may have always been present throughout the river and the Jolly indices from 1977 through 1980 were 

 inaccurately low to an unknown degree because only a portion of the population was sampled. This was 

 previously suspected for the 1980 survey, when a large percentage of relatively small and stationary winter 

 flounder were found and no samples were taken in the upper river unlike following years (NUSCo 1983b, 

 1985). Therefore, CPUE may have more realistically measured abundance from 1976 through 1980. 

 Examination of data from 1981 through 1986 showed that median CPUE values for station 1 and 2 in 

 the lower river were very similar to those from the upper river (within 4 fish per tow), except for 1983 

 (13.2 less in the lower river), so CPUE appears to be relatively consistent among areas. As surveys and 

 methods are presently more standardized, further comparisons of CPUE and the Jolly index may be made 

 in forthcoming years and the relative accuracies of the two measures of abundance can be reassessed. 



Harmonic regression models 



Another measure of winter flounder abundance throughout the MUlstone area was the development 

 of time-based harmonic regression models. I x)g-tr£insformed data from six stations of the trawl monitoring 

 program (see Fish Ecology section) were used to describe the fluctuations in abundance of winter flounder. 

 Models having data from October 1976 through September 1984 forecasted catches from October 1984 

 through September 1985 and the actual catches were then compared to those predicted (NUSCo 1986a). 

 Similar models were reported in NUSCo (1984, 1985) for 1982-83 and 1983-84 data. Results showed that 

 models for stations other than NR (Niantic River) were not satisfactory. Terms corresponding to a 

 sine-cosine function describing an 8-yr period were significant for most models in 1985, as were terms for 

 7 or 6 yr during 1984 and 1983. Since these terms represented the entire time-series of data at the time 

 of model development, this indicated either insufficient data or a lack of a repetitive pattern of abundance. 

 Terms of less than 1 yr were also found for most models and were probably related to aimual cycles of 

 abundance due to local movements and recruitment of juveniles into the trawl catch. Except for NR 

 (0.71), R values for each model remained low (0.38-0.53). However, this is a typical result for a species 

 taken year-round in samples. Forecast errors remained high in 1985 (65-279%; 128% for NR), although, 

 in most cases, improved each year since 1983. This was an indication that the models were perhaps 

 providing better predictability. 



Many factors have influenced trawl monitoring program catches, as they have in the Niantic River 

 abundance surveys discussed previously. Despite its relatively high abundance in the catch (43% of the 



35 



