French et al.: Strong relationship between catch of Hippoglossus hippoglossus and availability of habitat for juveniles 
115 
Table 1 
The 4 models of species distribution used to test different combinations of parameterization techniques 
used, in turn, to correct data biases in this study of the relationship between commercial catch of adult 
Atlantic halibut (Hippoglossus hippoglossus ) and availability of suitable habitat for juveniles in the 
northwest Atlantic Ocean. Correction parameters are those outlined by Fourcade et al. (2013). Cross- 
validation produced values for the area under the omission curve (AUC), which indicate a model’s 
proficiency in differentiating between presence and absence sites. 
Correction parameters 
Method 
AUC 
Default 
0.83 
Restricted background 
Select only background points that fall within 
the extent of the survey: use true absence locations 
0.81 
Restricted background 
+ bias file 
Select only background points that fall within the 
extent of the survey: use true absence locations + 
weight values by incorporating the raster layer that 
reflects the sampling effort or sampling probability 
0.80 
Restricted background 
+ split 
Select only background points that fall within the 
extent of the survey: use true absence locations + 
geographically splitting the data, compute the model 
separately for each area, then combine results 
0.89 (mean) 
subset the NAFO shapefile to include only the divisions 
that represent the extent of the Southern Grand Banks 
and Scotian Shelf stock. The NAFO Convention Area 
was further subdivided by 1) the Canadian EEZ, which 
is drawn 370 km (200 nautical mi) from shore to mark 
the extent of national jurisdiction over waters and the 
beginning of shared international resources, and 2) the 
Hague line, which delineates the border between wa¬ 
ters of Canada and the United States (Halliday and 
Pinhorn, 1990; Anderson, 1998) (Fig. 1). 
To measure adult abundance of Atlantic halibut, we 
used 3 data sets from the Atlantic halibut fishery. For 
recent values, we used commercial landings by divi¬ 
sion from 2010 through 2014 (DFO, seafisheries land¬ 
ings, available from website); these data reflect both 
data from directed fisheries and data on bycatch. We 
also used Butler and Coffen-Smout’s (2017) map of 
landings, by catch weight, to spatially represent the 
MARFIS data. This map shows a 5-year (2010-2014) 
composite of landings in kilograms per 2x2 minute 
lat.xlong. grid (Butler and Coffen-Smout, 2017). As a 
measure of historical adult abundance, we used his¬ 
torical fishery landings data gathered by McCracken 
(1958). He gathered information on landings from 
several governing agencies and was able to find suffi¬ 
cient location data to identify important areas (fishing 
grounds) in the Northwest Atlantic, and he expressed 
the landings as the annual percentage of shares of 
halibut landings per NAFO division. These historical 
landings (1953-1954) predate the intensive Atlantic 
cod (Gadus morhua) trawl fishery of the 1970s and 
1980s (Myers et al., 1996) and are considered to rep¬ 
resent a regulation-free fishery, when fishing crews 
were free to relocate operations as they pleased, and 
maximize their catch per unit of effort (Gillis et al., 
1993). Because the distribution of fishing vessels typi¬ 
cally achieved an ideal free distribution (Gillis et al., 
1993), the footprint of the unregulated fisheries is a 
good representation of historical spatial distribution of 
adult halibut. Supporting this notion, the abundance- 
based shares estimated by McCracken using unregu¬ 
lated fisheries data, along with his report of important 
and stable fishing grounds, were used to allocate fish¬ 
ing shares proportionally among NAFO divisions when 
the Canadian halibut fishery was first regulated in 
1988 (Neilson et al. 5 ). 
Analysis 
All statistical analyses were performed in RStudio, 
vers. 3.3.2 7 (RStudio, Inc., Boston, MA), and maps 
were produced in ArcMap, vers. 10.4.KESRI, Redlands, 
CA). We used functions available in the package dismo, 
vers. 1.1-4, in R, vers. 3.3.2 (R Core Team, 2016) to 
build 4 models of species distribution with maximum 
entropy. Each model tested a different combination 
of parameterization techniques that can help correct 
various types of sampling bias as outlined by Fourcade 
et al. 2013 (Table 1). Our models combined data from 
18 research surveys (Suppl. Table 1) (online only). Input 
consisted of the 1980 records of halibut presence, a 
random subset of 5000 records of absence, and raster 
7 Mention of trade names or commercial companies is for iden¬ 
tification purposes only and does not imply endorsement by 
the National Marine Fisheries Service, NOAA. 
