Davis et aL: Population assessment of Limulus polyphemus 



223 



ing survey will likely prove to be a valuable 

 source of information in the future. 



Negative correlations were present among 

 a relatively large number of Delaware Bay 

 surveys (Table 1) which are assumed to be 

 sampling the same population. This is a 

 common problem in fisheries stock assess- 

 ments (Richards, 1991; Schnute and Hilborn, 

 1993). The observed differences among these 

 surveys could be attributed to a number of 

 factors. Catches of horseshoe crabs are not 

 common, leading to small sample sizes and 

 high variability. Also, these surveys may 

 differ in location, time of year, and gear 

 selectivity. Future studies could employ an 

 analysis of variance (ANOVA) to attempt 

 to separate these factors from underlying 

 horseshoe crab abundance trends. 



The data sources included in individual 

 model runs also led to differing results. 

 Models incorporating fishery-dependent data 

 tended to present a slightly more optimis- 

 tic view of the population and the fishery. 

 predicting higher relative biomass and lower relative 

 fishing mortality. Although harvest data often have the 

 benefit of having been derived from large sample sizes 

 (and have resulting low variance estimates), there is 

 often a bias associated with fishery-dependent data. 

 Fisheries do not sample randomly because they target 

 areas of highest abundance, and thus biased indices 

 are produced (Quinn and Deriso, 1999). In addition, 

 the use of fishery-dependent abundance indices is often 

 complicated by changes in gear, regulations, or sam- 

 pling methods over time, any of which could affect catch 

 rates. Fishery-independent surveys are usually more 



2 °°1 Projected B/B^sy (model FI-1) 



1 75- 

 1 50 

 1 25- 



CO 1 00 



HI 



75- 



50 

 25- 

 00 



2004 



2006 



2008 



2010 2012 



Year 



2014 



Figure 9 



Example of projection results. Projected relative biomass iB/B^/^y) 

 in shown for model FI-1, with each line representing a harvest 

 level applied annually in the 15-year projections. The percentage 

 refers to the percent of the 2003 Delaware Bay regional landings 

 of 1356 t (i.e., 50% = 678 t). 



appropriate for assessments, assuming there is high 

 consistency in sampling methods among years (Chen 

 et al., 2003). 



Differences also existed among fishery-independent 

 surveys. Models FI-2 and FI-3 predicted a much low- 

 er population size than FI-1 or the fishery-dependent 

 models. Projections with these runs predicted that the 

 population was too depleted to recover in 15 years, even 

 in the absence of harvest. The models differed only in 

 the abundance indices included. In the trend analyses 

 conducted for ASMFC,- the most significant population 

 declines since 1997 were identified in the NJ ocean 



