34 



Fishery Bulletin 97(1), 1999 



It is clear that the size of the lobster stock off New 

 South Wales was low, with respect to its virgin bio- 

 mass, as was the 75% confidence. Recent stock bio- 

 mass (i.e. in 1995-96) has been between 15% and 

 slightly over 30% of the virgin biomass. However, it 

 seems that the decline in the biomass of the stock 

 has stopped in recent years, and that the stock is 

 perhaps in a period of recovery. 



0.30 



0.25 



0.20 



2 0.15 - 

 o 



0.10 



0.05 



0.00' 



;,oco 



500 1,000 1.500 2,000 2,500 



Stock biomass (t) 



Figure 8 



The distribution of stock biomass in 1996-97 estimated with the bootstrapped LMSE 

 method. 



Data observed in the two periods of time were 

 modeled separately in our study. An alternative, 

 perhaps better approach is to combine these two 

 time periods with information on catches landed 

 within the intervening period. This approach will 

 be used in the next stock assessment when good 

 estimates of catches from 1939 to 1968-69 become 

 available. 



Because the data were lim- 

 ited, the choice of models that 

 can be used to describe the 

 dynamics of lobster stock in 

 NSW is also limited. Two dis- 

 advantages of using the simple 

 production models are 1) an in- 

 adequate representation of the 

 fishery dynamics may result in 

 large biases in estimates of pa- 

 rameters and biomasses, and 2) 

 extra assumptions are needed 

 about fisheries (e.g. CPUE is a 

 good indicator of stock abun- 

 dance). However, these simple 

 models also have their advan- 

 tages. If the model fails, this fail- 

 ure can be seen easily. Because 

 of the simple mathematical 

 structure and lack of constraints 

 in the parameter estimation, es- 

 timates of parameters may not 

 necessarily be biologically rea- 

 sonable (e.g. as negative values), 

 an indication of the failure of 

 models or data (Hilborn and 

 Walters, 1992). It is now fash- 

 ionable to mimic biological and 

 fisheries realism by setting up 

 a model that is mathematically 

 and statistically complicated. 

 Although advantages are obvi- 

 ous, disadvantages associated 

 with this type of models may not 

 always be realized. In addition 

 to the requirement of extra in- 

 put data, such models tend to 

 have some undesirable attri- 

 butes in the estimation of pa- 

 rameters, such as overfitting 

 and high correlations among 

 estimates of parameters. Incor- 

 poration of diflferent biological 

 and fisheries processes (e.g. fish- 

 ing and recruitment processes) 

 into one model for parameter 

 estimation can also create some 



