98 



Fishery Bulletin 105(1) 



Fall 

 A Winter skate 



Spring 



D Little skate 



1 1 1 1 1 1 1 Ij^i I if 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 

 B Little skate 



h 1 1 1 1 1 1 1 1 ijj^l iTi«l 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 



E Silver hake 



U I I Ij^l I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I 



F Haddock 



1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 141 1 1*1 1 1 1 1 1 1 1 1 1 



1965 1970 1975 1980 1985 1990 1995 2000 



HI 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 4 



1965 1970 1975 1980 1985 1990 1995 2000 



-♦- Original 

 ^ Smoothed 



Year 



Figure 4 



Comparison of original time series of estimated total biomass on Georges 

 Bank from the fall (left column) and spring (right column) bottom trawl survey 

 (filled circles) and inverse-transformed, ARIMA-smoothed time series (open 

 triangles). For these time series, an ARIMA (0,1,1) model (i.e., a random-walk- 

 plus-uncorrelated-noise model) was used, although this was not the model 

 selected in our time series analysis. 



order of the observed time series beyond that expected 

 for uncorrelated noise. 



Of the nine species considered, only three (little skate, 

 Atlantic herring, and flounder) had models that exhib- 

 ited the same ARIMA order for both the fall and spring 

 surveys. Taken at face value, this would indicate that 

 the other six species exhibited substantially different 

 dynamical processes during the fall and spring that 

 influenced abundance on Georges Bank. One potential 

 mechanism for this could be differential seasonal migra- 

 tion patterns that result in changes in catchability that 

 have different autocorrelation structures. For example, 

 cod are distributed across the bank during the spring 

 survey, but are found only in deeper waters on the pe- 

 riphery of the bank during the fall where they may be 

 less available to the survey. Assuming that all Georges 



Bank cod are available to the spring survey, if the frac- 

 tion of cod available to the survey in the fall is density 

 dependent or is driven by autocorrelated environmental 

 conditions, then the fall survey abundance will exhibit 

 dynamical behavior different from that of the spring 

 survey abundance (note: if this were the case, it would 

 be inappropriate to smooth the fall time series for cod 

 with the ARIMA noise reduction approach applied in 

 our study because, as noted previously, one of the basic 

 assumptions with this approach is that the observation 

 noise is uncorrelated). Other plausible mechanisms can 

 be developed, as well. However, we feel it more likely 

 that the inconsistency in ARIMA order between spring 

 and fall surveys for the same species is an estimation 

 problem and indicates that even a 40-year time series 

 may not be long enough to reduce the variability inher- 



