Unwin: Survival of Chinook salmon in relation to spring and summer mainstem flows of the Rakaia River, New Zealand 
813 
losses were primarily due to problems with passage 
through hydroelectric dams rather than to low dis- 
charge per se. The effects of flow variability on sur- 
vival have also received little attention. Increased 
downstream movement of newly emerged salmonid 
fry following sudden increases in discharge has been 
well documented (e.g. Irvine, 1986; Saltveit et al., 
1995), but only rarely has flow variability been used 
as a predictor variable in population studies 
(Berggren and Filardo, 1993). 
Whenever brood year survival is estimated from 
stock-recruitment or similar data, a search for cor- 
relations between river flow and survival will usu- 
ally involve deriving a single flow index, such as the 
annual mean, for each cohort. Most such studies con- 
ducted to date have used flow averaged over periods 
from three months (Kope and Botsford, 1990) to one 
year (Beamish et al., 1994), but it is by no means 
obvious that these are the most informative or bio- 
logically meaningful parameters to use. A single cata- 
strophic flood during the incubation period may cause 
large-scale destruction of redds and loss of alevins 
through bed scour (Montgomery et al., 1996) with- 
out having much effect on the mean annual flow. 
Prior to smolting, fry may be susceptible to short- 
term floods that carry them prematurely into sea- 
water, when the same floods a few months later would 
have little impact. In addition, mean flow is not nec- 
essarily the most relevant statistic for characteriz- 
ing flow regimes; it is possible that in the two ex- 
amples given above, some other parameter (such as 
maximum flow or the coefficient of variation) might 
be more informative (e.g. Hvidsten and Hansen, 
1988). For example, in New Zealand, where high flow 
variability is a defining characteristic of riverine eco- 
systems (Biggs, 1995), statistics such as the propor- 
tion of the time the flow exceeds three times the 
median (Clausen and Biggs, in press) and the ratio 
of mean flow to median flow ( Jowett, 1990), have been 
successful in elucidating relations between flow re- 
gime and biological parameters. To explore fully the 
relation between flow and survival, therefore, it is 
necessary to consider not only the type of flow sta- 
tistic that is likely to be of interest but also the dura- 
tion and seasonal timing of the period over which 
the statistic is to be calculated. 
Since the introduction of fall-run Sacramento River 
stock to New Zealand in the early 1900’s (McDowall, 
1994a; Quinn et al., 1996), chinook salmon have 
maintained self-sustaining populations in all major 
rivers on the east coast of the South Island (McDowall, 
1990; Quinn and Unwin, 1993). Like most New Zealand 
rivers, these rivers (whose wide, braided shingle beds 
drain steeply mountainous catchments on the South 
Island main divide) are characterized by highly vari- 
able flows (Jowett and Duncan, 1990), flooding quickly 
whenever snow and ice melt in the headwaters is 
augmented by heavy orographic rainfall. These floods 
occur at any time of year but are particularly common 
in spring. In rivers such as the Rakaia they cause mas- 
sive bed movement ( Ibbitt, 1979) and reduce the abun- 
dance and diversity of invertebrate fauna (Sagar, 1986) 
for up to one month afterwards. The impact of these 
events on seaward-migrating juvenile chinook has gen- 
erated some debate. Several authors have remarked 
that survival of New Zealand chinook fry may be ad- 
versely affected during floods (McDowall, 1990; Flain 1 ). 
Other studies suggest that, although some fry may be 
lost during extreme floods, flow fluctuations in a more 
typical season do not have a serious negative impact 
on migration (Hopkins and Unwin, 1987). 
Chinook salmon spawning populations in Glen- 
ariffe Stream, a headwater spawning tributary of the 
Rakaia River (Fig. 1), have been monitored since 1965 
by means of an upstream counting fence (Quinn and 
Unwin, 1993). In this study I analyzed brood year 
survival, for chinook spawning in Glenariffe Stream, 
in relation to Rakaia mainstem discharge during 
spring and summer (August to January). My primary 
objectives were to examine various flow statistics as 
possible correlates with survival and to determine the 
sensitivity of any resulting correlations to changes in 
the interval used to calculate each statistic. A second- 
ary objective was to examine evidence that spring floods 
were detrimental to brood year survival. 
Chinook salmon in New Zealand 
New Zealand chinook salmon are broadly similar to 
their Sacramento ancestors in terms of both their 
general life history (Unwin, 1986) and genetic make 
up (Quinn et al., 1996). Present day stocks comprise a 
mixture of ocean- and stream-type fish (Gilbert, 1913; 
Healey, 1983), corresponding to juveniles that spend 
3-6 mo or 12-15 mo in fresh water before entering the 
ocean (Unwin and Lucas, 1993). In the Rakaia River, 
the most thoroughly studied of the major salmon pro- 
ducing rivers, ocean-type fish make up about two-thirds 
of the returning adults (Quinn and Unwin, 1993). 
The migration patterns of age-0+ juvenile chinook 
salmon in the Rakaia River and a key spawning tribu- 
tary, Glenariffe Stream (Fig. 1), have been studied 
in some detail, and are relatively well understood 
(Unwin, 1986; Hopkins and Unwin, 1987). From 
1 Flain, M. 1982. Quinnat salmon runs, 1965-1978, in the 
Glenariffe Stream, Rakaia River, New Zealand. Occasional 
Publ 28, N.Z. Ministry of Agriculture and Fisheries, Fisheries 
Research Div., 22 p. [Copy held at NIWA, Christchurch, New 
Zealand.] 
