Annex 09 Working documents

ICES WGNSSK REPORT 2016 | 931

Annex 09 Working documents

Alternate SAM models

DATRAS standard Q3 index

The standard DATRAS Q3 indices (DATRAS indies), ages 3-8, 1992-2015 were used

instead of the GAM generated indices in the SAM assessment model. Figure 1 shows

the indices and the internal consistencies. The standard indices do not include the

Skagerrak/Kattegat, but do include the southern North Sea (where saithe are no

found). The truncated GAM-derived Q3 index (no Skagerrak/Kattegat or southern NS

were compared with the DATRAS estimates for the expanded age range (Figure 2).

The GAM and standard DATRAS indices are (generally) very similar. However, ages

4 and 5, especially in the last 2 years, are over-estimated by the GAM (especially age

5).

Results of the SAM assessment are in Figure 3. Estimated SSB using the DATRAS in-

dices closely mirrors estimates from the cpue-only model until around 2010, unlike the

model with Q3 indices estimated with the GAM model. The DATRAS model shows

slightly lower SSB than the Q3 GAM indices in 2015. Retrospective patterns show that

SSB has been consistently underestimated, while fishing mortality has been mostly

over-estimated. The retrospective patterns in F4-7 are not as poor as the model with the

Q3 GAM indices. Residual plots are in Figure 5. Estimated catchabilities were very low

for the DATRAS model, compared to the indices estimated with the GAM model.

GAM Q3 index but without 2015

Results of the SAM assessment are in Figure 6. Omitting the 2015 Q3 data resulted in

SSB2015 estimates lying between those estimated by the cpue-only model and the GAM-

estimated Q3 (with 2015) model. Retrospective patterns are in Figure 7; retrospectives

are worse than the model including 2015 data (Figure 8). Residual patterns are in Fig-

ure 9.

GAM Q3 index but with stock weights=catch weights for ages 7-10+

Not finished – bounds for stock weights

Results of the SAM assessment are in Figure 10. Replacing stock weights with catch

weights for ages 7-10+ (where stock weights were greater than catch weights) made a

large difference in the SAM output. While SSB still increases in the last two years of the

series, SSB is lower for this model until 2014 than all other models. Retrospective pat-

terns are in Figure 11. Residual patterns are in Figure 12.

DATRAS Q3 index but with stock weights=catch weights for all ages

Results of the SAM assessment are in Figure 13; this is the model recommended as the

final model based on the external review in early June. Replacing stock weights with

catch weights for all ages had the effect of increasing SSB in comparison with the model

where stock weights were replaced for ages 7-10+. This is because for ages 3-6, catch

weights are higher than stock weights (Figure 14); these are the fish the make up the

dominant part of the catch for the targeted trawl fisheries. Retrospective patterns are

in Figure 15 and residuals are in Figure 16.

https://community.ices.dk/ExpertGroups/benchmarks/2016/wknsea/2014%20Meeting%20docs/04.%20Working%20documents/Saithe/_Amendment%20WD%208_SAM%20model/Analyses_June/Catchabilities%20of%20surveys%20in%20SAM.doc

932 | ICES WGNSSK REPORT 2016

Figure 1. Standard DATRAS indices for Q3, 1992-2015, ages 3-8 and internal consistencies.

Dotted lines are 95% confidence interval for the mean.

Log10 (Younger Age)

Lo

g10 (

Old

er

Ag

e)

r2

0.236

Age 3 vs 4

r2

0.492

Age 4 vs 5

r2

0.459

Age 5 vs 6

r2

0.573

Age 6 vs 7

r2

0.597

Age 7 vs 8


Figure 2. Standard DATRAS indices for Q3, 1992-2015, ages 3-8.

1995 2000 2005 2010 2015

01

23

45

Age group 1

Year

Index

1995 2000 2005 2010 2015

01

23

45

Age group 2

Year

Index

1995 2000 2005 2010 2015

01

23

45

Age group 3

Year

Index

1995 2000 2005 2010 2015

01

23

45

Age group 4

Year

Index

1995 2000 2005 2010 2015

01

23

45

Age group 5

Year

Index

1995 2000 2005 2010 2015

01

23

45

Age group 6

Year

Index

1995 2000 2005 2010 2015

01

23

45

Age group 7

Year

Index

1995 2000 2005 2010 2015

01

23

45

Age group 8

Year

Index

1995 2000 2005 2010 2015

01

23

45

Age group 9

Year

Index

1995 2000 2005 2010 2015

01

23

45

Age group 10

Year

Index


Figure 3. Trends in SSB, F4-7, and recruitment for the 4 models. Blue line: Q1 + GAM-estimated Q3

+ cpue index model; green line: cpue-only model; orange line: combined cpue + GAM-estimated

Q3 (truncated to exclude Skagerrak/Kattegat and southern NS); black line: DATRAS Q3 + cpue

model; orange/tan shaded region: 95% confidence interval for the DATRAS Q3 + cpue model; solid

grey line (grey dashed confidence intervals) are the previously saved base model (unknown at this

point).


Figure 4. Eight year retrospective pattern in SSB, F4-7, and recruitment. Model is combined cpue +

DATRAS Q3 and includes the discard revisions.


Figure 5. Residual patterns for the combined cpue + DATRAS Q3 assessment model. (left) Before

correlation taken into account between ages, within years in the Q3 index; (right) after accounting

for the correlation.




Q3 (truncated to exclude Skagerrak/Kattegat and southern NS); purple line: DATRAS Q3 + cpue

model; black line (orange/tan shaded region: 95% confidence interval): GAM-estimated Q3 indices

without 2015 + cpue model; solid grey line (grey dashed confidence intervals) are the previously

saved base model (unknown at this point).



GAM-estimated Q3 (without 2015, excludes Skagerrak/Kattegat and southern North Sea) and in-

cludes the discard revisions.



GAM-estimated Q3 (excludes Skagerrak/Kattegat and southern North Sea) and includes the dis-

card revisions.


Figure 9. Residual patterns for the combined cpue + GAM-estimated Q3 assessment model (no

2015). (left) Before correlation taken into account between ages, within years in the Q3 index; (right)

after accounting for the correlation.




Q3 (truncated to exclude Skagerrak/Kattegat and southern NS); purple line: DATRAS Q3 + com-

bined cpue; black line (orange/tan shaded region: 95% confidence interval): GAM-estimated Q3

indices + cpue model + stock weights=catch weights for ages 7-10+; solid grey line (grey dashed

confidence intervals) are the previously saved base model (unknown at this point).



GAM-estimated Q3 (excludes Skagerrak/Kattegat and southern North Sea) + stock weights=catch

weights for ages 7-10+ (includes discard revisions).


Figure 12. Residual patterns for the combined cpue + GAM-estimated Q3 + stock weights=catch

weights for ages 7-10+ assessment model. (left) Before correlation taken into account between ages,

within years in the Q3 index; (right) after accounting for the correlation.


Figure 13. Trends in SSB, F4-7, and recruitment for various models. Blue line: (benchmark model)

Q1 + GAM-estimated Q3 + combined cpue index model; green line: combined cpue-only model;

orange line: combined cpue + GAM-estimated Q3 (truncated to exclude Skagerrak/Kattegat and

southern NS); purple line: GAM-estimated Q3 indices + combined cpue model + stock

weights=catch weights for ages 7-10+; black line (orange/tan shaded region: 95% confidence inter-

val): DATRAS Q3 indices + combined cpue, stock weights=catch weights for all ages.

Figure 14. Stock weights (dashed lines) and catch weights (solid lines) for ages 1-10+. The left panel

shows age 3 (black lines) to age 6 (light blue lines), while ages 7-10+ are in the right panel. This

figure differs from the benchmark working document due to re-raising (InterCatch bug and chang-

ing of raising procedure for Norwegian discards).

2002 2004 2006 2008 2010 2012 2014

01

23

4 cw age 3

sw age 3

2002 2004 2006 2008 2010 2012 2014

02

46

81

0

cw

[, 5

]

cw age 7

sw age 7



DATRAS Q3 (excludes Skagerrak/Kattegat) + stock weights=catch weights for all ages (includes

discard revisions).


Figure 16. Residual patterns for the combined cpue + DATRAS Q3 (excludes Skagerrak/Kattegat)

(stock weights=catch weights for all ages) assessment model. (left) Before correlation taken into

account between ages, within years in the Q3 index; (right) after accounting for the correlation.


RBE

SIH Campagne

CGFS : Change of vessel from 2015

onwards and consequences on sur-

vey design and stock indices


1. Introduction 949

2. Adaptation of the sampling design 949

2.1. Rationale 949

2.2. Selected scenario 950

2.2.1. Red mullet 950

2.2.2. Plaice 950

3. From CPUE in number per hour fished to CPUE in number per km² 951

3.1. Differences in indices provided in 2015 and 2016 952

3.1.1. Comparison Old/New index for plaice 952

3.1.2. Comparison Old/New index for mullet 955

Annexe 1: Hauls kept in the new survey 957

https://community.ices.dk/ExpertGroups/WGNSSK/2016%20Meeting%20docs/03.%20Draft%20Report%202016/36%20WGNSSK%20Report%20-%20Annex%2009%20Working%20Documents/New_FR_GFS_survey.docx#_Toc449621554












Introduction

The Channel Ground Fish Survey (CGFS) has been conducted in the eastern English

Channel yearly in October since 1988 with a systematic fixed sampling program. The

CGFS was realized using a high opening (GOV) bottom trawl (20 mm meshsize coden)

and 30 minutes trawls using the same RV Gwen Drez since 1988.

The RV Gwen Drez was decommissioned in 2015 but given the international im-

portance of the CGFS it was decided to continue the time series using the RV Thalassa.

In order to allow for a continuation of the time series an intercalibration was realized

in 2014 by conducting paired tows, simultaneously with both vessels (see appendix of

the WGIBTS 2015 report for description of the intercalibration results).

Adaptation of the sampling design

Rationale

Based on the characteristics of the new RV Thalassa (bigger draught), and the vessel

time availability at this period of the year, three scenarios of reduction of the trawling

stations set have been tested. For each scenario, a selection of hauls was made among

the 89 hauls of original sampling scheme of the survey (Fig. 1) based on different crite-

ria. The relevance of these subsets of hauls was assessed by resampling on the historical

time series and computing the associated abundance indexes per age for plaice (Pleu-

ronectes platessa) and red mullet (Mullus surmuletus) which are assessed using this sur-

vey as tuning fleet in the ICES WGNSSK (Working Group on North Sea, Skagerrak and

Kattegat).

Figure 1 : Hauls of the original CGFS sampling scheme


Selected scenario

After a trial and error experiment on the haul selection, the selected sampling scheme

include the areas easily fishable by the RV Thalassa (69 hauls that are always more than

15 meters deep) and some of the shallower hauls (limited to the ones which the average

bathymetry over the time series is over 15 meters: 5 hauls). This selection allowed for

covering 74 hauls (i.e. 83% of the initial sampling scheme, Ann.1). It excludes the hauls

outside VIId which were not used.

To test the relevance of the selected hauls, the internal consistency of the indices was

tested for two species (plaice and red mullet).

Red mullet

Figure 2 : Internal consistency ; 2a : New index based on the subset of hauls, 2b : reference index

Correlation coefficients appeared to be higher with this selection of hauls than with the

original sampling scheme, improving the internal consistency of the index for red mul-

let.

Plaice

Figure 3 : Internal consistency ; 3a : New index based on the subset of hauls, 3b : reference index

2a 2b

3a 3b


The internal consistency of the original index was not completely satisfying. The new

sampling scheme resulting from a subset of the original stations do not deteriorate this

internal consistency further. The correlation coefficients are of the same order of mag-

nitude for ages 2/3, 3/4 and slightly increased for ages 4/5.

Figure 3 : New spatial coverage of the Channel Ground Fish Survey

From CPUE in number per hour fished to CPUE in number per km²

The original index provided was computed in number of fish per hour fished. In a first

step an index was computed per ICES square (the stratum in this survey) and then

elevated to the whole Eastern Channel to compute a number of fish per age class and

hour fished.

As the surface trawled differed between the two RV (difference in trawling speed and

width of the gear used (0.029 km² on average for the RV Gwen Drez over the period

2008-2014 against 0.052 km² for the RV Thalassa (average of the hauls realized in 2015))

a density index (number of fish per km²) was also tried in order to create a consistent

index over the whole time series. This is in line with the current effort led by the

IBTSWG to produce trawled surface and density indices for all the expert groups for

2017.

The index is then computed using the formula:

With :

sNmean abundance in the strata s, expressed in num-

ber/km²


s

s

s

ss

A

NA

N

.

sASurface of the strata s, in km²

As the vertical opening of the gear used by the RV Thalassa was higher than the pre-

vious one, and in order to take into account any vessel effect on catchability, the CPUE

were compared for all the species caught. Differences in CPUEs between the new and

the old survey setting were found for 9 species (mostly pelagic species). In the case of

plaice and red mullet, CPUEs were not significantly different, so no conversion factor

was applied to these two species.

Differences in indices provided in 2015 and 2016

During the process of automatizing the computation of the index, some errors were

found in the surface of some strata and ALK used for some species. These errors where

corrected and the new indices (expressed in number of fish per km² instead of number

of fish per hour fished) take these corrections into account.

In order to compare the “old” and “new” CGFS indices for plaice and red mullet they

were first plotted against each other to get a visual comparison of the index values at

age and assess the possible differences and inconsistencies. The correlations between

indices at age time series were then computed to check for consistency between these

two indices. The last step was to check the internal consistency to assess the impact of

the new calculation on the indices.

Comparison Old/New index for plaice

Index at age

Figure 4 : CGFS old (blue) and new (black) standardized index at age


The main trends of the CGFS index at age remain very similar. The main differences

are:

for age 1 in 1997 where the peak observed is no longer observed with the

new calculations;

a new peak in the age 1 in 2011 which is in line with what was observed by

the UK-BTS survey that year;

the main differences are observed for age 6, where the two indices seem to

be inconsistent ;

for age 0 in 2000 and age 1 in 2011, two new peaks appeared with the new

calculation.

In the assessment only the ages 1 to 6 are used.

Correlations between the two different indices at age time series

Figure 5 : Correlation between indices at age for the old and new indices

The correlations for ages 0, 2, 3, 4 and 5 are high, reflecting the coherence seen when

plotting the old and new surveys against each other. Correlations for ages 1 and 6 are

weaker, also reflecting the differences for some years for age 1 and a poor consistency

between new and old indices for age 6.


Internal consistency

Figure 6: Internal consistency for new and old indices

The internal consistency is globally improved. Correlation coefficients are increased

for ages 1/2 and 4/5 while they do not vary much for ages 2/3 and 3/4.


Comparison Old/New index for mullet

Index at age

Figure 7: CGFS old (black) and new (blue) standardized index at age

The main trends of the CGFS index at age are remaining very similar. The main differ-

ences are for age 2 in 2008 where the peak observed in the new calculations is higher

than the one from the old index. For Age 4 in 2004 and 2005, indices seem to be incon-

sistent with a decrease between 2005/2006 whereas the index increased with the old

index.


Correlations between the two different indices at age time series

Figure 8

Correlation between indices at age for the old and new indices

The correlations for all ages are high but with very few data points after age 4.

Internal consistency

Figure 9: Internal consistency for new and old indices


The main patterns are maintained from the old to the new index. The higher correlation

is between age 2 and 3 but increased with the new index.

Annex 1: Hauls kept in the new survey

TRAIT_SELECTION_CGFS_THALASSA_S3

sta_Recodage Latitude Longitude Trait2014

4M1 50.195 1.171667 62

4M2 50.01667 1.083333 67

5L1 50.385 0.7916667 84

5M1 50.305 1.171667 61

6J1 50.54 0.3533333

6K1 50.56167 0.7283334 57

6L1 50.53667 0.89 59

2D2 49.59 -1.118333 32

2E1 49.58333 -0.9666666 34

3D1 49.82833 -1.135 1

3E1 49.99833 -0.7683333 2

3F1 49.96 -0.6216667 17

3H1 49.90333 -0.1216667 36

3I1 49.84333 0.1633333 71

3J1 49.87833 0.4333333 70

3K1 49.895 0.5116667 69

3L1 49.98167 0.825 68

4C1 50.04 -1.275 96

4D1 50.09833 -1.265 4

4E1 50.02833 -0.905 3

4F1 50.075 -0.6183333 15

4G1 50.08833 -0.4566667 14

4H1 50.245 -0.05 10

4I1 50.01833 0.175 37

4J1 50.09333 0.3266667 83

4K1 50.11333 0.6016667 41

5D1 50.415 -1.166667 76

5E1 50.47667 -0.9066667 74

5F1 50.44333 -0.5966667 72

5H1 50.34667 -0.1583333 12

5I1 50.355 0.005 11

5J1 50.30167 0.4133333 86

5K1 50.35833 0.6366667 85




6E1 50.52333 -0.8833333 75

6F1 50.525 -0.71 73

6G1 50.57333 -0.4433333 80

1D1 49.42333 -1.058333 29

2D1 49.51167 -1.223333 97

2I1 49.64 8.166666E-02 88

2I2 49.60167 8.333334E-02 89

1E1 49.42333 -0.985 28

1E2 49.45 -0.9233333 27

1F1 49.46167 -0.675 26

1F2 49.41667 -0.5533333 25

1G1 49.45833 -0.43 24

1G2 49.47167 -0.325 23

2F1 49.66167 -0.6433333 35

2G1 49.55667 -0.3316667 20

2H1 49.65333 -0.145 19

3G1 49.84 -0.2533333 18

1H1 49.46333 -0.1433333 22

1H2 49.35833 -0.1766667 90

7O2 50.79 1.558333 49

7G1 50.76 -0.2833333 79

7K1 50.79167 0.5333334 54

7L1 50.87667 0.8366666 53

7L2 50.78167 0.84 56

7M1 50.97 1.085 51

6O1 50.655 1.541667 42

7O1 50.91333 1.61 48

7H1 50.755 -0.1216667 78

7N1 50.86666 1.346667 50

6H1 50.56 -0.1266667 81

6I1 50.635 7.666667E-02

3M1 50.00834 1.218333 65

4N1 50.2 1.39 63

4N2 50.09 1.37 64

5N1 50.47167 1.438333 46

5N2 50.41833 1.345 47

5O1 50.44833 1.526667

6M1 50.66 1.005 58




6N1 50.57667 1.43 44

6O2 50.56333 1.51 43

4L1 50.15667 0.9766667 60


Working Document to the ICES Working Group on the Assess-

ment of Demersal Stocks in the North Sea and Skagerrak

(WGNSSK), April 2016

Not to be cited without prior reference to the author

Plaice (Pleuronectes platessa) and sole (Solea solea) in the UK Beam

trawl survey in the Eastern English Channel (7d)

J. F. Silva

Centre for Environment, Fisheries and Aquaculture Science (CEFAS), Lowestoft Laboratory,

Pakefield Road, Lowestoft, Suffolk, NR33 0HT, UK

Abstract

The present document describes the calculation of the plaice and sole survey indices in

the Q3 UK beam trawl survey in the ICES division 7d. Further investigations were

made in relation to the survey data quality, selection of prime stations, age-at-length

keys and ultimately the indices calculations. Results for revisited index and previous

index are shown in the present document to facilitate a better comparison and further

inform on temporal differences to the year class abundance.

Survey indices

The present document describes the calculation of the plaice and sole indices in the UK

Q3 beam trawl survey in the Eastern English Channel and Southern North Sea. The

annual procedure is currently done automatically using the Cefas Fishing Survey Sys-

tem (FSS) and R software, and provides the index for the time-series since 1989 and

1996 (7d and 4c, respectively). Prior to 2016, survey indices were calculated using Cefas

FSS, SAS code and Microsoft Excel® outputs. Whilst re-writing the code from SAS to

R, discrepancies were found in the selection of valid primes used in the production of

age-length keys (ALKs) and length-distributions (LDs), with survey data within Cefas

FSS revisited and corrected accordingly, where possible. It should also be noted, cur-

rent survey biological sampling targets (otoliths) for both species are set by sector (7d

UK Inshore, 7d France Inshore and 4c North Sea), though previous indices had calcu-

lated ALKs by ICES rectangles.

Therefore, this document refers to an update of index calculations so that they are con-

sistent with current survey data collection protocols. Data prior to 2005 presented for

the 7d area should be viewed and used with some caution, since these data were not

revisited and reviewed in terms of their quality, and historical data collection proce-

dures may differ from the current one.

New results for survey area 4c were also provided to the 2016 ICES WGs, and although

are not discussed in the present document, should be viewed only as provisional, be-

cause further investigations are required on the survey data and historical prime selec-

tion when current primes were not fished.

A total of 75 primes (39 in the UK and 36 in the FR sector), were selected from 1989-

2015, with a few currently not fished, though historically relevant (Figure 1). Primes


used for the length-distributions (LD) and fishing effort are within the UK sector, 22–

27, 42–45, 47, 50–67, 73–75 and 94, with 1, 4, 6–12, 16–21, 29, 35–40, 68–72, 76–77 and 95

within the FR sector. Primes 2, 3, 5, 14, 15, 41 and 46 are included only in the age-length

key (ALK) as they have not been fished in recent years, though historically were part

of the survey primary grid. Similarly, only included in the ALKs calculations are

primes 200, 201, 202 and 203 within the UK sector. These are set as additional and no

longer fished since 2014, though historically, otoliths have been collected as part of the

survey target. It should be noted that the prevalence of static gear around prime 49,

currently on the main survey grid, has prevented the tow to be fished successfully in

recent years. Therefore, data collected for the latter prime has been excluded from LDs

and fishing effort, and only used as part of the ALK for the UK stratum.

R code procedures include an initial data retrieval from Cefas FSS, where data are elec-

tronically stored, for the relevant prime stations where fishing operations were consid-

ered valid. Numbers at length for each fishing station are standardized to 30 minute

tows, with the raising factor dependent on the actual tow duration. The total number

across stations within an ICES rectangle results in the LD for ICES rectangle within

sector (UK and FR). The ALK derived from the biological sampling at sea (otolith col-

lection) is produced separately by UK and FR sector and raised to the appropriate LDs,

resulting in an age-length composition for each ICES rectangle-sector-sex combination.

The ALKs and LDs for plaice are calculated by sex for all years, and for sole calculated

by sex when measured and biologically sampled by sex (1993–2009), and combined

when measured and biologically sampled unsexed (1989–1992, 2010–onwards). The

LDs used are only from valid stations; meanwhile ALKs use all stations within the

chosen primes, even when considered additional or invalid tows to accommodate the

occasional biological sampling occurrence. The total numbers across lengths by age

create the age composition (AC) for each ICES rectangle-sector-sex combination, with

the sum as the AC for the survey year. These are divided by the total number of valid

primes fished across UK and FR sectors, which may differ from the number of primes

with plaice and/or sole catches. The results are further raised and multiplied by four to

give the final index equivalent to one-hour tows with an 8-metre beam trawl (the factor

four is because stations are standardised to 30 minute tows and conducted with a 4-

metre beam trawl).

Furthermore, the R code is designed to reallocate, where possible, miss-matches where

a fish at a given length has no associated record in the ALK, with code reallocating

numbers at length (LD) up to a maximum of ± 2cm of the initial length. If there are age

records either above or below the initial length group, the fish are reallocated to those

respective length groups within the LD. However, if age records are found in both

lengths above and below the initial length group, the fish are split between those two

lengths groups, using the ratio of each value divide by the sum of both length groups.

If code is unable to reallocate fish, data are not used for further index calculations.

Results

The revised survey indices for plaice and sole in 7d area are presented on Table 1 and

3. Previous index provided to the WG is presented on Table 2 and 4. A comparison

between the two indices is presented on Figure 2 and 3 so as to better inform through

visualization if there are any substantial temporal changes on year class abundance for

fish aged one to six (ages currently used in the assessment).

Overall long-term trends for plaice are similar between the two indices for 1-year to 6-

year class (Figure 2). Meanwhile, for sole, although the main increases and declines are


being picked up by the two indices, there may be few discrepancies with historical data

(e.g. 1999) (Figure 3).

References

R Core Team (2014). R: A language and environment for statistical computing. R Foundation for

Statistical Computing, Vienna, Austria. URL http://www.R-project.org/

http://www.r-project.org/


Table 1 – Revised index for plaice in the UK-7D BTS (1989 – 2015)

AGE/YEAR 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002

0 4.39 1.30 0.00 0.00 0.00 0.20 0.00 24.14 0.98 43.19 1.38 1.59 2.73 1.31

1 3.79 9.24 16.80 22.37 4.59 9.35 14.48 22.09 48.17 30.59 12.82 19.53 27.90 37.86

2 15.84 9.39 14.53 21.31 20.18 8.54 6.24 17.26 28.55 37.93 10.67 30.19 20.27 25.86

3 28.93 11.13 11.47 6.60 7.99 10.07 3.80 1.73 10.97 12.06 28.77 18.75 14.12 12.51

4 31.66 11.73 8.68 6.64 2.79 5.95 5.68 1.03 1.25 4.98 4.62 20.47 9.82 5.46

5 4.00 12.59 8.64 7.17 2.87 1.98 2.22 2.00 1.57 0.63 1.61 4.99 14.84 2.62

6 1.72 1.53 4.60 5.41 2.38 0.61 0.75 1.29 0.51 0.60 0.31 1.27 2.74 5.28

7 1.65 0.96 1.83 3.20 3.05 0.97 0.75 0.57 0.56 0.65 0.19 0.73 0.78 0.98

8 0.63 1.23 1.08 0.54 3.42 1.73 1.48 0.38 0.36 0.32 0.26 0.38 0.45 0.20

9 0.31 1.02 0.11 0.28 0.62 1.78 1.17 0.66 0.20 0.30 0.13 0.44 0.32 0.17

10 + 1.75 0.63 1.14 0.79 0.65 0.80 1.36 4.13 1.84 2.03 1.01 2.04 1.79 0.90

Total

(ages 1-10+) 90.27 59.44 68.87 74.30 48.53 41.77 37.93 51.12 93.98 90.10 60.39 98.79 93.04 91.83

Age/Year 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015

0 3.20 15.97 0.34 5.58 0.23 0.13 8.76 1.36 12.30 0.00 0.22 0.52 0.00

1 10.62 52.93 15.62 30.06 53.11 39.58 77.73 64.24 115.07 24.69 32.26 145.33 37.99

2 39.70 22.48 36.18 28.85 28.90 40.58 39.53 64.70 112.22 81.10 61.02 156.47 178.70

3 9.81 20.72 12.80 16.80 12.17 10.51 20.92 17.74 39.55 55.98 88.19 50.67 63.19

4 4.42 4.75 10.04 5.94 6.21 4.29 5.87 9.15 10.28 18.65 45.04 62.13 30.15

5 2.28 1.15 3.19 4.27 3.17 3.84 3.23 3.12 7.00 4.24 10.24 26.75 33.42

6 1.14 0.26 1.07 1.31 2.90 1.80 2.27 1.72 2.85 3.30 3.41 8.95 15.69

7 2.67 0.84 0.64 1.08 0.82 0.90 0.77 1.27 1.09 1.06 1.13 1.96 3.30

8 0.81 1.27 0.43 0.59 0.59 0.67 1.30 0.18 0.34 0.90 1.08 1.82 1.21

9 0.20 0.23 0.99 0.33 0.19 0.16 0.33 0.35 0.70 0.66 0.13 0.92 0.27

10 + 0.47 0.55 0.98 0.94 1.59 0.39 1.19 0.99 1.05 0.95 0.92 1.20 0.44

Total

(ages 1-10+)

72.12 105.18 81.96 90.17 109.64 102.73 153.13 163.47 290.15 191.52 243.43 456.19 364.37


Table 2 – Previous index for plaice in the UK-7D BTS (1989 – 2014)

AGE/YEAR 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001

1 2.31 5.16 11.75 16.53 3.22 8.33 11.32 13.20 33.15 11.38 11.30 13.19 17.91

2 12.13 4.86 9.06 12.54 13.40 7.46 4.06 11.90 13.48 27.30 14.10 20.96 13.02

3 16.63 5.76 6.98 4.19 4.96 9.17 3.00 1.30 4.22 6.99 15.90 14.39 10.00

4 19.94 6.70 5.30 4.17 1.75 5.56 3.67 0.70 0.65 3.12 2.90 13.81 7.12

5 3.30 7.53 5.43 5.57 1.89 1.95 1.49 1.30 0.34 0.32 1.00 3.48 10.94

6 1.48 1.76 3.20 4.88 1.57 0.77 0.58 0.90 0.32 0.22 0.20 0.87 1.95

7 1.32 0.65 1.22 3.44 2.05 0.90 0.59 0.40 0.24 0.15 0.10 0.57 0.53

8 0.54 0.97 0.99 0.66 2.78 1.83 1.32 0.30 0.21 0.11 0.30 0.18 0.30

9 0.30 0.75 0.06 0.49 0.39 1.24 0.82 0.40 0.17 0.05 0.10 0.43 0.19

10 + 1.65 0.37 1.24 0.72 0.57 0.81 0.78 2.80 1.86 0.98 0.90 1.52 0.99

Total (ages 1-10+) 59.60 34.51 45.23 53.19 32.57 38.03 27.63 33.20 54.64 50.62 46.80 69.40 62.94

Age/Year 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014

1 20.66 6.18 36.18 10.84 17.21 42.61 30.28 71.62 65.25 105.55 23.23 34.33 153.63

2 15.95 22.79 14.97 31.21 16.11 18.81 26.52 42.88 63.83 95.31 76.07 59.27 140.96

3 7.73 6.00 13.15 13.77 9.22 8.70 7.20 19.15 17.27 35.70 45.26 87.99 50.67

4 3.55 2.94 3.44 10.28 3.35 3.87 2.97 5.74 8.90 9.25 12.73 45.47 55.50

5 1.80 1.61 0.91 2.95 2.64 1.75 2.32 3.20 3.04 6.68 3.53 10.58 25.08

6 3.46 0.79 0.16 1.17 0.77 1.95 1.11 2.17 1.90 2.82 1.61 3.54 9.13

7 0.72 1.77 0.66 0.77 0.57 0.80 0.50 0.78 1.38 1.40 0.42 1.03 2.32

8 0.14 0.60 1.16 0.42 0.31 0.30 0.41 1.24 0.30 0.19 0.41 1.37 1.88

9 0.11 0.11 0.17 0.86 0.14 0.10 0.09 0.37 0.36 0.57 0.43 0.14 1.01

10 + 0.61 0.28 0.17 0.65 0.46 1.11 0.25 1.31 0.89 0.95 0.12 0.20 1.36

Total (ages 1-10+) 54.71 43.06 70.97 72.91 50.79 80.01 71.66 148.46 163.10 258.41 163.82 243.92 441.55


Table 3 – Revised index for sole in the UK-7D BTS (1989 – 2015)

AGE/YEAR 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002

0 0.16 0.00 0.00 0.00 0.00 0.00 0.06 5.55 0.06 0.13 2.56 0.00 1.27 0.00

1 3.01 17.96 12.14 1.33 0.82 8.33 5.89 5.30 24.75 3.27 35.99 14.98 10.19 53.56

2 22.09 5.55 31.17 15.29 22.96 4.26 16.09 10.79 10.85 24.11 8.22 27.45 27.88 16.11

3 4.62 5.55 3.19 13.47 11.42 11.07 2.22 5.97 4.42 3.67 11.33 5.52 11.55 8.60

4 2.45 1.24 2.82 1.07 9.97 4.65 3.51 1.07 1.94 1.47 1.59 4.85 1.67 5.11

5 0.56 1.01 0.48 1.61 1.14 4.30 1.67 1.86 0.26 0.83 0.73 1.48 2.33 0.45

6 0.35 0.33 0.67 0.34 1.52 0.28 2.12 1.15 0.82 0.19 1.02 0.68 0.75 1.04

7 0.26 0.06 0.16 0.50 0.34 0.90 0.28 1.55 0.52 0.37 0.19 0.34 0.63 0.59

8 0.05 0.15 0.20 0.11 0.34 0.09 0.53 0.20 0.96 0.08 0.54 0.00 0.48 0.17

9 0.00 0.00 0.07 0.30 0.07 0.46 0.20 0.65 0.07 0.13 0.43 0.34 0.12 0.00

10 + 0.72 0.16 0.26 1.11 0.40 0.46 0.32 0.59 0.62 0.35 0.54 1.06 0.86 0.72

Total

(ages 1-10+)

34.11 32.00 51.14 35.15 48.98 34.80 32.84 29.14 45.21 34.48 60.59 56.70 56.46 86.36

Age/Year 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015

0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.45 0.00

1 11.03 12.67 43.27 10.84 2.57 3.77 51.25 16.59 13.66 1.75 0.72 25.39 25.24

2 45.65 11.81 6.91 42.62 28.97 7.35 19.16 30.76 28.60 9.72 8.91 16.35 21.36

3 5.87 10.97 3.50 4.51 15.45 9.14 7.10 5.14 14.70 7.51 15.09 12.38 6.04

4 3.20 2.08 5.18 2.68 1.47 5.82 5.81 1.66 1.66 3.53 9.72 11.92 2.29

5 2.05 2.02 1.90 2.59 1.04 0.40 5.02 2.70 0.54 0.92 3.23 5.09 4.51

6 0.42 1.34 1.15 0.55 1.56 0.68 0.44 2.73 2.62 0.39 1.12 2.73 2.08

7 0.55 0.41 0.71 0.47 0.44 0.37 0.31 0.33 0.77 0.78 0.51 1.08 2.20

8 0.27 0.64 0.08 0.66 0.21 0.37 0.63 0.06 0.24 0.67 0.89 0.32 0.20

9 0.03 0.26 0.36 0.00 0.55 0.25 0.26 0.49 0.19 0.00 0.78 0.20 0.00

10 + 0.92 0.88 0.35 0.40 0.53 0.26 0.59 0.31 0.12 0.70 0.17 0.70 0.67

Total

(ages 1-10+)

69.99 43.08 63.40 65.32 52.79 28.41 90.58 60.78 63.11 25.97 41.13 76.15 64.60


Table 4 – Previous index for sole in the UK-7D BTS (1989 – 2014)

AGE/YEAR 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001

1 2.60 12.10 8.90 1.40 0.50 4.80 3.50 3.50 19.00 2.00 28.14 10.49 9.09

2 15.40 3.70 22.80 12.00 17.50 3.20 10.60 7.30 7.30 21.20 9.44 22.03 21.01

3 3.40 3.40 2.20 10.00 8.40 8.30 1.50 3.80 3.20 2.50 13.17 4.15 8.36

4 1.70 0.70 2.30 0.70 7.00 3.30 2.30 0.70 1.30 1.00 2.51 4.24 1.20

5 0.60 0.80 0.30 1.10 0.80 3.30 1.20 1.30 0.20 0.90 1.73 1.03 1.91

6 0.20 0.20 0.50 0.30 1.00 0.20 1.50 0.90 0.50 0.10 1.28 0.58 0.54

7 0.20 0.10 0.10 0.50 0.30 0.60 0.20 1.10 0.40 0.30 0.16 0.28 0.57

8 0.00 0.20 0.20 0.10 0.20 0.10 0.30 0.10 0.90 0.00 0.93 0.03 0.35

9 0.00 0.00 0.10 0.20 0.00 0.30 0.20 0.50 0.00 0.10 1.07 0.24 0.04

10 + 0.70 0.00 0.10 0.60 0.40 0.30 0.30 0.40 0.70 0.30 0.47 1.20 1.01

Total (ages 1-10+) 24.80 21.20 37.50 26.90 36.10 24.40 21.60 19.60 33.50 28.40 58.89 44.28 44.09

Age/Year 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014

1 31.76 6.47 7.35 25.00 6.30 2.14 2.86 30.54 15.90 11.92 1.77 0.78 25.53

2 11.42 28.48 8.49 5.04 29.18 21.86 6.46 13.33 30.12 23.54 9.28 9.20 13.93

3 5.42 4.13 7.71 2.86 2.83 12.90 7.24 5.44 5.32 11.56 6.57 15.54 9.87

4 3.45 2.46 1.57 3.47 1.99 1.22 4.82 4.34 1.66 1.25 3.41 8.91 11.31

5 0.27 1.58 1.45 1.63 1.95 0.80 0.25 3.76 2.82 0.57 0.88 2.95 5.22

6 0.71 0.30 0.99 1.02 0.34 1.20 0.49 0.37 2.38 2.56 0.39 1.35 3.52

7 0.44 0.39 0.20 0.66 0.44 0.32 0.38 0.20 0.35 0.60 0.66 0.37 1.40

8 0.09 0.20 0.44 0.06 0.57 0.17 0.27 0.31 0.16 0.16 0.52 0.97 0.85

9 0.00 0.07 0.21 0.31 0.00 0.59 0.24 0.23 0.55 0.21 0.00 0.75 0.23

10 + 0.56 0.52 0.57 0.35 0.34 1.02 0.20 0.48 0.31 0.06 0.66 0.10 0.26

Total (ages 1-10+) 54.12 44.60 28.98 40.40 43.93 42.22 23.21 59.01 59.56 52.44 24.16 40.92 72.11


Figure 1 – Prime stations for Q3 UK beam trawl survey for survey index calculation (1989 – 2015)


Figure 2 – Long-term trends of plaice survey index in the UK – 7D BTS (revised and previous index) for 1-year to 6–year class.

0

20

40

60

80

100

120

140

160

180

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

Age_1

Revised Index Previous Index

0

20

40

60

80

100

120

140

160

180

200

19

89

19

90

19

91

19

92

19

93

19

94

19

95

19

96

19

97

19

98

19

99

20

00

20

01

20

02

20

03

20

04

20

05

20

06

20

07

20

08

20

09

20

10

20

11

20

12

20

13

20

14

20

15

Age_2


0

10

20

30

40

50

60

70

80

90

100

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

Age_3


0

10

20

30

40

50

60

70

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

Age_4


0

5

10

15

20

25

30

35

40

19

89

19

90

19

91

19

92

19

93

19

94

19

95

19

96

19

97

19

98

19

99

20

00

20

01

20

02

20

03

20

04

20

05

20

06

20

07

20

08

20

09

20

10

20

11

20

12

20

13

20

14

Age_5


0

2

4

6

8

10

12

14

16

18

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

Age_6



Figure 3 – Long-term trends of sole survey index in the UK – 7D BTS (revised and previous index) for 1-year to 6–year class.

0

10

20

30

40

50

60

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

Age_1

Previous Index Revised Index

0

5

10

15

20

25

30

35

40

45

50

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

Age_2


0

2

4

6

8

10

12

14

16

18

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

Age_3


0

2

4

6

8

10

12

14

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

Age_4


0

1

2

3

4

5

6

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

Age_5


0

2

4

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

Age_6



WD 8: SAM Assessment - AMENDMENT

WGNSSK had some concerns about the saithe assessment model.

Running the forecast with the benchmark-approved model resulted in un-

realistically high increases in TAC for the advice (119% increase, MSY ap-

proach). The working group asked to review the model with only the

standardized combined cpue index tuned to the exploitable biomass.

A model using both the standardized combined cpue index (FBI) and the

IBTS Q3 survey was put forward as an alternate model. Because this model

diverges from the cpue-only model, properties of the survey were investi-

gated (e.g., internal consistency, cross-consistency with other data, cover-

age).

This prompted a more thorough exploration of the survey data to:

Determine if spatial changes had occurred in the survey that could be

the result of fish moving in and out of the survey area (unrelated to stock

size).

Investigate the Q3 index models.

Include a ship effect to determine whether a newly added ship at

the end of the time series might be causing the problem (e.g., Dana

in Skagerrak).

Modify the spatial grid over which the indices are estimated so that

it is roughly representative of the population (do not include large

areas where there are almost no saithe). Two potential indices were

explored: one that removed the Skagerrak/Kattegat and southern

North Sea (south of 57° N); one that kept the Skagerrak but re-

moved the Kattegat and southern North Sea.

Investigate consistencies for each model option.

Determine the effect of various age groups.

There were questions regarding the use of SAM vs. XSA. Discussions via

email may have put this option to rest, but are summarized as:

When XSA and SAM are run with the same datasets, the XSA results fall

more-or-less within the confidence limits of SAM (Figure 1).

Reverting to XSA would actually hamper our ability to investigate the

uncertainties arising from the different datasets.

The 3 cpue indices get a very high weight and the IBTS q3 has hardly

any influence; this hides the issue that the assessment relies nearly en-

tirely on the commercial indices.

Would need to revert back to the age-based cpue indices because XSA

cannot handle the combined standardized index, that is fit to the ex-

ploitable biomass (within the model). This reverts back to the issue of

using the age information twice – once for the catch data, once for the

cpue tuning indices.

The XSA cannot handle the correlation between ages with years in the

survey indices; SAM can, as outlined in Berg & Nielsen (2016).

A bug in InterCatch resulted in the re-raising of discards for 2003, 2006, 2011, and 2014,

which were done following the procedure in WD 5; 2002 was also re-raised as it seemed

oddly high. After re-raising the data, several years still appeared to be atypical, so the

http://icesjms.oxfordjournals.org/content/early/2016/04/14/icesjms.fsw046.abstract%29.%3Chttp:/icesjms.oxfordjournals.org/content/early/2016/04/14/icesjms.fsw046.abstract


raising for all years was re-done following a modification to the rules used for the

benchmark:

No discard ratio >= 25% was used in the raising of any fleet. Previously, ra-

tios > 30% were omitted.

Norwegian trawler fleet discards were raised using German or French (or

both) discard information. Previously, they were raised with other

OTB_DEF fleets, using discard information from all OTB_DEF fleets for a

given area and quarter.

Results

Spatial changes in the surveys

Spatial plots of the catches (all ages combined) showed that, for the Q1 survey, saithe

were mainly on the shelf edges and the survey was unlikely to be sampling much of

the population (see Appendix: Q1 plots are catch weight per station per year, not age

specific). At the time of the benchmark, this was discussed, but it was thought that, for

the older ages, the amount of the population surveyed should be fairly consistent over

time. A month parameter had been added to the delta-GAM model to account for

changes in survey timing and any effect of fish movement in and out of the survey

area. However, closer inspection of the figures showed that, in some years, fish are

found further up on the shelf, while in other years, they are only along the shelf bound-

ary (200 m contour). This does call into question using the Q1 index in the assessment.

For the Q3 surveys, saithe are found on the northern part of the shelf, along the shelf

boundary, and in the Skagerrak (see Appendix: Q3 plots are catch weight per station

per year, not age specific). The amount of saithe found within the area differs, but the

distribution appeared fairly consistent. Stronger year classes are, for the most part, ap-

pearing in the survey when expected and persisting for at least 1 year (e.g., 1995, 2001,

2005).

Q3 index models

A ship effect was included in the index estimation. Sweden had begun using a new

vessel in 2011 in the Skagerrak. Including ship in the model resulted in a higher AIC

and BIC, and slightly worse internal consistencies (Tables 1, 2).

The spatial grid was truncated to a) exclude the area east of 8° E and south of 57° N,

i.e., Skagerrak/Kattegat and southern North Sea information were removed, and b) ex-

clude south of 57° N and the Kattegat (but include the Skagerrak). Saithe are not found

in the southern North Sea; excluding this area mainly truncates the zeros and keeps

the spatial spline of the GAM model from attempting to put fish where they are typi-

cally not found. Mainly young fish (the ages not included in the assessment model) are

found in the Skagerrak, but the German fleet fishes in this area; therefore, datasets in-

cluding and excluding this region were trialed. Ship was included in the final model.

Truncating the spatial area improved the model fit (Table 1). Removing the Skagerrak

improved the fit of the model the most, but the indices were larger for a given age class

and more variable for many of the age classes, especially at the beginning of the time

series (Figure 2). Average internal consistencies were higher for the model including

the Skagerrak, but the fit was not as good as the model excluding the Skagerrak (Tables

1, 2). Figure 3 shows the internal consistency plot, as given by FLR (note: correlations

are reported differently using FLR); there is no evidence in the internal consistencies


that something has gone wrong in the survey. The time series of indices by age (includ-

ing confidence intervals and comparison to the DATRAS indices) for the full survey

area, excluding the southern North Sea and Skagerrak/Kattegat, and excluding the

southern North Sea and Kattegat are in Figures 4-6.

The effect seen at the start of the time series cannot be due to ship; it would have been

captured within the model or also seen from 2001, when Sweden changed its research

vessel. The indices (all ages) with and without the Skagerrak show similar trends and

values.

Until 2003, Sweden did not take age samples, only lengths. This resulted in the age-

length key for the North Sea (subarea 4) being applied to the Skagerrak. Whether fish

in the Skagerrak were different from the North Sea was not thoroughly investigated,

so it is questionable whether the age-length key from the North Sea should be applied

to the Skagerrak. In addition, Sweden did not survey in 2000; this year had incomplete

coverage of the entire survey area. Finally, the Skagerrak was never included in the old

index estimation (in DATRAS). There is no documentation of why the Skagerrak was

included and the IBTSWG was unable to answer this question.

Survey properties

Internal consistencies

Internal consistencies for the Q3 survey are decent, although slightly poorer for age 3

vs. age 4 (Table 2, Figure 3). There is no evidence in the internal consistencies that

something has gone wrong in the survey.

Cross consistency with other data sources

Despite the Q1 survey having limited coverage of the stock, the external consistencies

between the Q3 and Q1 (in the following year and age), as well as catch numbers at

age, were used to see if tracking of cohorts was possible (Table 3). Cohorts can be

tracked between surveys (and ages). The external consistencies are not as strong when

comparing the catch numbers at age with the Q3 index, however, they still track co-

horts reasonably well. The external consistency for age 4, the age when fish are ex-

pected to be fully recruited to the fishery, is the lowest of all the age class comparisons.

Coverage

The amount of saithe found within the survey area differs between years, but the dis-

tribution has not changed over the time period. Stronger year classes are, for the most

part, appearing in the survey when expected and persisting for at least 1 year. The

increase in the last 2 years appears to be related to stronger recruitment.


Effect of age groups and research surveys on the assessment

Only the Q3 index was used to assessing the influence of the different age classes. The

decision was made that it is not appropriate to continue to include the Q1 indices in

the assessment model (see above).

Q3 indices without truncating spatial grid or including ship in the model

The assessment results when including only the Q3 + FBI indices show SSB in the final

years not as optimistic as the model including the Q1 index (Figure 7). It is, however,

much more optimistic than the FBI-only model or the model using the DATRAS-esti-

mated indices for ages 3-5. When looking at the effect of removing the oldest age clas-

ses one at a time, ages 5-8 have the largest effect on the assessment outcome (Figure 8).

Using only the age ranges 3-4 or 3-9 has a large effect on the estimated SSB; ages 3-9

result in a lower SSB over the entire time series, while using only ages 3-4 has a mixed

effect (lower SSB after 2010). The effect of changing the age range on Fbar and recruit-

ment are shown in Figure 9.

Q3 indices with truncation of spatial grid + including ship in model

Figure 10 shows the effect of the Q3 (without Skagerrak) index on assessment model

outputs. SSB and F are much closer to the DATRAS outputs and below that of the pre-

vious Q3 indices. Figures 11 and 12 detail the effects of changing the age range included

in the Q3 index on SSB, F4-7, and recruitment.

The effect of the Q3 with Skagerrak indices on the assessment model are in Figure 10.

Including the Skagerrak in the Q3 index resulted in output that was similar to the

model using the Q3 indices estimated from the entire North Sea dataset (Q3 + FBI

model). Figures 13 and 14 show the effect of changing the age range included in the

model on SSB, F4-7, and recruitment.

Discard estimation

The change in discard amounts are in Table 4. The years that had the greatest percent-

age difference due to the modifications noted above were the years that had very few

reported discards; Norwegian discards had to be estimated using poor data. Norway

takes 50% of the catch and this therefore resulted in high raised discards amounts. Be-

cause there is no information on the discarding practices of Norwegian fleets, the truth

is expected to lie somewhere between estimate (3) and estimate (2); these estimates

should be treated as upper and lower bounds on discards. It is doubtful that Norwe-

gian discards are at the low levels estimated in option 3. However, when low recruit-

ment is seen (2008–2010), discards should be low. This is seen in Table 3 using raising

option (3), but not in option (2). While raising option (3) may be under-estimating dis-

cards, it appears to be more likely than option (2).

The comparison of assessments (old raising procedure vs. option (3)) for the bench-

mark model (FBI + Q1 + Q3), FBI index-only model, and new Q3 model, where the

Skagerrak/Kattegat/southern North Sea were truncated from the spatial grid are in Fig-

ures 15-17. Results of all 3 models using revised discards data are in Figure 18.

Retrospectives using the newly estimated catch are in Figure 19 for the benchmark

model (FBI + Q1 + Q3) without discard revisions. Figure 20 is the benchmark model

including discard revisions, Figure 21 is the FBI-only model (including discards revi-

sions), and Figure 22 for the FBI+ new Q3 model (including discard revisions). The

retrospective pattern is much worse for the benchmark model with the revised catch


information. The retrospective pattern in F is particularly bad. The model with only

the exploitable biomass index shows the best performance in the retrospective analysis.

All models converge to approximately similar F and SSB values for the 2005-2010 pe-

riod (Figure 18). Therefore, by going back with the retro analysis before 2010 gives an

idea which assessment would have been more in line with the final converged values.

The assessment with FBI as only index would have assessed F around the converged

values for 2005-2010. The retrospective indicates all other models would have assessed

F well above the converged values for this period (with the FBI + new Q3 model being

the worst)In recent years the retro patterns became less, however each of the assess-

ments show F at a different level. It remains unclear whether the current FBI only as-

sessment will be again closer to the converged estimates in a few years. Reference

points and catch option tables are in the Appendix for the 3 models with revised catch

information.

Conclusions

The Q1 index should not be included as a tuning series because the survey does not

adequately cover the distribution of saithe. Saithe are spawning on the slope and their

movement into (or out of) the survey area does not appear to be linked to recruitment

or expected abundance.

For the Q3 index, the spatial distribution of saithe has not changed within the survey

area. Truncating the spatial grid to be remove the southern North Sea (where saithe

are not found) and the Skagerrak should be done. The arguments for excluding the

Skagerrak include: no age-length key in the Skagerrak until after 2003, incomplete cov-

erage of the survey area due to Skagerrak not surveyed in 2000, and exclusion of the

Skagerrak in the previous (DATRAS) index estimation (even though the reason is not

known).

Removing the Skagerrak and southern North Sea resulted in a less optimistic assess-

ment when compared to the benchmark model. The assessment using Q3 indices that

included the Skagerrak, but removing the Kattegat and southern North Sea, was (not

surprisingly) similar to the benchmark assessment. The data from the Skagerrak ap-

pears to be creating an issue with the index estimation. The reason for this is not clear

(biological or a survey effect, due to the lack of age information from this area). The

reason for the large discrepancy in the indices including/excluding the Skagerrak for

the beginning of the series should be investigated in the near future.

Because Norway lacks information on discards and takes 50% of the catch, the raising

of discards in InterCatch must be handled carefully. Raising discards for the Norwe-

gian trawlers based on reported discards from the French and German trawlers may

result in underestimating the discards, but it is the best information available at this

time. Germany, France and Norway have a targeted saithe fishery. Fisheries in coun-

tries like Scotland and Denmark are mixed demersal fisheries with higher discard rates

compared to the sampled fisheries targeting saithe.

The pre-benchmark assessment included the Q3 indices for ages 3-5. The internal con-

sistencies, coverage, and comparison with other data all show no reason to exclude the

survey from the assessment. It is only in the last two years that the assessment has

shown SSB is higher than the cpue-only model; prior to 2013, the cpue-only model had

consistently higher SSB (Figure 18). There is a lot of uncertainty in the assessment re-

gardless of the model chosen. The choice of survey data to include should be based on

the properties of that survey (e.g., internal consistency, cross-consistency with other

data, coverage).


The retrospective patterns, particularly for F, were very poor, especially for the assess-

ments with IBTS data included. This is worrying as it casts doubt on our ability to as-

sess the stock should conditions change again. Furthermore, the cause for the poor

ability to estimate F is unknown (and could occur again). There is some doubt that the

FBI + new Q3 model is the better model compared to the FBI-only model in light of the

retrospective patterns.

Keeping the stipulation from the EU-Norway management plan, where the TAC is not

allowed to deviate by more than 15% from the TAC in the previous year should protect

the stock from the uncertainty in the assessment. Furthermore, including catch options

based on probabilistic forecasts, e.g., 5% and 25% probability of being above FMSY and

Flim, is another option for dealing with the uncertainty in the assessment.


Table 1. Model diagnostics for the Q3 indices. The models are the benchmark model (no truncation

of spatial grid); benchmark model including Ship (no truncation of spatial grid); removing the

Skagerrak/Kattegat and southern North Sea and including Ship; removing the Kattegat and south-

ern North Sea and including Ship.

MODEL AIC BIC IC (ALL AGES)

Year+s(lon,lat)+s(Depth)+ HaulDur 34460 42834 0.3948

Year+Ship+s(lon,lat)+s(Depth)+HaulDur 34274 43476 0.4358

Truncated spatial range (57°N, 8°E):

Year+Ship+s(lon,lat)+s(Depth)+HaulDur, ages 1-

10 28122 36032 0.40527

Truncated spatial range (57°N, no Kattegat):

Year+Ship+s(lon,lat)+s(Depth)+HaulDur, ages 1-

10 32565 40590 0.4264


Table 2. Internal consistencies between ages classes for the four different Q3 indices.

MODEL/DATA IC

AVERAGE IC

ALL AGES

AVERAGE IC

AGES 3-8

Benchmark model:

Year+s(lon,lat)+s(Depth)+ HaulDur,

ages 0-10

Age 0 vs. 1 : 0.3231104

Age 1 vs. 2 : -0.1937066

Age 2 vs. 3 : 0.03960032

Age 3 vs. 4 : 0.4954253

Age 4 vs. 5 : 0.7447504

Age 5 vs. 6 : 0.7943942

Age 6 vs. 7 : 0.750217

Age 7 vs. 8 : 0.6407721

Age 8 vs. 9 : 0.4044236

Age 9 vs. 10 : -0.05130193

0.3948 0.6851

Benchmark model + Ship:

Year+Ship+s(lon,lat)+s(Depth)+HaulDur,

ages 0-10

Age 0 vs. 1 : 0.4646722

Age 1 vs. 2 : -0.1081123

Age 2 vs. 3 : 0.05023302

Age 3 vs. 4 : 0.4406976

Age 4 vs. 5 : 0.7406408

Age 5 vs. 6 : 0.8363853

Age 6 vs. 7 : 0.7676941

Age 7 vs. 8 : 0.5378916

Age 8 vs. 9 : 0.3850141

Age 9 vs. 10 : 0.2426996

0.4358 0.6647

Truncated spatial range (no

Skagerrak/Kattegat or southern North

Sea):


ages 1-10

Age 1 vs. 2 : 0.4287579

Age 2 vs. 3 : 0.1669562

Age 3 vs. 4 : 0.3777139

Age 4 vs. 5 : 0.759958

Age 5 vs. 6 : 0.7629555

Age 6 vs. 7 : 0.7211942

Age 7 vs. 8 : 0.6095779

Age 8 vs. 9 : 0.08241081

Age 9 vs. 10 : -0.262115

0.4053 0.6463

Truncated spatial range (no Kattegat or

southern North Sea):


ages 1-10

Age 1 vs. 2 : -0.3264555

Age 2 vs. 3 : -0.02738525

Age 3 vs. 4 : 0.4273828

Age 4 vs. 5 : 0.7532319

Age 5 vs. 6 : 0.8270072

Age 6 vs. 7 : 0.7994671

Age 7 vs. 8 : 0.6195003

Age 8 vs. 9 : 0.3514482

Age 9 vs. 10 : 0.4138057

0.4264 0.6853


Table 3. External consistencies between Q3 (ages 1-9, 1992-2014) and Q1 (year+1, age+1), and be-

tween catch numbers at age and Q3 (in the same year). This is identifying if cohorts can be tracked

from Q3 to the next survey in Q1. Numbers in bold refer to the ages included in the IBTS Q3 tuning

index in the assessment model.

EXTERNAL CONSISTENCIES: Q3 VS. Q1 EXTERNAL CONSISTENCIES: CATCH VS. Q3

Q3 Age 1 vs. Q1 Age 2 : 0.3218696

Q3 Age 2 vs. Q1 Age 3 : 0.4586471

Q3 Age 3 vs. Q1 Age 4 : 0.8203473

Q3 Age 4 vs. Q1 Age 5 : 0.8739198

Q3 Age 5 vs. Q1 Age 6 : 0.8839688

Q3 Age 6 vs. Q1 Age 7 : 0.7743481

Q3 Age 7 vs. Q1 Age 8 : 0.696888

Q3 Age 8 vs. Q1 Age 9 : 0.625716

Q3 Age 9 vs. Q1 10 : 0.4047001

Catch Age 1 vs. Q3 1 : -0.0165032

Catch Age 2 vs. Q3 2 : -0.1492027

Catch Age 3 vs. Q3 3 : 0.5044318

Catch Age 4 vs. Q3 4 : 0.3768049

Catch Age 5 vs. Q3 5 : 0.5894862

Catch Age 6 vs. Q3 6 : 0.5557922

Catch Age 7 vs. Q3 7 : 0.5059279

Catch Age 8 vs. Q3 8 : 0.4097457

Catch Age 9 vs. Q3 9 : 0.05872988

Catch Age 10 vs. Q3 10 : 0.4557988

Table 4. Amount of discards (estimated and reported) following 3 procedures: (1) as outlined in

WD-5 during the benchmark, (2) after fixing the bug in InterCatch (bolded years), and (3) after the

modification noted above. Differences are percentage.

YEAR

2015

ASSESSMENT

(1)

BENCHMARK

ESTIMATE

(2)

INTERCATCH

BUG

CORRECTION

(3)

MODIFICATION

TO NORWAY &

REDUCED

RATIO

ESTIMATE REPORTED

DIFFERENCE

2015 TO

(1)

DIFFERENCE

(1) TO (2)

DIFFERENCE

(2) TO (3)

2002 24812 21620 21544 21440 100 -13 0

2003 26377 12898 11438 11044 100 -51 -11

2004 9600 9656 8088 7850 100 1 -16

2005 8571 8571 8196 8072 100 0 -4

2006 15950 9498 8585 8340 100 -40 -10

2007 12050 12078 12413 11353 100 0 3

2008 9436 9436 8359 7891 100 0 -11

2009 14216 14216 4296 4170 100 0 -70

2010 10937 10937 4484 3009 100 0 -59

2011 12729 4951 4362 4285 100 -61 -12

2012 7585 9415 9415 9278 7471 24 0 -1

2013 8083 8173 8173 7777 7311 1 0 -5

2014 6289 6362 6356 6337 6068 1 0 0

2015 5060 5060 5003 4914 0 -1


Figure 1. Comparison of the 2015 assessments. Blue lines: XSA assessment results. Black lines: 95%

confidence interval of SAM assessment.

Figure 2. Comparison of IBTS Q3 indices, 1992-2015. Black lines: benchmark Q3 indices (no spatial

truncation, without ‘Ship’ in model); blue lines: truncated spatial grid (No Skagerrak) + ‘Ship’ in

model; red lines: truncated spatial grid (including Skagerrak) + ‘Ship’ in model.

Year

500

1000

1500

1995 2000 2005 2010 2015

age_6

0100

200

300

400

500

1995 2000 2005 2010 2015

age_7

50

100

150

200

1995 2000 2005 2010 2015

age_8

05000

10000

15000

20000

1995 2000 2005 2010 2015

age_3

2000

4000

6000

8000

10000

1995 2000 2005 2010 2015

age_4

01000

2000

3000

4000

5000

1995 2000 2005 2010 2015

age_5

benchmark Q3 indicesQ3 Spatial truncated No Skagerrak

Q3 Spatial truncated with Skagerrak


Figure 3. Internal consistencies as given by FLR. Note: FLR internal consistencies are estimated

differently from Berg et al. 2014, as given in the amendment to WD 8.

Dotted lines are 95% confidence interval for the mean.

Log10 (Younger Age)

Lo

g10 (

Old

er

Ag

e)

r2

0.143

Age 3 vs 4

r2

0.578

Age 4 vs 5

r2

0.582

Age 5 vs 6

r2

0.520

Age 6 vs 7

r2

0.372

Age 7 vs 8


Figure 4. IBTS Q3 indices, ages 0-10, 1992-2015. Comparing survey indices by age (and confidence

interval) to DATRAS indices for the full spatial range-no ship delta-GAM model (Q3 index as pre-

sented in the benchmark and WGNSSK).

1995 2000 2005 2010 2015

01

23

45

Age group 1

Year

Index

1995 2000 2005 2010 2015

01

23

45

Age group 2

Year

Index

1995 2000 2005 2010 2015

01

23

45

Age group 3

Year

Index

1995 2000 2005 2010 2015

01

23

45

Age group 4

Year

Index

1995 2000 2005 2010 2015

01

23

45

Age group 5

Year

Index

1995 2000 2005 2010 2015

01

23

45

Age group 6

Year

Index

1995 2000 2005 2010 2015

01

23

45

Age group 7

Year

Index

1995 2000 2005 2010 2015

01

23

45

Age group 8

Year

Index

1995 2000 2005 2010 2015

01

23

45

Age group 9

Year

Index

1995 2000 2005 2010 2015

01

23

45

Age group 10

Year

Index

1995 2000 2005 2010 2015

01

23

45

Age group 11

Year

IndexAge 8 Age 10

Age 3 Age 0


Figure 5. IBTS Q3 indices, ages 1-10+, 1992-2015. Comparing survey indices by age (and confidence

interval) to DATRAS indices (ages 1-6+) for the truncated spatial grid-with ship delta-GAM model;

this data excludes the Skagerrak-Kattegat and southern North Sea.

1995 2000 2005 2010 2015

01

23

45

Age group 1

Year

Index

1995 2000 2005 2010 2015

01

23

45

Age group 2

Year

Index

1995 2000 2005 2010 2015

01

23

45

Age group 3

Year

Index

1995 2000 2005 2010 2015

01

23

45

Age group 4

Year

Index

1995 2000 2005 2010 2015

01

23

45

Age group 5

Year

Index

1995 2000 2005 2010 2015

01

23

45

Age group 6

Year

Index

1995 2000 2005 2010 2015

01

23

45

Age group 7

Year

Index

1995 2000 2005 2010 2015

01

23

45

Age group 8

Year

Index

1995 2000 2005 2010 2015

01

23

45

Age group 9

Year

Index

1995 2000 2005 2010 2015

01

23

45

Age group 10

Year

Index


Figure 6. IBTS Q3 indices, ages 1-10+, 1992-2015. Comparing survey indices by age (and confidence

interval) to DATRAS indices (ages 1-6+) for the truncated spatial grid-with ship delta-GAM model;

this data includes the Skagerrak and excludes the southern North Sea and Kattegat.

Figure 7. Affect of different indices on SAM assessment: black line = benchmark model (Q3 + Q1 +

FBI indices); blue line = FBI index only (no surveys); green line = Q3 + FBI indices (no Q1); orange

line = DATRAS Q3 (ages 3-5) + FBI indices. The Q3 indices estimated from the delta-GAM are those

used in the benchmark meeting (no truncation of the survey area, without Ship in the model). Note:

this was conducted before the bug in InterCatch was found and discards had not been re-raised.

1995 2000 2005 2010 2015

01

23

45

Age group 1

Year

Index

1995 2000 2005 2010 2015

01

23

45

Age group 2

Year

Index

1995 2000 2005 2010 2015

01

23

45

Age group 3

Year

Index

1995 2000 2005 2010 2015

01

23

45

Age group 4

Year

Index

1995 2000 2005 2010 2015

01

23

45

Age group 5

Year

Index

1995 2000 2005 2010 2015

01

23

45

Age group 6

Year

Index

1995 2000 2005 2010 2015

01

23

45

Age group 7

Year

Index

1995 2000 2005 2010 2015

01

23

45

Age group 8

Year

Index

1995 2000 2005 2010 2015

01

23

45

Age group 9

Year

Index

1995 2000 2005 2010 2015

01

23

45

Age group 10

Year

Index

1970 1980 1990 2000 2010

1e+05

2e+05

3e+05

4e+05

5e+05

6e+05

year

SS

B

1970 1980 1990 2000 2010

0.2

0.4

0.6

0.8

1.0

year

F 4

-7

1970 1980 1990 2000 2010

1e+05

2e+05

3e+05

4e+05

5e+05

year

Recru

itment

benchmark model

FBI-only

Q3 only

DATRAS ages 3-5


Figure 8. Effect of adding an changing age range of the Q3 index on estimated SSB. The Q3 indices

were estimated using data from the entire North Sea. Note: this was conducted before the bug in

InterCatch was found and discards had not been re-raised.

Figure 9. Effect of adding an changing age range of the Q3 index on estimated (left) Fbar 4-7 and (right)

recruitment. Note: this was conducted before the bug in InterCatch was found and discards had

not been re-raised.

1990 1995 2000 2005 2010 2015

10

00

00

15

00

00

20

00

00

25

00

00

30

00

00

year

SS

B

ages 3-8

ages 3-4

ages 3-5

ages 3-6

ages 3-7

ages 3-9

1990 1995 2000 2005 2010 2015

0.0

0.2

0.4

0.6

0.8

1.0

year

F4

7

ages 3-8

ages 3-4

ages 3-5

ages 3-6

ages 3-7

ages 3-9

1990 1995 2000 2005 2010 2015

50

00

01

00

00

02

00

00

03

00

00

0

year

recru

its

ages 3-8

ages 3-4

ages 3-5

ages 3-6

ages 3-7

ages 3-9


Figure 10. Affect of different indices on SAM assessment: black line = benchmark model (Q3 + Q1

+ FBI indices); blue line = FBI index only (no surveys); green line = bm_Q3 + FBI indices (no Q1);

orange line = DATRAS Q3 (ages 3-5) + FBI indices; magenta line = new Q3 + FBI indices (no Q1);

brown line = Q3 including Skagerrak (without Kattegat or southern North Sea) + FBI. The bm_Q3

indices are those used in the benchmark meeting (no truncation of the survey area, without Ship

in the model), while the new Q3 indices include truncating the spatial grid + ship in the delta-GAM

model. Note: this was conducted before the bug in InterCatch was found and discards had not been

re-raised.

Figure 11. Effect of adding an changing age range of the Q3 index (truncated to remove the southern

North Sea and Skagerrak/Kattegat) on estimated SSB. Model bm_ includes the Q3 indices esti-

mated without truncation of the survey area or ship in the model. Note: this was conducted before

the bug in InterCatch was found and discards had not been re-raised.

1970 1980 1990 2000 2010

1e+05

2e+05

3e+05

4e+05

5e+05

6e+05

year

SS

B

1970 1980 1990 2000 2010

0.2

0.4

0.6

0.8

1.0

year

F 4

-7

1970 1980 1990 2000 2010

1e+05

2e+05

3e+05

4e+05

5e+05

year

Recru

itment

benchmark modelFBI-onlyQ3 w ithout SkagerrakDATRAS ages 3-5bm Q3Q3 w ith Skagerrak

1990 1995 2000 2005 2010 2015

10

00

00

15

00

00

20

00

00

25

00

00

30

00

00

year

SS

B

ages 3-8

ages 3-4

ages 3-5

ages 3-6

ages 3-7

ages 3-9

bm_ages 3-8


Figure 12. Effect of adding an changing age range of the Q3 indices (truncated to remove the south-

ern North Sea and Skagerrak/Kattegat) on estimated (left) Fbar 4-7 and (right) recruitment. Model

bm_ includes the Q3 indices estimated without truncation of the survey area or ship in the model.

Note: this was conducted before the bug in InterCatch was found and discards had not been re-

raised.

Figure 13. Effect of adding an changing age range of the Q3 indices (truncated to exclude the south-

ern North Sea and Kattegat) on estimated SSB. Model bm_ includes the Q3 indices estimated with-

out truncation of the survey area or ship in the model. Note: this was conducted before the bug in

InterCatch was found and discards had not been re-raised.

1990 1995 2000 2005 2010 20150

.00

.20

.40

.60

.81

.0

year

F4

7

ages 3-8

ages 3-4

ages 3-5

ages 3-6

ages 3-7

ages 3-9

bm_ages 3-8

1990 1995 2000 2005 2010 2015

50

00

01

00

00

02

00

00

03

00

00

0

year

recru

its

ages 3-8

ages 3-4

ages 3-5

ages 3-6

ages 3-7

ages 3-9

bm_ages 3-8

1990 1995 2000 2005 2010 2015

10

00

00

20

00

00

30

00

00

year

SS

B

ages 3-8

ages 3-4

ages 3-5

ages 3-6

ages 3-7

ages 3-9

bm_ages 3-8


Figure 14. Effect of adding an changing age range of the Q3 indices (truncated to exclude the south-

ern North Sea and Kattegat) on estimated (left) Fbar 4-7 and (right) recruitment. Model bm_ includes

the Q3 indices estimated without truncation of the survey area or ship in the model. Note: this was

conducted before the bug in InterCatch was found and discards had not been re-raised.

Figure 15. Effect of raising discards under assumption that Norway has low to zero discarding.

Comparison of benchmark model (Q1 + Q3 + FBI) before and after changing raising procedure.


Comparison of FBI index only model (no surveys) before and after changing raising procedure.

1990 1995 2000 2005 2010 2015

0.0

0.2

0.4

0.6

0.8

1.0

year

F4

7

1990 1995 2000 2005 2010 2015

50

00

01

00

00

02

00

00

03

00

00

0

year

recru

its

ages 3-8

ages 3-4

ages 3-5

ages 3-6

ages 3-7

ages 3-9

bm_ages 3-8

1970 1980 1990 2000 2010

1e+05

2e+05

3e+05

4e+05

5e+05

6e+05

year

SS

B

1970 1980 1990 2000 2010

0.2

0.4

0.6

0.8

1.0

year

F 4

-7

1970 1980 1990 2000 2010

1e+05

2e+05

3e+05

4e+05

5e+05

year

Recru

itment

benchmark model

benchmark model after re-raising discards

1970 1980 1990 2000 2010

1e+05

2e+05

3e+05

4e+05

5e+05

6e+05

year

SS

B

1970 1980 1990 2000 2010

0.2

0.4

0.6

0.8

1.0

year

F 4

-7

1970 1980 1990 2000 2010

1e+05

2e+05

3e+05

4e+05

5e+05

year

Recru

itment

FBI model

FBI model after re-raising discards



Comparison of new Q3 model (FBI + Q3 - spatial truncation excludes Skagerrak/Kattegat and south-

ern North Sea) before and after changing raising procedure.

Figure 18. Trends in SSB, F4-7, and recruitment for the 3 models. Blue line: Q1 + Q3 + cpue index

model; green line: cpue-only model; black line: Q3 + cpue model; orange/tan shaded region: 95%

confidence interval for the Q3 + cpue model; solid grey line (dashed): old Q3 + cpue model (95%

confidence interval). The old Q3 model was estimated without removing the southern North Sea

(where saithe are not found) and the Skagerrak (see amendment to WD 8 for details).

1970 1980 1990 2000 2010

1e+05

2e+05

3e+05

4e+05

5e+05

6e+05

year

SS

B

1970 1980 1990 2000 2010

0.2

0.4

0.6

0.8

1.0

year

F 4

-71970 1980 1990 2000 2010

1e+05

2e+05

3e+05

4e+05

5e+05

year

Recru

itment

New Q3 model

New Q3 model after re-raising discards


Figure 19. Five year retrospective pattern in SSB, F4-7, and recruitment. Model is FBI + Q1 + Q3

(untruncated spatial area) and does not include the discard revisions.

Figure 20. Eight year retrospective pattern in SSB, F4-7, and recruitment. Model is FBI + Q1 + Q3

(untruncated spatial area) and includes the discard revisions.


Figure 21. Eight year retrospective pattern in SSB, F4-7, and recruitment. Model is FBI index only (no

surveys) and includes the discard revisions.


Figure 22. Eight year retrospective pattern in SSB, F4-7, and recruitment. Model is FBI + new Q3

(excludes Skagerrak/Kattegat and southern North Sea) and includes the discard revisions.


APPENDIX

Reference Points and catch options

Reference Points estimated for the benchmark model, which includes the Q1, Q3

(untruncated spatially), and FBI indices are in Table A1; catch options are in Table A2

and basis for the catch options are in Table A3.

For the model that has only the FBI index (no surveys), reference points are in Table

A4, catch options are in Table A5, and basis for the catch options are in Table A6.

Table A7 contains reference points estimated from the assessment model that includes

the FBI and spatially truncated Q3 indices, where the Q3 indices do not include the

southern North Sea or Skagerrak/Kattegat. Catch options are in Table A8 and basis for

the catch options are in Table A9.

Table A1. Reference points estimated using the benchmark model (FBI + Q1+ Q3 no spatial trunca-

tion).

STOCK

Reference point Value

Blim 115 000

Bpa (1.4) 161 000

Bpa (sigma) 142 000

Btrigger 182 000

Flim 0.55

Fpa (1.4) 0.393

Fpa (sigma) 0.419

FMSY without Btrigger 0.359

FMSY lower without Btrigger 0.204

FMSY upper without Btrigger 0.492

New FP.05 (5% risk to Blim without Btrigger) 0.422

FMSY upper precautionary without Btrigger 0.393

FP.05 (5% risk to Blim with Btrigger) 0.534

FMSY with Btrigger 0. 396

FMSY lower with Btrigger 0.209

FMSY upper with Btrigger 0.694

FMSY upper precautionary with Btrigger 0.393

MSY (without HCR) 91 480

Median SSB at FMSY (without HCR) 220 827

Median SSB lower precautionary (median at

FMSY upper precautionary; without HCR) 195 709

Median SSB upper (median at FMSY lower;

without HCR) 420 907

Sigma (F) = 0.1653818, sigma (SSB) = 0.1300388.


Table A2. Saithe in Subareas 4 and 6 and Division 3a. The catch options. All weights in tonnes.

RATIONALE

TOTAL

CATCHES

2017 *

WANTED

CATCH

2017 *

WANTED

CATCH

3A & 4

2017

**

WANTED

CATCH

6 2017

** BASIS

F

(TOTAL

CATCH)

2017

F

(WANTED

CATCH)

2017

SSB

2018

% SSB

CHANGE

***

% TAC

CHANGE

WANTED

CATCH^

MSY

approach 133332 127631 115634 11997 FMSY 0.36 0.34 322434 -3 86

EU-Norway

management

strategy 82261 78824 71415 7409

Paragraph 5 of

management

strategy 0.21 0.20 374902 13 15

Precautionary

approach 298993 284409 257675 26734

SSB = min{1;

SSB2017/Btrigger} 1.08 1.03 161000 -51 315

Zero catch 0 0 0 0 F = 0 0 0 461461 39 -100

Other options

79730 76405 69223 7182 F2016 0.20 0.19 377503 14 11

71619 68601 62153 6448 TAC2016 0.18 0.17 386190 17 0

143773 137641 124703 12938 Fpa 0.39 0.38 311808 -6 101

Table A3. Saithe in Subareas 4 and 6 and Division 3a. The basis for the catch options.

VARIABLE VALUE SOURCE NOTES

F ages 4–7 (2016) 68601 t ICES

(2016a) TAC constraint (F=0.20)

SSB (2016) 284887 t ICES

(2016a) SSB in the intermediate year

SSB (2017) 331048 t ICES

(2016a) SSB at the beginning of the TAC year

Rage3 (2016) 101 billion ICES

(2016a)

Median recruitment resampled from 2003-

2015


(2016a)


2015

Total catch (2016) 71775 t ICES

(2016a) Assuming 2015 landings fraction by age

Commercial

landings (2016) 68601 t

ICES

(2016a) TAC 2015

Discards (2016) 3174 t ICES

(2016a) Assuming 2015 discard fraction by age


Table A4. Reference points estimated using the FBI index only model (no surveys).

STOCK

Reference point Value

Blim 115 000

Bpa (1.4) 161 000

Bpa (sigma) 151 000

Btrigger 161 000

Flim 0.506

Fpa (1.4) 0.361

Fpa (sigma) 0.364

FMSY without Btrigger 0.361 (was 0.405)
















Sigma (F) = 0.1651571, sigma (SSB) = 0.1997418.



RATIONALE

TOTAL

CATCHES

2017 *

WANTED

CATCH

2017 *

WANTED

CATCH

3A & 4

2017

**

WANTED

CATCH

6 2017

** BASIS

F

(TOTAL

CATCH)

2017

F

(WANTED

CATCH)

2017

SSB

2018

% SSB

CHANGE

***

% TAC

CHANGE

WANTED

CATCH^

MSY

approach 91749 85822 77755 8067 FMSY 0.36 0.34 221501 5 25

EU-Norway

management

strategy 84592 79134 71695 7439

Paragraph 5 of

management

strategy 0.33 0.31 228215 8 15

Precautionary

approach 157658 147368 133515 13853

SSB = min{1;

SSB2017/Btrigger} 0.71 0.68 161000 -24 115

Zero catch 0 0 0 0 F = 0 0 0 309384 47 -100

Other options

80442 75253 68179 7074 F2016 0.31 0.30 232171 10 10

73037 68601 62153 6448 TAC2016 0.28 0.26 240108 14 0

91749 85822 77755 8067 Fpa 0.36 0.34 221501 5 25



F ages 4–7 (2016) 68601 t ICES


SSB (2016) 199173 t ICES


SSB (2017) 211158 t ICES



(2016a)


2015


(2016a)


2015



Commercial

landings (2016) 68601 t

ICES

(2016a) TAC 2015




Table A7. Reference points estimated using the FBI + Q3 (spatially truncated, excludes southern

North Sea and Skagerrak/Kattegat) model.

REFERENCE POINT VALUE

Blim 121 000

Bpa (1.4) 169 000

Bpa (sigma) 155 000

Btrigger 170 000

Flim 0.514

Fpa (1.4) 0.367

Fpa (sigma) 0.376

FMSY without Btrigger 0.363
















Sigma (F) = 0.189643, sigma (SSB) = 0.1512602.



RATIONALE

TOTAL

CATCHES

2017 *

WANTED

CATCH

2017 *

WANTED

CATCH

3A & 4

2017

**

WANTED

CATCH

6 2017

** BASIS

F

(TOTAL

CATCH)

2017

F

(WANTED

CATCH)

2017

SSB

2018

% SSB

CHANGE

***

% TAC

CHANGE

WANTED

CATCH^

MSY

approach 114375 109057 98806 10251 FMSY 0.36 0.35 275385 0 59

EU-Norway

management

strategy 82766 78909 71492 7417

Paragraph 5 of

management

strategy 0.25 0.24 307208 12 15

Precautionary

approach 224186 212434 192465 19969

SSB = min{1;

SSB2017/Btrigger} 0.87 0.83 169000 -38 210

Zero catch 0 0 0 0 F = 0 0 0 391963 43 -100

Other options

79803 76084 68932 7152 F2016 0.24 0.23 310253 13 11

72198 68601 62153 6448 TAC2016 0.21 0.20 315995 15 0

115286 109922 99589 10333 Fpa 0.37 0.35 273885 0 60



F ages 4–7 (2016) 68601 t ICES


SSB (2016) 242142 t ICES


SSB (2017) 274310 t ICES



(2016a)


2015


(2016a)


2015



Commercial

landings (2016) 68601 t

ICES

(2016a) TAC 2015

