McElroy et al.: Comparison of length distributions from a longline survey and a trawl survey for 2 groundfish species 235 
[| Rough bottom 
[<<] Smooth bottom 
on ees 
Gulf of Maine 4 
24 
Figure 2 
A map of the strata of the bottom longline survey (LLS) (dashed black lines) and offshore strata of the bot- 
tom trawl survey (BTS) (thin gray lines, numbered in the center). The LLS strata include all of BTS strata 
26, 27, and 37 and portions of BTS strata 28, 29, and 36. The thick black line indicates the boundary between 
the exclusive economic zones of the United States and Canada. Both surveys are implemented by the NOAA 
Northeast Fisheries Science Center following a stratified random design, with stratification determined by 
depth and geographic region in the BTS and by depth, geographic region, and bottom type (rough or smooth) 
in the LLS. Sources for base map: Esri, Garmin, GEBCO, NOAA NGDC, and other contributors. 
included from LLS strata that were located on rough bot- 
tom where towing is difficult. Stations of the LLS on 
smooth bottom were omitted because of concerns about 
low catchability due to poor bait retention on soft bottom 
(McElroy et al., 2019). 
For each species, year, and season, we first calculated 
estimates of aggregate (i.e., all lengths combined) strat- 
ified mean abundance indices (for the BTS, see Cochran, 
1977; for the LLS, see McElroy et al., 2019). Second, 
we used the length frequencies from the raw data to 
derive estimates of stratified mean abundance indices at 
length (i.e., mean numbers at each length). The indices 
of abundance at length were then standardized to pro- 
duce estimates of the proportion of numbers at length 
as a representation of length distributions. The Bhat- 
tacharyya coefficient (BC) (Bhattacharyya, 1943) was 
calculated as a measure of overlap between the length— 
frequency distributions of the 2 surveys, for each species 
and season. This coefficient provides a robust measure 
of relative similarity in distribution ranging from 0.0, 
indicating no overlap, to 1.0, indicating identical distri- 
butions (Winner et al., 2018). 
Results 
Sampling coverage of the study area was comparable 
between the 2 gear types. The LLS data sets used in anal- 
ysis included data from 33-38 stations in each season 
in each year, 349 stations in total (Table 1). In 2017 and 
2018, modifications to the method of station allocation for 
the LLS resulted in increased sampling in rough-bottom 
habitat (McElroy et al., 2019). The BTS data sets used in 
analysis included at least 50 stations in the spring and 
fall for GOM Atlantic cod and white hake (Table 1) during 
2014-2017, for a total of 550 stations sampled. Because 
of operational difficulties, fewer stations were sampled in 
the BTS in 2018, resulting in a reduced sample size. 
On average, the species of interest were adequately rep- 
resented in the catches of both the LLS and BTS. During 
the LLS, catches of cod occurred at 73% of stations in the 
spring, and at 70% of stations in the fall, yielding lengths 
of 1603 individuals in the spring and of 1330 individuals 
in the fall (Table 1). For white hake, the rate of encoun- 
ter in the LLS averaged 50% in the spring and 74% in 
the fall, producing lengths of 882 individuals in the spring 
