While this is an interesting idea, there are some important details missing from the description of the 'data' that are used. Specifically, the RAM legacy stock assessment database consists almost entirely of stock assessment model output, not raw/real data. The SSB time series in particular are heavily filtered, smoothed, and interpolated, frequently exhibit monotonic or U-shaped trends, and have a very high degree of linear autocorrelation. Some are constant for long periods because they relax to steady state in the absence of real data to inform the model. Additionally, the recruitment estimates in many cases come directly from the internally fitted stock-recruit model, not from empirical estimates. It is highly likely that the analysis is not really recovering the biological dynamics, but rather the dynamics of the filtering models.

The difficult-to-believe finding that there is only one attractor for all the world's fish is better explained by the fact that what is being called data is actually model output, the vast majority of which come from a single modeling framework (with linear structure except in recruitment). This modeling framework is overwhelmingly likely to produce stable dynamics; more complex dynamics can only be obtained through external forcing which the filtered data have (by definition) had filtered out. (Note that the fact that EDM 'works' does not automatically imply that the dynamics are complex and nonlinear because EDM also 'works' in simple, linear systems; additional tests are required to assert this.)

The similarity of 'dynamics' by taxon/trophic level and geographic region can be explained by the fact that stock assessment scientists tend to model similar species using similar models, and that there are regional differences in assessment culture (managers from a given region tend to use similar models). For example, nearly all US west coast rockfish stock assessments are age-structured models with Beverton-Holt recruitment dynamics and a constant mortality rate. Given the confounding between biological information and modeling choices, it would make for a more convincing, interesting, and worthwhile study if direct observations of fish populations (such as the fishery independent survey data from around the globe, readily available in the FishGlob Database) were used instead of model output.

Thank you for considering my comment, Dr. Tanya Rogers