French et al.: Strong relationship between catch of Hippoglossus hippoglossus and availability of habitat for juveniles 
117 
50°N 
45°N 
40°N 
70°W 65°W 60°W 55°W 50°W 45°W 
Figure 2 
Map of areas in the northwest Atlantic Ocean characterized as suitable habitat for juvenile Atlantic hali¬ 
but <Hippoglossus hippoglossus) and estimated by using maximum entropy species distribution modeling 
and data from trawl surveys conducted from 2001 to 2013. This habitat suitability layer is a composite 
output from 3 maximum entropy models for 3 regions in the northwest Atlantic Ocean for which the larg¬ 
est cell value was retained. Habitat suitability is represented on a scale of 0-1, with 0.5 representing 
the relative habitat suitability and probability of presence of Atlantic halibut where environmental con¬ 
ditions are typical for presence. Percentages represent the total percentage of the area surveyed within 
each North Atlantic Fisheries Organization (NAFO) division that had suitable habitat. Values for the 
NAFO divisions in U.S. waters (5Y, 5Zw, and 5Ze) were grouped together. 
2014; Wilson et al., 2016). There is also the assump¬ 
tion that there are low levels of emigration from these 
areas. This assumption is consistent with tagging stud¬ 
ies, which have reported that the majority of recap¬ 
tures occur in the area of initial tagging (McCracken, 
1958; Jensen and Wise, 1961; Neilson et al. 5 ; Stobo et 
al., 1988; Kanwit, 2007; den Heyer et al., 2012; Seitz et 
al., 2016). Finally, the use of landings data as a proxy 
for abundance has been extensively debated; here we 
use the spatial distribution of landings as a surrogate 
for the spatial distribution of the stock (e.g., Pauly et 
al., 2013). 
Results 
Model output was a spatial representation of relative 
habitat suitability for predominantly juvenile Atlantic 
halibut. Overall, 22.5% of the total surveyed area was 
characterized as suitable habitat for juvenile halibut 
(Fig. 2) (we will refer to this as “suitable habitat”). 
Cross-validation on 20 iterations of the model produced 
average AUC values of 0.90 for NS and U.S., 0.85 for 
NF, and 0.92 in the GLS (on a scale from 0 to 1, with 1 
being the best possible score) (Suppl. Fig. 1, A-C) (online 
only). When we compared the habitat suitability values 
with the test data, locations with presence of Atlantic 
halibut were far more frequently associated with areas 
with high habitat-suitability values and absence with 
low values; this result also promoted confidence in the 
model (Suppl. Fig. 1, G-I) (online only). 
The greatest proportional amount of suitable habi¬ 
tat fell within the divisions 4X and 4W, which are ad¬ 
jacent NAFO divisions that cover a large section of the 
Scotian Shelf (Table 2; Fig. 2). Divisions 3Ps (off the 
southern coast of Newfoundland) and 5Y, 5Ze, and 5Zw 
(Maine) also contained proportionally large amounts of 
suitable habitat (Table 2; Fig. 2). The majority of the 
suitable habitat in the United States was concentrated 
near shore and along Georges Banks, whereas in 3Ps, 
it was found along the edge of the Laurentian Chan¬ 
nel and toward the deeper slopes (Fig. 2). The remain¬ 
ing NAFO divisions had smaller, although noteworthy, 
amounts of suitable habitat shares (Table 2). 
The southern Grand Banks (Newfoundland) fall 
within NAFO divisions 3L, 3N, and 30 and possess 
rich fishing grounds, but unlike the Scotian Shelf 
(where suitable habitat was predicted to be widely 
available), suitable habitat in this region was located 
mostly toward the slope (Fig. 2). A noteworthy amount 
of suitable habitat (18,162 km 2 ) was outside Canada’s 
EEZ on the southern Grand Banks and the Flemish 
Cap (Table 3). International waters are sampled less 
frequently than areas within the EEZ and therefore 
there is more uncertainty around abundances in these 
regions. 
Variable contributions describe how well a change 
in a covariate value contributes to changes in habitat 
