 160 g, 160 to 300 g, 300 to 500 g, and more than 500 g.  For each block where yellowfin sole were measured from both tows, the number of sole in the length subsample from each of these classes was expanded by both the length subsample fraction and the basket sampling fraction to provide estimates of the number of sole in each size class in the entire catch.  LnOTI values were calculated with these catch numbers for each size group.  Paired t tests between the LnOTI’s for adjacent size groups were then performed to test for selectivity differences.


The exempted fishery experiment to test the OTI was conducted from August 1 to 12, 1997.

Six vessels completed 138 blocks of paired tows in the eastern Bering Sea.  A total of 4,371 metric tons of groundfish were taken of the 4,500 set aside for the experiment.  Monitored bycatch species catches compared to the PSC limits allocated for the experiment are reported in Table 2.  In all prohibited species categories, catches were less than 50% of the upper limit quantity set by the Council and NMFS for the fishery.  Bycatch limits designated for the experiment were based on performance guidelines in the regular fishery for yellowfin sole.

Groundfish Forum administered the PSC limits in the experiment by dividing the available PSC quantities into individual allocations among vessels based on the tonnage of groundfish made available to each participating vessel.  This was done because prohibited species bycatch quantities for the experiment were binding constraints creating the possibility that the catches of one or a few vessels (or even a few careless tows) could have ended the experiment.  This would have been costly for participants who had incurred the gear expenses for the experiment. Groundfish Forum believes that the inherent accountability created from individual allocations of PSCs contributed greatly to our ability to conduct the experiment under a small PSC allocation and an even smaller usage of PSC.

The goal of including a range of towing methods and strategies was clearly achieved (Table 3).  Towing speeds, tow durations, proportions of day and night tows, average depth all varied between vessels, even though they were operating in the same general area most of the time.  All but 18 of the pairs were made within a 60 NM by 120 NM box on the central Bering Sea shelf, where depths ranged from 50 to 80 m and bottom temperatures from 1.3 to 3.8 degrees Celsius.  Outside of this box, seven pairs were carried out in 42 to 44 m depth by a vessel that moved northeast and 11 pairs were made by two vessels that moved south into depths between 75 and 105 m.  Bottom temperatures for the southern tows were in the same range as for the main area, while the northeastern area had temperatures between 6.3 and 6.6 degrees.

Examination of the light readings from the underwater dataloggers indicated that tows could be categorized into night and day by the average light level during the tow (Figure 1).  While not explaining all of the light variability, day/night differences were clearly the most important factor.  For those tows where light data was not available, comparison of the mean time of day during the tow to cut off times of 0700 and 2315 was used for day/night determination.

The distribution of experimental tows by vessel and day/night is presented in Table 4.  The blocks where both tows were during daylight were the most abundant with 81 pairs and at least 10 blocks for each vessel.  There were also a large number of mixed blocks and relatively few night blocks.

A set of t-tests comparing the catch rates of control tows between night and day indicated that pollock and cod were clearly more available to the trawls during the daytime, while there was a marginal indication of such a difference for flatfish (Table 5).  Therefore, the mixed blocks, the pairs that included both day and night tows, were not used in the catch rate analyses due to the potential for confounding of availability differences with the effects of the OTI.  Because so few night blocks were completed, it was decided to also exclude those from this initial analysis.  Day / night differences will be more thoroughly examined in later analyses.

The OTI significantly (p is greater than .001) lowered catch rates of pollock by 55% and cod by 46%.  Unfortunately, the reduction in the flatfish catch rates was also relatively large (44%).  The decrease in pollock catch was significantly greater than that for the flatfish catch at the p < 0.01 level.  However, these rates do not allow separation as well as those indicated in previous studies with smaller scale trawls used in the Gulf of Alaska and with open codends and smaller scale trawls in the Bering Sea (Rose, 1995).

Analysis with a general linear model detected differences between vessels in the effects of the OTI on species catch rates.  These differences were significant at levels from 3.9% (flatfish) to 6.7% (pollock) (Table 6). While these differences will be more rigorously examined in later analyses, the mean OTI effects presented here provide an indication of the variability between vessels.  The effects on pollock and flatfish catches were notably consistent (pollock -58% to -59%, flatfish -33% to -41%) between the large vessels (Browns Point, Arica and Ocean Peace), while they were highly variable between the smaller vessels (Rebecca Irene, Cape Horn and Legacy).  The OTI did not seem to work at all on the Rebecca Irene, where the flatfish loss was greater than that for pollock, with both loss rates relatively low, or the Cape Horn, where the loss rates for pollock and flatfish were high and equal.  The Legacy followed a similar pattern to the large vessels except that its catch rates of both pollock and flatfish were decreased by approximately an additional 10%.  Some factors that could be associated with this increased variability for the smaller vessels include the different OTI design used by the Cape Horn, the less constrained side panels of the Legacy’s OTI and low visibility in the intermediate sections associated with the low opening trawls used by these small vessels.  This last phenomenon is described below, in the section on video observations.

Paired t tests were used to test whether the effect of the OTI varied between size classes of yellowfin sole (Table 7).  These comparisons were based on only those daytime blocks where yellowfin sole size samples were taken for both tows.  Tows from the two vessels where the OTI did not separate pollock from flatfish (the Rebecca Irene and the Cape Horn) were also excluded.  No significant differences were detected.

Video observations were made on the OTI’s of all of the vessels participating in the experiment, except for the Cape Horn.  Time and weather conditions precluded getting the video system and operator to that final vessel.  Those videos have not been fully analyzed and the following observations are based on only preliminary examinations.  All of the OTI’s observed operated in the intended shape, forming a U-shaped trough for the length of their openings.  The rigidity of this shape varied, from the Ocean Peace, where all support lines were extremely tight, to the Legacy, where the side panels of the trough were relatively free to move in the water currents.

The most obvious difference between the observations during this experiment and those collected during previous tests of the OTI was visibility.  The intermediate sections seemed to be closer to the seafloor during these tests, putting them into a layer of particles suspended by the passage of the forward parts of the trawl.  This greatly reduced visibility in the intermediate, likely affecting the ability of fish to orient themselves in the net.  One potential cause of this lower position is the greater water resistance caused by the nets used for the EFP which were larger than those used in earlier research cruises.  Another contributing factor was the low openings of the two panel trawls used by the smaller vessels.  The intermediates of these trawls could only be observed up until the trawl was just coming into contact with the seafloor and as it was retrieved.  During the tow itself, only objects passing very close to the camera could be detected.  Observations of OTIs on the higher opening, four panel trawls used by the larger vessels, were not as fully obscured.  After these trawls had settled into towing configuration, only the bottom half of the intermediate was obscured.  This allowed observations of the escaping individuals, but not those fish which remained near the floor of the net.  As the catch accumulated and increased the pull of the codend, the intermediate lowered, allowing less and less visibility.  The entire OTI was usually obscured within the first 30 minutes of the tow.