North pacific research board project final report

Дата канвертавання20.04.2016
Памер223.03 Kb.
1   2   3   4

For the FIT data set, we have a large number of tag recoveries within each length group, by recovery fishing gear. The results indicate that longline and trawl fisheries recovered almost equal numbers of tags, while the pot fishery recovered a few less (Figure 10). The trends were similar when we considered only those recoveries with days at liberty less than a year (Figure 11). If we assume the standardized recovery curves to be an approximation of true selectivity curves, we can conclude that the selectivity curves for pots and trawls conform to a symmetrical dome shape, and that longlines tend to select the largest cod (Figures 12 & 13).



The four original data sets are too severely disjointed, either in temporally or geographically, to allow estimation of movement rates among regions of the Bering Sea or between the Gulf of Alaska and the eastern Berng Sea. No subsets could be identified that permitted the model (Hilborn, 1990; Anganuzzi et al, 1994) to be specified without irresolvable overparameterization. In order to apply our proposed movement model to quantify Pacific cod movement, we need to have independent estimates of exploitation rate and reporting rate by release batch, recovery area and period. In the movement model the exploitation, reporting, and movement rates are confounded. Without independent estimates of exploitation or reporting rates, the model will not be able to reliably estimate the movement rates. There are catch statistics but no abundance estimates by strata and season. Therefore, we cannot directly estimate exploitation rate by strata or season using a stock assessment program. However, fisheries independent abundance survey could obtain a snapshot of fish abundance or index of abundance by strata.

We also need direct estimates of tagging induced mortality and tag loss rate by release batch or area. There are two types of tagging induced mortality, i.e. acute and chronic. While acute tagging induced mortality effectively reduces number of tagged fish available for recovery, the chronic tagging related mortality would inflate model estimated natural mortality or negatively bias survival rate estimated by the model. The tag-loss rate affects model estimates in the same manner as tagging induced mortality. There are also acute and chronic components in tag loss. Data needed to obtain such independent estimates are not available or are incomplete.

In order to obtain independent estimates of exploitation, reporting, tagging induced mortality and tag loss rates, we propose a mark-recapture experiment to estimate movement rates. The experiment includes a joint tag release and abundance survey cruise, followed by a fishery independent tag recovery and abundance survey cruise. Included in the experimental design are mechanisms to ensure mixing of tagged fish in the population, coverage of all geographic strata, estimation of ancillary parameters (in particular, the exploitation rate), and means of defraying costs through directed research catches. The movement rate estimation model would be simplified using the data from such an experimental design, yet still remain close to being overparameterized. The assumption of a single, straight-line movement between mark and recapture remains but the experiment is designed for a short enough time period to allow this assumption to be valid. Within the overarching experiment, there will be several sub-experiments.

Before we discuss in details of each sub-experiment, we present a likelihood model based on a Halibut movement model (Skalski, FISH 558 lecture notes).

Where, l = 1, 2, 3, …, L, release strata.

t = 1, 2, 3, …, T, release time. In our proposed experiment, T = 1.
= multinomial coefficient for release Rlt.
i = 1, 2, 3, …, I, recovery strata.
j = 1, 2, 3, …, JT, recovery time. In our proposed experiment, JT = 1.
Rlt = number of tags released in stratum l at time t.
rlt = total number of tags recovered from releases Rlt.
ljtl = number of tags recovered from release Rlt in stratum i at time j.
, combined probability of movement rate from stratum l to i (mli), reporting rate  = Pr(tag reported|harvested), and survival rate (S) assumed to be constant over a short time period, utilization rate (Utij). Utij is Pr(harvested in time period j|escaped exploitation from release time t (t = 1, 2, 3,…, T) to recovery time j (j = 1, 2, 3, …, Jt), i.e. , Since in our proposed experiment T = 1 and JT = 1, . Vi is exploitation rate at stratum i, . Ci is total catch in stratum i and is population or abundance in stratum i and is estimated from survey CPUE and stock assessment model estimated overall abundance in entire eastern Bering Sea.

Sub-experiment 1: Direct estimate of exploitation rate by strata.

In this sub-experiment, we will carry out two cruises. The first cruise will be dedicated to releasing tags over the entire geographical distribution of Pacific cod in the eastern Bering Sea in combination with an abundance index survey in November. The second cruise will be in late March or early April and be dedicated to tag recovery over the entire eastern Bering Sea, with disproportionally more effort allocated to where no commercial fishery is present. As with the first cruise, we all also carry out an abundance index survey. We will get estimates of average catch per unit effort (or area) (CPUE) by cruise and stratum. Then we have:

Where, N = total abundance of Pacific cod in entire eastern Bering Sea from stock assessment.

= average CPUE of stratum i. = overall average CPUE of all strata in eastern Bering Sea, = area of stratum i and A = total area of all strata in eastern Bering Sea. Hence, we can estimate exploitation rates by strata.

Sub-experiment 2: Direct estimate of reporting rate by strata

During the tag release cruise in November, we propose to release a small number of high reward tags in each stratum. The number of high reward tags shall be proportional to number of regular tags released in each stratum. During tag recovery cruise, we shall seed tags in catches and derive direct estimate of tag reporting rate.

Sub-experiment 3: Direct observation of acute tagging induced mortality

During the tag release cruise, we will systematically hold tagged fish in a live tank for at least 24 hours to observed tagged fish survival rate, using a consistent culling criteria. We actually do not need a control sample (holding untagged fish). The main reason that no untagged fish are needed is that our goal is to estimate tagging induced mortality, which includes mortality due to tagging, barotrauma, and handling stress. Pacific cod mortality due to natural causes probably negligible over a few days.

Sub-experiment 4: Direct estimate of tag retention rate

Tagged fish held in live tanks for tagging induced mortality observation can also be used for acute tag loss observation.

We will double tag a sufficient number of fish in order to directly estimate chronic tag losses (Gulland (1963). Since funding and labor resources are limiting factor here, we will probably have to release as many tags as possible whether they are single or double tagged.

survival and exploitation rate

The estimated survival rate of Pacific cod varied from 0.3617 to 0.5384 and exploitation rate varied from 0.1612 to 0.3224 (Table 14). Model-predicted survival from 2002 to 2003 (0.4882 to 0.5879) appears significantly higher than that from 2003 to 2004 (0.3091 to 0.4179). The higher apparent survival from 2002 to 2003 probably resulted from later releases of tagged cod in 2002 (April, 2002). Therefore, the model predicted survival rate is for the period from late April through the end of 2002, rather than an estimate of annual survival rate.

Model-predicted exploitation rate is highest in 2003 (0.3224), and lowest in 2002 (0.1612), with 2004 in the middle (0.2587). The predicted 95% confidence limits for exploitation rates during these three years do not overlap (Table 14). Hence, it appears that the model-predicted exploitation rates are statistically different. However, this difference may actually be due to the timing of tag releases. The lower exploitation rate for 2002 probably resulted from later releases (April 2002). The model assumes that the tag releases were carried out at the same time and immediately before the fishing season each year. Nevertheless, the tagged cod in 2002 were not exposed to the A fishing season during winter 2002 (January to March). Historically the A season landings comprise almost 50% of annual landings in the Eastern Bering Sea. Therefore, the exploitation rate estimate for 2002 has substantially underestimated the annual exploitation rate and instantaneous fishing mortality. The 2003 estimate may have overestimated the real exploitation rate, because in 2003 tags were released during peak A fishing season and the release location was a major spawning ground, cod alley. Cod alley is a major winter cod trawl fishing ground, and cod released in February 2003 may have suffered a disproportionately high fishing pressure. The model-predicted exploitation rate of cod in 2004 is probably the least biased. The tagged cod were released in November 2003, which is about two months before the winter cod fishing season. During these two months, the cod fishery in the eastern Bering Sea was insignificant. Therefore, the tagged cod had sufficient time to mix with untagged cod before the fishing season. Due to the small number of tags released and substantially high tagging-induced mortality, the estimate has a relatively large standard error (Table 14).

To minimize the mixing and fishing effect on model-predicted population parameters, we suggest a mark-recapture experiment to estimate survival and exploitation rates. The experiment should consist of at lease three (years) releases. Each release should be carried out in November with equal number of tags released. If the reporting rate and survival rate (from tagging) are maintained at the current level, 2000 tags should be sufficient. As with the mark –recapture experiment for modeling Pacific cod movement, the experiment should include consistent culling practices, and on-deck live tank monitoring of tagging survival. If resources permit, a small number of double-tagged fish should be released in order to estimate tag retention rate.

The estimate of instantaneous natural mortality in 2002 is 0.4029 (SE = 0.0387), slightly lower than that of 2003 (0.5033 with SE = 0.0754). Based on the point estimates and their standard errors, the estimated natural mortality rates are not statistically different. On the other hand, the estimate of instantaneous fishing mortality in 2002 is 0.2162 (SE = 0.0122), less than half that of 2003 (0.5136 with SE = 0.0200). We believe that the fishing mortality in 2002 was significantly negatively biased, because tagged cod released in 2002 were not exposed to the winter cod fishing season. We also think that the fishing mortality of 2003 was positively biased due to their location (major trawl fishery fishing ground) and time (peak fishing season) of release.

Size specific recovery rate and approximation of selectivity curve

Recovery rates by size group show a dome shape curve with a peak at around 70 cm for FIT data. After scaling the recovery rate curve so that the maximum for each recovery curve equals 100%, these standardized recovery curves could be a reasonable approximation of the overall selectivity curve for the fisheries. However, this approximation was based on a suite of implied assumptions, i.e. size independent natural mortality, size independent tagging related mortality, size independent availability and spatial distribution, and size independent reporting rate.

The uncertainty related to some of the assumptions, such as size independent natural mortality, can be addressed through experimental design. If we release sufficient number of tags in each size (length) group, we can fit a joint Brownie model for each size group and predict survival and exploitation rates by size group.

Conclusion and Recommendation

From this Pacific cod tagging data analysis, we conclude that tagging experiments are critical to understanding Pacific cod biology and behavior as well as Pacific cod fishery. With proper study design, one can model Pacific cod movement, survival and exploitation rate. With careful design, one also can use tagging analysis to estimate fishery selectivity and/or fishing gear selectivity curves. A tagging study for estimating survival could also be implemented as a long term population monitoring program. Such a monitoring program would also accumulate a time series on cod survival. When it is executed with archival tags, it is possible to collect tempo-spatial environmental data at the same time, which can be applied to directly monitor the population impact of oceanographic and climate changes in the Bering Sea (decrease in sea ice, increase in water temperature, etc.).

Therefore, we recommend following studies:

  1. A mark-recapture experiment to estimate movement rates. The experiment includes a joint tag release and abundance survey cruise followed by a fishery-independent tag recovery and abundance survey cruise. If this experiment is successful, we further recommend multi-seasonal release and recovery cruises to ascertain seasonal migration patterns.

  2. A mark-recapture study to estimate natural and fishing mortality rates. The experimental design should utilize at least three releases, one in each year before the major fishing season begins (preferably in November before the winter cod fishery opening). Ideally, this mark-recapture study would become a multiannual stock and oceanographic monitoring program.

  3. Release tagged fish by size groups, in proportion to their population sizes so that the tag return data can be use to estimate size specific survival and selectivity.


Shi, Y.B., P. Munro, D.R. Gunderson. (in prep.) Estimating Movement, survival and exploitation rates of Pacific cod Gadus macrocephalus in the eastern Bering Sea and the Gulf of Alaska using mark-recapture methods. NOAA Processed Report


Dr. Libby Logerwell gave a talk at the Lowell Wakefield symposium in fall 2006 on "cod spawning, movement and effects of commercial fishing".

Yunbing Shi gave a talk at Pacific cod workshop in June 2007 on “Pacific cod movement and survival”.

We are continuing sending out tag rewards to fishermen who report and send recovered tags to us.

Mr. Peter Munro met with longline fishery representatives in September 2007 to discuss cooperative research opportunities.


We would like to thank Liz Conners and Sandi Neidetcher for their help with reward processing and data entry. We would to thank all fishermen and observers for reporting and returning tags. We also would like to thank NPRB for funding this project.

Table 1. Number of tags released by AFSC RACE summarized by release year and NMFS statistical area, number of tags recovered and recovery rate (%) by release year and release area (Shimada & Kimura, 1994).

Table 2. Number of tags released by ADF&G summarized by year and NMFS statistical area, number of tags recovered and recovery rate (%) by release year and release area (D. Urban).

Table 3. Number of tags released by AFSC FIT summarized by release month and NMFS statistical area, number of tags recovered and recovery rate (%) by release month and release area. For mortality study fish, only those alive at end of each study were released.

Table 4. Number of tags released by AFSC RACE summarized by release month and NMFS statistical area, number of tags recovered and recovery rate (%) by release month and release area (D. Nichol).

Table 5. Number of Pacific cod tagged, released by cruise and recovered by year in the waters adjacent to Unimak Island during 2002 and 2003. Tags used in mortality experiments are excluded in this analysis because those fish may experience different after release survival.


Number Released
Nt ()

Number Recovered, Rtj





Total (Rt.)

Apr. 2002

1758 (1737)






Feb. 2003

3403 (3311)






Nov. 2003

709 ( 513)






Total (Cj)












Table 6. Number of tags released and recovered by areas (BS, AI, or GOA) and percentage of recoveries from areas outside of area of releases.

Table 7. Number of tags released and recovered by areas shows great site fidelity as well as transoceanic migration capability of Pacific cod.

Table 8. Percentage of tag recovery by release and recovery area showed great site fidelity of Pacific cod in Alaska waters.

Table 9. Minimum distance (rounded to the nearest nautical miles) between release and recovery locations. Over 96% of recovered tags released by FIT were released in Bering Sea (BS). About 95% of recovered tags released by ADF&G were released in the Gulf of Alaska (GOA). 96% of recovered tags released during RACE I study were released in Bering Sea (BS). RACE II-G is for tags released in GOA and RACE II-B for BS.


















Standard Deviation












1st Quartile












3rd Quartile











1   2   3   4

База данных защищена авторским правом © 2016
звярнуцца да адміністрацыі

    Галоўная старонка