50 
Fishery Bulletin 107(1 ) 
Table 2 
History of transfers of Chinook salmon ( Oncorhynchus tshawytscha ) from the west coast of Vancouver Island, showing for each 
recipient population whether there was an indigenous stock, the enhancement status since transfers, the donor population, and 
the brood years transferred from donor to recipient sites. 
Recipient 
population 
Indigenous 
stock 
Enhancement status 
Donor 
population 
Brood years 
transferred 
Colonial 
Yes 
Not enhanced 
Marble 
1987-1989, 1991,1993,1998 
Thornton 
No 
Enhanced: river returns only 
Robertson 
1982-1984 
Toquart 
Yes 
Enhanced: river returns only 
Robertson 
Nitinat 
1989- 1990 
1990- 1997, 1999-2001 
Sooke 
Yes 
Enhanced: transfers plus river returns 
Nitinat 
1980-1984, 1987-1997, 
1999-2006 
Zeballos 
Yes 
Enhanced: river returns only 
Conuma 
1990-1998, 1999-2003 
to Sooke River (brood years 1980-2006, except 1995, 
1986, and 1998. Both Toquart River and Sooke River had 
existing Chinook salmon runs before hatchery releases, 
but populations were at a low abundance before trans- 
fers began. In addition, Marble River stock has been 
transferred to Colonial Creek, Goodspeed River (not 
sampled), and Coal Harbour, as well as various seapen 
release locations in Quatsino Sound. Other populations 
were either enhanced on-site or reared in a hatchery 
before being returned to the natal stream for release. 
For example, the Sarita River stock was reared in the 
Nitinat Hatchery, then returned to the Sarita River. 
Comparison of PORGS method 
with standard genetic methods 
To compare results obtained by the PORGS method with 
those from a standard hierarchical approach to cluster- 
ing genetic data, the unweighted pair-group method 
using arithmetic averages (UPGMA) was applied to both 
the simulated and Chinook salmon data sets to generate 
a tree with PHYLIP software (Felsenstein, 1989). The 
UPGMA approach uses successive agglomeration with 
average-linking (Sneath and Sokal, 1973). 
Results 
When PORGS was applied to the simulated data, the 
first partition ( k~2 ) occurred between P4 and P5 (Fig. 
2C). The next partition occurred between P7 and P8; 
with k = 4, P10 was separated from the P7-P9 grouping. 
The last two populations to split were P5 and P6. The 
values of the optimized cost function decreases mono- 
tonically as the number of groups k increased (Fig. 2D). 
The expected data from the reference distribution (boot- 
strapped proximity matrix) also decreased monotonically 
(Fig. 2D); but followed a less concave curve than that 
for the observed data. The optimum k value occurred 
where the observed data fell the farthest below the 
expected curve, at k = 3 (Fig. 2E), which corresponded 
to the groupings P1-P4, P5-P7, and P8-P10 (Fig. 2C). 
According to the relative positions of these populations 
along the line in Figure 2A, the group memberships 
and optimum number of groups determined by PORGS 
appears reasonable. These data were re-analyzed by 
using the bi-PORGS method, which generated the same 
cluster groupings as PORGS. 
These groupings are also consistent with the results 
depicted by an UPGMA tree (Fig. 3A). Figure 3A shows 
a vertical dashed line drawn to intersect branches that 
correspond to the three main clusters identified in the 
PORGS analysis. The UPGMA tree shows P10 with 
the longest branch length and P5-P6 with the shortest 
branch length, corresponding to both population loca- 
tion along the line (Fig. 2 A) and the results from the 
application of the PORGS method (Fig. 2C). 
The structure of the Chinook salmon data was evi- 
dent when populations were sorted as an anti-Robinson 
matrix (Fig. 4; Robinson, 1951). In Figure 4 the small- 
est dissimilarity values appeared close to the main 
diagonal, resulting in a grouping of the most similar 
populations. Four main clusters were apparent, and 
the two Quatsino Sound populations, Marble River and 
Colonial Creek, were the most distinctive of all the 
populations (0>O.O4). Three other groups lay along the 
main diagonal, where 0<O.O2, corresponding to northern 
(Nootka Sound), central (Clayoquot+Barkley sounds), 
and southwest Vancouver Island. The Toquart River and 
the Sarita River populations, although geographically 
part of Barkley Sound, clustered with southwest Van- 
couver Island populations. The Gold River population 
straddled the northern and central populations. The 
most genetically similar samples (0=0.002) were those 
from Somas River in the early 1970s and those from the 
more recently sampled Robertson Creek hatchery. 
Using the Chinook salmon data set, we were un- 
able, in an initial attempt, to find an optimum solution 
from random set partitions. Although some partitions 
occurred only a few times, others were more common 
(Fig. 5). After 5.0 xlO 8 iterations, no cluster combina- 
tions were evaluated where block sizes were less than 
2 or greater than 16. The expected number of occur- 
rences for each k (Stirling number of the second kind) 
