Recruitment estimates are extracted from stock assessments conducted within the last five years. Recruitment estimates were only extracted from assessments where recruitment deviations were estimated in the model. These assessments were all conducted using Stock Synthesis, which estimates annual recruitment in an integrated modeling framework, where recruitment is informed by multiple data sources to the model, but primarily age and length compositions.
The recruitment estimates we are using for this analysis are the log recruitment deviations around a Beverton-Holt stock-recruitment function, with their variability constrained by a parameter called SigmaR, which functions similarly to the standard deviation of a random effect. Recruitment deviations can either be estimated with a constraint where the annual deviations must sum to zero or where this constraint is lifted.
For our stocks, these stocks had a sum to zero constraint:
These stocks did not have a sum to zero constraint:
Recruitment is classified into boom, bust, and average categories for the purposes of visualization. These categories were produced by classifying any year with recruitment > 1 SD above the long-term mean a “boom” recruitment event, any year with recruitment < 1 SD below the long-term mean a “bust” recruitment event, and all years within 1 SD of the long-term mean an “average” year.
Autocorrelation in each of the recruitment time series is also visualized below. While persistent oceanographic conditions can lead to autocorrelation in recruitment deviations, autocorrelation may also result from aging error, which will “smear” out recruitment deviations and cause consecutive years to have autocorrelated recruitment deviations when in reality it is likely the signal from only one of these years being “smeared” across multiple years. This is particularly apparent in long-lived species, such as Yelloweye Rockfish (see below).
To explore whether the boom/bust distribution of recruitment deviations might be better represented as a mixture, a two distribution mixture mdoel was fit to the recruitment deviations for each stock. Additionally, Hartigan’s dip test was used to test if each distribution was significantly different from a unimodal distribution (spoiler alert, it was not for any stock).
| stock_name | dip.test.p.value |
|---|---|
| Black Rockfish - Northern CA | 0.134 |
| Black Rockfish - Central CA | 0.965 |
| Black Rockfish - Oregon | 0.993 |
| Black Rockfish - Washington | 0.815 |
| Canary Rockfish | 0.962 |
| Copper Rockfish - CA, North of Point Conception | 0.905 |
| Copper Rockfish - CA, South of Point Conception | 0.740 |
| Dover Sole | 0.304 |
| Lingcod - North of 40°10’N | 0.626 |
| Lingcod - South of 40°10’N | 0.991 |
| Petrale Sole | 0.551 |
| Quillback Rockfish | 0.991 |
| Rex Sole | 0.944 |
| Rougheye/Blackspotted Rockfish | 0.990 |
| Sablefish | 0.624 |
| Vermilion & Sunset Rockfish - Southern CA | 0.743 |
| Vermilion & Sunset Rockfish - Northern CA | 0.772 |
| Vermilion Rockfish - Oregon | 0.896 |
| Vermilion Rockfish - Washington | 0.714 |
| Yelloweye Rockfish | 0.934 |
| Yellowtail Rockfish | 0.679 |
## number of iterations= 315
## One of the variances is going to zero; trying new starting values.
## number of iterations= 36
## number of iterations= 102
## One of the variances is going to zero; trying new starting values.
## WARNING! NOT CONVERGENT!
## number of iterations= 1000
## number of iterations= 101
## number of iterations= 65
## number of iterations= 95
## number of iterations= 404
## number of iterations= 23
A mixture model could not be fit to this distribution.
## number of iterations= 88
## number of iterations= 95
## number of iterations= 210
## number of iterations= 35
## number of iterations= 450
## number of iterations= 56
## number of iterations= 12
## number of iterations= 53
## number of iterations= 34
## number of iterations= 58
## number of iterations= 73