Unwin: Survival of Chinook salmon in relation to spring and summer mainstem flows of the Rakaia River. New Zealand 
821 
1973 
I I 
1974 
Sep Oct Nov 
1983 
Sep Oct Nov 
Figure 5 
Hydrographs for the Rakaia River from September to November for three years of high 
mean flow to median flow (as calculated over the period indicated by dashed lines) and 
high survival (1973, 1974, 1983) and three years of low mean flow to median flow and low 
survival (1985, 1986, 1988). All hydrographs are drawn to a common vertical scale of 0- 
2,000 m 3 /s. 
The ratio of mean to median flow is a direct mea- 
sure of flow variability, but it does not necessarily 
follow that low values of Q correspond to low and 
stable flows. Flow regimes where Q is close to one 
(i.e. mean flow differs little from the median) can 
arise in two different ways, only one of which corre- 
sponds to an extended period of low flows (Fig. 5). 
The other possibility is a period of highly variable 
flows superimposed on a high base flow, so that even 
though flows vary greatly from day to day, the flow 
distribution is only moderately skewed. By contrast, 
high values of Q are more consistently associated 
with highly variable flows, tending to correspond to 
intervals when a lengthy period of low, stable base 
flows is punctuated by a relatively small number of 
short and sharp flood events. Both the 1973 and 1974 
seasons ( Qgoct-nNov^-^ and 1.55, respectively) 
were characterized by low and stable base flows dur- 
ing September, followed by two or three relatively 
short-lived flood events during October and Novem- 
ber. The 1985 and 1986 seasons ( Q 9Oc t-i 7 Nov = l-00 in 
both cases) were characterized by uniformly low and 
stable flows, with no floods over the six weeks from 
1 October. The effect of high base flows on flow vari- 
ability is illustrated by comparing the 1983 and 1988 
seasons ( Q 9 oct-i 7 Nov = l-46 and 1.15, respectively). 
Whereas base flows were low throughout 1983, so 
that a 2,000 m 3 /s event in late October generated a 
high figure for Q , the absence of a major flood dur- 
ing October and November 1988 produced a lower 
figure for Q despite high and rapidly fluctuating base 
flows over most of the period shown. 
The other flow parameters examined during the 
initial phase of this study, such as the number of days 
from August to January when flows exceeded a cer- 
tain level, showed no correlation with survival. Al- 
though in principle it would have been possible to 
subject these parameters to the same level of analy- 
sis used for Q , Q , and Q , this analysis becomes 
progressively less meaningful at shorter time scales. 
For example, given that N x 000 ranged from 0 to 8 
over a six month period (Table 2), the same statistic 
calculated over monthly intervals would typically 
take on only a few discrete integer values and would 
be inappropriate for correlation analysis. 
Mechanisms 
The topography of the correlations between Q and 
survival, as illustrated by the qualitative features of 
Fig. 3 A, coincides to a striking degree with several 
key events in the migration patterns of juvenile 
ocean-type fry in the Rakaia River. The period of 
highest positive correlation between flow variability 
