Draft - Norton Sound Red King Crab Stock Assessment April 30, 2013 1 Norton Sound Red King Crab Stock Assessment for the fishing year 2013/14 Toshihide Hamazaki 1 and Jie Zheng 2 Alaska Department of Fish and Game Commercial Fisheries Division 1 333 Raspberry Rd., Anchorage, AK 99518-1565 Phone: 907-267-2158 Email: [email protected]2 P.O. Box 115526, Juneau, AK 99811-5526 Phone : 907-465-6102 Email : [email protected]Executive Summary 1. Stock. Red king crab, Paralithodes camtschaticus, in Norton Sound, Alaska. 2. Catches. This stock supports three main fisheries: summer commercial, winter commercial, and winter subsistence fisheries. Of those, the summer commercial fishery accounts for more than 90% of total harvest. Summer commercial fishery started in 1977, and its catch quickly reached a peak in the late 1970s with retained catch of over 2.9 million pounds. Since 1982, retained catches have been below 0.5 million pounds, averaging 0.275 million pounds, including several low years in the 1990s. As the crab population rebounds, retained catches have been increasing. For past several years, retained catch is around 0.4 million pounds. 3. Stock Biomass. Estimated mature male biomass (MMB) shows an increasing trend since 1997, and an historic low in 1982 following a crash from the peak in 1977. However, uncertainty in historical biomass is great, which is in part by infrequent trawl surveys (every 3 to 5 years) and limited winter pot survey. 4. Recruitment. Model estimated recruitment was weak during the late 1970s and high during the early 1980s with a slight downward trend from 1983 to 1993. Estimated recruitment has been highly variable but on an increasing trend in recent years. 5. Management performance. Status and catch specifications (million lbs.) Year MSST Biomass (MMB) GHL Retained Catch Total Catch OFL ABC 2009/10 1.54 A 5.83 0.38 0.40 0.43 0.71 A 2010/11 1.56 B 5.44 0.40 0.42 0.46 0.73 B 2011/12 1.56 C 4.70 0.36 0.40 0.43 0.66 C 0.59 2012/13 1.78 D 4.59 0.47 0.47 0.47 0.53 D 0.48 2013/14
183
Embed
Norton Sound Red King Crab Stock Assessment April 30, 2013
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Draft - Norton Sound Red King Crab Stock Assessment April 30, 2013
1
Norton Sound Red King Crab Stock Assessment for the fishing year 2013/14
Toshihide Hamazaki1 and Jie Zheng 2 Alaska Department of Fish and Game Commercial Fisheries Division
1333 Raspberry Rd., Anchorage, AK 99518-1565 Phone: 907-267-2158
1. Stock. Red king crab, Paralithodes camtschaticus, in Norton Sound, Alaska.
2. Catches. This stock supports three main fisheries: summer commercial, winter commercial, and winter subsistence fisheries. Of those, the summer commercial fishery accounts for more than 90% of total harvest. Summer commercial fishery started in 1977, and its catch quickly reached a peak in the late 1970s with retained catch of over 2.9 million pounds. Since 1982, retained catches have been below 0.5 million pounds, averaging 0.275 million pounds, including several low years in the 1990s. As the crab population rebounds, retained catches have been increasing. For past several years, retained catch is around 0.4 million pounds.
3. Stock Biomass. Estimated mature male biomass (MMB) shows an increasing trend since
1997, and an historic low in 1982 following a crash from the peak in 1977. However, uncertainty in historical biomass is great, which is in part by infrequent trawl surveys (every 3 to 5 years) and limited winter pot survey.
4. Recruitment. Model estimated recruitment was weak during the late 1970s and high during
the early 1980s with a slight downward trend from 1983 to 1993. Estimated recruitment has been highly variable but on an increasing trend in recent years.
Notes: MSST was calculated as BMSY/2 A-Calculated from the assessment reviewed by the Crab Plan Team in May 2009 B-Calculated from the assessment reviewed by the Crab Plan Team in May 2010 C-Calculated from the assessment reviewed by the Crab Plan Team in May 2011 D-Calculated from the assessment reviewed by the Crab Plan Team in May 2012 E-Calculated from the assessment reviewed by the Crab Plan Team in May 2013 Biomass in millions of pounds
Candidate OFL and ABC 1000t. Parenthesis indicates standard deviation
Model B2013 BMSY B/ BMSY Legal male
biomass M FOFL OFL
ABC (0.9×OFL)
S3-1 2.27
(0.44) 1.86 1.22 1.61 (0.22) 0.18 0.18
0.26 (0.04)
0.24
S3-6 2.31
(0.44) 1.66 1.39 1.40 (0.23) 0.24 0.24
0.32 (0.05)
0.29
S3-7 2.20
(0.45) 1.73 1.30 1.42 (0.22) 0.30 0.30
0.367 (0.06)
0.33
Draft - Norton Sound Red King Crab Stock Assessment April 30, 2013
4
6. Probability Density Function of the OFL
S3-1 S3-6 S3-7
OFL profile. CV of the OFL was assumed to be 0.2.
7. The basis for the ABC recommendation For Tier 4 stocks, the default maximum ABC is based on P*=49% that is essentially identical to the OFL. Accounting for uncertainties in assessment and model results, the SSC chose to use 90% OFL (10% Buffer) for the Norton Sound red king crab stock in 2011. For 2013 analyses, we chose 90% OFL (10% Buffer) which was million lb because of remained uncertainties in the model.
8. A summary of the results of any rebuilding analyses. N/A
A. Summary of Major Changes in 2012
1. Changes to the management of the fishery:
In March 2012, the board of fish adopted a revised GHL: (1) 0% harvest rate of legal crab when estimated legal biomass < 1.25 million lbs; (2) ≤ 7% of legal male abundance when the estimated legal biomass falls within the range 1.25-2.0 million lbs; (3) ≤ 13% of legal male abundance when the estimated legal biomass falls within the range 2.0-3.0 million lbs; and (3) ≤ 15% of legal male when estimated legal biomass >3.0 million lbs.
Draft - Norton Sound Red King Crab Stock Assessment April 30, 2013
5
2. Changes to the input data
a. Data update: the 2011/12 winter pot survey, 2012 summer commercial fishery, 2011/2012 winter commercial and subsistence catch finalized
b. New Data: 2012 summer commercial fishery observer data, standardized commercial catch CPUE and CV.
c. Revised data: 1976-1991 NMFS survey NSRKC crab abundance estimates were revised based on original survey data.
d. Dropped data: 1981-85 summer pot survey data were dropped because of the lack of raw data and unverifiable abundance estimates.
3. Changes to the assessment methodology: Following model modification were evaluated
a. See Appendix A for model modification.
4. Changes to the assessment results.
B. Response to SSC and CPT Comments
CPT Review May 7-10, 2012 The team had the following comments:
1. Lack of bycatch data. The CPT requests that some data on bycatch be collected in conjunction with the NPRB project recently funded.
Author response: In 2012 limited summer commercial fishery observer data were collected. The data were included in the model. Further continuation of observer program depends upon availability of funding.
2. Length composition data have been downweighted, but there still is apparent conflict within the model. This is a possible indication of model mis-specification.
Author response: Since specifics of the “apparent conflict” were lacking, it is difficult to respond. See response for not fitting earlier abundance data.
Draft - Norton Sound Red King Crab Stock Assessment April 30, 2013
6
3. There is a need for better justification for the higher natural mortality on animals in the largest length bin (none of the models address dome vs. asymptotic M).
Author response: We re-examined dome shaped vs. constant M =0.24 and 0.30 (Scenario 0 vs. S1-6, S1-6). We did not evaluate asymptotic M. Constant M increased fit to trawl survey abundance and trawl length composition, and observer and winter pot length composition, but reduced fit to commercial catch length composition. Further, projected crab abundance was lower by 4-11%. Additional effects are lower molting probability for larger length classes, and reducing retrospective bias and error. As for calculation of OFL and ABC, constant M assumptions will increase of OFL proportion
))exp(1( OFLF For M = 0.18, it is 0.165, whereas it is 0.213 and 0.259, respectively for M =
0.24 and 0.30. This will result in an increase of OFL by 20-60% from that of M = 0.18.
4. Model does not fit early data, and it was suggested to start prospective analysis in 75, 76, 77, …
Author response: Possible reasons for the model not fitting earlier data are related to assessment crab population model structure: assumption of constant M and constant growth-per-molt. Recent (1996-2012) data show that the crab population has been increasing gradually (average 3% per year) at about 10% of harvest rate, and OFL harvest rate of 16.5%. However, during 1976-1982 periods, especially from 1979 to 1982 estimated total (CL > 73mm) crab abundance increased from 0.9 million to 2.09 million, more than 2 times, under the harvest rate of 50-60% (Table 3). This cannot be explained by the current population model structure. Consequently, the model estimated abundance was higher than observed during those years. This discrepancy can be partially solved by 1) further down weighting length composition data and thus increasing weight of trawl survey (scenario S1-3), 2) assuming less than 1.0 for survey Q (S1-4) (i.e., trawl survey underestimated true crab abundance), 3) time variant growth-per-molt matrix, 4) density-dependent growth/mortality, 5) inclusion of environmental changes, or combinations of all. In this iteration, we evaluated only the first two alternative scenarios. Those alternative model scenarios resulted in better fit of historical data; however, those still will meet an identical model criticisms on the lack of empirical/scientific justification of the model changes (e.g., is there any empirical evidence supporting that historical trawl surveys underestimated true abundance, that historical trawl length compositions are biased, that growth-per-molt changed over time, that density-dependent growth/mortality occurring, or that Norton Sound environmental conditions changed to affect crab population dynamics?). Due to the absence of historical data in Norton Sounds, those uncertainties could not be easily resolved. Simultaneously, prospective analyses show that those historical uncertainties about true historical population dynamics had little influences on population trends of 1996 to present. Another influence of a model better fitting to historical data is calculation of BMSY proc that is an average 82-present modeled biomass for the Norton Sound red king crab. For this, all model scenarios resulted in B/ BMSY proc > 1, suggesting that model misfits to historical data would not affect assessment of recent data.
Draft - Norton Sound Red King Crab Stock Assessment April 30, 2013
7
5. Use the derivative checker to verify that the objective function is differentiable.
Author response: We verified that the objective function is differentiable.
6. Plot histograms for effective sample sizes for the compositional data.
Author response: Implemented
SSC Review on June 4-6, 2012 The current model assumes that selectivity of the trawl survey follows a sigmoid function and Q was estimated 1.0 for length classes 3 through 5. The SSC asks the author to review this assumption given the results of recent studies of trawl survey Q for Bristol Bay red king crab, snow crab and Tanner crab. Author response: We relaxed the trawl survey selectivity function to estimate all length classes with 1.0 for length classes 5 and 6. However, this relaxation did not change the shape of trawl selectivity. Selectivity of all length classes was still estimated to 1.0 or closer. Under a direction of the CPT, we also included evaluated the assumption of survey Q = 1.0 for trawl surveys (i.e., estimated survey abundance is accurate). Holding survey Q=1 for NMFS resulted in greater than 1 Q for ADF&G surveys (i.e., ADF&G trawl surveys OVERESTIMATE crab abundance). Oppositely, holding survey Q=1 for AD&G survey resulted in less than 1 Q for NMFS surveys (i.e., NMFS trawl surveys UNDERESTIMATE crab abundance). Given that overestimation of ADF&G survey abundance is unlikely, we incorporated estimation of NOAA survey Q in the current assessment model.
C. Introduction
1. Species: red king crab (Paralithodes camtschaticus) in Norton Sound, Alaska.
2. General Distribution: Norton Sound red king crab is one of the northernmost red king crab populations that can support a commercial fishery (Powell et al. 1983). It is distributed throughout Norton Sound with a westward limit of 167-168o W. longitude with depths less than 30 m and summer bottom temperatures above 4oC. The Norton Sound red king crab management area consists of two units: Norton Sound Section (Q3) and Kotzebue Section
Draft - Norton Sound Red King Crab Stock Assessment April 30, 2013
8
(Q4) (Menard et al. 2011). The Norton Sound Section (Q3) consists of all waters in Registration Area Q north of the latitude of Cape Romanzof, east of the International Dateline, and south of 66°N latitude (Figure 1). The Kotzebue Section (Q4) lies immediately north of the Norton Sound Section and includes Kotzebue Sound. Commercial fisheries have not occurred regularly in the Kotzebue Section. This report deals with the Norton Sound Section of the Norton Sound red king crab management area.
3. Evidence of stock structure: Thus far, no studies have been made on possible stock separation within the putative stock known as Norton Sound red king crab.
4. Life history characteristics relevant to management: One of the unique life-history traits of Norton Sound red king crab is that they spend their entire lives in shallow water since Norton Sound is generally less than 40 m in depth. Distribution and migration patterns of Norton Sound red king crab have not been well studied. Based on the 1976-2006 trawl surveys, red king crab in Norton Sound are found in areas with a mean depth range of 19 ± 6 (SD) m and bottom temperatures of 7.4 ± 2.5 (SD) oC during the summer. Norton Sound red king crab are consistently abundant offshore of Nome.
Norton Sound red king crab migrate between deeper offshore waters during molting/feeding and inshore shallow waters during the mating period. Timing of the inshore mating migration is unknown; but is assumed to be during March-June. Offshore migration is considered to begin in May-July. Trawl surveys show that crab distribution is dynamic. Recent surveys show high abundance on the southeast side of the Sound, offshore of Stebbins and Saint Michael.
5. Brief management history: Norton Sound red king crab fisheries consist of commercial and subsistence fisheries. The commercial red king crab fishery started in 1977 and occurs in summer (June – August) and in winter (December – May) (Menard et al. 2011). The majority of red king crab are harvested by the summer commercial fisheries, whereas the majority of the winter harvest is in the subsistence fishery occurring near the coast (Table 2).
Summer Commercial Fishery
Summer commercial crab fishery started in 1977. A large-vessel summer commercial crab fishery existed in the Norton Sound Section from 1977 through 1990. No summer commercial fishery occurred in 1991 because there was no staff to manage the fishery. In March 1993, the Alaska Board of Fisheries (BOF) limited participation in the fishery to small boats. Then on June 27, 1994, a super-exclusive designation went into effect for the fishery. This designation stated that a vessel registered for the Norton Sound crab fishery may not be used to take king crabs in any other registration areas during that registration year. A vessel moratorium was put into place before the 1996 season. This was intended to precede a license limitation program. In 1998, Community Development Quota (CDQ) groups were allocated a portion of the summer harvest; however, no CDQ harvest occurred until the 2000 season. On January 1, 2000 the North Pacific License Limitation Program (LLP) went into effect for the Norton Sound crab fishery. The program dictates that a vessel which exceeds 32 feet in length overall must hold a valid crab license issued under the LLP by the National Marine Fisheries Service. Regulation changes and location of buyers resulted in harvest distribution moving eastward in Norton Sound in the mid-1990s. In the Norton Sound, a
Draft - Norton Sound Red King Crab Stock Assessment April 30, 2013
9
legal crab is defined as ≥ 4-3/4 inch carapace width (CW, Menard et al. 2011; equivalent to ≥ 124 mm carapace length [CL]). Since 2005, commercial buyers started accepting only legal crabs of ≥ 5 inch carapace.
Not all Norton Sound area is open for commercial fisheries. Since beginning of the commercial fisheries in 1977, inland waters near Nome area has been closed for summer commercial crab fishery, possibly to protect crab nursery grounds (Figure 2). Extent of closed water changed throughout history. Appendix E shows historical harvest by Stat area.
CDQ Fishery
The Norton Sound and Lower Yukon CDQ groups divide the CDQ allocation. Only fishers designated by the Norton Sound and Lower Yukon CDQ groups are allowed to participate in this portion of the king crab fishery. Fishers are required to have a CDQ fishing permit from the Commercial Fisheries Entry Commission (CFEC) and register their vessel with the Alaska Department of Fish and Game (ADF&G) before they make their first delivery. Fishers operate under authority of the CDQ group and each CDQ group decides how their crab quota is to be harvested. During the March 2002 BOF meeting, new regulations were adopted that affected the CDQ crab fishery and relaxed closed-water boundaries in eastern Norton Sound and waters west of Sledge Island. At its March 2008, the BOF changed the start date of the Norton Sound open-access portion of the fishery to be opened by emergency order and as early as June 15. The CDQ fishery may open at any time (as soon as ice is out), by emergency order. It is possible that the fishery starts BEFORE determination of OFL and ABC.
Winter Commercial Fishery
The Norton Sound winter commercial fishery is a small fishery using hand lines and pots through the nearshore ice. Approximately 10 permit holders participated in this fishery harvesting, on average 2,500 crabs during 1978-2009 (Menard 2011). The winter commercial fishery catch is influenced not only by crab abundance, but also by changes in near shore crab distribution, and ice conditions.
Subsistence Fishery
The Norton Sound subsistence crab fishery mainly occurs during winter using hand lines and pots through the nearshore ice. Average annual subsistence harvest was 5,300 crabs (1978-2007). Subsistence harvesters need to obtain a permit before fishing and record daily effort and catch. There is no size limit in the subsistence fishery. The subsistence fishery catch is influenced not only by crab abundance, but also by changes in distribution, changes in gear (e.g., more use of pots instead of hand lines since 1980s), and ice conditions (e.g., reduced catch due to unstable ice conditions: 1987-88, 1988-89, 1992-93, 2000-01, 2003-04, 2004-05, and 2006-07).
Draft - Norton Sound Red King Crab Stock Assessment April 30, 2013
10
6. Brief description of the annual ADF&G harvest strategy
Since 1997 Norton Sound red king crab have been managed based on a guideline harvest limit (GHL). Detailed history of GHL determination methods are unknown. Since 1999, GHL is determined by a prediction model and the model estimated predicted biomass: (1) 0% harvest rate of legal crab when estimated legal biomass < 1.5 million lbs; (2) ≤ 5% of legal male abundance when the estimated legal biomass falls within the range 1.5-2.5 million lbs; and (3) ≤ 10% of legal male when estimated legal biomass >2.5 million lbs.
In 2012 the Alaska Board of Fisheries adopted a revised GHL: (1) 0% harvest rate of legal crab when estimated legal biomass < 1.25 million lbs; (2) ≤ 7% of legal male abundance when the estimated legal biomass falls within the range 1.25-2.0 million lbs; (3) ≤ 13% of legal male abundance when the estimated legal biomass falls within the range 2.0-3.0 million lbs; and (3) ≤ 15% of legal male when estimated legal biomass >3.0 million lbs.
Year Notable historical management changes 1976 The abundance survey started 1977 Large vessel commercial fisheries began 1991 Fishery closed due to staff constraints 1994 Super exclusive designation into effect. The end of large vessel commercial fishery operation.
Participation limited to small boats. The majority of commercial fishery subsequently shifted to east of 164oW line.
1998 Community Development Quota (CDQ) allocation into effect 1999 Guideline Harvest Limit (GHL) into effect 2000 North Pacific License Limitation Program (LLP) into effect. 2002 Change in closed water boundaries (Figure 2) 2005 Commercially accepted legal crab size changed from ≥ 4-3/4 inch CW to ≥ 5 inch CW 2006 The Statistical area Q3 section expanded (Figure 1) 2008 Start date of the open access fishery changed from July1 to after June 15 by emergency order.
Pot configuration requirement: at least 4 escape rings (>4½ inch diameter) per pot located within one mesh of the bottom of the pot, or at least ½ of the vertical surface of a square pot or sloping side-wall surface of a conical or pyramid pot with mesh size > 6½ inches.
2012 Board of fisheries adopted a revised GHL
7. Summary of the history of the BMSY.
NSRKC is a tier4a crab stock, and direct estimation of the BMSY is not possible. BMSY is calculated as mean model estimated mature male biomass (MMB) from 1980 to present. Choice of this period was based on a belief that PDO shift occurred in 1976-77 could have changed the productivity.
D. Data
1. Summary of new information:
Draft - Norton Sound Red King Crab Stock Assessment April 30, 2013
11
1. Standardized summer commercial fishery cpue and standard deviation was calculated (Bishop 2013) and included in the model
2. Harvests of 2012 summer commercial fishery, and 2011/2012 winter commercial and subsistence fisheries, were updated. For winter 2012/13 harvest data, 2011/2012 winter harvest data were used.
3. 2012 summer commercial fishery observer discard data were included.
Abundance and proportion by length and shell condition
3,5, Appendix B, E
Summer pot survey 80-82,85 Abundance and proportion by length and shell condition
3, 8, Appendix C, Not used for assessment
Winter pot survey 81-87, 89-91,93,95-00,02-12
Proportion by length and shell condition
6
Summer preseason survey 95 Proportion by length and shell condition
Not used for assessment
Summer commercial fishery
76-90,92-12 Harvest, effort, standardized CPUE, and proportion by length and shell condition
1,4, Appendix E, Bishop et al (2013)
Observer data 87-90,92,94, 2012 Proportion by length and shell condition (sub-legal only)
7
Winter commercial and subsistence fishery
76-11 The Number of crab harvested 2
Tagging 80-11 Used to create a growth increment matrix
9
1. Summer commercial fishery and winter commercial and subsistence catch, (ADF&G 1976-2011) (Tables 1 and 2).
2. Discards of sublegal males (observer data) from the summer fishery (ADF&G 1987-90,
1992, 1994, 2012). The survey was opportunistic, and the number of crab discarded was not recorded. Continuation of summer commercial discards observer data depend upon future funding. Only catch-at-length and shell condition of sub-legal male were recorded (Table 8). In Norton Sound, no other crab, groundfish, or shellfish fisheries exist.
Fishery Data
availability Directed pot fishery (males) Summer commercial
Winter commercial/subsistence summer
commercial Directed pot fishery (females) None Bycatch in other crab fisheries Does not exist NA
Bycatch in ground pot Does not exist NA
Draft - Norton Sound Red King Crab Stock Assessment April 30, 2013
12
Bycatch in ground fish trawl Does not exist NA Bycatch in the scallop fishery Does not exist NA 3. Catch at length data for summer commercial fisheries (Table 4). 4. Survey abundance estimates:
1. Triennial Trawl Survey Triennial trawl surveys were conducted by the NMFS (1976-1991, 2010) and by the ADF&G (1996-2011) (Table 3). The NMFS survey was conducted using the 83-112 Eastern Otter Trawl, whereas the ADF&G survey was conducted using the 400 Eastern Otter Trawl. In both surveys, survey design was based on 10×10nm square, except for the NMFS survey in 2010 where survey grid was 20×20nm. Abundance of crabs were estimated by area-swept methods (Alverson and Pereyra 1969; see Appendix B). While the assessment model is based on crab population of ≥ 74mm CL, none of surveys have reported abundance of crab ≥ 74mm CL. NMFS survey reported crab abundance of ≥ 100 mm CL, 0-100 mm CL, and all sizes during 1991 periods, and ≥ 90 mm CL, 0 - 89 mm CL, and all sizes during 1985periods. On the other hand, ADF&G survey reported crab abundance of legal size (≥ 4 ¾ inch CW). To estimate abundance of ≥ 74mm CL crab, we re-estimated abundance of ≥ 74mm CL crab from original raw trawl survey data for ADF&G trawl survey. For 1976-1991 NMFS survey, we took two approaches for estimation of historical NMFS trawl abundances: 1) Indirect method based on published report (See Appendix B).
2. Summer pot survey
Summer pot surveys were conducted in 1980-1982 and 1985 by ADF&G. The pot survey crab abundance estimates were based on Petersen Mark-recapture method. Before fisheries season, crabs were captured throughout Norton Sound, tagged and released. The tagged crabs were recaptured by commercial crab fisheries, from which crab abundance was estimated. Details of procedures and estimates of abundances, however, were lacking except for 1985 (Brannian 1986), and original raw data were presumed lost. Hence, abundance of ≥ 74mm CL crab was estimated from published legal crab abundance and the proportion of length crab (See Appendix C). At the 2013 CPT meeting, it was discussed that all observed data should be reproducible from original data. Because the raw data do not exist, summer pot survey data were dropped from assessment model.
3. Survey catch-at-length data available include: Summer commercial catch (1977-2011) (Table
4), triennial Trawl survey (Table 5) and winter pot survey (Table 6). Other miscellaneous data include: summer commercial catch observer survey (1987-90, 92, 94) (Table 7), summer pot survey (1980-82, 85: Dropped) (Table 8), and summer preseason survey (1995) (Not included for the assessment model).
Draft - Norton Sound Red King Crab Stock Assessment April 30, 2013
13
4. Other miscellaneous data: None.
5. Growth-per-molt (Table 9), estimated from tagging data (1991-2007).
6. Proportion of legal size crab, estimated from trawl survey data (Table 10).
E. Analytic Approach
1. History of the modeling approach.
The Norton Sound red king crab stock was assessed using a length-based synthesis model (Zheng et al. 1998). The model was updated in 2009-2010 to provide information for the federal OFL. At the May 2010 CPT meeting, seven alternative models were presented: 1) based on 2009 model reviewed by Punt (University of Washington), 2) model 1 and including bycatch mortality, 3) model 2 with weight of fishing effort increased from 5 to 20; 4) model 3 with fishery selectivity for the last length group from 0.6 to being estimated from the model, 5) model 3 and reduce the maximum effective sample size for commercial catch and winter surveys from 200 to 100, 6) model 5 with M for the last length group increased from the default 0.18 to 0.288, and 7) model 6 with M increased to 0.34. The CPT and subsequent SSC recommended using the Model 6 for the 2010/11 iteration. During 2011 NPFMC meeting in June, SSC was concerned high hindcast prediction error and bias (i.e., model predicted crab abundance for assessment year tend to be higher than “actual/model reconstructed” abundance, which resulted in higher exploitation rate, than anticipated at the time of an assessment. The SSC, directed assessment authors to revise the model and reduce hindcast prediction error.
2. Model Description
a. Description of overall modeling approach:
The model is a male-only size structured model that combines multiple sources of survey, catch, and mark-recovery data using a maximum likelihood approach to estimate abundance, recruitment, catchability of the commercial pot gear, and parameters for selectivity and molting probabilities (See Appendix A for full model description).
b-f. See Appendix A. g. Critical assumptions of the model:
i. Male crab mature at CL length 94mm.
Bases for this assumption have not been located. No formal study has been conducted to test this assumption.
Draft - Norton Sound Red King Crab Stock Assessment April 30, 2013
14
ii. Instantaneous natural mortality M is 0.18 for all length classes, except for the last length group (> 123mm) where M =0.648 (0.18 × 3.6) (Zheng et al. 1998). M is constant over time.
This mortality is based on Bristol Bay red king crab, estimated with a maximum age 25 and the 1% rule (Zheng 2005), and was adopted for NSRKC by CPT. The assumption of the higher M for the last length group is based not on biological data, but rather a working hypothesis attempting to explain the lower than model predicted proportion of this group in summer commercial fisheries (Figures 10, 13). It is possible, that the last length group moved into areas inaccessible to commercial fisheries (CPT review 2010). However, this does not explain the low proportion observed in the summer trawl survey, when all of the Norton Sound Area was surveyed. In addition, lowering the catch selectivity did not result in lower log likelihood than increasing the mortality (CPT 2010).
2013 Model Alternatives:
M=0.24 for all length classes
M=0.30 for all length classes
The above alternative M were derived from likelihood profile analyses (Appendix D1)
iii. Trawl survey selectivity is a logistic function with 1.0 for length classes 5-6.
This assumption was not based on biological/mechanistic data and reasoning, but rather an attempt to improve model fit.
iv. Winter pot survey selectivity is a dome shaped function: 1.0 for length classes, a logistic function for length classes 1-4, 1.0 for length class 5, and model estimate for the last length group.
This assumption is based on a belief (but no empirical data) that very large crab less representative in near shore area where the winter surveys occur. This assumption improves the model fit and reduces the bias in the bubble plot.
v. Summer commercial fisheries selectivity is an asymptotic logistic function of 1.0 at the length class 5 and 6. It has two curves: before 1993, and 1993-present, reflecting changes in fishing vessel composition and pot configuration.
vi. Winter commercial and subsistence fishery selectivity and length-shell conditions are the same as those of the winter pot survey.
Winter commercial king crab pots can be any dimension (5AAC 34.925(d)). No data exists about crab pot configuration of commercial or subsistence crab fishery gears. However, because commercial fishers are also subsistence fishers, it is reasonable to assume that the commercial fishers used crab pots that they also used for subsistence harvest, and hence both fisheries have the same selectivity.
vii. Growth increments are is a function of length and are constant over time.
viii. Molting probabilities are an inverse logistic function of length for males.
ix. A summer fishing season for the directed fishery is short.
x. Discards handling mortality is assumed to be 20%. No empirical estimate is available.
Draft - Norton Sound Red King Crab Stock Assessment April 30, 2013
15
xi. Annual retained catch is measured without error.
xii. All legal size crabs (≥ 4-3/4 inch CW) are taken to the commercial dock.
xiii. Since 2005, all commercially acceptable size crabs ( ≥ 5 inch CW) are taken to
the commercial dock.
xiv. All sublegal size crab or commercially unacceptable size crab (< 5 inch CW, since 2005) are discarded.
xv. Length compositions have a multinomial error structure, and abundance has a log-normal error structure.
h. Changes of assumptions since last assessment:
Following model modifications were made since 2012:
1. Standardized commercial catch cpue and standard error (Bishop 2013) was incorporated into the model. Consequently, likelihood for effort was replaced with likelihood for standardized cpue.
2. Based on 2013 CPT modeling workshop. 1980-1985 summer pot survey was eliminated from the model.
i. Code validation. Model code is reviewed at CPT modeling workshop in 2013, and is available from the authors.
3. Model Selection and Evaluation
a. Description of alternative model configurations. See Appendix D for the rationale of selecting candidate alternative models and results
Three alternative model configurations were evaluated:
0. Baseline 2013 model: standardized CPUE data 1. S3-1: 1) No Summer pot survey (abundance, length comp) data, 2) estimate
survey q for 1976-1991 NMFS survey, 3) maximum sample size = 20. 2. S3-6: 1) No Summer pot survey (abundance, length comp) data, 2) estimate
survey q for 1976-1991 NMFS survey, 3) maximum sample size = 20, 4) M = 0.24 for all length classes
3. S3-7: 1) No Summer pot survey (abundance, length comp) data, 2) estimate survey q for 1976-1991 NMFS survey, 3) maximum sample size = 20, 4) M = 0.30 for all length classes
Draft - Norton Sound Red King Crab Stock Assessment April 30, 2013
Comparing alternative model scenarios, following trends were observed (See Appendix D1 D2 for details).
1) All alternative model scenarios (except for S1-4) decreased retrospective hindcast
error and bias. 2) Constant M = 0.24 and 0.3 increased fit to 1976-1979 trawl abundance, but resulted
in lower projected legal abundance. 3) Change of survey Q indicates that historical NMFS abundance is underestimated, or
that current ADF&G abundance is overestimated. 4) Lowering of effective sample size resulted in better fit of trawl abundance and CPUE. 5) Removing CPUE data generated in mixed results on model fit. 6) Regardless alternative model scenarios (except for S1-4), results of prospective
analyses were similar, or that uncertainties of historical abundance (1976-1991) do not seem affect model performance of recent (1996-2012) abundance trajectory.
d. Parameter estimates:
See Table 11
e. Model selection criteria.
Draft - Norton Sound Red King Crab Stock Assessment April 30, 2013
17
Results of alternative models were evaluated using following criteria: 1) lower total negative log-likelihoods, and 2) consistent prospective abundance results, and 3) low mean retrospective prediction bias and error.
Prospective analyses consist of systematically removing first n year’s data consecutively (e.g., 1976, 1976-1977, 1976-1978, …), fitting a model to the reduced data sets, and examine predicted model abundance (e.g., 2013 model predicted legal abundance). If the model is greatly influenced by a set of initial data, the predicted abundance would change greatly. On the other hand, retrospective analyses consist of systematically removing last n year’s data consecutively (e.g., 2012, 2012-2011, 2012-2010, …), fitting a model to the reduced data sets, and examine predicted model abundance (hindcast predicted abundance) (e.g., 2012, 2011, 2010, …) with that of complete data set (reconstructed abundance).
From hindcast predicted and reconstructed legal crab abundance, hindcast error for each year was calculated as
Ei = (ypi - yi)/ypi, mean hindcast error (ΣEi)/n
Mean hindcast bias (1-β) was calculated by regressing reconstructed legal crab abundance with hindcast predicted abundance as
yi = β ypi.
In these two measures, a better model should have lower mean hindcast error and mean bias (close to 0). Positive values indicate that predicted abundance tends to be higher than hindcast abundance.
f. Residual analysis. See Figures S3-1, S3-6, S3-7
g. Model evaluation: See Appendix D2
4. Results
1. Effective sample sizes and weighting factors.
Effective sample sizes were calculated as
2,,,, )ˆ()ˆ1(ˆly
llyly
lly PPPPn
Where lyP , and lyP ,ˆ are observed and estimated length compositions in year y and length
group l, respectively. Estimated effective sample sizes vary greatly overtime.
Following weights were used
Draft - Norton Sound Red King Crab Stock Assessment April 30, 2013
18
Data Weighting Factor
Recruitment 0.01
Maximum sample size for length proportion:
Survey data Sample size
Summer commercial, winter pot, and summer observer
minimum of 0.1× actual sample size or 10
Summer trawl and pot survey minimum of 0.5× actual sample size or 20
2. Tables of estimates.
Model Parameter estimates (Table 11).
a. Most of parameters were estimated with CV of around 30%. Notable exception was recruitment parameter for 1978 and 1979 (log_R78, log_R79), trawl selectivity parameter (log_ϕst and log_ωst), and winter pot survey selectivity (log_ωsw). For 1978 and 1979, estimates were close to zero reflecting extremely low proportion of < 94mm crab observed in 1979 trawl survey (Table 5, Figure 3,4). The high CVs for those selectivity parameters are an artifact because the estimated selectivity was 1.0 for those cases. In asymptotic logistic function, multitudes of parameter combinations can result in 1.0, so that model was not able to converge into single parameter. The parameter p4 hit the bound of 1.0. This shows that commercial buyer’s preference of purchasing only ≥ 5 inch CW legal crab (as opposed to 4 ¾ inch CW legal crab), did not seem to change fishing behavior (i.e., discarding <5 inch CW legal crabs).
b. Abundance and biomass time series are provided in Table 12 and Figure 4.
c. Recruitment time series are in Table 12 and Figure 4.
d. Time series of catch/biomass are in Table 13
e. Selectivities, molting probabilities, and proportions of legal crabs by length are provided in Table 10.
3. Graphs of estimates.
a. Estimated male abundances (recruits, legal, and total) are plotted in Figures 4.
b. Time series of catch and harvest rates are plotted in Figure 5.
c. Harvest rate are plotted against mature male biomass in Figure 6.
d. Estimated and observed catch effort was plotted in Figure 7.
e. Molting probability and catch selectivity in Figure 8
Draft - Norton Sound Red King Crab Stock Assessment April 30, 2013
19
4. Evaluation of the fit to the data
a. Fits to observed and model predicted catches. Not applicable. Catch is assumed to be measured without error
b. Model fits to survey numbers (Figure 4a).
The majority of model estimated abundances of total crabs were within the 95% confidence interval of the survey observed abundance, except for 1976 and 1979, where model estimates was higher than the observed abundance.
c. Model fits to catch and survey proportions by length (Figures S3-1-3 through S3-6-6, S3-6-3 through S3-1-6, S3-7-3 through S3-7-6).
A residual plot for the commercial catch showed that the model tended to overestimate catches of largest length class and thus underestimate crab sizes of (4 and 5). Residuals of winter pot survey showed the model tended to overestimate (negative residuals) the proportion of large length classes (>103 mm). However, during 1991-1995, the pattern was reversed.
Plots of summer trawl, pot, and observer data did not seem show noticeable patterns. Similar to the winter pot survey, the model tended to overestimate proportion of large length classes. This tendency was most prominent during the last 3 trawl surveys.
d. Marginal distribution for the fits to the composition data: (Figure 8 ). e. Plots of implied versus input effective sample sizes and time-series of implied effective
sample size (Figures S3-1-2, S3-6-2, S3-7-2). f. Tables of RMSEs for the indices:
h. QQ plots and histograms of residuals: Figure S3-1-1, S3-6-1, S3-7-1,
Draft - Norton Sound Red King Crab Stock Assessment April 30, 2013
20
5. Retrospective and prospective analyses.
See Figure 10 and 11
6. Uncertainty and sensitivity analyses.
F. Calculation of the OFL
1. Specification of the Tier level and stock status.
The Norton Sound red king crab stock is currently placed in Tier 4 (NPFMC 2007). It is not possible to estimate the spawner-recruit relationship, but some abundance and harvest estimates are available to build a computer simulation model that capture the essential population dynamics. Whereas tier 4 stocks are assumed to have reliable estimates of current survey biomass and instantaneous M, the estimates for the Norton Sound red king crab stock uncertain. Survey biomass is based on triennial trawl surveys with CVs ranging 15-42% (Table 4). The natural mortality of 18% adopted by the CPT (2010) is based on Bristol Bay red king crab with the maximum age 25 and the 1% rule (Zheng 2005); however, no data are available to support the assumption of a maximum age 25 for the Norton Sound red king crab.
The OFL is estimated by the FMSY proxy, BMSY proxy, and estimated legal male abundance and biomass:
where B is a mature male biomass (MMB), BMSY proxy is average mature male biomass over a specified time period. M = 0.18 and = 1.
For Norton Sound red king crab, MMB is defined as CL > 94 mm.
OFL was calculated for retained catch and total male catch. The retained OFL is based on legal crab biomass catchable to summer commercial pot fisheries (Legal_B):
lllsl,sl,sl
wmLSON=BLegal ,,, )(_
BLegalFOFL OFLretained _))exp(1(
Draft - Norton Sound Red King Crab Stock Assessment April 30, 2013
21
The total male OFL is
hmwmLSONFOFLOFL lllsl,sl,s
lOFLretainedtotalmales )1()())exp(1( ,,,
where Ns,l and Os,l are summer abundances of newshell and oldshell crabs in length class l in the terminal year, Ll is the proportion of legal males in length class l, Ss,l is summer commercial catch selectivity, wml is average weight in length class l and hm is handling mortality rate
For the selection of the BMSY proxy, default data used are survey MMB. However, for the Norton Sound red king crab stock, only available survey MMB data are triennial trawl surveys, 11 years of data during 37 years period. Instead, we used the model estimated MMB for calculation of BMSY proxy from 1980 to present.
Predicted legal male and mature male biomass in 2013 are:
Legal male biomass: Model S3-1: 3.55 million lb with a standard deviation of 0.48 million lb. Model S3-6: 3.31 million lb with a standard deviation of 0.50 million lb. Model S3-7: 3.12 million lb with a standard deviation of 0.49 million lb.
Mature male biomass: Model S3-1: 5.00 million lb with a standard deviation of 0.98 million lb. Model S3-6: 5.03 million lb with a standard deviation of 0.96 million lb. Model S3-7: 4.77 million lb with a standard deviation of 1.00 million lb.
Candidate OFL and ABC million lb. Parenthesis indicates standard deviation
In all three models, estimated mature male biomass of 2013 was higher than estimated BMSY proxy. Hence, FOFL was the same as M.
Retained OFL for legal male crab is Model S3-1: 0.58 million lb. Model S3-6: 0.71 million lb. Model S3-7: 0.81 million lb.
Draft - Norton Sound Red King Crab Stock Assessment April 30, 2013
22
Difference in OFL derived from difference in M.
G. Calculation of the ABC
1. Specification of the probability distribution of the OFL.
Probability distribution of the OFL was determined based on the CPT recommendation in January 2012 as follows: Tier 4 crab stocks Calculation of a distribution for the OFL for Tier 4 stocks involves repeating four steps (detailed below). The aim is to have the median of the distribution for the OFL equal the point estimate (so that P*=0.5 implies that the ABC equals to the point estimate of the OFL). The proposed steps are: (a) Sample current MMB from a normal distribution with mean given by the point estimate of current MMB and CV equal to the sampling CV. (b)The BMSY proxy is the average MMB over a pre-specified set of years. Uncertainty in the BMSY proxy only accounts for uncertainty in MMB for the years for which it is assumed the stock was “at BMSY” and not uncertainty in the years concerned. For each of the years used when defining the BMSY proxy, sample MMB from a distribution with mean given by its point estimate and CV equal to the sampling CV. The pseudo BMSY proxy is then the average of the samples values. (c)Sample M from a normal distribution with mean equal to the assumed M and CV equal to an assumed CV (e.g. 0.2). (d)Compute the OFL. Form a cumulative distribution for the OFL from the sampled values. Find the median of this distribution. Using normal quantiles to rescale the distribution so that the median equals the OFL (similar to a bias-corrected bootstrap).
For the Norton Sound red king crab, calculation of OFL was based on summer commercial retained legal male biomass. For calculation of the ABC, default percentile is P* = 49; however, for the Norton Sound Stock the NPFMC adopted 10% buffer of OFL (i.e., ABC = 0.9×OFL) in 2012.
Retained ABC for legal male crab is Model S3-1: 0.52 million lb. Model S3-6: 0.64 million lb. Model S3-7: 0.73 million lb.
H. Rebuilding Analyses
Not applicable
I. Data Gaps and Research Priorities The model suggests that historical NMFS survey underestimated historical crab abundances. In
Draft - Norton Sound Red King Crab Stock Assessment April 30, 2013
23
this assessment, NMFS abundances were estimated from published reports in ad hoc manner; however, it is more appropriate re-estimating the survey abundance original survey data. The major data gaps of the Norton Sound red king crab are: spatially and temporarily consistent estimate of abundance, length frequency of discards from fisheries, and estimates of the instantaneous natural mortality. In addition, life-history of the Norton Sound red king crab stock is poorly understood. This includes size at maturity, natural mortality rate, timing and locations of reproduction, location of females during summer.
Acknowledgments
We thank all CPT modeling workshop attendants for critical review of the assessment model and suggestions for improvements and diagnoses.
References
Alverson, D.L., and W.T. Pereyra. 1969. Demersal fish in the Northeastern Pacific Ocean - an
evaluation of exploratory fishing methods and analytical approaches to stock size and yield forecasts. J. Fish. Res. Board Can. 26:1985-2001.
Bishop, G., M.S.M. Siddeek, J. Zheng, and T. Hamazaki. 2013. Summary Report: Norton Sound red king crab CPUE standardization. Unpublished manuscript. Alaska Depart of Fish and Game, Division of Commercial Fisheries, Juneau.
Brannian, L. K. 1987. Population assessment survey for red king crab (Paralithodes camtschatica) in Norton Sound, Alaska, 1985. Alaska Department of Fish and Game, Technical Data Report No. 214, Juneau.
Fournier, D., and C.P. Archibald. 1982. A general theory for analyzing catch at age data. Can. J. Fish. Aquat. Sci. 39:1195-1207.
Fournier, D.A., H.J. Skaug, J. Ancheta, J. Ianelli, A. Magnusson, M.N. Maunder, A. Nielsen, and J. Sibert. 2012. AD Model Builder: using automatic differentiation for statistical inference of highly parameterized complex nonlinear models. Optimization Methods and Software 27:233-249.
Menard, J., J. Soong, and S. Kent 2011. 2009 Annual Management Report Norton Sound, Port Clarence, and Kotzebue. Fishery Management Report No. 11-46.
Methot, R.D. 1989. Synthetic estimates of historical abundance and mortality for northern anchovy. Amer. Fish. Soc. Sym. 6:66-82.
NPFMC/NMFS 2010. Environmental assessment for proposed amendments 38 and 39 to the fishery management plan for the Bering Sea and Aleutian Islands king and tanner crabs to
Draft - Norton Sound Red King Crab Stock Assessment April 30, 2013
24
comply with the annual catch limit requirements (Amendment 38) and to revise the rebuilding plan for the EBS snow crab (Amendment 39). NPFMC AGENDA C-3, October 2010.
Powell, G.C., R. Peterson, and L. Schwarz. 1983. The red king crab, Paralithodes camtschatica (Tilesius), in Norton Sound, Alaska: History of biological research and resource utilization through 1982. Alaska Dept. Fish and Game, Inf. Leafl. 222. 103 pp.
Schwarz, L. 1984. Norton Sound section of the Bering Sea 1983 king crab fishery report to the Board of Fisheries. Alaska Department of Fish and Game, Division of Commercial Fisheries, Region III: Shellfish Report No. 5, Anchorage.
Stevens, B.G., and R. A. MacIntosh. 1986. Analysis of crab data from the 1985 NMFS survey of the northeast Bering Sea and Norton Sound. National Marine Fisheries Service, Northwest and Alaska Fisheries Center, NWAFC Processed Report 86-16. September 1986.
Stevens, B.G. 1989. Analysis of crab data from the 1988 NMFS survey of Norton Sound and the northeast Bering Sea. National Marine Fisheries Service, Northwest and Alaska Fisheries Center, unpublished report. February 1989.
Stevens, B.G. 1992. Results of the 1991 NMFS survey of red king crab in Norton Sound. National Marine Fisheries Service, Alaska Fisheries Science Center, unpublished memorandum to the State of Alaska. May 1992.
Soong, J. 2007. Norton Sound winter red king crab studies, 2007. Alaska Department of Fish and Game, Fishery Data Series No. 07-53, Anchorage.
Soong, J. 2008. Analysis of red king crab data from the 2008 Alaska Department of Fish and Game trawl survey of Norton Sound. Alaska Department of Fish and Game, Fishery Data Series No. 08-58, Anchorage.
Wolotira, R.J., Jr., T.M. Sample, and M. Morin, Jr. 1977. Demersal fish and shellfish resources of Norton Sound, the southeastern Chukchi Sea, and adjacent waters in the baseline year 1976. National Marine Fisheries Service, Northwest and Alaska Fisheries Center, Processed Report. October 1977.
Zheng, J. 2005. A review of natural mortality estimation for crab stocks: data-limited for every stock? Pages 595-612 in G.H. Kruse, V.F. Gallucci, D.E. Hay, R.I. Perry, R.M. Peterman, T.C. Shirley, P.D. Spencer, B. Wilson, and D. Woodby (eds.). Fisheries Assessment and Management in Data-limite Situation. Alaska Sea Grant College Program, AK-SG-05-02, Fairbanks.
Zheng, J., G.H. Kruse, and L. Fair. 1998. Use of multiple data sets to assess red king crab, Paralithodes camtschaticus, in Norton Sound, Alaska: A length-based stock synthesis approach. Pages 591-612 In Fishery Stock Assessment Models, edited by F. Funk, T.J.
Draft - Norton Sound Red King Crab Stock Assessment April 30, 2013
25
Quinn II, J. Heifetz, J.N. Ianelli, J.E. Powers, J.F. Schweigert, P.J. Sullivan, and C.-I. Zhang, Alaska Sea Grant College Program Report No. AK-SG-98-01, University of Alaska Fairbanks
Table 1. Historical summer commercial red king crab fishery economic performance, Norton Sound Section, eastern Bering Sea, 1977-2012.
Guideline Commercial Harvest Harvest (lb) a,
b
Level Open Total Number (incl. Total Pots Season Length Year (lbs) b Access CDQ Harvest Vessels Permits Landings Registered Pulls CPUE SD Days Dates
a Prior to 1985 the winter commercial fishery occurred from January 1 - April 30. As of March 1985, fishing may occur from November 15 - May 15. b The winter subsistence fishery occurs during months of two calendar years (as early as December, through May). c The number of crab actually caught; some may have been returned. d The number of crab Retained is the number of crab caught and kept. e Information not available. f Confidential under AS 16.05.815. g Confidentiality was waived by the fishers. h Prior to 2005, permits were only given out of the Nome ADF&G office. Starting with the 2004-5 season, permits were given out in Elim, Golovin, Shaktoolik, and White Mountain.
Table 3. Summary of triennial trawl survey Norton Sound male red king crab abundance estimates. Trawl survey abundance estimate is based on 10×10 nmil2 grid, except for 2010 (20×20 nmil2). See Appendix for details
Table 4. Summer commercial catch size/shell composition. Sizes in this and Tables 5-10 and 12 are mm carapace length. Legal size (4.75 inch carapace width is approximately equal to 124 mm carapace length.
Table 9. Growth matrix (proportion of crabs molting from a given pre-molt carapace length range into post-molt length ranges) for Norton Sound male red king crab. Length is measured as mm CL. Results are derived from mark-recapture and winter tagging data from 1980 to 2007. Pre-molt Post-molt Length Class Length Class
Table 10. Estimated selectivities, molting probabilities, and proportions of legal crabs by length (mm CL) class for Norton Sound male red king crab. S3-1
Table 12. Annual abundance estimates (million crabs) and mature male biomass (MMB, million lbs) for Norton Sound red king crab estimated by length-based analysis from 1976-2011. S3-1
Table 13. Summary of catch and bycatch (million lbs) for Norton Sound red king crab. The bycatch (discards) is estimated from the model. Summer commercial catches are from ADF&G fish ticket database during 1985-2009 and from Menard et al. (2011) during 1977 to 1984. Winter commercial and subsistence catches are from ADF&G permit reporting and average weight of 2.5 lbs for the winter commercial catch and 2.0 lbs for the subsistence catch were assumed to estimate total weight. S3-1
Appendix A. Description of the Norton Sound Red King Crab Model a. Model description.
The model is an extension of the length-based model developed by Zheng et al. (1998) for Norton Sound red king crab. The model has 6 length classes with model parameters estimated by the maximum likelihood method. The model estimates abundances of crabs with CL 74 mm and with 10-mm length intervals because few crabs with CL <74 mm were caught during surveys or fisheries and there were relatively small sample sizes for trawl and winter pot surveys. The model was made for newshell and oldshell male crabs separately, but assumed they have the same molting probability and natural mortality. In this model year starts July 1st to June 30th of the following year.
Initial pre-fishery summer crab abundance on July 1st 1976
Abundance of the initial pre-fishery population was estimated as a stochastic process around the mean, B1:
),0(~, 211 1Bt NeBB t (1)
The length proportion of the first year was calculated as
)exp(1
)exp(1
11 for )exp(1
)exp(
1
1
1
1
1
1
n
ii
n
ii
n
n
ii
ii
a
ap
,..,n-ia
ap
(2)
Abundance of crab length class was is a multiplication of the first year abundance. In this it was assumed no oldshell crab exist for the first year.
11,, BpN ils (3) Where
Ns,l,1 , Os,l,1 : summer abundances of newshell and oldshell crabs in length class l in the first year. pn : proportion of the neswshell crab pn,l : conditional proportion of l-th length neswshell crab, pn,0 =0 po,l : conditional proportion of l-th length oldwshell crab, po,0 = po,1 =0
Summer crab abundance on July 1st
Summer crab abundance of the t-th year new and old shell of l-th length class before the summer commercial fishery, is the survivors of winter crab from fishery and natural mortality
e)PCPC-O(=O
e)PCPC-N(=Nl
l
M0.417-tl,optptl,owtwtl,wtl,s
M0.417-tl,nptptl,nwtwtl,wtl,s
ˆˆ
ˆˆ
1,,1,1,,,1,,
1,,1,1,,,1,,
(4)
where
Ns,l,t , Os,l,t : summer abundances of newshell and oldshell crabs in length class l in year t Nw,l,t, Ow,l,t :winter abundances of newshell and oldshell crabs in length class l in year t Cw,t, Cp,t : total winter and subsistence catches in year t, Pw,n,l,t, Pp,n,l,t : Length proportion of winter and subsistence catches for newshell crabs for length class l in year t Pw,o,l,t, Pp,o,l,t : length compositions of winter and subsistence catches for oldshell crabs in length class l in year t Ml : instantaneous natural mortality in length class l, constant for all sizes and shell conditions 0.417 : proportion of the year from Feb. 1 to July 1 is 5 months, or 0.417 Winter crab abundance on February 1st
Abundance of newshell crab of the t-th year and l-th length class (Nw,l,t ), is a population that molted to become l-th length class minus l-th length class harvested by summer commercial fishery and discards, the combined result of growth, molting probability, summer commercial harvests, mortality, and recruitment from the summer population:
R+]em)D)P+P(CeON(G[ = N tl,My-0.583-
tlt,lost,lnsts,My
t,lst,lsl,l
l=l
=ltl,w
lc
l
lc )(,',,,,,,, ˆˆ)(
1
(5)
Winter abundance of oldshell crabs Os,l,t is the non-molting portion of survivors of crabs from summer:
e)m-(1] D)P+P(CeON[ = O lc
l
lc My0.583--tltlostl,nsts,
Mytl,stl,stl,w
)(,,,,,,,,, ˆˆ)( (6)
where
Gl’, l : a growth matrix representing the expected proportion of crabs molting from length class l’ to length class l (independently estimated outside of the assessment model frame), Cs,t : total summer catch in year t (assumed to be accurate without error), Ps,n,l,t , Ps,o,l,t : Compositions of summer catch for newshell and oldshell crabs in length class l in year t, Dl,t : discards of length class l in year t, ml : molting probability in length class l, yc : the time in year from July 1 to the mid-point of the summer fishery 0.583: Proportion of the year from July 1 to Feb. 1 is 7 months, or 0.583 year Rl,t: recruitment into length class l in year t.
Discards
In summer commercial fisheries, sublegal males (<4.75 inch CW and <5.0 inch CW since 2005) are not retained, but are sorted and discarded. Those discarded crabs are subject to handling mortality. Due to lack of data, we assumed discards mortality to be 0.2.
Discards of length class l in year t from the commercial pot fishery were estimated as:
where hm: handling mortality rate assumed to be 0.2 Ll : the proportion of legal males in length class l. Reflecting the change of commercial acceptable crab size since 2005, proportion of legal males in the length class 4, was calculated as p4L4. Where p4 is the proportion of commercially acceptable crab among legal crab of the length class 4. p4
was estimated from the model. Ss,l : Selectivity of the summer commercial fishery. Molting Probability Molting probability for length class l, ml, was calculated using a reverse logistic function fitted as a function of length and time (Balsiger's 1974)
e+1
1-1 = m
i-l )( (8)
where and are parameters, and i is the mid-length of length class l. ml was re-scaled such that m1 = 1.
Trawl net and pot selectivity Selectivity of length class l for summer commercial fishery ( Ss,l ), summer trawl survey ( Sst,l ), summer pot survey (Sp,l ), winter pot survey (Sw,l ), and summer trawl survey were assumed to be an asymptotic logistic function with parameters and , where i is the mid-length of the length class l.
e+1
1 = S i-l )(
(9)
Selectivity of S1-4 were re-scaled such that S5 = S6 =1.
For summer commercial fisheries, two sets of parameters (1, 1), (2, 2) were estimated: 1) before 1993, and 2) 1933 to present reflecting changes in fisheries, and crab pot configurations.
For winter pot survey and winter harvest, selectivity (Sw,l) was estimated for the first 2 length classes, the length classes 3-5 were assumed to be 1, and Sw,6 was directly estimated from the model. This resulted in the model taking a dorm shaped selectivity.
Estimation of Recruitment
We modeled recruitment of year t, Rt, as a stochastic process around the mean, R0:
),0(~, 20 Rtt NeRR t (10)
Rt was assumed to come from only length classes 1 (R1,t) and 2 (R2,t) , and was calculated as
Rr)(1 = R
Rr=R
tt,
tt,
2
1 (11)
where r is a parameter with a value less than or equal to 1. Rl,t = 0 when l 3. Observation model Estimates of survey abundances Summer trawl survey abundance Abundance of t-th year trawl survey was estimated by subtracting population of July 1st abundance minus summer commercial fisheries harvested by before trawl survey, multiplied by selectivity of trawl.
llst
Mylslsst
l
Myytct,lost,lnsts,
My
lsttlstlstst
SeON=B
Se)PP+P(CeON=B
lst
lcstlc
,)(
1,,1,,1,
)(,,,,, ,,,,,,
)(ˆ
]ˆˆ)[(ˆ
(12)
Where yst : the time in year from July 1 to the mid-point of the summer trawl survey. (yst > yc: Trawl survey starts after opening of commercial fisheries) Pc,t : proportion of summer commercial crab harvested before the survey.
Summer pot survey abundance (Removed from likelihood components) Abundance of t-th year pot survey was estimated as
l
Mylptlstlstp SeON=B lp ])[(ˆ,,,,,, (13)
Where yp : the time in year from July 1 to the mid-point of the summer trawl survey.
Estimation of summer commercial cpue Summer commercial fishing cpue (ft) was calculated as a product of catchability coefficient q and mean exploitable abundance minus one half of summer catch, Ct.
)CA(qf ttt 5.0ˆ i (14)
Because fishing fleet and pot limit configuration changed in 1993 and 2008, q1 is for fishing efforts before 1993, q2 is from 1994 to present.
Estimates of length composition Winter commercial catch Length compositions of winter commercial catch (Pw,n,l,t, Pw,o,l,t) for length l in year t were estimated from the winter population, winter pot selectivity, and proportion of legal crabs for each length class as:
]LS)ON[(LSO=P
]LS)ON[(LSN=P
lllwtl,wtl,wllwtl,wtl,ow
lllwtl,wtl,wllwtl,wtl,nw
3,,,,,,,
3,,,,,,,
/ˆ
/ˆ
(15)
Winter subsistence catch Subsistence fishery does not have a size limit; however, crabs of size smaller than length class 3 are generally not retained. Hence, we assumed proportion of length composition l = 1 and 2 as 0, and estimated length compositions (l ≥ 3) as follows
3,,,,,,,
3,,,,,,,
/ˆ
/ˆ
llwtl,wtl,wlwtl,wtl,op
llwtl,wtl,wlwtl,wtl,np
]S)ON[(SO=P
]S)ON[(SN=P (16)
Winter pot survey
The above equations were also used to calculate length compositions of winter pot survey for newshell and oldshell crabs, Psw,n,l,t and Psw,o,l,t (l 1).
llwtl,wtl,wlwtl,wtl,osw
llwtl,wtl,wlwtl,wtl,nsw
]S)ON[(SO=P
]S)ON[(SN=P
,,,,,,,
,,,,,,,
/ˆ
/ˆ
(17)
Summer commercial catch Length compositions of the summer commercial catch for new and old shell crabs Ps,n,l,t and Ps,o,l,t, were calculated based on summer population, selectivity, and legal abundance;
ALSO =P
ALSN =P
tllstl,stl,os
tllstl,stl,ns
/ˆ
/ˆ
,,,,
,,,, (18)
Where At is exploitable legal abundance in year t, estimated as
l
llstl,stl,st ]LS)ON[(A ,,, (19)
Observer discards
Length/shell compositions of Observer discards in 87-90, 92, 94, and 2012 were estimated as
)]L(1S)ON[()L(1SO=P
)]L(1S)ON[()L(1SN=P
lllstl,stl,sllstl,stl,ob
lllstl,stl,sllstl,stl,nb
,,,,,,,
,,,,,,,
/ˆ
/ˆ (20)
Summer trawl survey
Some trawl surveys occurred during the molting period, and thus we combined the length compositions of newshell and oldshell crabs as one single shell condition, Pst,l,t, and were estimated as
Summer pre-season survey (1976) (Removed from likelihood)
The same selectivity for the summer commercial fishery was applied to the summer pre-season survey, resulting in estimated length compositions for both newshell and oldshell crabs as:
]S)ON[(SO =P
] S)ON[(SN =P
llstl,stl,slstl,stl,osf
llstl,stl,slstl,stl,nsf
,,,,,,,
,,,,,,,
/ˆ
/ˆ
(22)
This was not incorporated into likelihood calculation because of one year data.
Summer pot survey (1980-82, 85) (Removed from likelihood)
The length/shell condition compositions of summer pot survey were estimated as
llsptl,stl,slsptl,stl,osp
llsptl,stl,slsptl,stl,nsp
]S)ON[(SO =P
]S)ON[(SN =P
,,,,,,,
,,,,,,,
/ˆ
/ˆ
(23)
b. Software used: AD Model Builder (Fournier et al. 2012).
c. Likelihood components.
Under assumptions that measurement errors of annual total survey abundances and summer commercial fishing efforts follow lognormal distributions and each type of length composition
has a multinomial error structure (Fournier and Archibald 1982; Methot 1989), the log-likelihood function is:
1 triennial summer trawl survey 2 summer pot survey (1980-82, 85): Removed 3 annual winter pot survey 4 summer commercial fishery 5 observer bycatch during the summer fishery
ni: the number of years in which data set i is available Ki,t: the effective sample size of length/shell compositions for data set i in year t Pi,l,t : observed and estimated length compositions for data set i, length class l, and year t
In this, while observation and estimation were made for oldshell and newshell separately, both were combined for likelihood calculations.
: a constant equal to 0.001 CV : coefficient of variation for the survey abundance. CV for summer pot survey was assumed 0.34 Bi,k,t: observed and estimated annual total abundances for data set i and year t Wf : the weighting factor of the summer fishing effort ft : observed and estimated summer fishing cpue w2
t: extra variance factor WR : the weighting factor of recruitment. It is generally believed that total annual commercial crab catches in Alaska are fairly accurately reported. Thus, no measurement error was imposed on total annual catch. Variances for total survey abundances and summer fishing effort were not estimated; rather, we used weighting factors to reflect these variances.
e. Parameter estimation framework:
i. Parameters Estimated Independently
The following parameters were estimated independently: natural mortality (M =0.18),
proportions of legal males by length group, and the growth matrix.
Natural mortality was based on an assumed maximum age, tmax, and the 1% rule (Zheng 2005):
,
where p is the proportion of animals that reach the maximum age and is assumed to be 0.01 for the 1% rule (Shepherd and Breen 1992, Clarke et al. 2003). The maximum age of 25, which was used to estimate M for U.S. federal overfishing limits for red king crab stocks (NPFMC 2007) results in an estimated M of 0.18. Among the 199 recovered crabs from the tagging returns during 1991-2007 in Norton Sound, the longest time at liberty was 6 years and 4 months from a crab tagged at 85 mm CL. The crab was below the mature size and was likely less than 6 years old when tagged. Therefore, the maximum age from tagging data is about 12, which does not support the maximum age of 25 chosen by the CPT.
Proportions of legal males (CW > 4.75 inches) by length group were estimated from the ADF&G trawl data 1996-2011 (Table 8).
Mean growth increment per molt, standard deviation for each pre-molt length class, and the growth matrix (Table 8), were estimated from tagging surveys conducted in summer 1981-1985, and winter 1981-present. In summer 1981-1985 study legal and sublegal males captured by the survey pots were tagged, and in the1981-present winter survey, sublegal males were tagged. All tagged crabs were recaptured by summer and winter commercial/subsistence fisheries.
ii. Parameters Estimated Conditionally
Estimated parameters are listed in Table 5. Selectivity and molting probabilities based on these estimated parameters are summarized in Table 4 (also in the primary document).
A likelihood approach was used to estimate parameters, which include fishing catchability, parameters for selectivities of survey and fishing gears and for molting probabilities, recruits each year (except the first and the last years), and total abundance in the first year (Table 5).
Crabs usually aggregate, and this increases the uncertainty in survey estimates of abundance. To reduce the effect of aggregation, annual total sample sizes for summer trawl and pot survey data sets were reduced to 50% and all other sample sizes were reduced to 10%. Also, annual effective sample sizes were capped at 200 for summer trawl and pot surveys and 100 for the other data to avoid overweighting the data with a large sample size (Fournier and Archibald 1982). Weighting factors represent prior assumptions about the accuracy or the variances of the observed data or random variables. WR was set to be 0.01.
To reduce the number of parameters, we assumed that length and shell compositions from the first year (1976) summer trawl survey data approximated the true relative compositions. Abundances by length and shell condition in all other years were computed recursively from abundances by length and shell condition in the first year and by annual recruitment, catch, and model parameters. Initial parameter estimates were an educated guess based on observation
max/)ln( tpM
and current knowledge.
f. Definition of model outputs.
i. Mature Male Biomass (MMB): defined as those 94 mm carapace length and above (size classes 3 to 6). The mean weights for size classes 1-6 are 0.854, 1.210, 1.652, 2.187, 2.825 and 3.697 lbs.
ii. Projected Legal Male Biomass for OFL calculation: defined as the number of crab on July 1st 2012 of size class greater than 94mm (Nsl+Osl), multiplied by commercial pot selectivity(Ssl), proportion of legal crab (Ll), and mean weight lb (wml)
lllsl,sl,sl
wmLSON=BLegal ,,, )(_
iii. Recruitment: the number of males of the length classes 1 and 2.
Appendix B: Estimation of 1976-1991 NMFS and 1996-present trawl survey abundance
1976-1991 NMFS trawl survey abundance
In the indirect method, reported numbers were estimated from published reports and archived memos. Published abundance consisted of 0-100mm, 100-125mm, >125mm, and total for 1976-1979, and 0-90mm, 90-104mm, >104mm, and total for 1985-1991. For 1982, abundance was not reported formally, but estimates were produced for each length classes.
For 1976 – 1979 abundance of ≥74mm crab was estimated by adding abundance of ≥ 100mm and abundance of < 100mm multiplied by the proportion of 74-99mm among crabs of 0-99mm CL.
997410010074 PNNN
Where P74-99 is the proportion of 74-99mm among crabs of 0-99mm CL.
Similarly abundance of ≥74mm crab in 1985-1991 was estimated by adding abundance of ≥ 90mm and abundance of < 90mm multiplied by the proportion of 74-89mm among crabs of 0-89mm CL.
8974909074 PNNN
For 1982 abundance was calculated by summing abundance estimates of >74mm crabs.
CVs of the indirect method were substituted with those calculated from the original raw data.
abundance Proportion Estimated abundance
Year >100mm <100mm 74-99 mm ≥74 mm Data source
1976 3119.8 1171.2 0.939 4219.6 Wolotira et al 1977 1979 762.0 178.7 0.778 901.0 Sample and Wolotira 1985 1982 2325.0 Archived output file 1987
> 90 mm < 90mm 74-89mm 1985 2111.0 1354.0 0.587 2905.8 Stevens and MacIntosh 1986 1988 1607.0 1395.0 0.505 2311.5 Stevens 1989 1991 1771.0 1355.0 0.325 2211.4 Stevens 1992
1976
1979
1982
1985
1988
1991
1996-2011 ADF&G trawl survey abundance
In the ADF&G trawl survey, only one tow was conducted for each station. Second tow was conducted when the first tow caught more than 6 crabs or the first tow was unsuccessful. The survey stations were
Abundance of (CL > 73mm) red king crab at j-th station ( jN ) was estimated as:
j
jjj
a
AnN ^
. (1)
Where nj is the number of (CL > 73mm) crab captured, aj is a towed area computed by multiplying the width of the net mouth opening (0.00658 nmi) with the distance trawled (generally around 1.0 nmi), and Aj is an area of station (generally 100nmi2). Surveyed stations were stratified into single-towed and multiple-towed stations. For single-towed stations, the total crab abundance, sN , was estimated as the sum of estimated
station abundances:
^^ j
js NN . (2)
The variance of sN^
was estimated as:
1
)ˆˆ()
^(
2
n
NNnNV jj
s (3)
where n was the number of stations trawled.
For multiple-towed stations, stratum r, crab abundance per station, )(^
rjN , was estimated as the average abundance of tows for station j:
jrj NN ˆ^)(
. (4)
Total crab abundance for stratum r, ^
rN was estimated as the sum of estimated station abundances:
j
rjr NN )(ˆ^ . (5)
The variance of rN^
was estimated as:
)ˆ(^
)( )(j
rjr NVNV , (6)
where
1
)ˆˆ()
^(V
2)(
)(
n
NNnN rjj
rj
, (7)
where n was the number of tows at station j, and assuming independent estimates for each station. Total abundance of red king crab was a sum of abundance estimates for the 2 strata:
rs NNN^^^ .
(8)
Assuming independent estimates for the 2 strata, the variance of the estimated total red king crab abundance was estimated as:
)ˆ()ˆ(^
)( rs NVNVNV . (9)
Re-estimation 1976-1991 NMFS trawl survey data:
Under the direction of the CPT workshop, re-estimation of 1976-1991 trawl survey was directed. Despite both NMFS and ADF&G used identical survey data, re-estimated crab abundance differed considerably between ADF&G (using method above), and MMFS re-estimates. Overall, AD&G re-estimates are closer to original NMFS report, whereas NMFS re-estimate showed higher estimates for 1979, 1982, 1988, and 1991. At the moment of current assessment period, these discrepancies have not been resolved, and further investigations are needed. Under this circumstance, we elected to use ad hoc estimates of crab abundance from the original published report with CV of NMFS re-estimate.
Appendix C: Reconstruction of ADF&G Summer pot survey abundance
In the indirect method, reported numbers were estimated from published reports Brannian (1988)
The proportion of length class was estimated form original trawl length frequency data.
For 1980 – 1982 abundance, only legal crab abundance was available. Abundance of ≥74mm crab was estimated by expanding abundance of legal crab with the ratio of ≥74mm crab to legal crab, as follows
legal
sulegallegallegal P
PPNN 74,
74
Where Plegal is the proportion of legal crab and Psublegal,>74 is the proportion of sublegal crab of ≥74mm CL.
For 1985 abundance, abundance of both legal and sublegal crab was available. Hence, abundance of ≥74mm crab was estimated as
7474 PsNNN sublegallegal
where Ps>74 is the proportion of crab of ≥74mm CL among sublegal crab.
Estimated abundance of 1985 ≥74mm crab using the first method was reasonably closer.
Since CV of the pot survey were missing (1980-1982) or too small (1985), we employed cv = 0.34 that was an average CV of trawl survey
Abundance Proportion
Year Sublegal Legal Legal Sublegal >74mm
>74mm: Legal ratio
Estimated Abundance
1980 NA 1900.000 0.88 0.09 1.10 2092.303 1981 NA 1285.195 0.53 0.36 1.68 2153.407 1982 NA 353.273 0.28 0.64 3.23 1140.582 1985 1600.668 907.579 0.88 2320.381
0.43 0.50 2.16 1960.449
Appendix D1: Likelihood profile Analyses
The assumptions of high mortality and low trawl selectivity of the length class 6 (M=0.64) was intended to improve the model fit under the assumption of M= 0.18. At the reasonable range of M 0.1 – 0.5, total likelihood was minimized at M = 0.3-0.36, as well as commercial cpue, trawl length proportion, recruits. On the other hand, likelihood of trawl survey and winter pot survey was minimized at 0.22-0.26 range. Based on those results we assess M =0.24 and M =0.3 as an alternative model scenario.
Figure D1: weight sensitivity M changed 0.1 to 0.5 (x-axis) and corresponding likelihood components.
Appendix D2: Alternative Model Scenario Selection
For 2013 Assessment, we examined following model scenarios. All the model scenarios are in direct response to the CPT modeling workshop and SSC comments 2012.
0. Baseline 2013 model: Use standardized CPUE data, unrestricted net/pot selectivity functions, estimate of first year length composition
1: drop summer pot abundance & length comp data 2: drop CPUE data 3: estimate q of NMFS trawl surveys (1976-1991) 4: estimate q of ADF&G trawl surveys (1996-2011) 5: reduce length maximum n to 20 6: change M to 0.24
0.2 0.3 0.4 0.5
8090
100
110
Total negative log likelihood
0.2 0.3 0.4 0.5
1012
1416
Trawl survey
0.2 0.3 0.4 0.5
1.2
1.6
2.0
Pot survey
0.2 0.3 0.4 0.5
-16.
5-1
5.5
Commercial cpue
0.2 0.3 0.4 0.5
46
810
12
Trawl size
0.2 0.3 0.4 0.5
56
78
910
Summer Pot size
0.2 0.3 0.4 0.5
2527
2931
Winter Pot size
0.2 0.3 0.4 0.5
3035
4045
Summer Commercial size
0.2 0.3 0.4 0.5
0.3
0.4
0.5
0.6
recruits
0.2 0.3 0.4 0.5
11.5
12.5
13.5
14.5
Observer size
0.2 0.3 0.4 0.5
01
23
45
67
Total negative log likelihood
Trawl surveyCommercial cpueTrawl sizeWinter Pot sizeSummer Commercial sizerecruitsObserver size
7: change M to 0.30
Rationales
Scenario 1: Drop summer pot abundance and length composition data Scenario 2: Drop standardized CPUE data At the workshop, validity of input data was discussed, especially regarding to summer pot survey abundance data and standardized commercial crab catch CPUE. The scenario 1 and 2 examine the influence of those data on overall model fits.
Scenario 3: Estimate q of NMFS trawl surveys (1976-1991) Scenario 4: Estimate q of ADF&G trawl surveys (1996-2011) The model assumes that the trawl survey q be 1.0, validity of which was questioned. Especially, because survey coverage of NMFS surveys were generally larger than that of ADF&G (See Appendix E), validity of q = 1 for ADF&G survey was questioned.
Scenario 5: Reduce maximum sample size to 20 The 2012 assessment model lowered maximum sample size from 200 to 50; however, because the model did not fit well in earlier trawl abundance data and the majority of assessment data are length composition, further reduction of sample size was suggested. Scenario 6: Change M to 0.24 Scenario 7: Change M to 0.30 The 2012 assessment model assumes M = 0.18 for length classes 1-5 and 0.648 (i.e., 3.6 times higher) for the length class 6. This assumption was not based on biological evidence but rather an attempt to fit the model to data. CPT and SSC requested to review this assumption by conducting a likelihood profile analyses. Likelihood profile analyses (Appendix D) showed alternative M of 0.24 and 0.30 as candidate. Table 1. Alternative model configurations. Blank cell means that model configuration is the same as base model
Alt Summer
Pot survey
CPUE Q
NOAA Q
ADF&G
Max N
M
Base + + 1 1 50 0.18 S1-1 - S1-2 - S1-3 est S1-4 est S1-5 20 S1-6 0.24
S1-7 0.30 S2-1 - - S2-2 - est S2-3 - 20 S2-4 - 0.24 S2-5 - 0.30 S3-1 - est 20 S3-2 - est 0.24 S3-3 - est 0.30 S3-4 - 20 0.24 S3-5 - 20 0.30 S3-6 - est 20 0.24 S3-7 - est 20 0.30 S4-1 - - 20 S4-2 - - 20 0.24 S4-3 - - 20 0.30 S4-4 - - est S4-5 - - est 20 S4-6 - - est 20 0.24 S4-7 - - est 20 0.30
+: included, -: excluded, est: estimated. Among the seven alternative scenarios, scenarios S1-1, S1-3, S1-5, and S1-6 resulted in better fit to the trawl data. Removal of summer pot survey data and subsequent improvement of trawl abundance (TR) and length (TRL) data suggests data conflicts in length composition data between trawl and pot surveys. Estimating NMFS survey q (S1-3) resulted in estimate of q < 1, or that historical NMFS survey underestimated crab abundance. This may be caused by the fact that NMFS survey was conducted after the majority of fishery (though survey timing was incorporated in the model), or that NMFS trawl gear may not be as efficient as ADF&G trawl gear. Also, as expected, reduction of the maximum sample size (S1-5) increased fit of trawls abundance and CPUE. Changing M = 0.24 (S1-6) improved fit of trawl abundance, but this increased conflicts for summer commercial catch data. For other scenarios, removing CPUE data (S1-2) resulted lower fit of trawl data, and estimate of q for ADF&G (S1-4) resulted greater than 1 or that ADF&G trawl surveys overestimated crab abundance. Because this is unlikely, we dropped this scenario for further consideration. Finally, changing M=0.3 (S1-7) did not particularly improve fit of individual component; however, it had the best overall fit. Among each likelihood components, catch length composition (WPL, SCL, OBL) were not affected by choices of scenarios, except when M was fixed for all size classes (S1-6, S1-7). The other characteristics of constant M is an elimination of high recruits in 1976 and low molting probability that indicates slow growth (Figures ). Because the Norton Sound Red King crab is the northern most population, it is reasonable to assume that their growth rate is slower and their mortality is higher than those of southern population (e.g. Bristol Bay red king crab). Retrospective analyses showed that constant M scenario had better estimates. Overall, all model scenarios considered did not change fits of length composition, but resulted in improvement in fit of trawl survey data, especially on historical (1976-1991) trawl survey data. However, prospective analyses showed all model scenarios resulted in consistent and similar projected estimates. This suggests that majority of model improvements are made for fit of historical trawl data; however, those improvements had little impacts on fitting of recent (1996-2012) data. As for
differences of model scenarios on prediction bias and error, retrospective analyses suggests that constant M (M=0.24, 0.30) and removing CPUE seem to reduce bias/error. Considering all those above factors, we selected S3-1, S3-6, S3-7 as a candidate model.
Table 2. Likelihood components of alternative model scenarios. Bold type number show likelihood value 2 units lower than baseline or among the scenario groups.
rho––a non-parametric measure of statistical dependence), and frequency histograms were plotted to
evaluate collinearity between explanatory variables, strength and linearity of relationships between
response and explanatory variables, and normality of distributions. Note that although Spearman’s rho is
appropriate only for continuous and ordinal numeric variables, R nonetheless calculates it for strictly
Norton Sound red king crab CPUE standardization
8
categorical variables (Vessel, Permit Fishery, and Modified Statistical Area) as well; these results must be
disregarded. Second, Akaike Information Criterion (AIC) (Akaike 1974) were calculated for pairwise
relationships of the response with each explanatory variable to assess their relative predictive ability.
Theoryandequations
Let denote the observed CPUE, U0 the reference CPUE, Pij a factor i at level j, and let Xij take a value
of 1 when the jth level of the factor Pij is present and 0 when it is not. If observation error ijk, of k is
normally distributed with mean 0 and standard deviation σ, then the lognormal distribution of (Quinn
and Deriso 1999), can be denoted as:
∏ ∏ , (1)
or
ln ln ∑ ∑ ln .
By substituting ln to β0 and ln(Pij ) to βij, we then obtain an additive GLM lognormal error
distribution of :
ln ∑ ∑ . (2)
For selection of the best model, we used a forward step-wise selection procedure. After selection of the
best model, we calculate standardized CPUE. To do this we first divide coefficients by their geometric
mean to obtain canonical coefficients:
. (3)
We then exponentiate the result to obtain the non-log space canonical coefficients:
. (4)
Norton Sound red king crab CPUE standardization
9
Finally, we subtract the year coefficient reference level to obtain standardized CPUE Uj for each year
level j as:
. (5)
Eliminating all factors but Year in the GLM, but otherwise following Equations 2 and 3, 4, and 5 above
gives an estimate of the base year CPUE index.
If we let denote the catch and the effort, in pot lifts, for each delivery i in year y, then the
arithmetic CPUE can be calculated as:
∑
∑ , (6)
the geometric mean of the arithmetic CPUE as:
∏ , (7)
and the scaled arithmetic CPUE for each year level j as:
′ . (8)
Analyses
A forward stepwise GLM fitting algorithm was used to select explanatory variables from the full variable
set to the model. We assumed the null model to include only Year, and the full variable set to include the
six factor variables: Year, Vessel, Permit Fishery, Month of Year, and Modified Statistical Area and Week
of Year. First, a GLM was fit for each explanatory variable against the natural log of CPUE and an AIC
generated for the fit. The explanatory variable whose fit produced the lowest AIC was then added to the
model. This was repeated, accumulating explanatory variables and increasing the model degrees of
freedom until the increase in R2 for the final iteration was less than 0.01.
Norton Sound red king crab CPUE standardization
10
Several approaches to dealing with interactions have been used in fitting GLMs designed to produce
annual estimates of standardized fishery CPUE. The first and most common is to ignore interactions
altogether (Starr 2012; Vignaux 1994), a second is to include interactions in the selection process, but to
sequentially remove variables which exhibit strong interactions with Year and re-run the selection process
(Zuur et al. 2010). A third approach is to average CPUE over the interacting variable (Maunder and Punt
2004), and finally, if it is thought that random processes are responsible for the interaction, interactions
with Year can be included as random terms in a generalized linear mixed model (GLMM) (Maunder and
Punt 2004). In this analysis the first approach was taken. However, for the preferred models only, a two-
stage process was used to determine the necessity of including interactions in future improved models.
The first stage consisted of the stepwise variable selection described above and the second stage of
offering the variables selected in the first stage to the stepwise selection process along with their second-
order interactions.
After the GLM was fitted, generalized variance inflation factors (GVIF) (Faraway 2006) were calculated
for preferred models. GVIF indicate the degree to which variance of model parameter estimates is being
inflated as a result of multicollinearity. A threshold of GVIF > 3 was used to decide when to remove
multicollinear variables from the model.
Diagnostics were conducted (for models without interactions only) to check assumptions, look for
outliers, and check model choice. First, Pearson residuals were plotted against each variable in or out of
the model to assess independence of observations. Second, residuals were plotted against the linear
predictor to assess homogeneity of variance and goodness of fit of the model. Thirdly, a QQ plot was
made to test the assumption of normal distribution of residuals. QQ plots are constructed by plotting
residuals against their theoretical computed value if they were normally distributed, and should coincide
with or be parallel to the line y = x. Plots were constructed to allow graphic examination of interaction of
each explanatory variable with Year. Component + Residual plots were constructed to show the
independent influence of each explanatory variable on CPUE. Finally, four plots were made to detect
influential outliers. First hat values, DFFits, and Cooks distance (Faraway 2006) were plotted versus fitted
values and then studentized residuals were plotted against hat values, where data point diameter is
proportional to Cooks Distance.
Finally, interannual trends in standardized, base year, and scaled arithmetic CPUE were plotted.
Norton Sound red king crab CPUE standardization
11
We coded in the open-source programming language R to process data, and used two R scripts obtained
from Paul Starr for stepwise model selection and CPUE index calculation (Appendix E). R notation is
used in some of the tables and text to describe models.
ChangessinceFebruary2013crabmodelingworkshop
This work was first presented at a Crab Modeling Workshop of the Scientific and Statistical Committee
(SSC) of the Crab Plan Team of the North Pacific Fisheries Management Council in February 2013.
Based upon recommendations by the SSC at this meeting, the following changes have been implemented
in this report.
1.) Previously imputed data for effort and number of crab has been removed.
2.) Data for 1977 has been added.
3.) In order to add 1977 data, it was necessary to remove Commercial Fisheries Entry Commission
(CFEC) data (because this data set begins in 1978). This had provided the variables Owner and
Length Overall. However, this loss was not considered important as these variables are collinear
with Vessel and Vessel has more explanatory power. Furthermore, removing the join to the CFEC
dataset increased the null degrees of freedom as several Vessel had been being eliminated as a
result of having missing Owner data.
4.) We removed the link to Management Data for Season Start Date and ceased calculating the
variable Day of Season based upon SSC recommendation. This variable was also highly collinear
with the stronger Week of Year and had not been selected to the model.
5.) We removed the 1978–2012 time series per SSC recommendation because of excessive
management changes over this time period.
6.) We added more stringent filtering criteria (3 deliveries in 5 years and 3 deliveries in 7 years) for
the 1993–2012 time series.
7.) We included additional diagnostics from the ‘Car’ package including interaction plots,
component plus residual plots, and influence plots. However, we have not as of yet incorporated
the (http://projects.trophia.com/projects/influ/repository/entry/influ.R.) diagnostics package.
8.) We included tables with stepwise selection of second-order interactions for the preferred models.
The next step should be to incorporate significant substantial interactions into the model using
GLMM.
9.) An error in the data subsetting algorithm was corrected resulting in fewer vessels being included
in subsets.
Norton Sound red king crab CPUE standardization
12
Results
Preliminarydataprocessing
The large management change in 1993 resulted in vessels making more and smaller deliveries each year
and in harvest shifting to the southeast (Figure 2). The subset criteria of vessels having two deliveries a
year for three years was chosen for the 1977–1992 time series because more stringent criteria retained
insufficient data for a well-balanced design matrix (Appendix A2, Appendices B1 and B2). For the 1993–
2012 time series, more stringent subset criteria of three deliveries a year for three, five or seven years
could be applied while still retaining a high proportion of the harvest (Appendix A2, Appendices B1 and
B2).
CPUE did not differ from the full data for most subsets of the 1977–1992 time series, but were very
slightly greater for most subsets of the 1993–2012 time series (Appendix B3).
There was no significant effect of subsetting on the statistical area composition of harvest for the 1977–
1992 time series (Appendix A5, Appendix B4). However, for the 1993–2012 time series, subsetting data
resulted in overrepresenting the Inner and underrepresenting the Outer Modified Statistical Area;
however, the strength of the association, measured by Cramers V, was weak (Appendix A6, Appendix
B4).
Scatter plots for the two time series analyzed exhibited many visual patterns among explanatory variables
and a few with the response variable, log(CPUE). For the 1977–1992 time series the most notable
relationships for log(CPUE) are with Year and with Modified Statistical Area; while Year exhibits
patterns of varying strength with all other explanatory variables. It is also evident that Month of Year and
Week of Year are highly collinear and there also appears to be some collinearity of Month of Year and
Week of Year with Modified Statistical Area (Appendix B5). For the 1993–2012 time series the most
notable relationships for log(CPUE) are with Year and Modified Statistical Area; while Year again
exhibits trends with all other explanatory variables. Once again, Month of Year and Week of Year are
highly collinear (Appendix B6). The many covariations with Year are important because they hinder the
ability to unambiguously extract interannual trends in CPUE. Although Spearman’s rank correlation
coefficient (Spearman’s rho) exceeded the “folk lore” threshold of 0.7 (Briand et al. 2004) only for
correlations between Month of Year, and Week of Year, (Appendices B6 and B7), Spearman’s rho is only
a useful of measure of the strength of relationships for continuous (CPUE) and ordinal numeric (Year,
Norton Sound red king crab CPUE standardization
13
Month of Year, and Week of Year) variables. Thus, patterns with Year can only be visually assessed for
categorical variables (Vessel, Permit Fishery, and Modified Statistical Area).
The three variables having the lowest AICs for both data sets consistently included Year and Vessel; the
third variable was respectively, Week of Year for 1977–1992, and Modified Statistical Area for 1993–
2012 (Appendices A7 and A8).
The value of GVIF ^(1/(2*df)) did not exceed a threshold value of three for any of the variables selected
to the model for either time series (Appendices A9 and A10).
Analyses
1977–1992Dataset
When all six variables, but no interactions, were offered to the stepwise selection Year, Vessel, Modified
Statistical Area, and Week of Year were selected, producing an R2 of 0.702 (Tables 4 and 5). Review of
GVIF found no multicollinearity (Appendix A9) and 13 of 31 or 41.9% of model coefficients were
significant (Appendix C1). Offering these variables and their second order interactions to the model
resulted in addition to the above model of the interaction variables Vessel:Week of Year, Year:Week of
Year, Week of Year:Modified Statistical Area, and Year:Modified Statistical Area and an increase in R2 to
0.886 (Tables 4 and 6).
Consistent with the relatively high R2, model diagnostics indicate little departure from normality, no
excessively influential large outliers, independence of observations, and homogeneity of variance
(Figures 3–6). However, some unexplained variation remains in Year, Vessel, and Week of Year. The
interaction plots are roughly parallel. No Cooks Distance exceeding 1.0 were observed so no influential
data points were eliminated (Figures 3–6).
Standardized CPUE deviated little from scaled arithmetic or base year CPUE except in 1978. CPUE
declined from 1978 through 1982, after which it bounced somewhat erratically (Figure 7). Standard errors
are moderate ( = 0.36) (Appendix D1).
1993–2012Dataset
When all six variables, but no interactions, were offered to the stepwise selection procedure Year, Vessel,
Week of Year, and Modified Statistical Area were selected to the model for all three data subsets (Tables
Norton Sound red king crab CPUE standardization
14
4, 7, 8, and 9); however, review of diagnostics revealed residual patterns with Permit Fishery so it was
added. The model R2 was highest for the data subset having three deliveries for three years and lowest for
the subset of three deliveries for seven years (Tables 4, 7, 8, and 9); however, review of diagnostics
(Figures 8–22) revealed that the subset having three deliveries for five years produced the best fit and it
was selected as the preferred model. Review of GVIF found no multicollinearity (Appendix A10) and, 38
of 79 or 48.1% of coefficients were significant (Appendix C2). Offering the selected variables and their
second order interactions to the model resulted in addition to the model of the interaction variables
Year:Vessel, Vessel:Week of Year, and Year:Week of Year, the loss of Modified Statistical Area and an R2
increase to 0.699 (Tables 4 and 10).
Model diagnostics suggest no significant departure from normality, no excessively large and influential
outliers, homogeneous error distribution, and independence of observations (Figures 15–19). Despite the
relatively low R2, there is very little unexplained variability in the residuals. Interaction plots are messy
but largely parallel. There are no data points with Cooks Distances exceeding one, so no outliers were
removed from the model.
In addition to exhibiting the best fit, the standardized CPUE extracted from the preferred model exhibited
the best smoothing (Figures 12, 17, and 22). Standardized CPUE increased modestly over the 1993–2012
time series and was slightly greater than the base year and scaled arithmetic CPUE prior to 2003 and
slightly less after 2003 (Figure 20). Standard errors were very small ( = 0.06) (Appendix D2).
CPUEmeasurecomparisons
Norton Sound summer commercial red king crab fishery standardized CPUE was only slightly more
useful than arithmetic CPUE in predicting trawl survey legal male population size (Table 11). Likewise
there were no differences in the variance of standardized, base year, and scaled arithmetic CPUEs for
either data set: 1977–1992 (Bartlett’s K2 = 0.8161, df = 2, p = 0.665), 1993–2012 (Bartlett’s K2 = 1.2, df =
2, p = 0.542).
Discussion
We were able to successfully fit generalized linear models for two time series, 1977–1992 and 1993–
2012, identifying useful sets of explanatory variables and producing the first standardized indices of
CPUE for the Norton Sound red king crab summer commercial fishery. Although this standardized index
Norton Sound red king crab CPUE standardization
15
was not less variable than the arithmetic index, it was more closely related to trawl survey population
estimates, suggesting that its use will improve the Norton Sound stock synthesis model.
These benefits are somewhat diminished by the presence of significant and substantial interactions of
explanatory variables with Year, most notably of Week of Year, and Vessel. These interaction terms could
not be included in the loglinear generalized linear model, as this would prevent the extraction of
interannual trends in CPUE. The significance of these interaction terms may be a result of missing
explanatory variables. For example, the significance of the Year:Vessel interaction term may be due to
changes in Vessel catchability due to interannual changes in vessel size, pot configuration, soak time, and
electronics. Likewise, the significance of the Year:Week of Year interaction term is likely a result of the
large interannual variation in season start date. Although anecdotal reports suggest the existence of
information on historic pot configuration (Charlie Lean, personal communication, December 13, 2012),
neither pot configuration nor soak time data were available to this analysis. Pot configuration data could
be collected during registration and soak time through mandatory logbooks. Registration forms do not
currently note pot configuration; however, and logbooks are not required in regulation. Changes in
registration forms would require a simple modification of the form and spreadsheet, but any logbook
requirement would need to be promulgated through an action of the BOF, and might be considered
onerous for this small-boat fishery. Another explanatory variable whose addition could be investigated to
improve model fit is the 2002 change in the size of crab accepted by processors. This would simply
require the inclusion of a dummy variable to denote this change.
Besides acquiring additional explanatory variables, interactions might also be dealt with analytically.
Interactions of explanatory variables with year are a common problem in CPUE standardization and other
investigators have solved the problem by averaging CPUE over interacting variables (Maunder and Punt
2004) or by incorporating interactions as a random effects variable in a Generalized Linear Mixed Model
(Brandão et al. 2004; Ortiz and Arocha 2004).
The interaction of Year with Modified Statistical Area for the 1977–1992 data subset is most likely
caused by interannual shifts in harvest distribution. Because red king crab aggregate (Dew 1990; Taggart
et al. 2008), changes in their distribution are often a result of changes in population size. The fact that
historic survey grounds have also contracted (Soong and Hamazaki 2012) supports this hypothesized
explanation. This suggests that there may have been a larger change in population size than what we
currently describe. This question might be addressed either by using a spatially discrete method to
Norton Sound red king crab CPUE standardization
16
standardize CPUE (Campbell 2004; Quinn et al. 1982), or by expanding the survey to include historic
grounds.
In summary, we identified explanatory variable sets, and developed first-generation models to standardize
Norton Sound red king crab summer commercial fishery data. We also identified data needs and analytic
methods to improve the modeling process.
Acknowledgements
We are thankful for conversations with Joyce Soong, Jim Menard, and Jenefer Bell, who provided
valuable insights into the model, survey, biology, and fishery for red king crab in Norton Sound, for
editorial input from Chris Siddon, and for statistical advice from February 2013 Model workshop
members.
Norton Sound red king crab CPUE standardization
17
Literaturecited
Akaike, A. 1974. A new look at the statistical model identification. IEEE Transactions on Automatic Control. AC-19: 716-723.
Brandão, A., D. S. Butterworth, S. J. Johnston, and J. P. Glazer. 2004. Using a GLMM to estimate the somatic growth rate trend for male South African west coast rock lobster, Jasus lalandii. Fisheries Research. 70(2-3): 339-349.
Briand, G., S. C. Matulich, and R. C. Mittelhammer. 2001. A catch per unit effort - soak time model for the Bristol Bay red king crab fishery, 1991-1997. Canadian Journal of Fisheries and Aquatic Sciences. 58(2): 334-341.
Campbell, R. 2004. CPUE Standardisation and the construction of indices of stock abundance in a spatially varying fishery using general linear models. Fisheries Research. 70: 209-227.
Dew, C. B. 1990. Behavioral ecology of podding red king crab, Paralithodes camtschatica. Can. J. Fish. Aquat. Sci. 47: 1944-1958.
Faraway, J.J. 2006. Extending the Linear Model with R. Chapman & Hall/CRC.Boca Raton, FL. Hamazaki, T., and J. Zheng. 2011. Norton Sound red king crab stock assessment for the fishing year
2011/12. Alaska Department of Fish and Game, Division of Commercial Fisheries, Anchorage. Howard, K. G., and T. Hamazaki. 2012. Norton Sound red king crab harvest strategy. Alaska Department
of Fish and Game, Division of Commercial Fisheries, Special Publication #12-02, Anchorage. Johnson, A. B. 1985. Statistical analysis of the effect of pot soak-time on the catch of king crab
(Paralithodes camtschatica) and Tanner crab (Chionoecetes bairdi) in Chiniak Gully near Kodiak Island, Alaska. State of Alaska Department of Fish and Game, Informational leaflet #249, Juneau.
Lean, C., and B. Brennan. 1997. 1996 Norton Sound district shellfish report to the Alaska Board of Fisheries. Alaska Department of Fish and Game, Commercial Fisheries Management and Development Division, Regional Information Report #3A97-09, Anchorage.
Maunder, M. N., and A. E. Punt. 2004. Standardizing catch and effort data: a review of recent approaches. Fisheries Research. 70: 141-159.
Maunder, M. N., and P. J. Starr. 1995. Rock lobster standardised CPUE analysis, New Zealand Fisheries Assessment Research Document #95/11, Wellington, New Zealand.
Menard, J., J. Soong, and S. Kent. 2012. 2010 Annual Management Report Norton Sound, Port Clarence, and Kotzebue, Fishery Management Report #12-31, Anchorage.
Miller, R. J. 1977. Saturation of crab traps: reduced entry and escapement. Journal du Conseil international pour L'Exploration de la Mer. 38: 338-345.
Neter, J., W. Wasserman, and M. H. Kutner. 1990. Applied linear statistical models: regression, analysis of variance, and experimental designs. R. T. J. Hercher, editor. Irwin, Burr Ridge, IL.
Ortiz, M., and F. Arocha. 2004. Alternative error distribution models for standardization of catch rates of non-target species from a pelagic longline fishery: billfish species in the Venezuelan tuna longline fishery. Fisheries Research. 70(2-3): 275-297.
Pengilly, D., and D. Tracy. 1998. Experimental effects of soak time on catch of legal-sized and nonlegal red king crabs by commercial king crab pots. Alaska Fishery Research Bulletin. 5(2): 81-87.
Powell, G. C., R. Peterson, and L. Schwarz. 1983. The red king crab, Paralithodes camtschatica, (Tilesius) in Norton Sound, Alaska: History of biological research and resource utilization through 1982. Alaska Department of Fish and Game, Division of Commercial Fisheries, Informational Leaflet #222, Juneau.
Punsley, R., and H. Nakano. 1992. Analysis of variance and standardization of longline hook rates of bigeye Thunnus obesus and yellowfin Thunnus albacares tunas in the eastern Pacific Ocean during 175-1987. Inter-American Tropical Tuna Commission Bulletin. 20: 167-184.
Quinn, T. J., II, and R. B. Deriso. 1999. Quantitative Fish Dynamics. Oxford University Press, New York, NY.
Norton Sound red king crab CPUE standardization
18
Quinn, T. J. I. 1987. Standardization of catch-per-unit-effort for trends in catchability. Natural Resource Modeling. 1: 279-296.
Quinn, T. J. I., S. H. Hoag, and G. M. Southward. 1982. Comparison of two methods of combining catch-per-unit-effort from geographic regions. Canadian Journal of Aquatic and Fisheries Science. 39: 837-846.
Rodríguez-Marín, E. 2003. Standardization of bluefin tuna, Thunnus thynnus, catch per unit effort in the baitboat fishery of the Bay of Biscay (Eastern Atlantic). ICES Journal of Marine Science. 60(6): 1216-1231.
Shirley, T. C., and S. M. Shirley. 1989. Temperature and salinity tolerances and preferences of red king crab larvae. Marine Behavioral Physiology. 16: 19-30.
Soong, J. 2008. Analysis of red king crab data from the 2008 Alaska Department of Fish and Game trawl survey of Norton Sound. Alaska Department of Fish and Game, Division of Commercial Fisheries, Fishery Data Series #08-58, Anchorage.
Soong, J., and T. Hamazaki. 2012. Analysis of red king crab data from the 2011 Alaska Department of Fish and Game trawl survey of Norton Sound. Alaska Department of Fish and Game, Division of Commercial Fisheries, Fishery Data Series #12-06, Anchorage.
Starr, P. J. 2012. Standardised CPUE analysis exploration: Using the rock lobster voluntary logbook and observer catch sampling programmes, New Zealand Fisheries Assessment Report .
Taggart, S. J., J. M. Mondragon, A. G. Andrews, and J. K. Nielsen. 2008. Spatial patterns and movements of red king and Tanner crabs: Implications for the design of marine protected areas. Marine Ecology Progress Series. 365: 151-163.
Tsou, T. S., and S. Y. Yeh. 1991. Studies on selection of standard years and abundance trneds of the South Atlantic albacore based on 1967-1988 Taiwanese longline fishery data. ICCAT, Coll. Vol. Sci. Papers. 34(SCRS/90/48): 123-127.
Vignaux, M. 1994. Catch per unit effort (CPUE) analysis of west coast South Island Cook Strait spawning hoki fisheries, 1987-93, New Zealand Fisheries Association Research Document #94/11, Wellington, New Zealand.
Zar, J. H. 1996. Biostatistical analysis, Third edition. Simon & Schluster, Upper Saddle River, NJ. Zheng, J., G. Kruse, and L. Fair. 1998. Using multiple data sets to assess red king crab in Norton Sound,
Alaska: length-based stock synthesis. pp. 591-612, In: Fishery Stock Assessment Models. Alaska Sea Grant College Program, AK-SG-98-01, Anchorage.
Zhou, S., and T. C. Shirley. 1997. A model expressing the relationship between catch and soak time for trap fisheries. North American Journal of Fisheries Management. 17: 482-487.
Zuur, A. F., E. N. Ieno, N. J. Walker, A. A. Saveliev, and G. M. Smith. 2010. Mixed Effects Models and Extensions in Ecology with R, First edition. Springer, Breinigsville, PA.
Tables
Norton Sound red king crab CPUE standardization
19
Table 1. Commercial and subsistence harvest (lbs) of red king crab in Norton Sound by fishery, 1977–
Table 2. Timeline of management actions for the Norton Sound summer commercial red king crab fishery
(Hamazaki and Zheng 2011; Menard et al. 2012).
Year Management action
1977 Summer commercial fishery begins.
1991 Fishery closed due to staff constraints.
1992 Pot limit of 100 becomes effective.
1993 Pot limit 40 for vessels <125 ft and to 50 for vessels >125 ft becomes effective.
1994 Norton Sound red king crab superexclusive designation becomes effective.
1996 Vessel moratorium in preparation for Federal license limitation program effective.
1998 Community Development Quota (CDQ) allocation becomes effective.
1999 Guideline Harvest Limit becomes effective.
1999 Alaska Board of Fisheries (BOF) promulgates new management strategy.
2000 North Pacific License Limitation Program (LLP) becomes effective. Vessels exceeding 32 ft in LOA must hold LLP issued by National Marine Fisheries Service (NMFS).
2000 CDQ groups begin to take a portion of summer harvest quota.
2002 BOF adjusts season start dates for CDQ fishery and provides for a second “clean-up” CDQ fishery.
2002 BOF changes closed water boundaries.
2002 Commercially accepted legal crab size changed from ≥ 4 ¾-in to ≥ 5-in CW.
2006 Norton Sound Section expanded but waters of Norton Sound Section above latitude of Cape Prince of Wales closed.
2008 BOF makes 4 ½ -in escape rings or 6 ½ -in stretch mesh mandatory.
2008 BOF changes start date of open access fishery from June 15 to any time on or after June 15, requirement for herring fishery to be completed before crab season starts removed.
2008 BOF changes commercial size limit for blue king crab in Norton Sound from 5 ½ to 5 inches carapace width.
2010 NMFS closes area above latitude of Cape Prince of Wales.
2012 BOF adjusts harvest strategy.
Norton Sound red king crab CPUE standardization
21
Table 3. List of 16 variables used to calculate other variables, or offered to the generalized linear model
stepwise selection procedure for standardization of Norton Sound summer commercial red king crab
CPUE data through generalized linear modeling.
Variable Description
ADF&G Number Unique 5-digit number assigned to all commercial fishing vessels in the State of Alaska, recoded here to preserve confidentiality, and renamed “Vessel.”
CPUE Number of legal male red king crabs harvested per pot lift, calculated as: Number of Crab divided by Effort.
Effort Number of pot lifts.
Fish Ticket Number Unique 6-digit number assigned to all fish tickets, which are legally required records of deliveries of commercially harvested fish or invertebrates to a processor by a vessel.
Landing Date Date crabs were delivered to a processor.
Mean Crab Weight Calculated as Pounds of Crab divided by Number of Crab.
Modified Statistical Area One of four groups of spatially proximate Statistical Areas, used only for analyses described in this document. See Appendix A 4 for definitions.
Month of Year Month of Landing Date.
Number of Crab Number of legal male red king crab harvested.
Permit Fishery One of six permit categories. See Appendix A 3 for definitions.
Permit Number Unique 5-digit number associated with a permit.
Pounds of Crab Harvested whole pounds of legal male red king crab, includes both commercial and that retained for personal use while commercial fishing. Legally required information on all fish tickets.
Statistical Area Unique 6-digit number associated with a spatially discrete marine area. Legally required information on all fish tickets.
Vessel ADF&G Number recoded to achieve confidentiality, see above.
Week of Year Week of Landing Date, ranging from 1 to 53; a week starts on Sunday and ends on Saturday.
Year Calendar year, ranging from 1977–2012, measured from January 1–December 31.
Norton Sound red king crab CPUE standardization
22
Table 4. Final generalized linear model formulae and associated R2 selected for Norton Sound summer
commercial red king crab fishery from varous data subsets. The dependent variable is ln(CPUE) in
numbers. Preferred models are shown in bold. Notation from the open source programming language R is