Unwin: Survival of Chinook salmon in relation to spring and summer mainstem flows of the Rakaia River, New Zealand 
819 
18 October and 15 November, of five to 
nine weeks duration, the correlation be- 
tween Q and log S averaged 0.493. Taken 
as an isolated result, this correlation cor- 
responds to an average significance level 
of P < 0.01 and an average coefficient of 
determination (r 2 ) of 0.243. There was 
some evidence of a weaker and more tran- 
sient period of negative correlation be- 
tween Q and log S earlier in the season. 
For Q calculated over periods of seven 
to nine weeks duration and centered on 
the fortnight from 6-19 September, the 
correlation with log S averaged -0.404, 
corresponding to an r 2 of 0.16. The sym- 
metric upright “V” shape apparent in the 
contours of Fig. 3 A, centered on the be- 
ginning of November, is an artifact 
caused by the tendency for data sets rep- 
resenting Q over periods of similar du- 
ration centered on the same date to be 
highly correlated. 
Maximum flow ( Q ) and mean flow ( Q) 
were generally only weakly correlated 
with survival, and the few periods dur- 
ing which correlations stronger than ±0.4 
were recorded showed little tendency to 
persist over temporal scales of more than 
a few days (Fig. 3, B and C). In addition, 
these correlations tended to become 
stronger at shorter time scales, suggest- 
ing that for Q , and possibly for Q , the cor- 
relations apparent in Table 3 were a tran- 
sient effect of a fortuitous sequence of 
flood events. By contrast, the persistence 
of a positive correlation between Q and 
log S for flows averaged over periods of 
up to 90 days suggests that the relation 
is much more robust and hence likely to 
be of biological significance. The correla- 
tions between Q , Q , and Q are also con- 
sistent with this interpretation. Whereas 
Q nov and Q nov were highly correlated 
(r=0.84), Q nqv was only moderately cor- 
related with both parameters (r=0.61 in 
both cases), confirming that Q captured 
information on flow variation not mea- 
sured by either Q or Q. 
The relation between Q and log S was 
moderately influenced by the 1973 brood 
(for which the survival rate was unusu- 
ally high) but was not dependent on it. 
The 1965, 1974, 1982, and 1983 broods 
(for which survival was also relatively 
high) and the 1985, 1986, and 1988 
A 
B 
C 
Figure 3 
Correlations between log-transformed survival data for Glenariffe Stream 
chinook salmon and Rakaia River mainstem discharge expressed as the 
ratio of mean to median discharge (A), maximum discharge (B), and mean 
discharge (C), as a function of the period used to calculate each flow 
statistic. For each point, the locations on the x- and y-axes represent the 
mid-point of the period, and the duration, respectively. The points “a” 
and “b” correspond to the two intervals represented in Figure 4. 
