Issue setting up recdevs and initial catch, short TS of bio data

[Gwladys Lambert]

Hi dear SS3 users,

I tried to post a message last night but it may not have worked so trying again (ignore previous if it shows up as I made further investigations since).

I am exploring a new SS3 model for scallops (relatively sessile, simultaneous hermaphroditic species), with a short time series of indices and length/age, and having issues with setting up the initial catch equilibrium and rec devs.

 

I don't have much info on pre-2000 catch levels but however I set it (0, or some value with a tight or large s.e., with lambda =0 or 1 for the log-likelihood) makes a difference into scaling and direction of SSB trajectory.

And it interacts with how I set up the rec devs. Option 1 and 2 (for the deviation vector) give me conflicting results – always on the verge of/ raising a convergence flag.

Option 1 (rec devs centered on 0) leads to completely ignoring LPUE and starts SSB pretty high, while Option 2 fits the LPUE, but leads to positive rec devs  (not sure why it would do that rather than scale R0 up since it is being estimated?) and it also increases dynamic B0 (suggesting it wants to point to a regime change?) 

Depending on the set up, I either get a realistic SSB in the final years (I have a direct biomass estimate for the stock), or it ends up at about half the size I'd expect. 

Length and ALKs always reasonably well fitted. SRR seems inexistant – SSB range is relatively narrow and no information re functional form of SRR (likely largely environmentally driven) so using a bevholt and steepness of 0.9.

 

Copying the data plot here for info, as well as catch and LPUE

 

I’d be after recommendations on what to set the equilibrium catch to, what year to start main and early recdevs (data include mostly scallops age 2-6 for the survey, and 3-6 for the fishery so for now I am using 2017-2022 as main recdevs and 2012-2016 as early rec devs - playing with that also have an impact on the scaling and direction of SSB trajectory), and with what dev option, as I am going round in circles now... Anything else I need to consider that I may have overlooked?

 

Let me know if further info required. And many thanks in advance if anyone has any pointers.

 

All the best,

Gwladys

 

 

[Gwladys Lambert]

Trying to re-post images, didn't seem they were showing for me

[Ian Taylor - NOAA Federal]

Hi Gwladys,

Thanks for sharing your problem and successfully attaching the images to the most recent email.

Do you have confidential data, or can you attach the SS3 input files? It's hard to know what advice to give based on your description.

I know I've seen other SS3 models for stocks in the NE Atlantic or Mediterranean that have a similar time series length and similar questions about setting up the initial conditions, so hopefully folks with experience in those cases could post some comments here.

-Ian

--
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/b3a31c77-714d-499f-9216-987037dc6de0n%40googlegroups.com.

[Richard Methot - NOAA Federal]

Hi Gwaldys,
You can send your draft input files to Ian and I at nmfs.stock...@noaa.gov

When I respond, I'll respond to the whole forum because, as Ian says, this is a common situation so best if all can see the ideas.

Rick

[Gwladys Lambert]

Hi Ian, 

Thanks for your response - yes it's a difficult problem to explain in a few words, lots of intricacies. I probably didn't do the best job of wording it! There are no confidential data there, just the caveat that some data have not all gone through proper QC and are unpublished, subject to small changes, so I'd rather email them to anyone interested in taking a look rather than put them on a public forum if that is ok.

Cheers,

Gwladys

[Gwladys Lambert]

Yes that's completely fine. I made use of other discussions here so more than happy to support if anything of value in my issues and questions!

[Richard Methot - NOAA Federal]

Here's a general response and I'll respond directly to you separately after studying the files and the report that you shared.

1.  Initial catch is very important in situations where there is little contrast in the time series of catch and abundance.  It is analogous to the initial depletion parameter in a biomass dynamics model.  Without time series contrast, SS3 is often most informed by the earliest composition data in the time series. Initial catch helps the model estimate initial depletion, which relates to initial Z and the age/size composition of the stock when the first composition data are collected.   SS3 is essentially doing a catch curve analysis that manipulates available parameters to match that early comp data, and many parameters interact!  M, R0, early recruitment trend, and domed selectivity all conspire to make this difficult.  Whether or not you can reasonably assume asymptotic selectivity for at least one fleet or survey will have a big effect on how informative the comp data are.  A catch curve analysis with freely domed selectivity will not be very informative.

2.  The data show an increasing trend in catch and LPUE over the 25 year time series.  This could be due to several possible causes:  a.  expansion of fishing grounds; b.  environmentally driven increase in recruitment and stock abundance; c.  historical depletion before 2000 and recovery of stock due to lower exploitation after 2000; d.  market driven increase in demand with increased LPUE due to increased fishing efficiency.  SS3 can help you explore option c, but only if you can establish a reasonable bound on catch before 2000.  Given the region, I cannot imagine that it was near 0.

3.  the dredge surveys seem to have a calibrated efficiency, providing a direct estimate of biomass.  I would use this directly in SS3.  Given that the stock seems to be increasing, SS3 or any pop.dy. model will have a very difficult time estimating absolute stock abundance without more information.  The Q of the dredge survey could be the info you need.

4.  Steepness of the SRR will be very difficult and very dependent on the initial depletion.  The trends you see are for increasing catch and LPUE.  This means that if early catch was high enough to deplete the SSB/R then reduced catch in early 2000s could have allowed for SSB/R to increase, and if steepness is low then the increased SSB/R will allow increased R also.  So, your initial catch estimate will affect steepness estimation because steepness affects the speed of stock recovery.

5.  With the comp data not starting until recently and being mostly for ages 2-6, I suggest starting the main recdevs a few years before the comp data begin and keeping the early recdevs fixed at 0.0.  With early recdevs estimated, I think the model will have too much flexibility to do strange things in order to match the early comp data.  

Rick

Richard  D. Methot Jr.

Stock Assessment Research Scientist (ST)

Northwest Fisheries Science Center

NOAA Fisheries | U.S. Department of Commerce

Office: (425) 666-9893

Mobile: (301) 830-2454

www.fisheries.noaa.gov 

To view this discussion visit https://groups.google.com/d/msgid/ss3-forum/99157aa9-9ce8-4e8b-9210-0a9670dfc4e2n%40googlegroups.com.