818 
Fishery Bulletin 95(4), 1997 
monthly means varied from 71 m 3 /s (Septem- 
ber 1977) to 734 m 3 /s (December 1979). The 
mean maximum August to January flow (effec- 
tively the mean annual spring and summer 
flood) was 1,488 m 3 /s; values for individual sea- 
sons ranged from 630 m 3 /s in 1980 to 2,800 m 3 / 
s in 1979. Flow variability was lowest during 
August and September, although highly vari- 
able flows ( Q > 1.6) were recorded in all months. 
Hydrographs for 1977 (a low-flow season) and 
1984 (an above average season) illustrate the 
sharp peaks and rapid recession typical of Rakaia 
floods (Fig. 2). In the November 1984 event, dis- 
charge increased by a factor of 10 (from 172 m 3 /s 
to 1,710 m 3 /s) over 48 h and then fell from 1,960 
m 3 /s to 455 m 3 /s over 72 h. Despite the contrast 
between the two seasons, both hydrographs also 
show a common period of low and relatively stable 
base flows in August and September, followed by 
more frequent floods as the season progresses. 
Correlation analysis 
Of the 32 flow parameters listed in Table 2, 29 showed 
no correlation with survival rates for Glenariffe 
Stream chinook (Table 3). Correlation coefficients for 
these indices ranged from -0.223 to 0.283, none of 
which differed significantly from zero (P>0.16 in all 
cases). The three exceptions were the mean flow Q , 
the maximum flow Q , and the ratio of the mean to 
median flow Q , for the month of November. All three 
measures were positively correlated with survival 
(Q N0V ,P = 0.045; Q N0V ,P = 0.003; Q N0V , P = 0.006). 
By contrast, there was no correlation between sur- 
vival and the same set of flow variables for the adja- 
cent months of October and December. 
The strongest and most consistent set of correla- 
tions involved the ratio of mean to median flow, cal- 
culated over periods of 30 to 90 d duration centered 
on or about November 1 (Fig. 3 A). Over much of this 
range the correlation between Q and log S exceeded 
0.5, with a maximum value of 0.667 for Q calculated 
over the 40-day period from 9 October to 17 Novem- 
ber. More generally, survival tended to be high for 
broods corresponding to years when flow variability 
during October, November, and early December was 
high. For periods centered on the four weeks between 
Table 3 
Correlations between log-transformed brood year survival rates for Glenariffe Stream chinook and 32 indices of Rakaia mainstem 
flows, 1965-90. See Table 2 for definitions of symbols. Asterisks denote correlations significant at the 95% level (*) and 99% level 
Flow parameter 
Period 
Q 
Q 
Q ^500 ^1,000 
2 , 
© 
o 
Q 10max 
February-January (year) 
-0.091 
August-January (spring and summer) 
-0.105 
0.109 
0.182 -0.062 0.261 
0.165 
0.164 
August-October (spring) 
-0.195 
-0.140 
-0.218 
November-January (summer) 
0.046 
0.205 
0.269 
August 
-0.068 
-0.054 
-0.066 
September 
-0.082 
-0.191 
-0.159 
October 
-0.223 
-0.053 
0.283 
November 
0.397* 
0.558** 
0.520** 
December 
-0.214 
-0.112 
-0.088 
January 
-0.070 
0.053 
0.141 
