Shepherd et al. (1984) and Sissenwine (1984) 

 cautioned in the use of data exploration exercises 

 not based on plausible a priori hypotheses and 

 pointed out the risks of finding spurious correla- 

 tions from empirical studies. Fundamentally dif- 

 ferent empirical models may also be indistinguish- 

 able because they account for virtually the same 

 proportion of variability in recruitment. However, 

 even without clear evidence for causal relation- 

 ships, circumstantial evidence may be adequate 

 and empirical models can serve as the basis for 

 the creation and testing of future hypotheses 

 (Garrod 1982; Shepherd et al. 1984). Definitive 

 explanations do not yet exist to explain why Feb- 

 ruary water temperatures were correlated with 

 winter flounder recruitment. This period encom- 

 passes much of the period of egg deposition, in- 

 cubation, and hatching. Most likely, temperature 

 is a surrogate for a complex set of physical and 

 biological interactions. Winter flounder egg in- 

 cubation time and hatching success (Scott 1929; 

 Williams 1975; Rogers 1976) and larval growth 

 (Laurence 1975; NUSCO 1987) were related to 

 temperature. Temperature may also be correlated 



with other ecological factors affecting reproductive 

 success. Most importantly, it probably affects the 

 timing of the match of larval production with 

 that of their prey and predators, thereby strongly 

 influencing the success of a year-class (Gushing 

 1973). 



Accordingly, the difference between each an- 

 nual February mean water temperature from the 

 1 1-year mean of 2.4°C was used as a third vari- 

 able in the Ricker stock and recruitment model. 

 The result was an improved fit (R^ = 0.78) to the 

 observed data (Fig. 15). Estimated values of re- 

 cruitment using the three-parameter model fol- 

 lowed the predictions relatively closely (Fig. 16). 

 February water temperatures helped to explain 

 the previously mentioned diflerence between 1978 

 and 1984; the former year was among the coldest 

 and the latter the warmest of the series. The 1978 

 year-class was also reported by Gibson (1987) to 

 be exceptionally large in Rhode Island. This was 

 not unexpected nor were the high correlations 

 found among abundance indices for the Niantic 

 River stock and others in the region. Climatic 



Q 15 



R = 3.342 * P * exp(-0.012 * P) * exp(-0.197 * Temp) (R^ = 0.78) 



• = actual 

 • ^^ y •» = predicted 



15 20 25 30 



PARENT STOCK INDEX (P) 



Fig. 1 5. Temperature-dependent stock -recruitment model (ihree-parameter) for Nianlic River winter flounder. 



Winter Flounder Studies 175 



