242 
Fishery Bulletin 1 14(2) 
Table 2 
Annual mean fishing effort, catch, and catch per unit of effort (CPUE), by gear type, in Guam during the 2 periods 1985- 
1990 and 2007-2012. Data were collected during shore-based creel surveys conducted by the Guam Division of Aquatic 
and Wildlife Resources and expanded by the Western Pacific Fisheries Information Network of the NOAA Pacific Islands 
Fisheries Science Center. Catch and effort values are based on expanded data and exclude nonreef-associated fishes (e.g., 
pelagic, freshwater, and intertidal fish species). Values of CPUE (with standard errors [SEs] in parentheses) are based on 
unexpanded interview data (not on a ratio of tabled catch and effort). P-values are from the Welch’s 2-sample f-test (which 
allowed unequal variances), in which CPUE was compared between the 2 time periods. NA represents cases when less than 
3 interviews per year per gear type were available (see Appendix 3 Table 1). Effort is given in gear hours (gh), and catch is 
given in kilograms. 
Gear type 
1985-1990 
2007-2012 
P-value 
Mean effort (gh) 
(% of total) 
Mean catch 
(kg) 
CPUE (SE) 
(kg/gh) 
Mean effort (gh) 
(% of total) 
Mean catch 
(kg) 
CPUE (SE) 
(kg/gh) 
Hook and line 
127,219 (58.5%) 
22,464 
0.15 (0.02) 
145,309 (70.6%) 
17,828 
0.08 (0.02) 
0.015 
Cast net 
31,913 (14.7%) 
15,88 
00.36 (0.04) 
29,555 (14.4%) 
4,108 
0.10 (0.02) 
<0.001 
Gill net 
22,647 (10.4%) 
28,259 
1.25 (0.11) 
10,918 (5.3%) 
9,807 
0.56(0.18) 
0.01 
Surround net 
1470 (0.7%) 
4128 
1.36 (0.67) 
552 (0.3%) 
555 
0.67 (0.08) 
NA 
Snorkel spear 7 
18,453 (2.9%) 
16,507 
0.76 (0.15) 
11,736 (5.7%) 
2,137 
0.19 (0.06) 
0.013 
Scuba spear 7 
1553 (0.7%) 
988 
0.75 (0.10) 
155 (0.1%) 
58 
0.34(0.04) 
NA 
Hooks and gaffs 
6396 (2.9%) 
2228 
0.30 (0.05) 
5554 (2.7%) 
2,139 
0.36 (0.16) 
NA 
Drag net 
901 (0.4%) 
1377 
1.18 (0.53) 
141 (0.1%) 
133 
0.97 (0.35) 
NA 
Other gear 
6868 (3.2%) 
8454 
0.55 (0.29) 
1930 (0.9%) 
181 
0.07 (0.01) 
0.10 
Total 
217,420 
100,285 
205,850 
36,945 
7 Catch and effort values for 2007-2012 are underestimated because most spear fishermen refused to participate in interviews 
after 2005 (Lindfield et al., 2014). 
jacks) that can be hyper-abundant in shallow, sandy 
bays that were not part of the sampling domain for 
the fishery-independent visual surveys from which we 
took data. Appendix 4 shows the number of times each 
of the exploited fish groups defined in Table 1 were 
caught by fishermen interviewed for the creel survey 
program; data were broken out by gear type and pooled 
into 3-year time periods. 
During the interviews for the creel surveys, fishing is 
reported per region per day type (i.e., weekday or week- 
end days; for a map of the regions used for creel surveys, 
see Appendix 2 Figure). Therefore, we first compared 
annual average CPUE by region and day type, using a 
Welch 2-sample /-test for cases in which we had at least 
6 years of CPUE data with at least 3 observations per 
year (Appendix 3 Table 2). On the basis of the results of 
these tests (Appendix 5), we decided to aggregate CPUE 
data over regions and day types for subsequent analy- 
ses. Even with such treatment, for some gear types, 
interview data were sparse, especially for surround net 
fishing, scuba spearfishing, and drag-net fishing (Appen- 
dix 3 Figure). Moreover, for drag-net fishing and scuba 
spearfishing, some values were missing. We calculated 
CPUE for missing data as the average of the 3 previous 
years (Appendix 3 Table 2). 
We based our reconstruction of historical relative 
fish biomass on the assumption that CPUE (/), itself 
expressed as catch (C) divided by effort (E), is linearly 
related to biomass ( B ) in year t: 
Cf/ E t — I t — q * B t . (2) 
We further assumed that the gear-specific catchability 
coefficient (q) was constant. Therefore, although catch- 
ability varies depending on the experience of the fish- 
erman and other factors, we assumed that there is no 
overall change in catchability through time. We also 
assumed that the observational errors of CPUE were 
lognormal, were multiplicative, and had constant vari- 
ance. Although these conditions are unlikely — given the 
small extent of actual interview data for many taxa and 
strata, with the resulting high incidence of zeros and 
Poisson-like distributions with outliers — these assump- 
tions were adopted to allow a simple approximation. If 
there are n years in an analysis, the best estimate of the 
constant q is the geometric average of the time series 
of ratios of CPUE and biomass ( I t /B t ), and it can be 
calculated with the following equation (Haddon, 2001): 
This method is well established for pelagic fisheries 
(Haddon, 2001), and we further assumed that it is ap- 
plicable for the complex reef fishery when we parsed 
out catches and effort by gear type. We used the fishery- 
dependent statistics to estimate CPUE per gear type per 
year and fishery-independent data to estimate biomass 
(as detailed previously). However, we had reliable bio- 
mass estimates only for 2011. Because we required an n 
