450 
Fishery Bulletin 95(3), 1997 
0.20 -r 
Number of years in SVPA 
Figure 3 
Root-mean-square retrospective difference of fishing mortality esti- 
mates for age-1 Atlantic menhaden from SVPA’s with three to ten 
years of catch at age. 
0.13 (33% relative difference) with back-calculated 
values of fully recruited F and increased to 0. 18 (45% 
relative difference) with catch-curve estimates 
(Fig. 4). 
Retrospective differences in age-1 abundance 
ranged from -2.4 billion to 11.5 billion individuals 
in terminal years (Fig. 5). There was no significant 
bias in log-transformed differences, and the RMS 
difference was 1.2 billion recruits (n=34). The RMS 
relative difference for estimates of age-1 abundance 
was 46%. 
Retrospective differences in terminal SSB esti- 
mates ranged from -72,000 t to 448,000 t (Fig. 6). 
The large positive differences in 1962 and 1963 SSB 
were primarily due to large negative differences in 
fully recruited F (Fig. 2). Log- transformed retrospec- 
tive differences were not significantly biased, and the 
RMS difference was 9,000 t SSB (n=34). The 
RMS relative difference for estimates of SSB was 
33%. Retrospective differences in %MSP ranged from 
-5.5 to 19.5 in terminal years (Fig. 7). 
The RMS difference was 4.7 %MSP, and there was 
no significant bias in log-transformed differences 
(n=34). The RMS relative difference for estimates of 
%MSP was 106%, because inconsistencies were larger 
than the estimated level of %MSP. Retrospective dif- 
ferences in %MSP were negatively correlated with ret- 
rospective differences in fully recruited F (r=-0.79). 
Discussion 
This case study illustrates how retrospective com- 
parisons can provide useful data for analytical deci- 
sions and reveal important insights for management 
advice, especially in situations where statistical es- 
timates of uncertainty are not available. For example, 
SVPA with seven years of catch data clearly provided 
more consistent results than SVPA of longer or 
shorter time series (Fig. 3). A time period of seven 
years appears to be long enough to smooth annual 
variation in partial recruitment, while including only 
years which represent the current schedule of F at 
age. By including more years in the analysis, there 
is a likelihood that catches from substantially dif- 
ferent exploitation patterns will be incorporated. 
Performance of alternative intervals of catch 
data was judged according to general conditions over 
three decades. Although such guidance is valuable, 
specific SVPA runs should be examined to confirm the 
assumption of separability. For example, targeting spe- 
cific cohorts, such as the superabundant 1958 year class, 
may change fishing patterns. Abrupt changes should 
be reflected in patterns of log catch-ratio residuals 
(Pope and Shepherd, 1982) and may necessitate a 
longer or shorter time series of catch at age for SVPA. 
Retrospective inconsistency can result from a host 
of systematic problems, including errors in the cur- 
