Sagarese et al.: Spatiotempora! overlap of Squalus acanthias and commercial fisheries within the northeast U.S. marine system 
105 
use of the program PERMANOVA, vers. 1.6 (former- 
ly called NPMANOVA; Anderson, 2005). This method 
tested for the multivariate null hypothesis of no rela- 
tionship between modes of fishing (i.e., fisheries versus 
survey) (Anderson, 2001). A multivariate test statistic 
analogous to Fisher’s E-ratio was calculated directly 
from a dissimilarity matrix with the P-value obtained 
by both permutation and Monte Carlo randomization 
(Anderson, 2001). 
For comparisons between the bottom trawl survey 
and each fishery, the mode of fishing was treated as 
a fixed factor, the dependent variables latitude and 
longitude defined centers of spiny dogfish abundance, 
and annual values served as observations. Significance 
was determined by 9999 permutations of the raw data 
and an a priori significance level of a=0.05. If the re- 
sult was statistically significant, a posteriori pairwise 
comparisons were conducted with 9999 permutations 
to determine which modes of fishing differed signifi- 
cantly (Anderson, 2001). All a posteriori significance 
levels (a=0.05) were adjusted through the use of the 
Bonferroni correction method (a a dj=0.0167) to reduce 
the potential for type-I errors during multiple compari- 
sons (Crawley, 2007). 
Spatial analyses 
Grid determination with semivariograms A comparable 
grid scheme of spatially identical grid cells enabled di- 
rect comparison and geostatistical modeling to quan- 
tify the spatiotemporal overlap between spiny dogfish 
distribution and commercial fisheries. Empirical semi- 
variograms (y[/z] ) estimated the range (a) or the asymp- 
totic distance beyond which samples were spatially in- 
dependent (Matheron, 1971). For each year, survey and 
fishery CPUE of spiny dogfish were log e transformed 
(logJCPUE >0]+0.05) to meet the normality require- 
ment for semivariogram modeling and to account for 
zero values. Semivariograms were fitted both annually 
and overall (i.e., all years combined). After multiple 
theoretical models (nugget, spherical, Gaussian, and 
exponential) were tested, the optimal model given the 
data was selected on the basis of the lowest Akaike’s 
information criterion (Webster and McBratney, 1989) 
and of model weights (Wagenmakers and Farrell, 
2004). All models were run in R software with the gstat 
package (Pebesma, 2004). Annual range estimates av- 
eraged across years were compared to range estimates 
obtained from all the data. Further details regarding 
semivariogram modeling are provided in the Appendix. 
Spatial overlap Seasonal spatial overlap was quanti- 
fied on an annual basis to examine how spiny dogfish 
distribution was related to both commercial fishery 
effort and catch throughout the NEFOP time series 
(1989-2009). First, station data for the NEFSC bot- 
tom trawl survey and each fishery were converted into 
rasters with the raster package (Hijmans et al., 2012) 
in R to summarize data points and interpolate catch 
(survey and fishery) in grid cells that were not directly 
sampled and, therefore, to enable estimation of spatial 
overlap outside the spatial domain. Grid cells for which 
CPUE >0 (i.e., positive catch of spiny dogfish by fish- 
ery) were identified and used in spatial overlap analy- 
ses concerning fishery catch. The magnitude of CPUE 
was not considered in any spatial analyses. 
We assessed the amount of direct spatial overlap 
(Brodeur et al., 2008) between spiny dogfish distribu- 
tion and 2 aspects of each commercial fishery: 1) effort, 
indicative of where a fishery fished (fleet presence), 
and 2) catch, indicative of spiny dogfish presence on 
the fishing grounds. The percentage of spatial overlap 
of spiny dogfish distribution with commercial fisher ef- 
fort (SOe) was calculated with the following equation: 
SO E (%)= ^ c ’ F -xl00, (3) 
a fe 
where N^c ,FE = the number of grid cells that contained 
both survey catch of spiny dogfish 
and commercial fishery effort; and 
Np e = the number of surveyed grid cells repre- 
senting areas that were fished. 
This metric described how each commercial fishery was 
operating in relation to distribution of spiny dogfish 
and served as a proxy of spiny dogfish availability to 
the OT and SGN fisheries. Low overlap indicated that 
fishing crews were infrequently encountering spiny 
dogfish (whether directly or as bycatch), indicating re- 
duced availability to the fishery. The footprint of the 
SOe metric was equivalent to the fishing grounds be- 
cause only fished areas contributed toward the denomi- 
nator (sites that were surveyed but not fished had no 
bearing on SOe). 
The percentage of spatial overlap of spiny dogfish 
distribution with commercial fishery catch (SOq) was 
calculated with an equation similar to Equation 3: 
SO c (%)= iVsc ’ FC xlOO, (4) 
iV FC 
where Nqq fq = the number of grid cells that contained 
both survey and fishery catch of 
spiny dogfish; and 
Npc = the number of surveyed grid cells where 
commercial fishing crews caught 
spiny dogfish. 
This metric represented the similarity between the 
commercial fishery and the bottom trawl survey in en- 
countering spiny dogfish. Here, low overlap indicated 
a spatial mismatch between the area where the fish- 
ery and the survey caught spiny dogfish. The footprint 
of the SOq was based on fishing grounds where spiny 
dogfish were encountered. As explained previously for 
SOe, sites that were surveyed but not fished had no 
bearing on SOq. 
In addition to estimation of direct spatial over- 
lap with station data, estimation of spatial overlap 
was done with interpolated survey and fishery catch. 
Semivariograms were used in conjunction with or- 
dinary kriging (Oliver and Webster, 1990; Reese and 
