I used a wider variety of methods in 2025 than I did in 2024. They fall roughly into four categories. 1) Trends and 2) Patterns were treated similarly to 2024. For 2025, I have added 3) Consideration of the Age Composition information (informally called “Comps”) from the Brood Tables containing details of the runs. Finally, I have added a simplified approach to considering 4) Environmental Influences on predicted run size. I used a factor analysis to reduce the PACEA set of six environmental variates to a single combined environmental index, which accounted for 69% of the total variability in the original six variates. I then searched the spectrum of correlations of run size (or an individual run component) over a range of time lags up to 40 months earlier (May is time=0). In most cases peak correlations occurred sharply at lags of 24 and 36 months, which seems biologically plausible. When Comps and the Environmental Index both appeared to be informative, I considered them simultaneously by means of multiple linear regression, usually on log-transformed values. In a few cases where the predicted run size was outside the recent historical experience, I Winsorized the prediction toward the mean and adopted a recently experienced extreme value. The mix of methods varied for each run depending on my subjective impression of the reliability of each approach, and the details are documented in run-specific spreadsheets.

The models are mostly linear or linear under log-transform, and are documented in the individual spreadsheets.