ICES Journal of Marine Science (2017), 74(1), 170–179. doi:10.

1093/icesjms/fsw128

Original Article
Age-specific differences in the seasonal spatial

Downloaded from https://academic.oup.com/icesjms/article/74/1/170/2669550 by guest on 20 May 2025

distribution of butterfish (Peprilus triacanthus)
Charles F. Adams*
National Marine Fisheries Service, Northeast Fisheries Science Center, 166 Water Street, Woods Hole, MA 02543-1026, USA
*Corresponding author: tel: þ 1 508 495 2217; fax: þ 1 508 495 2393; e-mail: [email protected].
Adams, C. F. Age-specific differences in the seasonal spatial distribution of butterfish (Peprilus triacanthus). – ICES Journal of Marine
Science, 74: 170–179.
Received 4 February 2016; revised 22 June 2016; accepted 24 June 2016; advance access publication 31 July 2016.

The spatial distribution of butterfish (Peprilus triacanthus) in the Northwest Atlantic Ocean was investigated using a suite of spatial indicators
based on Northeast Fisheries Science Center spring and fall bottom trawl survey data, 1982–2013. In the spring, ages 2 and 3 were found far-
ther northeast and deeper than age 1 butterfish, while in the fall, age 3 butterfish were found farther northeast and deeper than ages 0 and 1.
There was no significant northward movement of butterfish in spring or fall over the course of either time-series. However, there was a signifi-
cant increase in the area occupied by ages 1–3 in the spring that was correlated with surface temperature. This illustrates that responses to
climate change may be manifested as range expansions, rather than poleward movement of the centre of gravity (i.e. bivariate weighted mean
location of the population). Two changes were observed over the course of the fall time series, both for ages 1 and 2: increased spatial disper-
sion; and a decrease in depth. The former result would have been masked, while the latter would have been erroneously generalized to all age
classes, if an age-specific analysis had not been done. This study demonstrates the importance of an age-based and seasonal analysis. It is also
shown how a spatial distribution analyses can inform stock assessments by providing insights into diverging survey indices and availability to
surveys in general. Similarly, spatial distribution analyses can be used to verify the spatial equilibrium assumption for the calculation of biolog-
ical reference points.
Keywords: area occupancy, butterfish, centre of gravity, Peprilus triacanthus, spatial distribution.

Introduction the Fishery Conservation and Management Act of 1976

Butterfish (Peprilus triacanthus) in the Northwest Atlantic Ocean (Murawski and Waring, 1979). Foreign landings were completely
between Cape Hatteras and the Gulf of Maine (Figure 1) are con- phased out by 1987 (Adams et al., 2015). From 2002 to 2012 there
sidered to be a unit stock for management purposes (Adams was no directed fishery, and landings, primarily as bycatch in the
et al., 2015). Butterfish begin schooling around 6 cm (Collette small mesh (<10.2 cm) bottom trawl longfin squid fishery,
and Klein-MacPhee, 2002). They are a short lived, fast growing dropped to a low of 400 mt in 2005. However, a directed fishery
species, overwintering offshore, and then moving inshore and was re-established in 2013, and harvest limits were increased
northwards in the summer (Cross et al., 1999). Spawning occurs (NMFS, 2015) following the most recent stock assessment
from May to September, but peaks in June and July (O’Brien (Adams et al., 2015).
et al., 1993). They are fully recruited by their third summer at age The Northeast Fisheries Science Center (NEFSC) conducts spring
2 (DuPaul and McEachran, 1973). and fall bottom trawl surveys along the northeastern continental
Historically, butterfish catch peaked in 1973 at 40 000 mt, pri- shelf of the United States (Politis et al., 2014). One of the concerns
marily because of foreign fleets targeting longfin squid raised in the most recent butterfish stock assessment was conflicting
(Doryteuthis pealeii) in offshore areas (Adams et al., 2015). survey trends: the spring series has generally been increasing over
Butterfish catch declined sharply following the implementation of time, while the fall series has been decreasing. Although the spring

Published by Oxford University Press on behalf of International Council for the

Exploration of the Sea 2016. This work is written by a US Government employee
and is in the public domain in the United States.
Butterfish spatial distribution 171

Downloaded from https://academic.oup.com/icesjms/article/74/1/170/2669550 by guest on 20 May 2025

Figure 1. Northeast Fisheries Science Center bottom trawl survey strata from Cape Hatteras, NC, USA to Nova Scotia, Canada. Strata used in
this study include the butterfish stock assessment offshore strata (cross hatch), as well as Gulf of Maine and outer Georges Bank strata (white).

series tracked cohorts more clearly through the age structure, but- survey data (Murawski and Mountain, 1990). However, there was
terfish are more widely distributed throughout the survey area dur- a significant effect of abundance when spring NEFSC survey data
ing the fall. Thus, fall survey trends are thought to more accurately were restricted to the Mid-Atlantic Bight (Figure 1), 1980–1989
represent patterns in overall abundance. Accordingly, only the fall (Mountain and Murawski, 1992). Thus, this secondary objective
survey data were used in the assessment. Research into the spatio- serves as a re-evaluation of environmental and density-dependent
temporal distribution of butterfish may provide insights into these effects on the spatial distribution of butterfish.
divergent trends (Adams et al., 2015).
The primary objective of this study was to quantify the spatio- Material and methods
temporal distribution of butterfish by age and season, from 1982 Data sources
to 2013. The most recent butterfish stock assessment (Adams The NEFSC has conducted spring and fall bottom trawl surveys on
et al., 2015) used a statistical catch-at-age model that relies in the continental shelf of the Northeast United States since 1968 and
part on NEFSC survey abundance indices and age composition. 1963, respectively. Butterfish otoliths were first collected in the
Thus, NEFSC survey-based spatial indicators for butterfish were NEFSC survey in 1982. Thus, data used in this analysis were from
calculated at age to inform future stock assessments. Spatial indi- 1982 to 2013. Exact survey dates are given in Supplementary Table
cators, such as the centre of gravity (CG), can be used to detect S1. The survey employs a random stratified design. Strata are defined
changes over time in the distribution of fish stocks (Woillez et al., primarily by depth, and the number of stations allocated to each
2007). The CG has been used in fisheries for several decades, stratum is proportional to stratum area. Sampling originally oc-
where it is also referred to as the centroid of distribution or centre curred at depths between 27 and 366 m, but shallower strata were
of mass (e.g. Koslow et al., 1985; Heath and MacLachlan, 1987; added in 1972 and 1979. In spring 2009, the survey vessel FRV
Murawski and Finn, 1988; Kendall and Picquelle, 1989). Albatross IV (AIV) was replaced by the FSV Henry B. Bigelow (HBB).
A secondary objective of this study was to examine environmen- These and other changes in gear and protocols over the course of the
tal (i.e. temperature and salinity) and density-dependent effects on time-series are documented in Johnston and Sosebee (2014).
the spatial distribution of butterfish. The CG has also been used to Because of the deeper draft of the HBB only strata with
link changes in fish distribution to climate change (e.g. Nye et al., depths >18 m have been surveyed since 2009 (Johnston and
2009). In the case of butterfish, there was no effect of abundance, Sosebee, 2014). During the most recent stock assessment for but-
bottom temperature or surface temperature on the weighted mean terfish (Adams et al., 2015) these strata were referred to as the off-
latitude in the spring (1968–1990) or fall (1967–1989) NEFSC shore strata. To maintain the same footprint over the course of
172 C. F. Adams

the time-series, a choice had to be made between including the sampling, spatial indicators are weighted with an area of influence
shallow strata and ending the time-series in 2008, or including (Bez et al., 1997; Woillez et al., 2007, 2009). Given the random
data from 2009 forward and restricting the footprint to the off- stratified survey design (as opposed to a grid), a Dirichlet tessella-
shore strata. Given the observed changes in the distribution of tion (Legendre and Legendre, 1998) was used as a non-subjective
many species in response to climate change (e.g. Nye et al., 2009), method to calculate areas of influence, with areas along the edge
the latter was chosen so as to incorporate the most recent avail- of the study area clipped to the boundary of the strata. Prior to
able data. Additionally, several of the offshore strata that were not analysis, the CG of sample locations (unweighted by zi) was cal-
sampled consistently throughout the time-series were omitted. culated to verify that changes in the CG over time were not be-
To investigate possible northward shifts in distribution, Gulf of cause of changes in sampling design (Woillez et al., 2009).
Maine strata, as well as several outer Georges Bank strata, were Abundance weighted mean depth was also calculated with (1)
added to provide more reliable estimates of the distribution cen- (Faraj and Bez, 2007).
tres over time (Brown et al., 2011). Assessment offshore strata
and the strata used in this spatial analysis are listed in Inertia
Supplementary Table S2 and shown in Figure 1.

Downloaded from https://academic.oup.com/icesjms/article/74/1/170/2669550 by guest on 20 May 2025

The inertia (variance) describes how dispersed the population
Age determination methods for butterfish are documented in
is around its CG:
Dery (1988). Age–length keys from each cruise were used to
transform the length frequencies observed at each trawl station Xn
ðx 
i¼1Xi
CGÞ2 wi zi
into age frequencies. Butterfish abundance at each station was I¼ n (2)
disaggregated into age 0 to 4þ for the fall, and age 1 to 4þ for the i¼1
wi zi
spring. Data for 2009–2013 were converted to AIV units using the
length-based calibration in Miller (2013) to account for the afore- I can be decomposed into two orthogonal axes describing the
mentioned vessel and gear changes. maximum and the minimum components of the inertia. The
Conductivity, temperature and depth data are collected at all square root of I for a given axis gives the standard deviation of
NEFSC trawl stations (Politis et al., 2014). However, salinity data the respective axis. As I has units of square kilometres, axes of in-
are currently only available in the NEFSC survey database going ertia are plotted in CG maps as the standard deviation, which has
back to fall 1997, and most 2008 data are also not available. units of kilometres. After back transformation to longitude and
Additionally, in spring 2005 salinity measurements were not re- latitude, the axes may no longer appear orthogonal in CG maps
corded at > 5% of stations. Altogether this reduced the number (Bez, 2007; Faraj and Bez, 2007).
of years with sufficient hydrographic data to n ¼ 14 for the spring,
and n ¼ 16 for the fall. Thus, the environmental analysis was Positive area
restricted to 1998–2013 for the spring (minus 2005 and 2008),
The positive area (PA) is the area (in square kilometres) occupied
and 1997–2013 for the fall (minus 2008).
by fish abundances greater than zero:

Spatial indicators X
n

The CG characterizes one property of the spatial distribution of a PA ¼ wi ½zi > 0 (3)
i¼1
fish population. This and other properties of the spatial distribu-
tion of a fish population have been formalized into a suite of spa-
tial indicators related to transitive geostatistics by Bez et al.
(1997) and Woillez et al. (2007, 2009). In this section it is only Intra-season analysis
described how each indicator is calculated in practice. As a preliminary, basic age-specific differences in spatial distribu-
tion within each of the two seasons were characterized. This was
done with a Kruskal–Wallis test for each spatial indicator, fol-
Geographical referencing lowed by a non-parametric multiple comparison test (Siegel and
Distances between points must be computed in a Euclidean refer- Castellan, 1988). For both tests, significance was set at a ¼ 0.05.
ence system (Bez, 2007). This was done by setting the minimum Age 4þ butterfish were omitted from this analysis as there was a
longitude and latitude of the strata used in this analysis number of years in the time-series when this age class was not ob-
(75 480 W, 35 090 N) as (0, 0) and converting all coordinates to served (n ¼ 10 in the spring; n ¼ 21 in the fall).
kilometres according to Rivoirard et al. (2000). The cosine of the
midpoint latitude (39 490 N) was used to convert longitude.
Geographically referenced longitude and latitude are hereafter re- Spatial distribution over time
ferred to as the X- and Y-components of the CG. To quantify the spatial distribution of butterfish over time, each
of the spatial indicators for each age class within a season was re-
gressed as a function of year. A Durbin–Watson test was used to
Centre of gravity check each linear model for serial correlation and, if present, a
The CG is the mean location of the surveyed population: first-order autoregressive model was fit instead (Neter et al.,
Xn 1996).
xi wi zi
CG ¼ Xi¼1
n (1)
i¼1
wi zi Environmental and density-dependent effects
To examine environmental and density-dependent effects on the
where xi is location (XCG or YCG), wi is the area of influence, spatial distribution of butterfish, each of the spatial indicators for
and zi is the number of butterfish. In the case of irregular each age class within a season was fit to a multiple linear regression
Butterfish spatial distribution 173

model. Candidate predictor variables were: stratified mean number Spatial indicators also showed interannual variation in the fall
per tow from the survey (hereafter referred to as stratified mean for all age classes (Figure 4). There was no change in the CG over
number); bottom temperature, surface temperature, bottom salin- the course of this time-series either. In this case, however, there
ity, and surface salinity. Note that all hydrographic parameters was a significant increase in inertia for ages 1 and 2 butterfish.
were also calculated as the stratified mean. Stratified mean number There was also a significant decrease in depth for the same two
was log transformed for analysis. Prior to multiple linear regres- age classes. Finally, there was no change in area occupancy over
sions, a variance inflation factor analysis (Neter et al., 1996) was the course of the fall time series.
used to detect collinearity among the candidate predictors. A con- A simple post hoc analysis was done to examine whether age-
servative cut-off of 5 was used (e.g. Puerta et al., 2014). A backward specific effects observed in the fall would have been masked using
elimination procedure (Neter et al., 1996) was then used to reduce the total number of fish. As noted above, significant changes for
each model to predictor variables that were significant at the level inertia and depth over time were observed for ages 1–2, but not
of a ¼ 0.05. Residuals were tested for autocorrelation with a age 0 butterfish. Thus, regressions were also run for these two
Durbin–Watson test, and, if present, a first-order autoregressive cases using the sum of all ages. There was no change in inertia
model was fit instead (Neter et al., 1996). over the course of the fall time series (b ¼ 468.70, t(30) ¼ 1.42,

Downloaded from https://academic.oup.com/icesjms/article/74/1/170/2669550 by guest on 20 May 2025

p ¼ 0.166). This indicates that the significant slopes for ages 1 and
Software 2 shown in Figure 4 would have been masked. Conversely, there
Several R (R Core Team, 2015) packages were used in this analy- was a significant decrease in depth for the total number of butter-
sis: spatial indicators were calculated in RGeostats (Renard et al., fish (b ¼ –0.51, t(30) p ¼ 0.032), when this was actually only true
2014); Dirichlet tessellae were calculated in spatstat (Baddeley for ages 1 and 2 (Figure 4).
and Turner, 2005); non-parametric multiple comparison tests
were done in pgirmess (Giraudoux, 2014); and variance inflation
factor analysis was done in usdm (Naimi, 2015). Autoregressive Environmental and density-dependent effects
models were calculated in PROC AUTOREG (SAS Institute, Inc., There were clear environmental relationships with the spatial dis-
Cary, NC, USA). tribution of butterfish in the spring (Table 1). There was a highly
significant correlation between surface temperature and area oc-
Results cupancy for all age classes. There was also a correlation between
Intra-season analysis surface temperature and the YCG for age 1 butterfish. The nega-
The spatial distribution of butterfish varied by age class in the tive relationship between surface salinity and depth for age 2 but-
spring. There was a significant difference between age classes for terfish was barely significant (p ¼ 0.047) and thus should be
XCG (H ¼ 25.73, d.f. ¼ 2, p < 0.001), YCG (H ¼ 20.68, d.f. ¼ 2, viewed with caution. There were no significant density-
p < 0.001), depth (H ¼ 29.10, d.f. ¼ 2, p < 0.001) and PA dependent relationships.
(H ¼ 35.71, d.f. ¼ 2, p < 0.001). Multiple comparison tests re- Environmental relationships with the spatial distribution of
vealed that ages 2 and 3 butterfish were significantly farther butterfish in the fall were more complicated (Table 1). For age 0
northeast and deeper than age 1 butterfish (Figure 2). These age- recruits, there was a highly significant correlation between bot-
specific differences in distribution can also be visualized by tog- tom temperature and the XCG; and a correlation between surface
gling through the CG maps in the online supplementary material salinity and the YCG. Area occupancy of recruits was positively
(Supplementary Figure S1). Multiple comparisons tests also correlated with bottom temperature and negatively correlated
showed that ages 1 and 2 butterfish had a significantly larger PA with surface temperature. The YCG of age 1 butterfish was posi-
than age 3 butterfish (Figure 2). tively correlated with bottom temperature. Depth of age 1 butter-
The spatial distribution of butterfish also varied by age class in fish was positively correlated with bottom temperature and
the fall. There was a significant difference between age classes for negatively correlated with surface temperature. For ages 2 and 3
XCG (H ¼ 27.30, d.f. ¼ 3, p < 0.001), YCG (H ¼ 15.04, d.f. ¼ 3, butterfish there was a positive relationship between bottom tem-
p ¼ 0.002), depth (H ¼ 19.52, d.f. ¼ 3, p < 0.001) and PA perature and depth. There was also a correlation between the
(H ¼ 80.37, d.f. ¼ 3, p < 0.001). Multiple comparison tests for abundance of age 3 butterfish and area occupancy. The relation-
XCG revealed that ages 1–3 were significantly farther east than age ship between the stratified mean number of age 4þ butterfish and
0 butterfish, while age 3 butterfish were significantly farther north the YCG was barely significant (p ¼ 0.047) and may be an artefact
than ages 0 and 1 butterfish (Figure 2). In this case, age-specific dif- of the low sample size (n ¼ 7).
ferences are best observed in the maps (Supplementary Figure S1)
by noting that the CGs for age 0s are generally more inshore.
Multiple comparison tests found that ages 1–3 were significantly Discussion
deeper than age 0 butterfish (Figure 2). Finally, age 0 butterfish The primary objective of this study was to quantify the spatio-
had a significantly larger PA than ages 1–3, while ages 1 and 2 had temporal distribution of butterfish by age and season, from 1982
a significantly larger PA than age 3 butterfish (Figure 2). to 2013. The intra-season analysis revealed age-specific differences
for all spatial indicators of butterfish distribution in both seasons.
Spatial distribution over time The clearest signal observed during the time-series analysis was
Spatial indicators showed interannual variation in the spring for increased area occupancy for ages 1–3 in the spring. The second-
all age classes (Figure 3). There was a significant increase in area ary objective of this study was to examine environmental and
occupancy for ages 1, 2, and 3 butterfish over the course of the density-dependent effects on the spatial distribution of butterfish.
time-series. However, there was no change in the CG, inertia or This revealed a highly significant relationship between the in-
depth for any age class in the spring. creased spring area occupancy and surface temperature.
174 C. F. Adams

Downloaded from https://academic.oup.com/icesjms/article/74/1/170/2669550 by guest on 20 May 2025

Figure 2. Notched box plots showing medians of selected spatial indicators for spring ages 1–3 and fall ages 0–3 butterfish. Hinges are the
first (Q1) and third (Q3) quartiles. Whiskers are within 1.5 times the range between Q1 and Q3. Outliers are shown as circles. Within each sea-
son, black boxes are significantly different from white boxes; grey boxes (i.e. PA for fall ages 1 and 2) are significantly different from black and
white boxes; while light grey boxes (i.e. latitude for fall age 2) are not significantly different from other boxes. Significance level for all multiple
comparisons was a ¼ 0.05. Note that Kruskal–Wallis tests were done on XCG and YCG, not the back-transformed longitude and latitude,
which are shown here to aid interpretation.

Intra-season analysis deeper than age 0s. This more complicated spatial segregation
This study revealed age-specific differences in the CG of butter- may be because the water column is warm and stratified during
fish from 1982 to 2013. In the spring, ages 2 and 3 were found the fall, and butterfish associations with fronts are weak or ab-
farther northeast and deeper than age 1 butterfish. This is likely sent (Manderson et al., 2011). Area occupancy decreased with
a response to upwelling conditions, as butterfish are associated age in both spring and fall. In both seasons, the PA was signifi-
with fronts on the outer continental shelf during winter and cantly less for age 3 butterfish than younger age classes. Woillez
early spring (Manderson et al., 2011). In the fall, age 3 butterfish et al. (2007) found that the PA for European hake (Merluccius
were found farther northeast and deeper than ages 0 and 1, merluccius) in the Bay of Biscay was relatively stable until age 3,
while ages 1 and 2 were found farther east (XCG only) and and then dropped for ages 4 and 5þ.
Butterfish spatial distribution 175

Downloaded from https://academic.oup.com/icesjms/article/74/1/170/2669550 by guest on 20 May 2025

Figure 3. Spring time series of spatial indicators for butterfish ages 1 to 4þ. Solid lines indicate a significant linear fit at the level of a ¼ 0.05. Note
that linear models were tested on XCG and YCG, not the back-transformed longitude and latitude, which are shown here to aid interpretation.

Spatial distribution over time time. An increase in inertia with age has also been observed in
The finding of no northward movement of butterfish in spring or European hake (Woillez et al., 2007). The other change over the
fall over the course of the 32-year time-series is consistent with course of the fall time-series is that age 1–2 butterfish have occu-
another recent analysis. Using an alongshelf measure, Walsh et al. pied shallower habitat. Walsh et al. (2015) also found that adult
(2015) also found no poleward movement of adult butterfish in butterfish were shallower in the fall during the period 1999–2008
the spring or fall during the period 1999–2008 when compared when compared with 1977–1987.
with 1977–1987. However, they did report that adult butterfish The importance of using age-specific indices was illustrated
were found more inshore (e.g. cross shelf) in both spring and fall with the post hoc fall time-series analysis. In this case, age-
in the Mid-Atlantic Bight during 1999–2008. A comparable result specific changes in spatiotemporal distribution would have been
would have been observed in the present study as a decrease in masked or misrepresented if the data were not disaggregated by
the XCG over time (i.e. westward movement). Although none of age. The increased dispersion of ages 1 and 2 butterfish over the
the slopes was significant, this was the general trend for fall ages course of the fall time-series would have been masked if only the
1–3 butterfish (Figure 4). total number of butterfish was analysed. Conversely, it would
There were two changes in spatial distribution over the course have been erroneously reported that butterfish have occupied
of the fall time-series. The increase in inertia for fall ages 1–2 in- shallower depths in the fall, when this was actually only true for
dicates that these age classes have become more scattered over ages 1 and 2.
176 C. F. Adams

Downloaded from https://academic.oup.com/icesjms/article/74/1/170/2669550 by guest on 20 May 2025

Figure 4. Fall time series of spatial indicators for butterfish ages 0 to 4þ. Solid lines indicate a significant linear fit at the level of a ¼ 0.05. Note
that linear models were tested on XCG and YCG, not the back-transformed longitude and latitude, which are shown here to aid interpretation.

Environmental and density-dependent effects 2004 illustrate the typical spring distribution of butterfish along the
The highly significant relationship between surface temperature and shelf edge; whereas in 2012 the north and shoalward distribution
area occupancy was the primary environmental effect in the spring. predicted by Murawski and Mountain (1990) is observed. Although
Given the lack of northward movement during this time, this sug- average sea surface temperatures on the northeastern continental
gests that, in the spring, the response of butterfish to shelf warming shelf of the USA in 2012 were the highest in the 160-year record
will be a range expansion, with the CG remaining in the Southern (Fratantoni et al., 2013), this example illustrates a possible range ex-
New England/Mid-Atlantic Bight region. Murawski and Mountain pansion for butterfish under a shelf warming scenario.
(1990) hypothesized that, if shelf warming results primarily in The only other significant effect was that surface temperature
warmer fall and winter conditions, then species such as butterfish had a positive correlation with the YCG for age 1 butterfish.
should be found north and shoalward of their present winter and Previous analyses found no relationship between surface tempera-
early spring distributions, with perhaps some northward extension ture and the weighted mean latitude of total number of butterfish
of their summer range. The present analysis supports their hypothe- (Murawski and Mountain, 1990; Mountain and Murawski,
sis. Positive area maps comparing warm vs. cold years help to visu- 1992). These conflicting results may be because of the use of
alize what such a range expansion might look like (Figure 5). The more recent data and/or because an age-specific effect was re-
stratified mean surface temperature (for the strata used in this anal- vealed in the present analysis. Future studies with a longer time-
ysis) was 4.3 and 8.2  C in 2004 and 2012, respectively. PA tiles for series of environmental data may resolve this issue.
Butterfish spatial distribution 177

Table 1. Slope (b), standard error, t-value, degrees of freedom and p-value for multiple linear regressions of butterfish spatial indicators as a
function of abundance and hydrographic parameters.
Spring Fall

Indicator Predictor b s.e. t-value d.f. p-value Indicator Predictor b s.e. t-value d.f. p-value
Age 0

XCG bottemp 85.92 18.06 4.76 14 <0.001

YCG surfsalin 190.61 45.41 4.20 14 0.001
Depth
PA bottemp 18365.54 7181.13 2.56 13 0.024
surftemp 15863.07 6687.19 2.37 13 0.034
Age 1 Age 1

XCG XCG

Downloaded from https://academic.oup.com/icesjms/article/74/1/170/2669550 by guest on 20 May 2025

YCG surftemp 73.49 25.90 2.84 12 0.015 YCG bottemp 65.96 27.16 2.43 14 0.029
Depth Depth bottemp 6.52 2.30 2.84 13 0.014
surftemp 7.19 2.14 3.36 13 0.005
PA surftemp 21883.31 3899.04 5.61 12 <0.001 PA

Age 2 Age 2

XCG XCG
YCG YCG
Depth surfsalin 34.07 15.43 2.21 12 0.047 Depth bottemp 9.75 3.54 2.76 14 0.015
PA surftemp 17382.32 2572.99 6.76 12 <0.001 PA

Age 3 Age 3

XCG XCG
YCG YCG
Depth Depth bottemp 15.52 5.02 3.09 11 0.010
PA surftemp 15526.80 2207.73 7.03 12 <0.001 PA logage3 21999.00 5391.00 4.08 11 0.002

Age 41 Age 41

XCG XCG
YCG YCG logage4 109.93 42.02 2.62 5 0.047
Depth Depth
PA surftemp 12853.10 2867.98 4.48 8 0.002 PA
Spatial indicators and associated units are: geographically referenced longitude and latitude of the centre of gravity (XCG and YCG, respectively; km), depth
(m) and positive area (PA; km2). Predictor variables are: stratified mean bottom temperature (bottemp), surface temperature (surftemp), and surface salinity
(surfsalin); as well as log transformed stratified mean number of butterfish per tow for ages 3 and 4þ (logage3 and logage4, respectively). Predictors were se-
lected using a backward elimination procedure with significance level set at a ¼ 0.05. Only significant predictors are shown.

The spatial distribution of age 0 recruits appears to be driven temperature and depth for age 1–3 butterfish. Conflicting trends
by environmental conditions. There was a positive correlation be- between bottom and surface temperature and the depth of age 1
tween hydrographic parameters and the CG. The seemingly con- butterfish are less readily resolved. This serves as a reminder that
flicting correlations of bottom and surface temperature with area biological factors, such as predators and prey, can affect the verti-
occupancy can be explained as follows. Bottom temperature used cal distribution of fish (e.g. Murawski and Finn, 1988).
in the multiple linear regression analysis ranged from 10.1 to The only clear density-dependent effect was between the abun-
12.9  C, whereas surface temperatures ranged from 15.3 to dance of fall age 3 butterfish and area occupancy. Lange and
18.3  C. A previous analysis showed that a histogram of the pro- Waring (1992) reported a negative relationship between the pro-
portion of positive tows for butterfish using NEFSC fall survey portion of zero tows and abundance of large (>12 cm) butterfish
data, 1963–1997, peaked at 12  C (Cross et al., 1999). Thus, the in the fall NEFSC survey data, 1976–1985. Frisk et al. (2011)
negative relationship with surface temperature in the present found a positive relationship between abundance and area occu-
analysis would correspond to the right tail of the histogram in the pancy of butterfish on Georges Bank, using fall NEFSC survey
previous analysis. This interpretation is consistent with Colton data, 1963–2006. Further research is needed to determine whether
(1972), who reported a contraction of the northern and eastern the findings in these previous studies are driven by age 3
limits of butterfish during a downward trend in temperatures butterfish.
during the period 1953–1967. Ideally some measure of fishing pressure would have been in-
Environmental effects on older age classes in the fall appear to cluded in the multiple linear regressions as a predictor variable.
be restricted to bottom temperature. There was a positive correla- Landings were not used in this analysis because butterfish have
tion between bottom temperature and the YCG for age 1 butter- been caught primarily as bycatch in the directed longfin squid
fish; and an increasingly strong relationship between bottom fishery over the last decade (Adams et al., 2015). Given the
178 C. F. Adams

Downloaded from https://academic.oup.com/icesjms/article/74/1/170/2669550 by guest on 20 May 2025

Figure 5. Positive area maps for age 1 butterfish in spring 2004 and spring 2012, when stratified mean surface temperatures were 4.3 and
8.2  C, respectively. Black polygons indicate positive tows.

resumption of a directed fishery in 2013, future analysis could in- distribution analyses can be used to verify the spatial equilibrium
clude an exploitation index. assumption for the calculation of biological reference points.

Supplementary data
Conclusions Supplementary material is available at the ICESJMS online
This study can inform future butterfish assessments in several version of the manuscript.
ways. With respect to the fall NEFSC survey index (which was
used in the most recent assessment), there has been no change in
the CG or PA through 2013. This indicates that the assumption Acknowledgements
of constant habitat suitable for production is being met within I thank Josh Dayton for expediting the aging of the fall 2013 oto-
the boundaries of the strata used in this analysis. This conclusion liths. Lisa Hendrickson and Gary Shepherd provided helpful dis-
should be verified using the assessment strata. In terms of the di- cussions throughout the analysis. John Manderson reviewed an
verging spring and fall indices, increased area occupancy in the early draft of the manuscript. Comments from the editor and
spring suggests that availability to the spring survey is increasing. four anonymous reviewers greatly improved the final version of
Decreasing coefficients of variation for the NEFSC spring survey this manuscript.
in recent years (Adams et al., 2015) support this interpretation.
In the spring NEFSC survey index (which was not used in the
most recent assessment), increased area occupancy suggests that
References
butterfish should be more broadly distributed over the shelf as Adams, C. F., Miller, T. J., Manderson, J. P., Richardson, D. E., and
Smith, B. E. 2015. Butterfish 2014 stock assessment. NEFSC
shelf warming continues. Under this scenario, the spring index Reference Document 15-06. 110 pp. http://www.nefsc.noaa.gov/
would be a candidate to be included in the assessment model. publications/crd/crd1506/ (last accessed 31 August 2015).
Finally, potential range expansion into the Gulf of Maine (Figure Baddeley, A., and Turner, R. 2005. spatstat: an R package for analyz-
5) illustrates that the inclusion of these strata should be re- ing spatial point patterns. Journal of Statistical Software, 12: 1–42.
examined. Bez, N. 2007. Transitive geostatistics and statistics per individual: a
This study has demonstrated the importance of an age-based relevant framework for assessing resources with diffuse limits.
and seasonal spatial distribution analysis. Increased area occu- Journal de la Société Française de Statistique, 148: 53–75.
pancy was observed only in the spring, whereas changes in depth Bez, N., Rivoirard, J., Guiblin, P., and Walsh, M. 1997. Covariogram
and dispersion in the fall were restricted to specific age classes. and related tools for structural analysis of fish survey data. In
Increased area occupancy in the spring was correlated with sur- Geostatistics Wollongong ’96, Vol. 2, pp. 1316–1327. Ed. by E. Y.
Baafi and N. A. Schofield. Kluwer Academic Publishers,
face temperature, demonstrating that responses to climate change Dordrecht. 665 pp.
may be manifested as range expansions, rather than poleward Brown, C. J., Schoeman, D. S., Sydeman, W. J., Brander, K., Buckley,
movement of the CG. Spatial distribution analyses can inform L. B., Burrows, M., Duarte, C. M., et al. 2011. Quantitative
stock assessments by providing insights into diverging survey in- approaches in climate change ecology. Global Change Biology, 17:
dices and availability to surveys in general. Furthermore, spatial 3697–3713.
Butterfish spatial distribution 179

Collette, B. B., and Klein-MacPhee, G. 2002. Bigelow and Schroeder’s to their environment, 1980–1989. ICES Marine Science
Fishes of the Gulf of Maine, 3rd edn. Washington, DC: Symposium, 195: 424–432.
Smithsonian Institution Press. 748 pp. Murawski, S. A., and Finn, J. T. 1988. Biological bases for mixed-
Colton, J. B. Jr., 1972. Temperature trends and the distribution of species fisheries: species co-distribution in relation to environ-
groundfish in continental shelf waters, Nova Scotia to Long mental and biotic variables. Canadian Journal of Fisheries and
Island. Fishery Bulletin, 70: 637–657. Aquatic Sciences, 45: 1720–1735.
Cross, J. N., Zetlin, C. A., Berrien, P. L., Johnson, D. L., and McBride, Murawski, S. A., and Mountain, D. G. 1990. Climate change and ma-
C. 1999. Essential fish habitat source document: butterfish, rine fish distributions: analogies from seasonal and annual vari-
Peprilus triacanthus, life history and habitat characteristics. NOAA ability. ICES Document CM 1990/C: 36. 24 pp.
Technical Memorandum NMFS-NE-145. 42 pp. http://www.
Murawski, S. A., and Waring, G. T. 1979. A population assessment of
nefsc.noaa.gov/publications/tm/tm145/tm145.pdf (last accessed
butterfish, Peprilus triacanthus, in the Northwestern Atlantic
15 February 2013).
Ocean. Transactions of the American Fisheries Society, 108:
Dery, L. M. 1988. Butterfish, Peprilus triacanthus. In Age 427–539.
Determination Methods for Northwest Atlantic Species, pp. 85–
92. Ed. by J. Penttila and L. M. Dery. NOAA Technical Report Naimi, B. 2015. usdm: uncertainty analysis for species distribution
models. R package version 1.1–15. https://cran.r-project.org/web/

Downloaded from https://academic.oup.com/icesjms/article/74/1/170/2669550 by guest on 20 May 2025

NMFS 72. 132 pp. http://www.nefsc.noaa.gov/publications/clas
sics/penttila1988/penttila1988.pdf (last accessed 12 August 2014). packages/usdm/index.html (last accessed 16 March 2016).
DuPaul, W. D., and McEachran, J. D. 1973. Age and growth of the NMFS (National Marine Fisheries Service) 2015. Fisheries of the
butterfish, Peprilus triacanthus, in the lower York River. Northeastern United States; Atlantic mackerel, squid, and butter-
Chesapeake Science, 14: 205–207. fish fisheries: specifications and management measures. Federal
Faraj, A., and Bez, N. 2007. Spatial considerations for the Dakhla Register, 80: 14870–14875.
stock of Octopus vulgaris: indicators, patterns, and fisheries inter- Neter, J., Kutner, M. H., Nachtsheim, C. J., and Wasserman, W.
actions. ICES Journal of Marine Science, 64: 1820–1828. 1996. Applied Linear Statistical Models, 4th edn. McGraw-Hill,
Fratantoni, P. S., Holzwarth-Davis, T., Bascu~ nan, C., and Taylor, M. Boston. 1408 pp.
H. 2013. Description of the 2012 oceanographic conditions on the Nye, J. A., Link, J. S., Hare, J. A., and Overholtz, W. J. 2009.
Northeast U.S. continental shelf. NEFSC Reference Document 13- Changing spatial distribution of fish stocks in relation to climate
26. 40 pp. http://www.nefsc.noaa.gov/publications/crd/crd1326/ and population size on the Northeast United States continental
(last accessed 10 June 2015). shelf. Marine Ecology Progress Series, 393: 111–129.
Frisk, M. G., Duplisea, D. E., and Trenkel, V. M. 2011. Exploring the O’Brien, L., Burnett, J., and Mayo, R. K. 1993. Maturation of nine-
abundance–occupancy relationships for the Georges Bank finfish teen species of finfish off the northeast coast of the United States,
and shellfish community from 1963 to 2006. Ecological 1985–1990. NOAA Technical Report NMFS 113. 66 pp. http://
Applications, 21: 227–240. spo.nmfs.noaa.gov/tr113.pdf (last accessed 26 April 2016).
Giraudoux, P. 2014. pgirmess: data analysis in ecology. R package Politis, P. J., Galbraith, J. K., Kostovick, P., and Brown, R. W. 2014.
version 1.5.9. https://cran.r-project.org/web/packages/pgirmess/in Northeast Fisheries Science Center bottom trawl survey protocols
dex.html (last accessed 8 December 2014). for the NOAA Ship Henry B. Bigelow. NEFSC Reference
Heath, M. R., and MacLachlan, P. 1987. Dispersion and mortality of Document 14-06. 138 pp. http://www.nefsc.noaa.gov/publica
yolk-sac herring (Clupea harengus L.) larvae from a spawning tions/crd/crd1406/index.html (last accessed 6 January 2015).
ground to the west of the Outer Hebrides. Journal of Plankton Puerta, P., Hidalgo, M., Gonzalez, M., Esteban, A., and Quetglas, A.
Research, 9: 613–630. 2014. Role of hydro-climatic and demographic processes on the
Johnston, R., and Sosebee, K. 2014. History of the United States bot- spatio-temporal distribution of cephalopods in the western
tom trawl surveys, NAFO subareas 4–7. Northwest Atlantic Mediterranean. Marine Ecology Progress Series, 514: 105–118.
Fisheries Organization SCR Document 14-024.
R Core Team. 2015. R: a language and environment for statistical
Kendall, A. W. Jr., and Picquelle, S. J. 1989. Egg and larval distribu- computing. R Foundation for Statistical Computing, Vienna,
tions of walleye pollock Theragra chalcogramma in Shelikof Strait, Austria. https://www.R-project.org/ (last accessed 8 December
Gulf of Alaska. Fishery Bulletin, 88: 133–154. 2014).
Koslow, J. A., Brault, S., Dugas, J., and Page, F. 1985. Anatomy of an Renard, D., Bez, N., Desassis, N., Beucher, H., and Ors, F. 2014.
apparent year-class failure: the early life history of the 1983 RGeostats: geostatistical package. R package version 10.0.8. http://
Browns Bank haddock Melanogrammus aeglefinus. Transactions rgeostats.free.fr/ (last accessed 8 December 2014).
of the American Fisheries Society, 114: 478–489.
Rivoirard, J., Simmonds, J., Foote, K. G., Fernandes, P., and Bez, N.
Lange, A. M., and Waring, G. T. 1992. Fishery interactions between
2000. Geostatistics for Estimating Fish Abundance. Blackwell
long-finned squid (Loligo pealei) and butterfish (Peprilus triacan-
Science, Oxford. 206 pp.
thus) off the Northeast USA. Journal of Northwest Atlantic
Fishery Science, 12: 49–62. Siegel, S., and Castellan, N. J. Jr. 1988. Nonparametric Statistics for
Legendre, P., and Legendre, L. 1998. Numerical Ecology, 2nd edn. the Behavioral Sciences, 2nd edn. McGraw-Hill, Boston. 399 pp.
Elsevier, Amsterdam. 853 p. Walsh, H. J., Richardson, D. E., Marancik, K. E., and Hare, J. A.
Manderson, J., Palamara, L., Kohut, J., and Oliver, M. J. 2011. Ocean 2015. Long-term changes in the distributions of larval and adult
observatory data are useful for regional habitat modeling of spe- fish in the Northeast U. S. shelf ecosystem. PLoS ONE, 10:
cies with different vertical habitat preferences. Marine Ecology e0137382.
Progress Series, 438: 1–17. Woillez, M., Poulard, J.-C., Rivoirard, J., Petitgas, P., and Bez, N.
Miller, T. J. 2013. A comparison of hierarchical models for relative 2007. Indices for capturing spatial patterns and their evolution in
catch efficiency based on paired-gear data for US Northwest time, with application to European hake (Merluccius merluccius)
Atlantic fish stocks. Canadian Journal of Fisheries and Aquatic in the Bay of Biscay. ICES Journal of Marine Science, 64: 537–550.
Sciences, 70: 1306–1316. Woillez, M., Rivoirard, J., and Petitgas, P. 2009. Notes on survey-
Mountain, D. G., and Murawski, S. A. 1992. Variation in the distri- based spatial indicators for monitoring fish populations. Aquatic
bution of fish stocks on the northeast continental shelf in relation Living Resources, 22: 155–164.

Handling editor: Manuel Hidalgo