significant at some stations. Cooling-water flow was a significant 

 multiplicative regressor in the impingement models. The 6-yr cycle 

 found in impingement and four out of the six trawl models could be an 

 artifact of the 6-yr database (1976-82). The shorter 6-mo and 4-mo 

 cycles were evidence of consistent fluctuations within a year. The 

 impingement models had high R^ values and explained 80% of the variation 

 (Table 11). Even though these models described observed patterns, the 

 1983 data exceeded the upper 95% confidence limit (Fig. 21) . Among the 

 9 0- 



/\ A , - *** 



6 0H -' O; 



3 0- 



3 e.e- r- 



-3 0- 



OCT DEC FEB APR JUN AUG OCT DEC FEB APR JUN AUG 



7 5- 



5 0- 



?! 2.5- 



C 0- 



-2 5- 



-. — I 1 — I . — I — 1 1 . 1 — . — I . — 1 . — I . 1 . 1 . r 



OCT DEC FEB APR JUN AUG OCT DEC FEB APR JUN AUG 



1982 data 



1983 data 



Figure 21. Forecast (- 



-) and 95% confidence limits ( ) for S_. 



aquosus impinged at Unit 1 (top) and trawled at BR (botTom) . 

 Both models were fit to data from January 1976 through 

 September 1982. 



37 



