820 
Fishery Bulletin 95(4), 1997 
broods (for which survival was low) conform to the 
general trend irrespective of the duration of the pe- 
riod used to calculate Q (Fig. 4). With the exception 
of the 1969 brood, the value of Q during October 
and November was not highly sensitive to the choice 
of interval. 
Discussion 
Correlations and flow parameters 
This study shows that, although fry-to-adult survival 
of Glenariffe Stream chinook salmon is correlated 
with the occurrence of springtime flood events in the 
Rakaia River, both the magnitude and the direction 
of the observed correlation depend strongly on the 
flow parameter used and the period over which this 
parameter is calculated. Survival was most strongly 
correlated with flow variability (as measured by the 
ratio of mean to median flow), the correlation being 
moderately negative for flows averaged over the pe- 
riod from mid-August to mid-October, and rather 
more strongly positive from mid-October to Novem- 
ber. Similar but weaker correlations were apparent 
between survival and maximum flow, but mean flow 
was a poor predictor of survival irrespective of the 
time interval used. 
The correlations reported here have two key fea- 
tures. First, although quite strong in a biological con- 
text, they are nevertheless relatively weak, account- 
ing for at most 25-30% of the observed variation in 
log survival. Even if the maximum positive correla- 
tion (r=0.667) is taken at face value, its predictive 
power allows years to be categorized only as “above 
average” or “below average,” at best (Prairie, 1996). 
This result is consistent with an earlier finding that 
annual survival rates for New Zealand chinook are 
primarily determined by marine rather than fresh- 
water influences (Unwin, 1997; see also Bradford, 
1995). Second, the tendency for survival to be posi- 
tively correlated with flow variation but uncorrelated 
with mean flow suggests that increased flow vari- 
ability (in the sense illustrated in Fig. 5) at the ap- 
propriate time of year is beneficial to survival. This 
is in sharp contrast to the generally held view that 
spring floods have a detrimental impact on chinook 
fry in the Rakaia and other New Zealand rivers 
(Waugh, 1980; McDowall, 1990; Flain 1 ). 
The three key flow parameters used in this study — 
mean flow, maximum flow, and ratio of mean flow to 
median flow — -are by no means the only ones pos- 
sible and can only capture some of the information 
contained in the hydrograph for a given time period. 
Mean flow essentially measures the total volume of 
water passing through the river over the interval in 
question, without conveying any information about 
the magnitude or distribution of floods. For example, 
a 90-day mean of 200 m 3 /s could arise from 90 con- 
secutive days at exactly 200 m 3 /s each, or from 89 days 
at 180 m 3 /s and one day at 1,980 m 3 /s. Maximum flow 
characterizes peak flood intensity, but not flow vari- 
ability, so that a flow period with one 2,000 m 3 /s flood 
will outrank another period with ten 1,900 m 3 /s floods. 
Mean flow/median flow 
Figure 4 
The relation between log-transformed survival data for Glenariffe Stream chinook salmon and 
the ratio of mean to median discharge in the Rakaia River mainstem for two periods near local 
maxima in Figure 3A. Twelve points corresponding to broods with extreme survival or flow 
indices for at least one of the two periods shown are identified by year. 
