Ref. Ares(2018)3458869 - 29/06/2018
Ref. Ares(2019)2387732 - 04/04/2019
NOTAT
Til
UM, Fiskeripolitisk kontor
Vedr. Discard survival of plaice (
Pleuronectes platessa) caught in the bottom
otter trawl (OTB) demersal mixed fishery in Skagerrak during summe r
2017 and winter 2018
Fra
DTU Aqua
23/03 2018
J.nr: 18/05964
Ref: JKA/ESAV/THNO/JD
Request
DTU Aqua has received a request from UM to present the results on the discard survival of plaice col-
lected in the bottom otter trawl fishery during summer 2017 and winter 2018 under the EMFF-project
COPE (grant no. 33113-B-16-086).
Summary
The North Sea-Skagerrak stock of plaice is considered to have full reproductive capacity and to be
sustainably harvested.
Discard survival was investigated for the bottom otter trawl (OTB) fisheries during Aug-Oct 2017 and
Mar-Apr 2018 in Skagerrak. The assessments were done onboard a commercial vessel, and accord-
ing to guidelines made by ICES WKMEDS. In the summer, two commercial 90 mm diamond codends
representative for the mixed demersal fishery were used in a twin rig to target plaice. In the winter, in
addition to the standard commercial 90 mm diamond codend, a modified horizontally divided codend
with 120 mm square mesh upper compartment and 60 mm square mesh lower compartment to sepa-
rate fish from
Nephrops and limit surface damages was tested. All other operational factors of the ex-
periment were representative of commercial practices in Danish waters.
Regarding the commercial standard codend (90mm diamond), the mean survival rate for undersized
plaice was higher in the winter 75% (95%-confidence interval including variability from the captivity ex-
periment, haul and fish uncertainty: 67-83%) than in the summer, 44 % (37-52%). The mean survival
rate for undersized plaice commercially caught when targeting
Nephrops during winter was 41% (28-
57%), i.e. lower than when targeting plaice in the same season, and similar to when targeting plaice
during summer. The larger amount of
Nephrops in the catch caused more damages to the fish due to
contact between the hard, spiny surface of the
Nephrops and the soft skin of the plaice, leading to
higher mortalities. In Skagerrak, the highest discard survival was found when the amount of discarded
individuals in the fishery is the highest.
In the summer when targeting plaice, discard survival was affected by air exposure duration, dropping
to 8% (CI: 2-31%) if released after 60 min of air exposure. This was, however, not the case in the win-
ter. The air exposure times used in the experiment were within commercial practice, but it is not known
if air exposure time are higher at the fleet level. The length range of the sampled fish was limited in the
Danmarks Tekniske Universitet
Charlottenlund Slot
Tlf. 35 88 33 00
xxxx@xxxx.xxx.xx
Institut for
Jægersborg Allé 1
Fax 35 88 33 33
www.aqua.dtu.dk
Akvatiske Ressourcer
2920 Charlottenlund
summer and larger in the winter, explaining why this biological factor had an effect on discard survival
in the winter only.
The upper compartment of the modified codend (120mm square) showed a better discard survival, but
also less undersized (and commercial) individuals due to a higher selectivity. The lower compartment
of the modified codend (60mm square) did not seem to improve discard survival compared to the
standard commercial codend. An ongoing project is aiming at improving its selectivity for flatfish.
The survival of the four control groups were high, but there might be some influence of transportation
on survival.
Species and stock status
Plaice (
Pleuronectes platessa) has no swim bladder and is considered robust with respect to surviving
the fishing process, partly due to its sedentary life style that likely has evolved towards enhanced met-
abolic adaptation to hypoxia (Benoît et al., 2013; Morfin et al., 2017a). It is therefore a candidate spe-
cies for obtaining an exemption from the landing obligation. Plaice in the Skagerrak has been as-
sessed together with the North Sea stock since 2015 (ICES Advice 2017). The stock is considered to
have full reproductive capacity and to be sustainably harvested (Table 1, Fig. A1). At the stock level,
the proportion of unwanted catch is on average 57% (years 2011-2016, ICES Advice 2017).
Table 1. Plaice in North Sea and Skagerrak. State of the stock and fishery relative to reference points
(ICES Advice 2017).
Fig. 1. Plaice in North Sea and Skagerrak. Summary of the stock assessment. Shaded areas (F, SSB)
indicate ±2 standard errors (approximately 95% confidence intervals) (ICES Advice 2017).
2
Methods
Study design, vessel, and fishing gear
The survival rate and vitality of plaice under the MCRS of 27 cm in the trawl fishery in Skagerrak was
investigated during summer (August, September and October) 2017 and winter (March and April)
2018. The study was done onboard the commercial vessel S84 ‘Ida Katrine’ chosen in collaboration
with the Danish Fishermen Organisation DFPO (Fig. 2). The trawler represents the mixed demersal
fishery targeting fish (including plaice
Pleuronectes platessa) and
Nephrops (
Nephrops norvegicus),
with a length of 15.1m and a power of 221kW, working in a twin rig.
Fig. 2. The commercial vessel used in the study to represent the bottom otter trawl fleet in the demer-
sal mixed fisheries.
In the summer, two commercial 90 mm diamond codends representative for the mixed demersal fish-
ery were used to target plaice. A 90mm mesh size was chosen to account for the ‘worst case sce-
nario’, but fishermen commonly use a 120mm diamond codend instead when targeting plaice. To-
gether with improving size selectivity, a larger mesh size in the codend is expected to reduce potential
damages in fish and therefore improve discard survival.
In the winter, in addition to the standard commercial 90 mm diamond codend, a modified experimental
codend was tested (Fig. 3). This horizontally divided codend with 120 mm square mesh upper com-
partment and 60 mm square mesh lower compartment had previously been used to separate fish from
Nephrops (Karlsen
et al., 2015), and therefore seemed promising to reduce catch damage by limiting
frictions in the codend. In the winter, half of the hauls targeted plaice and half of the hauls targeted
Nephrops.
3
Fig. 3. A conceptual drawing of the horizontally divided codend used as one of the codends in the
twin-rig of the trawler in the winter.
Data collection
Data was collected for a total of 10 days divided on five sub-cruises conducted from 17 August to 10
October 2017 and from 08 to 27 March 2018 on commercial fishing grounds north of Hirtshals (Fig. 4).
Due to space limitations during transport and holding on land, repetition of the experiment allowed col-
lecting a higher number of individuals to increase the robustness of the survival estimates. The two
codends were fished on each side of the twin-trawl rig.
Fig. 4. Data collection at commercial fishing grounds in Skagerrak. The five sub-cruises were con-
ducted in the summer (August, September and October 2017, on the left) and the winter (March and
April 2018, on the right).
When using the modified codend to target
Nephrops in the winter (second sub-cruise in 2018), clog-
ging of the lower compartment at the point of the second frame was observed for some of the hauls.
This had not been experienced when using the same gear during previous trials (Karlsen
et al., 2015).
It is expected that clogged individuals would suffer higher levels of damages and therefore show lower
survival than expected in the absence of clogging. Additionally, few individuals experienced rough
handling accidentally caused by a sudden degradation in weather conditions at the end of the second
sub-cruise in 2018. Potential negative effects was tested for as part of the data analysis.
4
In the summer, the catch of both codends was hauled on deck, emptied together into the pounder, and
sorted by the crew according to normal commercial practices. Six fish were sampled at five time inter-
vals during the sorting process to cover the entire air exposure time of the catch sorted normally by
the crew.
In the winter, each of the first two codends hauled onboard were, one after the other, emptied into the
pounder and collected from the sorting belt into separate tanks on deck while the catch of the last
codend stayed in the pounder/on the sorting belt while being sorted. All three catches were sampled
and assessed in parallel. The catches were held separately at all times until the individuals were
tagged and identifiable. The handling order of the three catches was alternated at each haul. The ef-
fect of being hauled first on discard survival was tested in the analysis. Four fish were collected at four
time intervals during the sorting process. The total sorting time was decided together with the crew ac-
cording to usual practices, i.e., about 1 h when targeting plaice and up to 2.5 h when targeting
Nephrops.
Fish were assessed for vitality, length measured and tagged for individual recognition. Fish were
stored in custom-made survival units to minimize the effects of handling, holding and transportation on
mortality. The survival units were continuously supplied with running seawater, and oxygen and tem-
perature were monitored. Fish were transported to the close-by holding facilities at DTU-Aqua and
transferred into 1x1m tanks in a common garden set-up to prevent a tank-effect on mortality. The
tanks had a semi-circulated water supply and the bottom was covered with a 2 cm sand layer. Mortal-
ity was assessed and water parameters monitored for 14 days. After the first week, the fish were fed
each day.
Controls
Four control groups were used to control for the effect of handling, assessing, transporting and holding
the fish, i.e. all the experimental steps which took place after the fish would normally be discarded in
commercial fisheries. Plaice in control groups 1 and 2 were caught prior to the study using the trawler
R/V Havfisken. These fish were allowed to acclimatise before entering the study. Control group 1
(land) was used to control for the land-based holding facilities. Plaice in control group 2 (HV) were
brought onboard the commercial vessel, and thus underwent the transportation to and from the fishing
ground, and vitality assessment, length measurement and tagging. This group controlled for transport
and assessment when held up against control group 1. Plaice in control group 3 (S84) were caught
with the commercial trawler (short hauls) and entered the experiment without acclimatisation. This
group controlled for the same as control group 2 in addition to the fishing process and commercial
handling. A fourth control group was added during the winter sub-cruises to disentangle the effects of
transportation and fish assessment. Plaice in control group 4 (land+tag) were caught by Havfisken and
acclimatized beforehand, and experienced the assessment and tagging procedure, but no transporta-
tion process.
Analysis
A Weibull mixture model was used to estimate survival probabilities including uncertainty from the fish
selection when appropriate, i.e., when the covariate of interest was dependent on individual fish, the
haul selection and the conditions of the captivity experiment, and investigate the effect of air exposure,
bottom temperatures, and fish length on survival (Benoît et al., 2012; Benoît et al., 2015; Morfin et al.,
5
2017a, see Annex for more details). Information on other operational and environmental factors, i.e.,
haul duration, fishing depth, cloud cover, sea state, wind force and wind direction, were collected but
not included in the modelling approach as data exploration showed no relationship or high correlation
with the already chosen explanatory variables.
Results
Data collected
The operational conditions during the experimental trials are given in Table 2.
Table 2. Characteristics of the control and experimental hauls, separated by season, target species
and haul type (control, experimental). Values shown as mean (min-max).
Condition Hauls
Haul
duration
Catch weight
Bottom temp. Number of Fish length
(min)
(kg)
(°C)
individuals (sampled)
sampled
(cm)
Summer
Plaice
Control
6
15 (13-16)
47 (30-60)
14 (10-17)
60 22.4 (14-27)
Experimental
12
141 (37-185) 387 (65-1509)
14 (10-17)
333 23.4 (17-26)
Winter
Nephrops
Control
2
18 (16-19)
38 (30-45)
6 (6-7)
10 22.3 (20-24)
Experimental
4 210 (180-239) 375 (200-500)
7 (7-7)
274 22.2 (11-26)
Winter
Plaice
Control
2
18 (16-19)
4 (2-5)
6 (6-7)
10 22.3 (20-26)
Experimental
6 181 (177-185) 150 (100-200)
7 (6-7)
279 22.1 (13-26)
6
Survival of the control groups
The survival of the four control groups were high, but there might be some influence of transportation
on survival (Table 3).
Table 3. Survival of the control groups, separated by season and target species.
Season Target
Control
group
Number of individuals Observed survival
Control 1 (land)
50
1.00
Summer Plaice
Control 2 (HV)
60
0.92
Control 3 (S84)
60
0.87
Control 1 (land)
16
1.00
Control 2 (HV)
10
1.00
Nephrops
Control 3 (S84)
10
1.00
Control 4 (land+tag) 16
0.94
Winter
Control 1 (land)
10
1.00
Control 2 (HV)
10
1.00
Plaice
Control 3 (S84)
16
1.00
Control 4 (land+tag) 16
1.00
Overall survival rates by season and target species for the commercial standard codend
Regarding the commercial standard codend (90mm diamond), the mean survival rate for undersized
plaice was higher in the winter than in the summer, respectively 44% (95%-confidence interval: 37-52)
and 75% (67-83%) (Table 4). A lower survival at higher temperatures was observed in previous stud-
ies. The mean survival rate for undersized plaice commercially caught when targeting
Nephrops was
lower than when targeting plaice, as observed in the winter, reaching survival rates similar to those
when targeting plaice in the summer, i.e., 41 (28-57) % (Table 4). The larger amount of
Nephrops in
the catch caused more damages to the fish by friction in the codend, leading to higher mortalities.
Caution must be made when doing direct comparisons. Mean discard survival (with uncertainty esti-
mates) are limited by the conditions during the trials, especially by the factors found to affect the sur-
vival rates.
7
Table 4. Estimated overall survival rates in % with 95%-confidence interval (* including uncertainty
from the haul selection and the conditions of the captivity experiment when the chosen covariates did
not depend on the fish selection, ** including uncertainty from the fish selection, the haul selection and
the conditions of the captivity experiment) of undersized plaice in the Skagerrak for the OTB targeting
plaice and
Nephrops in the summer and winter for the standard commercial codend.
Target:
Plaice
Target:
Nephrops
Summer 44
(37-52*,
n=333) -
Winter 75
(67-83**,
n=142) 41
(28-57*,
n=123)
Effects of operational factors on discard survival for the commercial standard codend
In the summer when targeting plaice, discard survival was affected by air exposure duration (Table 5).
This was not observed in winter, also when targeting
Nephrops, as discard survival was primarily
driven by damages/loss of reflexes in an overall cold/mild environment.
The length range of the sampled fish was limited in the summer but larger in the winter, explaining
why this biological factor had an effect in the winter only (Table 5).
8
Table 5. Effects of operational, environmental and biological covariates on the parameters of the fitted
survival function and mixture proportion for discard survival of undersized plaice caught by a Danish
otter trawler targeting plaice and
Nephrops with a standard commercial codend in the summer and
winter. Only the mixture proportion affects the overall survival estimate (as observed at the end of the
experiment when an asymptote is reached).
Target
Season
Survival function (
α, γ)
Mixture proportion (
π)
Summer -
Operational: Air exposure
Plaice
Operational: Sorting order
Winter
Biological: Fish length
Biological: Fish length
Operational: Sorting order, fail due to bad
Nephrops
Winter
-
weather condition
Note the caution mentioned in the above section when comparing overall mean survival rates. Be-
cause overall survival rates estimated above are, for some, dependent on the number of observed fish
for each level of the selected covariates, we also predicted survival rates for given values of the se-
lected operational covariates independently, within the ranges of the experimental data, i.e., air expo-
sure from 0 to 62 min for OTB targeting plaice in the summer (Fig. 5).
Fig. 5. Discard survival as a function of air exposure (black) with 95% confidence intervals estimated
by parametric bootstrap accounting for variability from the captivity experiment (grey) for undersized
plaice caught by the OTB targeting plaice in the summer.
9
Effect of the modified codend on discard survival
The upper compartment of the modified codend (120mm square) showed a better discard survival, but
also less undersized (and commercial) individuals due to a higher selectivity. The lower compartment
of the modified codend (60mm square) did not seem to improve discard survival compared to the
standard commercial codend. An ongoing project is aiming at improving its selectivity for flatfish.
Discard survival in the context of the Danish demersal mixed otter trawl fishery
Fleet description: number, size and power of the vessels in the Skagerrak and North Sea
The OTB fleet in the MCD fishery in Skagerrak counts 102 vessels in the size range 11.00-19.99 m
and power range 67-365 kW (2017, logbook database). The same fleet segment in the North Sea
counts only 11 vessels (size and power ranges of 11.00-16.99 m and 126-365 kW, respectively; 2017,
logbook database) (Fig. 8).
Fig. 8. Number of Danish vessels in the OTB fleet by length category in m by area and mesh size
(2017, logbook database). The dashed black line represents the length of the vessel used in the ex-
periment (S84). In brackets in the legend is the average vessel length for each area and mesh size.
10
Fig. 9. Number of Danish vessels in the OTB fleet by power category in kW by area and mesh size
(2017, logbook database). The dashed black line represents the power of the vessel used in the ex-
periment (S84). In brackets in the legend is the average vessel power for each area and mesh size.
Catch data: catch pattern and seasonality
Plaice and
Nephrops are caught year round both in the Skagerrak and the North Sea (Fig. 10 and 11).
However, fish and
Nephrops are usually caught on separate fishing operations (Fig. 10), which should
be highlighted as the presence of
Nephrops in the catch can increase damages and therefore fish
mortality (Karlsen et al. 2015). I.e. when
Nephrops dominates the catch the proportion of plaice is low
and vice versa.
11
Fig. 10. Proportion of plaice and
Nephrops in the total catch when targeting plaice (high proportion of
plaice) and
Nephrops (high proportion of
Nephrops) by month for the Danish OTB fleet separated by
area (2015-2017, logbook database).
In the Skagerrak, the largest landings of plaice take place in the autumn and winter. In the North Sea,
the largest landings take place in the summer, but are all year round at least as big as in the Skager-
rak (Fig. 11). Although discard ratios of plaice are usually higher for smaller mesh sizes, i.e. often tar-
geting
Nephrops in all seasons except for autumn in the Skagerrak (Fig. 11), absolute numbers of dis-
carded plaice are usually higher when the proportion of plaice in the total catch is larger, i.e., using
larger mesh sizes. The proportion of unwanted catch of plaice is on average 60.4% in volume with 90-
119 mm mesh size and 7.4% with >120 mm mesh in the Skagerrak, and 6.4% in volume with 90-119
mm mesh size and 3.4% with >120 mm mesh in the North Sea (data from the Data Collection Frame-
work database from 2015-2017).
12
Fig. 11. Total landed catch in tons (light grey), plaice landed catch in tons (dark grey) and discard ratio
(boxplot) by month for the Danish OTB fleet by area and mesh size (2015-2017, logbook database,
Data Collection Framework database).
Biological and operational factors influencing discard survival
All biological and operational factors of the experiment were representative of commercial practices in
Danish waters. All our sampled plaice were representative of the biological conditions at the time of
the experiment, i.e., in line with the length distribution of the fish discarded in the fishery between 2015
and 2017 (Table 2, Table 6).
Air exposure is in close relation to sorting time. The sorting times during the experimental trials were
within commercial practices, as discussed with the crew and the DFPO. There is no data available on
the sorting times at the fleet level from which we could assess the proportion of hauls with sorting
times within the range of sorting times included in our study. The sorting time depends on catch weight
(and thus also vessel size) and composition, and the size of the crew onboard the vessel. Experience
from DTU-Aqua observers at sea programme suggests that in commercial conditions, sorting time is
up to 1 h depending on catch weight when plaice is the main target species, and up to 2.5 h when
Nephrops is the main target species. A proxy for sorting time is catch weight. For hauls, conducted be-
tween 2015 and 2017 in the Skagerrak, the average catch weight per haul for trawlers using mesh
sizes ≥120 mm (i.e. mainly targeting plaice or roundfish) was 674 (53-2957) kg (Table 6). For trawlers
using mesh sizes <120 mm (mainly targeting
Nephrops), it was 559 (121-2236) kg (Table), i.e.
catches of our experiment (Table 2) are within the range of these values.
13
Table 6. Characteristics of commercial hauls conducted between 2015 and 2017 (Data Collection
Framework database). Values shown as mean (min-max).
Area
Mesh size Haul duration (min) Catch weight (kg) Length of plaice discarded (cm)
<120 mm
248 (142-300)
559 (121-2236)
23 (11-37)
Skagerrak
≥120 mm
215 (75-300)
674 (53-2957)
25 (13-39)
<120 mm
296 (290-299)
985 (226-1932)
26 (18-40)
North Sea
≥120 mm
258 (34-300)
1643 (175-4949)
26 (17-39)
Discard survival rates with respect to the amounts discarded in the fishery
Discard survival of undersized plaice caught by a standard commercial codend (90mm diamond)
by a Danish otter trawler was higher in winter and when targeting plaice. In the Skagerrak, this is
also when the amount of discarded individuals is the highest compared to when targeting
Nephrops or in the summer. In the North Sea, the discard ratio is low when targeting plaice.
14
References
Benoît HP, Hurlbut T, Chassé J (2010) Assessing the factors influencing discard mortality of demersal
fishes using a semi-quantitative indicator of survival potential. Fisheries Research, 106, 436-447.
Benoît HP, Plante S, Kroiz M, Hurlbut T (2013) A comparative analysis of marine fish species suscep-
tibilities to discard mortality: effects of environmental factors, individual traits, and phylogeny. ICES
Journal of Marine Science 70(1):99-113. doi: 10.1093/icesjms/fss132
ICES (2014) Report of the Workshop on Methods for Estimating Discard Survival (WKMEDS), 17–21
February 2014, ICES HQ, Copenhagen, Denmark. ICES CM 2014/ACOM:51. 114 pp.
ICES (2017) ICES Advice on fishing opportunities, catch, and effort. Greater North Sea Ecoregion.
Plaice (
Pleuronected platessa) in subarea 4 (North Sea) and subdivition 20 (Skagerrak). doi:
10.17895/ices.pub.3529
Karlsen JD, Krag LA, Albertsen CM, Frandsen RP (2015) From Fishing to Fish Processing: Separation
of Fish from Crustaceans in the Norway Lobster-Directed Multispecies Trawl Fishery Improves
Seafood Quality. PLoS ONE 10 (11): e0140864. doi:10.1371/journal.pone.0140864
Morfin M, Méhault S, Benoît HP, Kopp D (2017a) Narrowing down the number of species requiring de-
tailed study as candidates for the EU Common Fisheries Policy discard ban. Marine Policy 77:23-
29. doi: 10.1016/j.marpol.2016.12.003
Morfin M, Kopp D, Benoît HP, Méhault S, Randall P, Foster R, Catchpole T (2017b) Survival of Euro-
pean plaice discarded from coastal otter trawl fisheries in the English Channel. Journal of Environ-
mental Management 204: 404-412. doi: 10.1016/j.jenvman.2017.08.046
15
Annex
Parametric Weibull mixture distribution model
A parametric Weibull mixture distribution model was used, allowing some proportion of individuals to
survive (Benoît et al., 2012; Benoît et al., 2015; Morfin et al., 2017a). The probability that a fish was
mortally affected by capture, handling and discarding is π. For those affected fish, according to the
shape of the non-parametric Kaplan-Meier curves (Kaplan and Meier, 1958), a reasonable model for
the survival function is a two-parameter Weibull distribution, with parameters α (the scale, with α>0)
and γ (the shape, with γ >0). Natural mortality is considered negligible at the time scale of the observa-
tion period, and therefore the survival rate is expected to eventually converge to an asymptote 1 – π
(for further detail, see Benoît et al., 2012; Benoît et al., 2015; Morfin et al., 2017a).
Explanatory variables (e.g., air exposure, fish length, bottom temperatures) were tested as covariates
on the three parameters describing the survival model, i.e., α, γ and π (for further detail, see Benoît et
al., 2012; Benoît et al., 2015; Morfin et al., 2017a).
Model parameters were estimated by a maximization of the model likelihood using a quasi-Newton op-
timization algorithm (Byrd et al., 1995).
Model selection and validation
An information-theoretic approach was used to identify which of the covariates were important deter-
minants of survival probability using Akaike Information Criterion (AIC) (Akaike, 1981; Burnham and
Anderson, 2002). Models with a relative difference in AIC less than two with the model with the lowest
AIC could be interpreted as having similar support in the data, while larger values suggested less sup-
port for the competing model (Burnham and Anderson, 2002). Among all models with a relative differ-
ence in AIC less than two, the simpler model was then selected as the best model. Model fit was as-
sessed visually by superimposing the predicted survival curves on the non-parametric Kaplan-Meier
curves (Morfin et al., 2017a; Morfin et al., 2017b).
Model estimation and confidence intervals providing with an overall survival rate for the observed fish-
eries. Confidence intervals of the survival rates were estimated by a parametric bootstrap based on
Monte Carlo simulation with 5000 iterations (Benoît et al., 2012; Benoît et al., 2015; Morfin et al.,
2017a). At each iteration, based on asymptotically normal behavior of the maximum likelihood estima-
tors, the regression parameters were simulated according to a multivariate Gaussian distribution (for
further detail, see Benoît et al., 2012; Benoît et al., 2015; Morfin et al., 2017a). Uncertainty due to the
selection of hauls was estimated by randomly re-sampling m hauls with replacement from the m ob-
served hauls. The observed fish were re-sampled to capture the variability due to the selection of fish
in each haul, only when the covariates in the chosen best model depended on the sampled fish, i.e.,
fish length. These steps were repeated 5000 times. The overall survival rate was given as the median,
and its 95%-confidence interval as the range between the 5th and the 95th centile.
Model prediction and confidence intervals for assessing operational covariate effects on survival rate
Overall survival rates for each gear estimated above are dependent on the number of observed fish
for each level of the selected covariates. Thus, we also predicted survival rates for given values of the
16
selected operational covariates independently, within the ranges of the experimental data, i.e., air ex-
posure from 0 to 62 min for OTB targeting plaice in the summer. Survival was estimated at the asymp-
tote, i.e., calculated as 1- π. As previously, confidence intervals of the survival rates were estimated by
a parametric bootstrap based on Monte Carlo simulation with 5000 iterations, but we accounted only
for the variation of the regression parameter π, similarly simulated according to a multivariate Gauss-
ian distribution. Survival rate was also given as the median, and its 95%-confidence interval as the
range between the 5th and the 95th centile of the 5000 iterations.
References
Akaike, H. 1981. Likelihood of a model and information criteria. Journal of Econometrics, 16: 3-14.
Benoît, H. P., Capizzano, C. W., Knotek, R. J., Rudders, D. B., Sulikowski, J. A., Dean, M. J., Hoff-
man, W., et al. 2015. A generalized model for longitudinal short- and long-term mortality data for com-
mercial fishery discards and recreational fishery catch-and-releases. ICES Journal of Marine Science,
72: 1834-1847.
Benoît, H. P., Hurlbut, T., Chassé, J., Jonsen, I. D. 2012. Estimating fishery-scale rates of discard
mortality using conditional reasoning. Fisheries Research, 125-126: 318-330.
Byrd, R. H., Lu, P., Nocedal, J., Zhu, C. 1995. A Limited Memory Algorithm for Bound Constrained Op-
timization. SIAM Journal on Scientific Computing, 16: 1190-1208.
Kaplan, E. L., Meier, P. 1958. Nonparametric Estimation from Incomplete Observations. Journal of the
American Statistical Association, 53: 457-481.
Morfin, M., Kopp, D., Benoît, H. P., Méhault, S., Randall, P., Foster, R., Catchpole, T. 2017a. Survival
of European plaice discarded from coastal otter trawl fisheries in the English Channel. Journal of Envi-
ronmental Management, 204: 404-412.
Morfin, M., Méhault, S., Benoît, H. P., Kopp, D. 2017b. Narrowing down the number of species requir-
ing detailed study as candidates for the EU Common Fisheries Policy discard ban. Marine Policy, 77:
23-29.
17