Here is some Willapa information.


I have worked on the Willapa Bay Chinook baseline analysis but havenít had time to think hard about the results. Ken Warheit helped with the analyses by running a population mean centered PCA and an assessment of mixture analysis power using a program he wrote that implements a method described by Eric Anderson and others.

Analysis of the data set we generated from the Willapa Bay tributaries Chinook samples that were sent to the Molecular Genetics Laboratory suggests that there is little population structuring among the donor populations. For these analyses we combined the samples from each of the three hatcheries with the samples taken from the adjacent streams.

Pairwise Fst s were all well below 0.01 (table below) and indicate that most of the genetic variance is within populations rather than between them.

FST (theta)

ForksCr Naselle Nemah


NorthRiv 0.00429 0.00346 0.00521

ForksCr 0.00068 0.00212

Naselle 0.00116

We (Ken Warheit and I) ran exploratory principle component analyses (PCA) to see if there was enough variance structure in the data set to separate samples into more-or-less discrete population clusters. The analysis that I ran was naÔve to sampling locations and showed perhaps a weak offset to the North River/Fall River group (red) from the other three groups.

Ken ran a PCA that was aware of population membership and was based on population means, an approach that is similar to Discriminant Analysis of Principle Components (dapc) in the R-package adegenet. Kenís scatterplot shows significant overlap between the population clusters but the North River and Forks populations show greater dispersion along the first and second axes respectively. The Nemah and Naselle samples form tighter clusters that overlap substantially.

The first plot looks at individuals distributed around a study-wide central value while the second plot looks at individuals distributed around population mean values. Confusing?

We also looked at assignment accuracy in two ways. Ken ran a program he wrote that implements a method developed by Eric Anderson and others to estimate the accuracy of genetic mixture analysis stock assignment the four populations. The plot below shows the results from 100 simulated mixtures. Simulated fish from North River/Fall River misassigned (assigned to one of the other 3 populations) about 20% of the time and the other populations had higher misassignment rates.

None of these analyses address misassignment rates of tule Fall Chinook to any of the Willapa Bay populations Ė which might be a bigger issue than misassignments within the Willapa Bay population complex.

I will be out of town until August, but we can talk about this when I get back. Maybe I will have additional insight when I get back after letting this bounce around in my head for a bit.

Sewall F. Young
Molecular Genetics Laboratory
Washington Dept. of Fish and Wildlife
600 Capitol Way N.
Olympia, WA 98501-1091

Phone: 360.902.2773

Edited by Rivrguy (02/23/18 10:58 AM)
Dazed and confused.............the fog is closing in