Content uploaded by Josep Alós
Author content
All content in this area was uploaded by Josep Alós on Oct 07, 2014
Content may be subject to copyright.
Recreational-angling intensity correlates with alteration of vulnerability to fishing in a 1
carnivorous coastal fish species 2
3
Josep Alós
1,2*
, Miquel Palmer
1
, Pedro Trías
1
, Carlos Díaz-Gil
1
& Robert Arlinghaus
2,3
4
5
1. Instituto Mediterráneo de Estudios Avanzados, IMEDEA (CSIC-UIB). C/ 6
Miquel Marqués 21, 07190, Esporles, Illes Balears, Spain 7
2. Department of Biology and Ecology of Fishes, Leibniz-Institute of Freshwater 8
Ecology and Inland Fisheries, Müggelseedamm 310, 12587 Berlin, Germany 9
3. Division of Integrative Fisheries Management, Faculty of Life Sciences and 10
Integrative Research Institute for the Transformation of Human-Environmental 11
Systems, Humboldt-Universität zu Berlin, Invalidenstrasse 42, 10155 Berlin, 12
Germany 13
14
*To whom correspondence should be addressed: 15
Dr. Josep Alós (email: alos@igb-berlin.de) 16
Leibniz-Institute of Freshwater Ecology and Inland Fisheries 17
Department of Biology and Ecology of Fishes 18
Müggelseedamm 310, 12587 Berlin 19
Tel: +49 (0)30 641 81 612 and Fax: +49 (0)30 641 81 750 20
21
22
23
24
Page 1 of 38
Can. J. Fish. Aquat. Sci. Downloaded from www.nrcresearchpress.com by LEIBNITZ-INSTITUT FUR on 10/07/14
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
Abstract 25
Increased timidity is a behavioral response to exploitation caused by a combination of 26
learning and fisheries-induced selection favoring shy fish. In our study, the potential for 27
angling-induced change in fish behavior was examined in two marine coastal fish 28
exploited by boat recreational fishing in the Mediterranean (Mallorca, Spain). It was 29
expected that the average vulnerability to capture of surviving individuals would differ 30
across a gradient of previous exposure to recreational angling, and that this effect would 31
be present in multiple species. The prediction received partial empirical support. 32
Recreational angling intensity was correlated with enhanced gear-avoidance behavior in 33
only one of the two study species, the carnivorous painted comber (Serranus scriba). By 34
contrast, the omnivorous fish species in our study, the annular seabream (Diplodus 35
annularis), did not differ in its behavior towards hooks in exploited compared to 36
unexploited sites. These results suggest that recreational angling may contribute to 37
patterns of hyperdepletion in catch rates due to increased timidity and associated 38
reduced vulnerability to fishing gear in some exploited species. Such effects would lead 39
to erroneous interpretations about the status of the fish stocks when assessed by fishery-40
dependent data and would negatively affect catch rates and quality of the fishery in the 41
affected species. 42
43
Key words: boldness, catch rates, catchability, fisheries-induced selection, fishery-44
dependent stock assessment, recreational fisheries, shyness, survival analysis 45
46
47
Introduction 48
Page 2 of 38
Can. J. Fish. Aquat. Sci. Downloaded from www.nrcresearchpress.com by LEIBNITZ-INSTITUT FUR on 10/07/14
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
Fish behavior plays a key role in determining and moderating the impact of fishing on 49
wild populations, inter alia, because it affects the vulnerability of individual fish to 50
fishing gear (Walters and Martell 2004). The vulnerability of all vulnerable individuals 51
in an exploited stock to fishing effort is subsumed in fisheries literature in the 52
catchability coefficient (Arreguín-Sánchez 1996). Any change induced by fishing in the 53
population-level catchability as a function of alteration of individual-level vulnerability 54
to fishing will affect fishing quality because, for example, recreational anglers derive 55
satisfaction from high catch rates (Arlinghaus 2006). Moreover, fishing-induced 56
changes in vulnerability-related behavior, and consequently catchability, will affect 57
fishery-dependent assessments due the potential for de-coupling true fish abundance and 58
catch rates (Pauly et al. 2013; Pine et al. 2009; Walters 2003). Despite its importance, 59
the behavioral dimension of selective fisheries has been largely unexplored (Arlinghaus 60
et al. 2013; Olsen et al. 2012; Parsons et al. 2011). 61
Foraging arena theory provides a suitable framework for the mechanistic study 62
of the consequences of any behavior-based change in response to fishing (Ahrens et al. 63
2012). It is assumed in foraging arena theory that predation risk caused by natural 64
predators or fishing is one of the major selective forces of selection leading to 65
behavioral adaptation. Accordingly, to avoid predation, in any moment fish populations 66
cluster into two mutually exclusive states of being vulnerable or invulnerable to 67
predation (or fishing) (Ahrens et al. 2012; Walters and Martell 2004). The risk-sensitive 68
behavioral decisions of fish (for example the decision to stay in safe refuges) ultimately 69
will determine the proportion of fish that are vulnerable to fishing gear. The transition 70
rate between the vulnerable to invulnerable pools in relation to risk of being harvested 71
has been shown to have a genetic component in fish (Ariyomo et al. 2013; Biro and Post 72
2008; Philipp et al. 2009), but can also be strongly affected by phenotypically plastic 73
Page 3 of 38
Can. J. Fish. Aquat. Sci. Downloaded from www.nrcresearchpress.com by LEIBNITZ-INSTITUT FUR on 10/07/14
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
responses to a range of ecological processes (e.g., availability of refuges, presence of 74
predators, Abrams et al. 2011; Inoue et al. 2005; Matsuda and Abrams 2004). Cox and 75
Walters (2002) applied foraging arena theory to recreational angling, theorizing that 76
foraging arenas also exist in relation to the fish reactions to the threat of angling. 77
Indeed, similar to hunting (Ciuti et al. 2012) recreational fisheries may create a 78
“landscape of fear” (Januchowski-Hartley et al. 2013a) that increasingly moves risk-79
sensitive fish into invulnerable pools. Assessing the proportions of vulnerable and 80
invulnerable fish in different moments in time and studying how the proportions vary 81
with fishing intensity therefore seems a suitable starting point to furthering our 82
understanding about how fishing may alter behavior of fish. 83
Two not mutually exclusive mechanisms can affect the flow of fish from 84
vulnerable to invulnerable pools in response to fishing as a form of human-induced 85
predation threat (Côté et al. 2014): (i) an evolutionary (i.e., genetic) response favoring 86
invulnerable behavioral phenotypes caused by fisheries selectively capturing bold 87
genotypes that actively forage outside refuges (Uusi-Heikkilä et al. 2008; Sutter et al. 88
2012; Wohlfarth et al. 1975), and (ii) acquisition of gear avoidance behavior through 89
individual or social learning from previous experiences (Beukema 1968; 1970; Raat 90
1985; van Poorten and Post 2005). Selective capture of certain behavioral types has 91
recently received some scientific attention due the growing evidence that fish, like many 92
other vertebrates, show consistent individual differences in their behavioral patterns 93
across time and contexts (i.e., personality or behavioural type, Conrad et al. 2011; 94
Mittelbach et al. in press). The assumption is that there is genetic variance associated 95
with behavioral variation and that bolder and more explorative or active fish are more 96
easily captured by many fishing gears than less bold, explorative or active behavioral 97
types (Biro and Post 2008; Côté et al. 2014; Sutter et al. 2012, but see Wilson et al. 98
Page 4 of 38
Can. J. Fish. Aquat. Sci. Downloaded from www.nrcresearchpress.com by LEIBNITZ-INSTITUT FUR on 10/07/14
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
2011). The second plasticity mechanism instead focuses on the ability of fish to quickly 99
adjust their behavior due to learning, either by observing other fish being attacked by 100
predators or from own encounters with predators, which may include encounters with 101
fishing gear as a form of human-induced predation threat (Brown et al. 2013; Klefoth et 102
al. 2011; Warburton 2003). Both processes would leave behind individuals that are 103
intrinsically harder to catch, thereby increasing the proportion of fish in invulnerable 104
pools as fishing intensity magnifies. 105
The potential for species-specific behavioral responses to recreational angling 106
gear in the wild and how these changes alter the proportion of vulnerable and 107
invulnerable pools with increasing fishing pressure remain open questions. The 108
objective of the present work was to provide empirical evidence in relation to these 109
open questions by analyzing the vulnerability to recreational angling gear of two fish 110
species with contrasting foraging ecology. We empirically tested the hypothesis that 111
human predation risk (as induced by recreational boat angling) induces a change in 112
behavior (Klefoth et al. 2011) that alters the proportion of fish in vulnerable and 113
invulnerable pools in two exploited coastal fish species in the Mediterranean Sea. To 114
test our hypotheses, we contrasted the fishes’ risk-taking behavior as determined using 115
an autonomous underwater video recording device in the wild and estimated the 116
proportion of fish in vulnerable and invulnerable pools across a gradient of fisheries-117
induced risk in exploited and unexploited sites. 118
119
Material and Methods 120
Study species 121
We studied the popular recreational angling fishery above the Posidonia oceanica 122
seagrass meadows in the Mediterranean Sea in Mallorca, Spain (March et al. 2014; 123
Page 5 of 38
Can. J. Fish. Aquat. Sci. Downloaded from www.nrcresearchpress.com by LEIBNITZ-INSTITUT FUR on 10/07/14
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
Seytre and Francour 2014). The habitat supports a species-rich fish community that is 124
primarily based on small-bodied species with confined home ranges such as the annular 125
seabream, Diplodus annularis, the painted comber, Serranus scriba, the Mediterranean 126
rainbow wrasse Coris julis or the comber, Serranus cabrilla (see for details of the 127
species composition in the fishery and its size structure in the catch, Alós and 128
Arlinghaus 2013). The fishery is a low-skilled fishery that is based on the use of natural 129
baits (shrimp) fished from anchored boats where the anglers distributes themselves over 130
patches of seagrass, which are known from independent studies to concentrate the 131
targeted species (March et al. 2014). Two of the most targeted fish species are two 132
similar-body sized species with different feeding ecology: D. annularis (Sparidae), and 133
S. scriba (Serranidae) (Morales-Nin et al. 2005). The seagrass is the preferred habitat 134
for both species, offering refuge against large-bodied predators, while the use of refuge-135
free sand habitat is rare (March et al. 2010; 2011). Although the preferred prey of the 136
two study species overlaps somewhat (Stergiou and Karpouzi 2001), both species tend 137
to forage on different prey types within the seagrass habitat. Based on stable isotope 138
studies, S. scriba is a carnivorous fish that primarily feeds on mobile prey, such as small 139
fish or decapods (Stergiou and Karpouzi 2001). By contrast, D. annularis primarily 140
feeds on small sessile prey, including algae and small bivalves (Pinnegar and Polunin 141
2000; Stergiou and Karpouzi 2001). Note that S. scriba does not feed on D. annularis 142
directly as both species have similar body sizes. Because the two species differ in their 143
feeding ecology while using the same habitat for refuge and foraging, we selected them 144
as models of an omnivorous and a carnivorous exploited fish in the present work. 145
146
Study site and fishing pressure index 147
Page 6 of 38
Can. J. Fish. Aquat. Sci. Downloaded from www.nrcresearchpress.com by LEIBNITZ-INSTITUT FUR on 10/07/14
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
Along the 20 m isobath of the coastline of the inner Palma Bay in Mallorca (N39 34, E2 148
38), NW Mediterranean (Figure S1), we randomly selected 54 sampling sites over the 149
seagrass meadows of P. oceanica. Sites were separated from each other by a minimum 150
distance of 250 m. The sampled area is regularly frequented by local anglers (Alós and 151
Arlinghaus 2013; Alós et al. 2014a). 152
The probability for an individual fish of encountering a human predator can be 153
considered an index of underlying predation risk (Lima and Dill 1990). For several 154
reasons, the number of angling boats per area constituted a suitable surrogate of the 155
number of encounters between the rather immobile species studied here and the mobile 156
anglers in our study system. First, a larger number of recreational fishing boats per area 157
should correlate with the angling gear density in the water body, thereby increasing the 158
probability of encounter and capture. Second, as the fishery is characterized by low-skill 159
techniques and is carried out in a particular habitat (i.e., seagrass of P. oceanica) located 160
close to the shore, we expected the spatial distribution of the anglers and the typology of 161
anglers (e.g., high and low skilled) to be uncorrelated. Hence, all sites should be fished 162
by the same angler types because most of them use low-skill techniques in habitats that 163
are accessible and well within the boat distance of harbors. Third, the fishery is 164
predominantly located in the shallow seagrass habitat, and a larger number of boats per 165
area should increase noise levels. This process can generate a change in fish behavior 166
associated with the recreational fishing activity by signaling the presence of an angler 167
(Holles et al. 2013). Hence, one angler-boat unit is likely to constitute a valid fishing 168
pressure unit. Note that the number of angler per boat did not vary significantly among 169
the high and low exploited fishing areas we surveyed (ANOVA, F = 1.403 p-value = 170
0.245), so that an angler-boat unit seemed an appropriate effort index. It should be noted 171
Page 7 of 38
Can. J. Fish. Aquat. Sci. Downloaded from www.nrcresearchpress.com by LEIBNITZ-INSTITUT FUR on 10/07/14
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
that there were few if any pleasure boats in the area, and it is thus easy to identify angler 172
boats during censuses (see below for details). 173
For the reasons given, we approximated the number of encounters between fish 174
and anglers in each sampling site through visual census of recreational fishing boats 175
(Cabanellas-Reboredo et al. 2014). To that end, we considered an area of 1 km
2
around 176
each sampling site, which corresponded to the average home range size of the two study 177
species (March et al. 2010; 2011), and we visited each field sampling station at least 178
once a month over a two-year period (2009 and 2010). We first counted the total 179
number of fishing boats on the 54 the sampling sites and calculated the total number of 180
fishing boats per km
2
per census-day. Second, we determined the average number of 181
fishing boats per km
2
for each sampling site as a fishing predation risk index. We 182
categorized the fishing predation risk of each sampling site as either low or high based 183
on a median split of average site-specific fishing boats per km
2
per day. Fish exposed to 184
the high fishing predation risk were exploited by a mean ± s.d. (range) of 1.3 ± 0.6 185
(0.41-3.04) fishing boats per km
2
per day, whereas the sampling locations categorised as 186
having low fishing risk had a mean ± s.d. of 0.16 ± 0.13 (0-0.39) fishing boats per km
2
187
per day. Accordingly, the fishing predation risk was on average 87 % lower in the low 188
risk sites. 189
190
Assessment of fish behavior in the field 191
We used an autonomous underwater video recording device to record the behavior of 192
the fish when they were exposed to baited hooks (Figure S2) that represented 193
conventional fishing gear used by recreational fishers in the study area. Underwater 194
video has previously been successfully used to record the behavior of wild marine fish 195
around baited hooks (e.g., Løkkeborg et al. 1989; Mallet and Pelletier 2014). We 196
Page 8 of 38
Can. J. Fish. Aquat. Sci. Downloaded from www.nrcresearchpress.com by LEIBNITZ-INSTITUT FUR on 10/07/14
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
measured vulnerability-related behavior using the latency time to attack a natural bait as 197
a potential food object presented above the seagrass habitat. Variants of “open field 198
tests” where a bait is offered outside a refuge are sometimes considered a measure of 199
feeding under risk of predation in laboratory trials (Carter et al. 2012; Réale et al. 2007). 200
In our case the seagrass was both refuge and foraging habitat for both study species 201
(Deudero et al. 2008). The choice of offering the bait above the seagrass or short 202
vicinity to the seagrass was simply convenience to receive good visual enumeration of 203
attack frequencies and latencies, and we cautiously did not interpret the results as a 204
measure of boldness (i.e., feeding under risk of predation) per se. Instead we called our 205
behavioral measure “vulnerability to fishing”. 206
The experimental protocol was based on simultaneously deploying three 207
different camera devices and baited hooks in each of the 54 sampling stations. The three 208
cameras were identical; they were deployed 50 m apart to ensure that there would be no 209
overlap between the cameras. The cameras continuously recorded (in full high 210
definition) over the seagrass for a period of 10 min. In each 10 minute video, all 211
identifiable individuals of D. annularis and S. scriba were continuously monitored. No 212
cameras were deployed when anglers were present in the sampling site. We recorded the 213
latency time as the duration in seconds from the time at which a focal fish appeared in 214
the video field to the time until the fish potentially approached and ingested one of five 215
baited hooks with a piece of shrimp, Penaeus vannamei (the commonly bait and gear 216
used in the fishery, Alós and Arlinghaus 2013). The hook shank was cut to prevent 217
hooking the fish. Instead, we assumed a “theoretical capture” event whenever an 218
individual fish ingested the bait and the individual was not tracked in the video for 219
longer. We considered the individuals that did not ingest the bait to be right-censored 220
data. The use of underwater video cameras usually don’t allow for identification of the 221
Page 9 of 38
Can. J. Fish. Aquat. Sci. Downloaded from www.nrcresearchpress.com by LEIBNITZ-INSTITUT FUR on 10/07/14
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
individual (Mallet and Pelletier 2014). Therefore we couldn’t discard the existence of 222
some bias of measuring the latency time twice in an individual that left the field of the 223
camera (right-censored data) and re-entered in the field after. We accounted for this 224
potential bias by testing trough ANOVA if the number of fish measured per video was 225
independent of the fishing pressure (high or low). We found a non-significant effect of 226
the fishing pressure in the number of fish measured in both species (D. annularis: 227
ANOVA, F = 0.267 p-value = 0.61 and S. scriba: ANOVA, F = 0.127 p-value = 0.724). 228
We therefore discarded that the number of potential individuals measured for more than 229
one time affected the intra-specific individual behavior differences in low- and high-230
fishing intensity environments. We performed a survival analysis on the latency times 231
of non-captured and hypothetically captured fish (see below). All of the S. scriba (n = 232
62 fish) sampled and a random representative sample of the whole of the sampling 233
stations of D. annularis (n = 119 fish) were analysed (Figure 1). 234
Once the three experimental trials in one site station were completed, we visited 235
another sampling site until all of the sites were sampled at random. For logistical 236
reasons, it was impossible to sample all 54 sites in one day. Therefore, we structured the 237
sampling into four days (sampling time from 9:00 AM to 1:00 PM). On each sampling 238
day, we visited a number of different sampling stations (n = 12, 14, 15 and 13 sampling 239
sites per day), but the order of sampling within days was fully randomised. This meant 240
that we visited sampling stations spaced across the entire field site and covering both 241
harvesting pressure sites on each sampling day. Overall, a total of 162 experimental 242
trials (videos of 10 min duration) were collected. 243
244
Data analysis 245
Page 10 of 38
Can. J. Fish. Aquat. Sci. Downloaded from www.nrcresearchpress.com by LEIBNITZ-INSTITUT FUR on 10/07/14
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
Two possible confounding variables were assessed and subsequently controlled in the 246
modelling process: the effect of habitat characteristics and the density of potential 247
competitors for food. First, although all of the experimental trials were conducted in 248
seagrass meadows, we explicitly considered the presence of microhabitats (e.g., the 249
presence of a rock in the seagrass, patchy sandy or muddy sediment within the seagrass, 250
Figure S3). We categorized the habitat as absence-presence (0 or 1) of seagrass, rocks, 251
sand or muddy and obtained a matrix of multiple habitat combinations (up to 8 252
combinations of absence and presence of the specific habitats). We reduced these 253
multiple combinations of habitats using a principal component analysis (PCA, Figure 254
S3). The first two axes of the habitat-specific PCA explained 82% of the total 255
variability. One main microhabitat gradient was identified for each of the axes [Figure 256
S3, PC1: gradient involving the presence of rocks (negative scores) and PC2: presence 257
of mud (positive scores) in the seagrass]. We used the first two PCA components 258
instead the 8 combinations of habitats in all further analyses by adopting the PCA 259
scores of the first and second axes as variables. Second, using the video footage, we 260
also measured the density of potential competitors by counting the total number of fish 261
of the same and any others species located in the sampling area at the moment a focal 262
individual of either one of the two species appeared in the camera field. Both potentially 263
confounding variables were considered as co-variates in further data analyses. 264
We used survival analysis on latency time to ingest a bait as surrogate for being 265
vulnerable to harvest to investigate the factors that affected the time at which the 266
particular event occurred (Hougaard 1999). Survival analysis of the sort tackled here 267
has to deal with an important challenge: not all fish ingested the bait within the duration 268
of an experimental trial. Therefore, the exact latency time is unknown for some fish that 269
did not reach the endpoint of the event (i.e., theoretical capture). These partially missing 270
Page 11 of 38
Can. J. Fish. Aquat. Sci. Downloaded from www.nrcresearchpress.com by LEIBNITZ-INSTITUT FUR on 10/07/14
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
data are called right-censored data (Hougaard 1999). To account for this, specific 271
likelihood functions have been developed in past survival analysis applications 272
(Crawley 2007). From these, we chose a Cox regression model for describing the 273
probability of non-capture (survivorship) against a set of explanatory variables. The 274
sample unit for a survival analysis was each individual fish censored, and the full 275
survival model included the fixed properties of the treatments (i.e., harvesting pressure, 276
habitat type using the PCA scores and the abundance of conspecifics or heterospecifics) 277
and the random effect of day. We performed two different types of analyses: (i) a test 278
for assessing the existence of inter-species differences (S. scriba vs. D. annularis) in 279
relation to the intrinsic vulnerability to capture based on a scenario of low harvesting 280
pressure (which was assumed to represent a more natural situation with less intensive 281
human-induced disturbances) and (ii) two intra-species models for assessing the effects 282
on capture probability attributable to the harvesting pressure (high vs. low) and all 283
relevant covariates. We used the coxph function in the survival library of the R package 284
(developed by T. Therneau and T. Lumleyat; http://cran.r-285
project.org/web/packages/survival/survival) to estimate the model parameters of the 286
minimally adequate model [Akaike information criterion (AIC)-based stepwise-287
selection using the function step] and the likelihood ratios of the model. The predicted 288
capture rates at different times and for different factors (species or harvesting pressure) 289
were estimated using the function survfit from the same library to visualize the results. 290
The stabilization of the survival probability over time was useful to explore the 291
proportion of fish that remained invulnerable to the fishing gear, and this information 292
was used to test the clustering of fish into invulnerable and vulnerable components, as 293
predicted by foraging arena theory (Ahrens et al. 2012). 294
Page 12 of 38
Can. J. Fish. Aquat. Sci. Downloaded from www.nrcresearchpress.com by LEIBNITZ-INSTITUT FUR on 10/07/14
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
To assess the relationship of harvesting pressure and true fish abundance and start to 295
appreciate whether alterations of fishing vulnerability may decouple catch rates from 296
abundance, we derived an index of relative abundance to compare the number of fish of 297
each species that appeared on the videos in relation to the two types of fishing pressure 298
while controlling for all potentially confounding variables. We used all fish (S. scriba 299
and D. annularis) that appeared on the video tape during the 10 min and treated this 300
measure as a fishery-independent measure of the relative abundance (number of fish per 301
10 min). We obtained a total of 162 measures (i.e., 162 videos that were 10 min in 302
length) of relative abundance to explore patterns of abundance in relation to harvesting 303
pressure (high vs. low). Differences in the index of relative abundance at sites with 304
different harvesting pressure were explored via GLMM (Zuur et al. 2009). A Poisson 305
distribution was assumed because the relative abundance was expressed as count data 306
(number of fish in 10 min). The relative abundance values of S. scriba and D. annularis 307
were considered dependent variables. The sample unit was the video, and the harvesting 308
pressure (low vs. high), habitat type (PC1 and PC2) and depth (m) were included as 309
explanatory variables (fixed factors). The design was nested and fully balanced: the 3 310
replicates (3 videos) per site obtained were incorporated as random nested factors 311
(videos nested in site and day because the whole of the sites were sampled over four 312
different days) in the model. The effects of the fixed and random factors were 313
eventually included in the minimally adequate model following a forward step-by-step 314
approach by comparing the model with and without the factors using a likelihood-ratio 315
test (Zuur et al. 2009). The parameter estimates were generated using the lme4 library 316
(by D. Bates and M Maechler; http://cran.r-project.org/web/packages/lme4) in the R 317
package. 318
319
Page 13 of 38
Can. J. Fish. Aquat. Sci. Downloaded from www.nrcresearchpress.com by LEIBNITZ-INSTITUT FUR on 10/07/14
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
Results 320
Inter-specific comparison of behavior in low-fishing intensity environments 321
In the sites which experienced low fishing (harvesting) pressure, S. scriba (n = 37) 322
showed a shorter average latency time to approach and ingest a baited hook compared to 323
D. annularis (Table 1). The fraction of non-captured S. scriba individuals declined 324
within the first 10 s of the experiment trial, indicating vulnerable individuals. Over 60% 325
of the S. scriba individuals were “captured” within a few seconds after the presentation 326
of the baited stimulus (Figure 2). The probability of avoiding capture stabilized at a 327
value of approximately 10% (i.e., only one-tenth of S. scriba were invulnerable to 328
harvest within 10 min of gear deployment; Figure 2). 329
The behavioral response of D. annularis (n = 60) to baited gear in low-fishing-330
induced predation risk environments was different; it was characterized by a longer 331
average latency time and a smoother decrease in the non-capture probability compared 332
with S. scriba (Table 1, Figure 2). Only approximately 30% of the D. annularis 333
individuals were captured within the first 10 s of gear deployment, and the probability 334
of avoiding capture stabilized at 50%, meaning that half of the D. annularis remained 335
invulnerable to harvest within 10 minutes of gear deployment (Figure 2). The 336
percentage of D. annularis captured in the first 10 s of gear deployment was much less 337
than in S. scriba, which confirmed the higher intrinsic capture vulnerability of S. scriba. 338
Environmental variables exerted insignificant effects in the inter-specific 339
comparison of behavior in low-fishing intensity environments. Although the habitat 340
characteristics defined by the PC1 remained in the final model, the final effect on the 341
latency time of this variable was not significant (Table 1). The habitat characteristics 342
defined by the PC2, the number of competitors of the same or other species, and the 343
Page 14 of 38
Can. J. Fish. Aquat. Sci. Downloaded from www.nrcresearchpress.com by LEIBNITZ-INSTITUT FUR on 10/07/14
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
random variance of the day of sampling had no effect on the latency time in both 344
species, and were thus dropped from the final model (Table 1). 345
346
Intra-specific individual behavior in low- and high-fishing intensity environments 347
Individuals of D. annularis inhabiting either low- with high-fishing-induced predation 348
risk environments did not differ in their latency times, and no evidence for fishery-349
related effects on behavior towards baited hooks was found (Table 1, Figure 3). Only 350
the habitat characteristics affected the behavior of D. annularis (Table 1). The presence 351
of sand and the absence of rocks or mud in the seagrass, increased the latency times 352
(i.e., lowered vulnerability) (Table 1). The number of competitors of the same or other 353
species and the random variance of the day of sampling had no effect on the latency 354
time in both species, and these variables were thus dropped from the final model (Table 355
1). 356
In S. scriba both fishing-induced predation risk and habitat characteristics had a 357
significant effect on the vulnerability to capture (Table 1). As elaborated before, the 358
behavior of individuals inhabiting low harvesting sites was characterized by short 359
latency times and large probabilities to be rapidly captured (Figure 3). In the high-360
fishing intensity sites, however, the behavior of S. scriba was characterized by long 361
latency times (low vulnerability to fishing) and a smaller overall probability to be 362
captured (Table 1 and Figure 3). The results revealed that the pools of vulnerable and 363
invulnerable S. scriba varied among sites with contrasting fisheries intensity. Although 364
the vulnerable pool of S. scriba in low risk sites represented 70% of the population, only 365
20% of the individuals were in the vulnerable pool in the high risk sites at the moment 366
of sampling (Figure 3). Therefore, we accepted the hypothesis that heavily harvested 367
populations of S. scriba were dominated by invulnerable fish. Although PC1 and PC2 368
Page 15 of 38
Can. J. Fish. Aquat. Sci. Downloaded from www.nrcresearchpress.com by LEIBNITZ-INSTITUT FUR on 10/07/14
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
remained in the final model of S.scriba after AIC-based model reduction, only the 369
presence of mud in the field, denoted by positive values of PC2 had a significant effect 370
by increasing the latency time and hence decreasing the vulnerability to angling (Table 371
1). Both the number of competitors of the same or other species and the random 372
variance of the day of sampling had no effect on the latency time, and were also 373
dropped from the final model (Table 1). 374
375
Relative abundance of S. scriba and D. annularis in low- and high-fishing intensity sites 376
Minimal adequate GLMMs only retained variables related to the habitat type to explain 377
the relative abundance of both study species as revealed by video recordings (Table 2). 378
In the case of D. annularis, only the habitat variable PC2 (the presence of muddy 379
sediment in the seagrass) was significant, and the abundance of fish decreased with an 380
increasing presence of mud (Table 2). By contrast, the presence of rocks in the seagrass 381
favoured the presence of S. scriba (Table 2). Neither depth (m), fishing intensity nor the 382
random variance of days affected the abundance of either species. 383
384
Discussion 385
In our comparative field study, we found evidence of correlation of risk-taking behavior 386
in S. scriba in relation to angling intensity. In particular, we found the proportion of fish 387
that were vulnerable to angling to be different in sites varying by fishing intensity 388
levels, while controlling for relevant environmental variables related to habitat structure, 389
depth and competitor density (of conspecifics or heterospecifics). By contrast, no 390
evidence of such variation in behavior was found in D. annularis. Irrespective of the 391
exact mechanism (selection or learning) that could cause the alteration in fishing 392
vulnerability in S. scriba, our results join previous studies (e.g., Biro and Post 2008; 393
Page 16 of 38
Can. J. Fish. Aquat. Sci. Downloaded from www.nrcresearchpress.com by LEIBNITZ-INSTITUT FUR on 10/07/14
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
Philipp et al. 2009; Sutter et al. 2012) that suggest that at least some exploited fish 394
species respond to fishing by becoming more risk averse. Such change in behavior 395
increases the pool of invulnerable fish and may contribute to the decoupling of catch 396
rates and fish abundance. The latter statement received some support in our work 397
because we did not detect any differences in relative abundances of either species 398
among fishing risk levels despite the differences in vulnerability to fishing gear detected 399
in S. scriba. 400
A growing body of literature documents the existence of personality and behavioral 401
types in fish, defined as consistent individual differences in behaviors (Sih et al. 2012). 402
There is also growing evidence that fishing selects for some of these behavioral traits; 403
usually bold and aggressive individuals were found to be more vulnerable to capture 404
than shy fish ( Alós et al. 2014b; Biro and Post 2008; Klefoth et al. 2012). Accordingly, 405
harvesting with recreational hook-and-line can be expected to generally select for shy 406
and less active phenotypes (Alós et al. 2014a), traits that may also be associated with 407
lower life-history productivity (Biro and Stamps 2008). In our study, we were not able 408
to directly assess activity in low and highly exploited sites, and we were also unable to 409
generate an independent measure of boldness. Moreover, we did not measure 410
repeatability and consistency of the “latency to bite” measure across different ecological 411
contexts, which prevented us from interpreting our behavioral measure as personality 412
trait. However, it is undisputed that we found vulnerability to fishing to be substantially 413
different in S. scriba that inhibited highly exploited sites compared to individuals living 414
in lowly exploited areas. In this respect, our findings were consistent with previous 415
experimental studies on fishing vulnerability (Alós et al. 2012; Biro and Post 2008; 416
Sutter et al. 2012) and selection of vulnerable behavioral types could explain the results 417
we obtained. 418
Page 17 of 38
Can. J. Fish. Aquat. Sci. Downloaded from www.nrcresearchpress.com by LEIBNITZ-INSTITUT FUR on 10/07/14
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
Learning from previous experiences is the other mechanism which may contribute 419
to the patterns we found in S. scriba (Anderson and LeRoy Heman 1969; Askey et al. 420
2006; Young and Hayes 2004). Experiential learning to avoid capture certainly is an 421
important contributor to the expression of risk-taking behavior in fish in response to 422
predation risk by humans (Januchowski-Hartley et al. 2013a, Klefoth et al. 2013). 423
Moreover, in addition to plasticity, learning ability has a genetic basis in fishes 424
(Huntingford and Wright 1992). Askey et al. (2006) demonstrated that an exploited 425
catch-and-release fishery of rainbow trout, Oncorhynchus mykiss, contained a group of 426
fish that quickly learned to avoid hooks in just one week of exploitation, while others 427
continued to be readily captured. Overall, catch rates drastically dropped when rainbow 428
trout, Oncorhynchus mykiss, fishing started (Askey et al. 2006), mirroring findings 429
previously reported for pike, Esox lucius fished with lures (Beukema 1969) and carp, 430
Cyprinus carpio (Raat 1985). Therefore, declining angling catch rates with increasing 431
angling effort can be expected for some species even when the number of fish remains 432
constant (Askey et al. 2006; Klefoth et al. 2013). Although fisheries-induced evolution 433
of behavior as well as plastic learning might be involved in the results we reported, 434
without common garden experiments we cannot definitively determine whether fishery-435
induced behavioral change in S. scriba was caused by genetic selection acting on risk-436
taking behavior trait directly, by genetic selection of learning ability or by plastic 437
learning from previous hooking or through the observations of conspecifics being 438
hooked and possibly removed. However, irrespective of the mechanism, our work 439
suggests that exploitation can drive populations of S. scriba to become more risk-440
averse, which will decouple angling catch rates and fish abundance as the proportion of 441
invulnerable fish increases. 442
Page 18 of 38
Can. J. Fish. Aquat. Sci. Downloaded from www.nrcresearchpress.com by LEIBNITZ-INSTITUT FUR on 10/07/14
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
Based on our study, no generalization across species in terms of fishery-induced 443
decrease in risk-taking behavior is possible. In fact, it is expected that fishery-induced 444
behavior-based changes may be species-specific by reflecting the evolutionary history 445
of fishes (Blowes et al. 2013; Feary et al. 2011; Januchowski-Hartley et al. 2011). We 446
did not find the same responses in D. annularis (an omnivorous fish) as we did in S. 447
scriba (a carnivorous fish) and the vulnerable proportion of D. annularis fish was 448
generally low, independent of angling risk. Species-dependent results could be 449
explained by the feeding ecology of the species (Stoner 2004), assuming that the more 450
aggressive carnivorous fish species would be intrinscally more vulnerable to fishing and 451
hence show stronger responses to the omnivorous species. However, Wilson et al. 452
(2011) noted that shy, omnivorous bluegill, Lepomis macrochirus, were preferentially 453
harvested by angling compared with the bolder individuals of the population. By 454
contrast, Klefoth et al. (2012; 2013) studied the omnivorous carp, Cyprinus carpio, 455
revealing a positive relationship between boldness and vulnerability. Therefore, in the 456
three omnivorous fish studied so far varying scenarios of angling-induced adaptive 457
change in behavior were reported, involving scenarios of increasingly bold (bluegill), 458
shy (carp) or unaltered (D. annularis) behavioral phenotypes. There seems to be limited 459
room for generalization as to which behavioral response to expect to fishing in this 460
feeding guilt. By contrast, the available evidence is more consistent in carnivorous fish 461
species, such as largemouth bass, Micropterus salmoides (Cooke et al. 2007) or rainbow 462
trout, Oncorhynchus mykis (Biro and Post 2008). All these studies documented that 463
bolder, more aggressive and more active fish to be preferentially harvested, similar to 464
the outcomes shown in S. scriba in the present work. 465
We also found species-specific differences in terms of how abundance and 466
latency time to bite varied along microhabitats. Uniform seagrass microhabitats and 467
Page 19 of 38
Can. J. Fish. Aquat. Sci. Downloaded from www.nrcresearchpress.com by LEIBNITZ-INSTITUT FUR on 10/07/14
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
those scattered with patches of sand decreased the latency time to bite (e.g., theoretical 468
capture was quicker) in D. annularis, while also increasing its abundance compared to 469
seagrass microhabitats with a presence of rocks (PC1) or mud (PC2). Uniform seagrass 470
habitat constitutes to the preferred foraging habitat of D. annularis, presumably because 471
it offers optimal conditions for feeding on small non-mobile prey types, such as small 472
crustaceans , or epiphytes attached to P. oceanica (Giakoumi 2013). The habitat effects 473
on abundance and vulnerability to fishing were different in the carnivorous S. scriba. In 474
this species, the presence of rocks breaking the uniformity of the seagrass habitat 475
increased the abundance of S. scriba, which was consistent with the literature reporting 476
more structured seagrasses to be the preferred habitat of this species (Giakoumi 2013). 477
Interestingly, decreased latency time to attack the bait was found in the less preferred 478
seagrass habitat with muddy patches, presumably because the less structured muddy 479
habitat promoted predation risk assessment, in turn fostering more rapid attacks after 480
deployment of the bait. Our data collectively showed that vulnerability to fishing is not 481
only a function of the fishing pressure and the intrinsic biology of exploited species, but 482
is moderated by habitat features in a species-specific fashion. In light of these findings, 483
further studies in the two species outside P. oceanica might be worthwhile to analyze 484
whether the differential fishing pressure found in the present study holds for even less 485
structured habitat. Obviously, also more studies with other species are a worthwhile 486
endeavor. 487
Our work has four relevant limitations that should be mentioned and ideally 488
addressed in future work. First, we only assessed two species and inferences expressed 489
towards behavioral change across species in our study must be treated with extreme 490
caution and as tentative at best. Second, our experimental design was designated to 491
asses the behavior of fish above seagrasses. Further work is needed to fully understand 492
Page 20 of 38
Can. J. Fish. Aquat. Sci. Downloaded from www.nrcresearchpress.com by LEIBNITZ-INSTITUT FUR on 10/07/14
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
the fish behaviors in more open habitat (e.g., rocky or sandy habitats). Thirdly, inter-493
specific behavioral differences should be studied across time and different contexts to 494
infer whether the behavioral change truly has a personality basis. This is an interesting 495
question that has to be addressed in further investigations by incorporating different 496
habitat (ecological context) in the proper assessment of behavioral and temperamental 497
traits, ideally replicating the assessments on individually marked fish over time. The 498
fourth limitation is that we have focused our work on a specific rod-and-reel fishery. 499
Similar work could be done with other passive gear where the same behavioral traits 500
may play a key role determining the fate of the fish. This includes other recreational 501
fishing gears that uses artificial baits, but also commercial fisheries such as long-line or 502
trammel nets. We are however unsure how fishes would respond to more actively 503
operated gear or how other species inhabiting different habitats (e.g., open water) 504
responds in front of the fishing gears. Our work hopefully stimulate other groups to 505
perform field experiments such as ours to better understand the role of fishing in the 506
alteration of the fish behavior and how this translates to catch rates and hyperdepletion 507
(Hilborn and Walters 1992) and therefore the reliability of stock assessments that are 508
based on fishery-dependent data. 509
Despite these limitations, we can draw three implications of our study. First, 510
increasing proportions of fish in invulnerable pools in response to fishing in some 511
species can have implications for population dynamics, food web interactions, the 512
productivity of the fishery and individual fitness. Second, reliably inferring population 513
abundance data from hook-and-line-based catch rate indices will be a challenge with 514
some species, and this challenge likely holds true for other passive harvesting 515
techniques where the capture success strongly depends on fish behavior (e.g., longline, 516
gill netting or trapping fish). An example from our own work shall illustrate the issue. 517
Page 21 of 38
Can. J. Fish. Aquat. Sci. Downloaded from www.nrcresearchpress.com by LEIBNITZ-INSTITUT FUR on 10/07/14
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
We did not find any differences in the abundances of the two species in the low- and 518
high-fishing intensity sites. This finding calls into question our own previous 519
conclusions about the conservation value offered by partial marine protected areas in the 520
Mediterranean (Alós and Arlinghaus 2013). Indeed, we reported earlier that the 521
abundances of S. scriba in exploited areas were lower than in protected areas, but we 522
inferred these results from angling-based catch rate indices. Based on the present work, 523
we might have wrongly equated catch rates to an index of underlying abundance, at 524
least in S. scriba. Third and finally, because behavior and life history traits such as 525
growth will often be correlated (Biro and Stamps 2008), one should pay attention to the 526
potential for sampling bias caused by the preferential capture of certain behavioral types 527
that carry phenotypic traits of interest (e.g., growth, Ricker 1969). Active sampling 528
methods may avoid some of this bias but can also suffer from trait-selective sampling 529
(e.g., with respect to swimming speed or schooling tendency). Because this key 530
uncertainty cannot be resolved without fish tracking studies in the wild, we echo 531
Walters and Bonfil (1999) and recommend better experimental studies to further our 532
understanding about the exchange processes between vulnerable and invulnerable 533
arenas. Adding to this challenge, studies are needed to improve our understanding 534
regarding how vulnerability arenas change over time and how this varies across fish 535
species. Shedding light on these questions is not only of academic interest but has 536
important implications for fisheries and its assessments. 537
538
Acknowledgements 539
The data-acquisition of this study was financed by the research project REC
2
540
(grant#CTM2011-23835) funded by the Spanish Ministry of Economy and 541
Competiveness (MINECO). JA was supported with a Marie Curie Post Doc grant (FP7-542
Page 22 of 38
Can. J. Fish. Aquat. Sci. Downloaded from www.nrcresearchpress.com by LEIBNITZ-INSTITUT FUR on 10/07/14
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
PEOPLE-2012-IEF, grant # 327160) and the finalization of the manuscript was made 543
possible by the B-Types project funded through Leibniz Competition (grant # SAW-544
2013-IGB-2). We specially thank the researchers involved in the field work and the 545
comments made by Alecia Carter, two anonymous reviewers and the associate editor on 546
an earlier version of the manuscript. RA received additional funding from the German 547
Federal Ministry for Education and Research (BMBF) through the Program for Social-548
Ecological Research for the project Besatzfisch (grant # 01UU0907, www.besatz-549
fisch.de) and the University of Florida, School of Forest Resources and Conservation 550
during a sabbatical stay during which this manuscript was predominantly drafted. RA 551
acknowledges the inspiring discussions with Carl Walters, Robert Ahrens and Mike 552
Allen, and the hospitality of Mike and Mendy. 553
554
References 555
Abrams, P.A., Ruokolainen, L., Shuter, B.J., and McCann, K.S. 2011. Harvesting 556
creates ecological traps: consequences of invisible mortality risks in predator-prey 557
metacommunities. Ecology 93(2): 281-293. 558
Ahrens, R.N.M., Walters, C.J., and Christensen, V. 2012. Foraging arena theory. Fish 559
Fish. 13(1): 41-59. 560
Almany, G.R., Peacock, L.F., Syms, C., McCormick, M.I., and Jones, G.P. 2007. 561
Predators target rare prey in coral reef fish assemblages. Oecologia 152(4): 751-761. 562
Alós, J., and Arlinghaus, R. 2013. Impacts of partial marine protected areas on coastal 563
fish communities exploited by recreational angling. Fish. Res. 137(0): 88-96. 564
Alós, J., Palmer, M., and Arlinghaus, R. 2012. Consistent selection towards low activity 565
phenotypes when catchability depends on encounters among human predators and fish. 566
PLoS ONE 7(10): e48030. 567
Page 23 of 38
Can. J. Fish. Aquat. Sci. Downloaded from www.nrcresearchpress.com by LEIBNITZ-INSTITUT FUR on 10/07/14
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
Alós, J., Palmer, M., Catalán, I.A., Alonso-Fernández, A., Basterretxea, G., Jordi, A., 568
Buttay, L., Morales-Nin, B., and Arlinghaus, R. 2014a. Selective exploitation of 569
spatially structured coastal fish populations by recreational anglers may lead to 570
evolutionary downsizing of adults. Mar. Ecol. Prog. Ser. 503: 219-233. 571
Alós, J., Palmer, M., Linde-Medina, M., and Arlinghaus, R. 2014b. Consistent size-572
independent harvest selection on fish body shape in two recreationally exploited marine 573
species. Ecol. Evol. 4(11): 2154-2164. 574
Anderson, R.O., and LeRoy Heman, M. 1969. Angling as a factor influencing 575
catchability of Largemouth Bass. Trans. Am. Fish. Soc. 98(2): 317-320. 576
Ariyomo, T.O., Carter, M., and Watt, P.J. 2013. Heritability of boldness and 577
aggressiveness in the zebrafish. Behav. Genet. 43(2): 161-167. 578
Arlinghaus, R. 2006. On the apparently striking disconnect between motivation and 579
satisfaction in recreational fishing: the case of catch orientation of German anglers. N. 580
Am. J. Fish. Manag. 26(3): 592-605. 581
Arlinghaus, R., Cooke, S.J., and Potts, W. 2013. Towards resilient recreational fisheries 582
on a global scale through improved understanding of fish and fisher behavior. Fish. 583
Manag. Ecol. 20(2-3): 91-98. 584
Arreguín-Sánchez, F. 1996. Catchability: a key parameter for fish stock assessment. 585
Rev. Fish Biol. Fish. 6(2): 221-242. 586
Askey, P.J., Richards, S.A., Post, J.R., and Parkinson, E.A. 2006. Linking angling catch 587
rates and fish learning under catch-and-release regulations. N. Am. J. Fish. Manag. 588
26(4): 10. 589
Beukema, J.J. 1968. Angling experiments with Carp (Cyprinus Carpio L.). Neth. J. 590
Zool. 19(4): 596-609. 591
Page 24 of 38
Can. J. Fish. Aquat. Sci. Downloaded from www.nrcresearchpress.com by LEIBNITZ-INSTITUT FUR on 10/07/14
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
Beukema, J.J. 1970. Acquired hook-avoidance in the pike Esox lucius L. fished with 592
artificial and natural baits. J. Fish Biol. 2(2): 155-160. 593
Biro, P.A., and Post, J.R. 2008. Rapid depletion of genotypes with fast growth and bold 594
personality traits from harvested fish populations. Proc. Natl. Acad. Sci. U. S. A. 595
105(8): 2919-2922. 596
Biro, P.A., and Stamps, J.A. 2008. Are animal personality traits linked to life-history 597
productivity? Trends Ecol. Evol. 23(7): 361-368. 598
Blowes, S.A., Pratchett, M.S., and Connolly, S.R. 2013. Heterospecific aggression and 599
dominance in a guild of coral-feeding fishes: the roles of dietary ecology and 600
phylogeny. Am. Nat. 182(2): 157-168. 601
Brown, C., Jones, F., and Braithwaite, V. 2005. In situ examination of boldness–shyness 602
traits in the tropical poeciliid, Brachyraphis episcopi. Anim. Behav. 70(5): 1003-1009. 603
Brown, G.E., Ferrari, M.C.O., and Chivers, D.P. 2013. Adaptive forgetting: why 604
predator recognition training might not enhance poststocking survival. Fisheries 38(1): 605
16-25. 606
Cabanellas-Reboredo, M., Alós, J., March, D., Palmer, M., Jordà, G., and Palmer, M. 607
2014. Where and when will they go fishing? Understanding fishing site and time choice 608
in a recreational squid fishery. ICES J. Mar. Sci. (published online). 609
Carter, A.J., Heinsohn, R., Goldizen, A.W., and Biro, P.A. 2012. Boldness, trappability 610
and sampling bias in wild lizards. Anim. Behav. 83(4): 1051-1058. 611
Ciuti, S., Muhly, T.B., Paton, D.G., McDevitt, A.D., Musiani, M., and Boyce, M.S. 612
2012. Human selection of elk behavioral traits in a landscape of fear. Proc. R. Soc. 613
Lond. B Biol. Sci. 279(1746): 4407-4416. 614
Page 25 of 38
Can. J. Fish. Aquat. Sci. Downloaded from www.nrcresearchpress.com by LEIBNITZ-INSTITUT FUR on 10/07/14
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
Conrad, J.L., Weinersmith, K.L., Brodin, T., Saltz, J.B., and Sih, A. 2011. Behavioral 615
syndromes in fishes: a review with implications for ecology and fisheries management. 616
J. Fish Biol. 78(2): 395-435. 617
Cooke, S.J., Suski, C.D., Ostrand, K.G., Wahl, D.H., and Philipp, D.P. 2007. 618
Physiological and behavioral consequences of long-term artificial selection for 619
vulnerability to recreational angling in a teleost fish. Physiol. Biochem. Zool. 80(5): 620
480-490. 621
Côté, I.M., Darling, E.S., Malpica-Cruz, L., Smith, N.S., Green, S.J., Curtis-Quick, J., 622
and Layman, C. 2014. What doesn't kill you makes you wary? effect of repeated culling 623
on the behavior of an invasive predator. PLoS ONE 9(4): e94248. 624
Cox, S.P., and Walters, C. 2002. Modeling exploitation in recreational fisheries and 625
implications for effort management on British Columbia rainbow trout lakes. N. Am. J. 626
Fish. Manag. 22(1): 21-34. 627
Crawley, M.J. 2007. The R Book. John Wiley & Sons, Ltd, New York. 628
Deudero, S., Morey, G., Frau, A., Moranta, J., and Moreno, I. 2008. Temporal trends of 629
littoral fishes at deep Posidonia oceanica seagrass meadows in a temperate coastal 630
zone. J. Mar. Syst. 70: 182-195. 631
Feary, D.A., Cinner, J.E., Graham, N.A.J., and Januchowski-Hartley, F.A. 2011. Effects 632
of customary marine closures on fish behavior, spear-fishing success, and underwater 633
visual surveys. Conserv. Biol. 25(2): 341-349. 634
Giakoumi, S. 2013. Effects of habitat and substrate complexity on shallow sublittoral 635
fish assemblages in the Cyclades Archipelago, North-eastern Mediterranean Sea. Med. 636
Mar. Sci. 14: 10. 637
Hilborn, R., and Walters, C. 1992. Quantitative fisheries stock assessment: choice, 638
dynamics and uncertainty. Chapman and Hall Inc., New York. 639
Page 26 of 38
Can. J. Fish. Aquat. Sci. Downloaded from www.nrcresearchpress.com by LEIBNITZ-INSTITUT FUR on 10/07/14
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
Holles, S., Simpson, S.D., Radford, A.N., Berten, L., and Lecchini, D. 2013. Boat noise 640
disrupts orientation behavior in a coral reef fish. Mar. Ecol. Prog. Ser. 485: 295-300. 641
Hougaard, P. 1999. Fundamentals of Survival Data. Biometrics 55(1): 13-22. 642
Huntingford, F.A., and Wright, P.J. 1992. Inherited population differences in avoidance 643
conditioning in three-spined Sticklebacks, Gasterosteus aculeatus. Behavior 122(3-4): 644
264-273. 645
Inoue, M., Miyayoshi, M., and Sone, S. 2005. Foraging modes of stream benthic fishes 646
in relation to their predation effects on local prey density. Ecol. Res. 20(2): 151-161. 647
Januchowski-Hartley, F.A., Cinner, J.E., and Graham, N.A.J. 2013a. Fishery benefits 648
from behavioral modification of fishes in periodically harvested fisheries closures. 649
Aquatic Conservation: Marine and Freshwater Ecosystems. (online version). 650
Januchowski-Hartley, F.A., Graham, N.A.J., Cinner, J.E., and Russ, G.R. 2013b. 651
Spillover of fish naïveté from marine reserves. Ecol. Lett. 16(2): 191-197. 652
Januchowski-Hartley, F.A., Graham, N.A.J., Feary, D.A., Morove, T., and Cinner, J.E. 653
2011. Fear of fishers: human predation explains behavioral changes in coral reef fishes. 654
PLoS ONE 6(8): e22761. 655
Klefoth, T., Kobler, A., and Arlinghaus, R. 2011. Behavioral and fitness consequences 656
of direct and indirect non-lethal disturbances in a catch-and-release northern pike (Esox 657
lucius) fishery. Know.Manage. Aquat. Ecosyst. (403): 11. 658
Klefoth, T., Pieterek, T., and Arlinghaus, R. 2013. Impacts of domestication on angling 659
vulnerability of common carp (Cyprinus carpio L.): the role of learning, foraging 660
behavior and food preferences. Fish. Manag. Ecol. 20: 12. 661
Klefoth, T., Skov, C., Krause, J., and Arlinghaus, R. 2012. The role of ecological 662
context and predation risk-stimuli in revealing the true picture about the genetic basis of 663
boldness evolution in fish. Behav. Ecol. Sociobiol. 66(4): 547-559. 664
Page 27 of 38
Can. J. Fish. Aquat. Sci. Downloaded from www.nrcresearchpress.com by LEIBNITZ-INSTITUT FUR on 10/07/14
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
Lima, S.L., and Dill, L.M. 1990. Behavioral decisions made under the risk of predation: 665
a review and prospectus. Can. J. Zool. 68(4): 619-640. 666
Løkkeborg, S., Bjordal, Å., and Fernö, A. 1989. Responses of cod (Gadus morhua) and 667
haddock (Melanogrammus aeglefinus) to baited hooks in the natural Environment. Can. 668
J. Fish. Aquat. Sci. 46(9): 1478-1483. 669
Mallet, D., and Pelletier, D. 2014. Underwater video techniques for observing coastal 670
marine biodiversity: a review of sixty years of publications (1952–2012). Fish. Res. 671
154(0): 44-62. 672
March, D., Alós, J., and Palmer, M. 2014. Geospatial assessment of fishing quality 673
considering environmental and angler-related factors. Fish. Res. 154(0): 63-72. 674
March, D., Alos, J., Palmer, M., and Grau, A. 2011. Short-term residence and 675
movement patterns of the annular seabream Diplodus annularis in a temperate marine 676
reserve. Est. Coast. Shelf Sci. 92: 581-587. 677
March, D., Palmer, M., Alós, J., Grau, A., and Cardona, F. 2010. Short-term residence, 678
home range size and diel patterns of the painted comber Serranus scriba in a temperate 679
marine reserve. Mar. Ecol. Prog. Ser. 400: 195-206. 680
Matsuda, H., and Abrams, P.A. 2004. Effects of predator–prey interactions and adaptive 681
change on sustainable yield. Can. J. Fish. Aquat. Sci. 61(2): 175-184. 682
Mittelbach, G.G., Ballew, N.G., and Kjelvik, M.K. in press. Fish behavioral types and 683
their ecological consequences. Can. J. Fish. Aquat. Sci. 684
Morales-Nin, B., Moranta, J., Garcia, C., Tugores, M.P., Grau, A.M., Riera, F., and 685
Cerda, M. 2005. The recreational fishery off Majorca Island (western Mediterranean): 686
some implications for coastal resource management. ICES J. Mar. Sci. 62(4): 727-739. 687
Page 28 of 38
Can. J. Fish. Aquat. Sci. Downloaded from www.nrcresearchpress.com by LEIBNITZ-INSTITUT FUR on 10/07/14
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
Olsen, E.M., Heupel, M.R., Simpfendorfer, C.A., and Moland, E. 2012. Harvest 688
selection on Atlantic cod behavioral traits: implications for spatial management. Ecol. 689
Evol. 2(7): 1549-1562. 690
Pauly, D., Hilborn, R., and Branch, T.A. 2013. Fisheries: does catch reflect abundance? 691
Nature 494(7437): 303-306. 692
Parsons, D.M., Morrison, M.A., McKenzie, J.R., Hartill, B.W., Bian, R., and Francis, 693
R.I.C.C. 2011. A fisheries perspective of behavioural variability: differences in 694
movement behaviour and extraction rate of an exploited sparid, snapper (Pagrus 695
auratus). Can. J. Fish. Aquat. Sci. 68(4): 632-642. 696
Philipp, D.P., Cooke, S.J., Claussen, J.E., Koppelman, J.B., Suski, C.D., and Burkett, 697
D.P. 2009. Selection for vulnerability to angling in largemouth bass. Trans. Am. Fish. 698
Soc. 138(1): 189-199. 699
Pine, W.E., Martell, S.J.D., Walters, C.J., and Kitchell, J.F. 2009. Counterintuitive 700
responses of fish populations to management actions. Fisheries 34(4): 165-180. 701
Pinnegar, J.K., and Polunin, N.V.C. 2000. Contributions of stable-isotope data to 702
elucidating food webs of Mediterranean rocky littoral fishes. Oecologia 122(3): 399-703
409. 704
Raat, A.J.P. 1985. Analysis of angling vulnerability of common carp, Cyprinus carpio 705
L., in catch-and-release angling in ponds. Aquac. Res. 16(2): 171-187. 706
Réale, D., Reader, S.M., Sol, D., McDougall, P.T., and Dingemanse, N.J. 2007. 707
Integrating animal temperament within ecology and evolution. Biol. Rev. 82(2): 291-708
318. 709
Ricker, W.E. 1969. Effects of size-selective mortality and sampling bias on estimates of 710
growth, mortality, production, and yield. J. Fish. Res. Board Can. 26(3): 479-541. 711
Page 29 of 38
Can. J. Fish. Aquat. Sci. Downloaded from www.nrcresearchpress.com by LEIBNITZ-INSTITUT FUR on 10/07/14
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
Seytre, C., and Francour, P. 2014. A long-term survey of Posidonia oceanica fish 712
assemblages in a Mediterranean marine protected area: emphasis on stability and no-713
take area effectiveness. Mar. Freshw. Res. 65(3): 244-254. 714
Sih, A., Cote, J., Evans, M., Fogarty, S., and Pruitt, J. 2012. Ecological implications of 715
behavioral syndromes. Ecol. Lett. 15(3): 278-289. 716
Stergiou, K.I., and Karpouzi, V.S. 2001. Feeding habits and trophic levels of 717
Mediterranean fish. Rev. Fish Biol. Fish. 11(3): 217-254. 718
Stoner, A.W. 2004. Effects of environmental variables on fish feeding ecology: 719
implications for the performance of baited fishing gear and stock assessment. J. Fish 720
Biol. 65(6): 1445-1471. 721
Sutter, D.A.H., Suski, C.D., Philipp, D.P., Klefoth, T., Wahl, D.H., Kersten, P., Cooke, 722
S.J., and Arlinghaus, R. 2012. Recreational fishing selectively captures individuals with 723
the highest fitness potential. Proc. Natl. Acad. Sci. U. S. A. 109(51): 20960-20965 . 724
van Poorten, B.T., and Post, J.R. 2005. Seasonal fishery dynamics of a previously 725
unexploited Rainbow Trout population with contrasts to established fisheries. N. Am. J. 726
Fish. Manag. 25(1): 329-345. 727
Uusi-Heikkilä, S., Wolter, C., Klefoth, T., and Arlinghaus, R. 2008. A behavioral 728
perspective on fishing-induced evolution. Trends Ecol. Evol. 23(8): 419-421. 729
Walters, C. 2003. Folly and fantasy in the analysis of spatial catch rate data. Can. J. 730
Fish. Aquat. Sci. 60(12): 1433-1436. 731
Walters, C., and Martell, S.J.D. 2004. Fisheries ecology and management. Princeton 732
University Press, Princeton, New Jersey, USA. 733
Walters, C.J., and Bonfil, R. 1999. Multispecies spatial assessment models for the 734
British Columbia groundfish trawl fishery. Can. J. Fish. Aquat. Sci. 56(4): 601-628. 735
Warburton, K. 2003. Learning of foraging skills by fish. Fish Fish. 4(3): 203-215. 736
Page 30 of 38
Can. J. Fish. Aquat. Sci. Downloaded from www.nrcresearchpress.com by LEIBNITZ-INSTITUT FUR on 10/07/14
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
Wilson, A.D.M., Binder, T.R., McGrath, K.P., Cooke, S.J., and Godin, J-G.J. 2011. 737
Capture technique and fish personality: angling targets timid bluegill sunfish, Lepomis 738
macrochirus. Can. J. Fish. Aquat. Sci. 68(5): 749-757. 739
Wohlfarth, G., Moav, R., Hulata, G., and Beiles, A. 1975. Genetic variation in seine 740
escapability of the common carp. Aquaculture 5(4): 375-387. 741
Young, R.G., and Hayes, J.W. 2004. Angling pressure and trout catchability: behavioral 742
observations of brown trout in two New Zealand backcountry rivers. N. Am. J. Fish. 743
Manag. 24(4): 1203-1213. 744
Zuur, A.F., Leno, E.N., Walker, N.J., Saveliev, A.A., and Smith, G.M. 2009. Mixed 745
Effects Models and Extensions in Ecology with R. Springer-Verlag New York Inc., 746
New York, USA. 747
748
Table captions 749
Table 1: Survival analysis. Cox regression coefficients (coef), standard error (se), z-750
value and p-value of the minimal adequate survival models fitted to explore inter- and 751
intra-species differences in the survivorship (i.e., no-capture) and multiple predictors. 752
The full model included species in the case of inter-species comparison, harvesting 753
pressure in the case of intra-species comparison, habitat characteristics [PC1: gradient 754
from the presence of rocks (negative score) to the presence of sand (positive scores) in 755
the seagrass, and PC2: the presence of mud in the seagrass (positive scores)] and the 756
density of competitors. The table shows the fixed factors that are included in the 757
minimally adequate model as well as the likelihood ratio test. In all cases, random 758
effects were not significant and not included in the model. 759
Table 2: Estimates, standard error (se), z-values and p-value of the minimally adequate 760
generalised linear mixed models fitted to explore differences between the relative 761
Page 31 of 38
Can. J. Fish. Aquat. Sci. Downloaded from www.nrcresearchpress.com by LEIBNITZ-INSTITUT FUR on 10/07/14
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
abundance (fish count per 10 min) and multiple predictors in D. annularis and S. scriba. 762
The full model included depth (in m), harvesting pressure (high vs. low), habitat 763
characteristics [PC1: gradient from the presence of rocks (negative score) to the 764
presence of sand (positive scores) in the seagrass, and PC2: the presence of mud in the 765
seagrass (positive scores)]. The table shows the fixed and random factors included in 766
the minimally adequate model as well as the likelihood ratio test. The number of groups 767
denoted by the random effects included in the model is shown as well as its variance 768
(σ
2
). 769
770
Page 32 of 38
Can. J. Fish. Aquat. Sci. Downloaded from www.nrcresearchpress.com by LEIBNITZ-INSTITUT FUR on 10/07/14
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
Figure captions 771
Figure 1: Histogram of the latency times (in seconds) observed in S. scriba (left panel, n 772
in low fishing intensity sites = 37 and n in high fishing intensity site = 25) and D. 773
annularis (right panel, n in low fishing intensity sites = 60 and n in high fishing 774
intensity sites = 59). The inner panels show the proportion of fish captured (black) and 775
non-captured (grey) for both species defining the group (pool) sizes of vulnerable and 776
invulnerable fish in high and low fishing intensities. Note the decrease in the proportion 777
of captured fish in high fishing intensity sites in S. scriba. 778
Figure 2: Predicted survivorship of the Cox regression fitted to explore changes in the 779
latency time (seconds) and the probability to theoretical survival (no-capture) in the two 780
species studied in low fishing intensity sites. The solid lines show the survival 781
distribution and the shaded lines show the confidence intervals (± 95%) in S. scriba 782
(grey, n = 37) and D. annularis (black, n = 60). Note the minimum overlap of the 783
confidence intervals (95%), which indicates significant differences. 784
Figure 3: Predicted survivorship of the Cox regression results fitted to explore changes 785
in the latency time (seconds) and the probability to theoretical survival (no-capture) 786
between the two fishing predation risks: low (black) and high (grey). The solid lines 787
show the survival distribution, and the shaded lines show the confidence intervals (± 788
95%) in D. annularis (upper panel) and S. scriba (lower panel). 789
790
Page 33 of 38
Can. J. Fish. Aquat. Sci. Downloaded from www.nrcresearchpress.com by LEIBNITZ-INSTITUT FUR on 10/07/14
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
Table 1: 791
792
coef exp(coef)
se(coef) z-value Pr(>|z|)
Between-species comparison (n = 93)
Species (S. scriba) 0.945 2.572 0.306 3.091 0.002
PC1 -0.294 0.745 0.156 -1.882 0.060
Likelihood ratio test= 13.59 on 2 df, p=0.0011
Within-species comparison
D. annularis (n = 119)
PC1 -0.461 0.631 0.162 -2.897 0.004
PC2 0.250 1.285 0.140 2.061 0.039
Likelihood ratio test= 8.13 on 2 df, p=0.017
S. scriba (n = 62)
Fishing predation risk (high) -1.756 0.173 0.653 -2.690 0.007
PC1 -0.968 0.380 0.517 -1.870 0.061
PC2 -3.521 0.030 1.030 -3.420 0.001
Likelihood ratio test= 28.42 on 3 df, p <0.001
793
794
795
796
797
Page 34 of 38
Can. J. Fish. Aquat. Sci. Downloaded from www.nrcresearchpress.com by LEIBNITZ-INSTITUT FUR on 10/07/14
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
Table 2: 798
799
Estimate se(estimate) z-value Pr(>|z|)
Diplodus annularis (n = 162, groups = 54)
Fixed effects
(Intercept) -1.251 0.395 -3.170 0.0015
PC1 -0.307 0.182 -1.694 0.090
PC2 -0.978 0.226 -4.325 p <0.001
Random effects
Sampling station (σ
2
= 6.47)
Likelihood ratio test= 825.84 on 2 df, p <0.001
Serranus scriba (n = 162, groups = 54)
Fixed effects
(Intercept) -3.544 0.539 -6.581 p <0.001
PC1 -1.318 0.282 -4.667 p <0.001
Random effects
Sampling station (σ
2
= 5.966)
Likelihood ratio test= 187.8 on 1 df, p <0.001
800
801
Page 35 of 38
Can. J. Fish. Aquat. Sci. Downloaded from www.nrcresearchpress.com by LEIBNITZ-INSTITUT FUR on 10/07/14
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
Figure 1 802
803
804
Page 36 of 38
Can. J. Fish. Aquat. Sci. Downloaded from www.nrcresearchpress.com by LEIBNITZ-INSTITUT FUR on 10/07/14
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
Figure 2 805
806
807
Page 37 of 38
Can. J. Fish. Aquat. Sci. Downloaded from www.nrcresearchpress.com by LEIBNITZ-INSTITUT FUR on 10/07/14
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
Figure 3 808
809
Page 38 of 38
Can. J. Fish. Aquat. Sci. Downloaded from www.nrcresearchpress.com by LEIBNITZ-INSTITUT FUR on 10/07/14
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
