Best practices for setting starting value of initial equilibrium catch

[Laura Lee]

I’m wondering about best practices for setting the initial equilibrium catch starting value. I have come across a number of different approaches including using the minimum of the catch time series (and sometimes divided by 2), average of the entire catch time series, average of the beginning of the catch time series (e.g., 5 to 10 years), average of historical catch time series not used in the modeling, etc. In my current project, I have catches dating back to 1867 but catches prior to 1978 are highly uncertain so I am not currently including them in the model; however, the highest catches of the time series occur in the historical (pre-1978) period.

I am interested in how others develop their starting values for initial equilibrium catch. Is there currently a ‘best practice’ I should be aware of?

Also, in what circumstances would I set the lambda associated with initial equilibrium catch to 0? I remember seeing documentation about this, but I can’t find it.

Thank you for your time.

  

Best,

 Laura

[Nathan Vaughan - NOAA Affiliate]

Hey Laura, 

         In the Southeast/Caribbean we have started shifting towards setting very large uncertainty on initial equilibrium catch and allowing the model to freely estimate it because we usually have poor information in this period. I think setting lambda equal to zero would achieve something similar. Our primary concern is that this value is often very influential on the final stock status and sustainable yield estimates which is not something we want from a poorly informed value. We also intend to look into other methods of decoupling initial depletion from sustainable yields (using an offset on initial equilibrium R0, or adding a dummy depletion fleet in the first model year) but haven't landed on an optimal solution yet. Looking forward to hearing what approaches others have tried.

Cheers,

Nathan

[Richard Methot - NOAA Federal]

Good topic Laura.  I do not think there are universal best practices.  We have captured some ideas in this short essay on the SS3 website.

https://nmfs-ost.github.io/ss3-website/qmds/start_year_considerations.html

I'll add a few more comments.

1. If you see an increasing trend in the population index, then SS3 may try to increase initial equilibrium F in order to deplete the starting population and to decrease steepness in order to allow that initial F to delete the initial population further.
2. If the age/length composition early in the time series has a low % of old/large fish, then SS3 will look to either have dome-shaped selectivity or a high F (conditional on M).  This is the same consideration as occurs in a data-limited assessment that does a catch curve analysis to find the M+F that best matches a sample of the population's age or length comp.  
3. So, if you treat initial equilibrium catch as unknown, then SS3 will estimate the initial equilibrium F to help address the two points above.  If you provide data on initial equilibrium catch, the possible range of solutions will be refined.

Rick

[Laura Lee]

Thanks so much for the feedback, Nathan and Rick! Would love to hear from others as well.

--
You received this message because you are subscribed to the Google Groups "SS3 - Forum" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ss3-forum+...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/ss3-forum/1f2c4517-d8ca-4450-afa8-114e021c8b0an%40googlegroups.com.

[Shanae Allen-Wertz]

Hi Laura,

Rick, Nathan, and the start year vignette already covered most of the key considerations for setting the start year and initial equilibrium catch, but I’ll add a few additional thoughts.

As Rick mentioned, if the initial equilibrium catch is treated as unknown (λ = 0), the model will lean heavily on early age/length compositions, natural mortality, and selectivity assumptions. Based on the vignette, it may be best to begin the model in an unfished or lightly fished state, even if early landings are highly uncertain. You could then capture that uncertainty by running a bootstrap routine.

I’m currently testing the effects of these model configurations in a simulation exercise, which so far suggests that the fit to the stock recruit curve and MSY reference points are more sensitive to initial equilibrium conditions than SPR proxies, although the full results aren't ready for primetime just yet.. One possible diagnostic is to profile on the initial equilibrium catch values (λ = 1). I’m happy to share my code if that would be helpful. Also, if your model is sensitive to whether steepness is used in the equilibrium, that may indicate that historical landings were high enough to reduce recruitment—which could also suggest the need for an earlier start year.

All the best,

Shanae

To view this discussion visit https://groups.google.com/d/msgid/ss3-forum/CAMUj-RCYuX9LtE5BoN8Xj3mzu3YbPJZb%3DkNBE1VNm4uCuMCM%3Dg%40mail.gmail.com.

[Mark Maunder]

We have discussed this several times in the past. After a quick look, I could not find a summary of the discussion hopefully someone has one. My preference is to treat the initial conditions as simply reducing the number of parameters to represent the numbers at age and not assign meaning to it by fitting to equilibrium catch. I also prefer short term models to long term models for which there is no data (comps and/or index) in the early years, which is consistent to not fitting to equilibrium catch. I also prefer dynamic target reference points which lessens the importance of initial conditions. I know that there are others that disagree with this point of view.    

To view this discussion visit https://groups.google.com/d/msgid/ss3-forum/CAAeO-dxiSgJE3C%3DyEfHDT3FWwDDfoa6FJdDMUAF4brG4pJDQTg%40mail.gmail.com.