Unwin: Survival of Chinook salmon in relation to spring and summer mainstem flows of the Rakaia River, New Zealand 
817 
such as CV or skewness, which involve raising flows 
to the second and third power, respectively. I also 
calculated several indices related to the incidence of 
flood peaks, including the number of days when the 
daily mean discharge exceeded 500 m 3 /s, 1,000 m 3 /s, 
and 1,500 m 3 /s, and the mean of the ten highest flows 
over the six months from August to January. The 
complete set of flow parameters used is summarized 
in Table 2. 
For each parameter, I looked for evidence of a re- 
lation with the log-transformed Glenariffe Stream 
survival data by calculating the correlation coeffi- 
cient for the paired data sets over the 26 years of 
record. I examined bivariate scatter plots and re- 
sidual plots for any data sets showing a significant 
relation (P<0.05). For these preliminary results I did 
not correct for the effect of multiple tests (i.e. the 
possibility of finding an artificially inflated correla- 
tion with one of the 32 flow parameters purely by 
chance); therefore P-values for each correlation over- 
estimated their true significance (Walters and Col- 
lie, 1988; Kope and Botsford, 1990). For these pa- 
rameters, my next level of analysis was to recalcu- 
late the appropriate flow statistic for periods rang- 
ing in duration from one week to four months, dat- 
ing from 1 June to 31 January (representing 805 
periods in total), and to recalculate the correlation 
with the survival data for each choice of date and 
duration. I then constructed contour plots depicting 
variations in the correlation coefficient as a function 
of starting date and duration and examined these 
“surfaces of correlation” for the presence of local ex- 
trema. My motivation for this analysis was not to 
identify a single period that maximized the correla- 
tion, but rather to gauge the sensitivity of the corre- 
lation to small changes in interval, and hence to iden- 
tify seasonal periods for which significant correla- 
tions between flow indices and survival persisted over 
biologically meaningful time scales. 
All statistical calculations were performed with 
version 6.0 of SYSTAT software (Wilkinson, 1996). 
Contour and surface fits were accomplished with 
version 6 of SURFER for Windows’ implementation 
of Kriging smoothing (Keckler, 1994) applied to a grid 
of correlation coefficients calculated at 5-day inter- 
vals on both the date and period axes. 
Results 
River flows 
Over the period covered by this study (August 1965 
to January 1991), monthly mean discharge ranged 
from 146 m 3 /s to 306 m 3 /s (Table 2). Individual 
Table 2 
Flow parameters used for correlation analysis, together 
with their mean and range over the period 1965 to 1990. 
Parameter 
Symbol 
Mean 
Range 
Measures of flow volume (m 3 /s) 
Mean annual 
flow, Feb-Jan Q Annua , 
209 
156-277 
Mean flow, Aug-Jan 
Q Spring/Summer 
237 
157-329 
Mean flow, Aug-Oct 
Q Spring 
190 
86-348 
Mean flow, Nov-Jan 
Q Summer 
283 
186-456 
Mean flow, Aug 
Q Aug 
146 
80-293 
Mean flow, Sep 
Q Sep 
181 
71-570 
Mean flow, Oct 
Q Oct 
243 
106-493 
Mean flow, Nov 
Q Nov 
282 
131-457 
Mean flow, Dec 
Q Dec 
306 
160-589 
Mean flow, Jan 
Q Jan 
262 
148-416 
Measures of flood peaks (flows in m 3 /s) 
Maximum flow, 
Aug— Jan Q Spring/Summer 
1,488 
630-2,800 
Maximum flow, 
Aug-Oct 
Q Spring 
975 
167-2,470 
Maximum flow, 
Nov-Jan 
Q Summer 
1,279 
540-2,800 
Maximum flow, Aug 
Q Aug 
385 
86-2,470 
Maximum flow, Sep 
Q Sep 
502 
83-2,230 
Maximum flow, Oct 
QOct 
Q Nov 
755 
133-2,030 
Maximum flow, Nov 
783 
203-1,960 
Maximum flow, Dec 
Q Dec 
1,006 
282-2,800 
Maximum flow, Jan 
Q Jan 
806 
222-2,470 
Mean of 10 highest 
flows, Aug-Jan 
Q lOmax 
854 
443-1,410 
Number of days 
Q > 500 m 3 /s, 
Aug-Jan 
N 
1 o00 
13 
3-26 
Number of days 
Q > 1,000 m 3 /s, 
Aug-Jan 
N 
A> 1,000 
2 
0-8 
Number of days 
Q > 1,500 m 3 /s, 
Aug-Jan 
N 
-^1,500 
1 
0-4 
Measures of flow variability 
Mean/median, 
Aug-Jan Q Spring/Summer 
1.33 
1.04-1.73 
Mean/median, 
Aug-Oct 
Q Spring 
1.32 
1.01-1.92 
Mean/median, 
Nov-Jan 
Q Summer 
1.29 
1.09-1.52 
Mean/median, Aug 
Q Aug 
1.14 
0.98-2.25 
Mean/median, Sep 
Q Sep 
1.15 
1.00-1.91 
Mean/median, Oct 
Qoct 
1.22 
0.94-1.64 
Mean/median, Nov 
Q Nov 
1.23 
1.02-1.75 
Mean/median, Dec 
Q Dec 
1.27 
1.04-2.11 
Mean/median, Jan 
Q Jan 
1.27 
1.02-2.03 
