weight-at-age models and discarding

[chantell.wetzel]

I am working on a model with stationary growth where retention and selectivity are estimated with a fixed discard mortality rate.  Due to issues in estimating growth (e.g., non-trivial departures between internally estimated and externally estimated growth) and observed variable growth at young ages by year, we would like to explore a weight-at-age approach in SS3.  However, retention is not fully implemented for weight-at-age models, so we are considering options of how best to move to a SS3 model that does not incorporate retention.

A general approach could be to estimate discard mortality outside the model and then add these values to the landings to pass catches to SS3.  However, if there is size/age dependent discarding occurring, the catches removed from the modeled population would differ from those estimated from a model with discarding and retention.

Another idea that has been considered is to define separate fleets that represent either landings or externally estimated discard mortality.  This approach can be useful to account for the size/age of fish discarded if one does not want (or can't) to estimate discarding and retention in the model. The only discard composition data we have for this species are length data of observed discarded fish.  My understanding in a SS3 with growth estimation it internally converts the length data into catch-at-age estimates via growth parameters.  Since weight-at-age models in SS3 bypass the growth parameters, I am not sure if the translation of length data would be possible.  

I have the following questions:  

1.  Is it correct to assume that length data in an SS3 model that uses weight-at-age could not be used?

2.  Are there any general suggestions from the user-group from those that may have dealt with a similar situation?

Thank you,

Chantel

[Massimiliano Cardinale]

Dear Chantell,

 

I think 1. You are correct, if you use weight at age data, the growth part of SS is ignored

 

2. You are correct that retention is not fully implemented for weight-at-age models (or at least this is the message we get), however, we initially used a model with discard retention in weight at age models for North Sea sole and it seems to work as intended. At the end, for North Sea sole, we opted for a model with a discard fleet, so that we skipped discard retention, but we had discard at age data by fleet, which you don’t seem to have if I got you right.

 

Cheers

 

Max

 

 

 

“I had never been salaried and felt that the research should be its own reward”

Gorillas in the Mist

 

Dian Fossey

 

 

Massimiliano Cardinale
Associate Professor

Sveriges lantbruksuniversitet
Swedish University of Agricultural Sciences

Department of Aquatic Resources (SLU Aqua)
Institute of Marine Research
Turistgatan 5, 45 330 Lysekil
Phone: +46 761 268 005
massimilian...@slu.se www.slu.se/akvatiskaresurser
SLU Aqua on Facebook

Our vision: Viable fish stocks in healthy waters

Please consider the environment before printing this e-mail

 

 

From: 'chantell.wetzel' via SS3 - Forum <ss3-...@googlegroups.com>
Sent: 24 March 2025 18:04
To: SS3 - Forum <ss3-...@googlegroups.com>
Subject: [SS3] weight-at-age models and discarding

 

CAUTION: This email originated from outside of the organization. Do not click links or open attachments unless you recognize the sender and know the content is safe.

 

--
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/e6320de6-e0c0-4db6-a045-65e759d7f9bbn%40googlegroups.com.

---
När du skickar e-post till SLU så innebär detta att SLU behandlar dina personuppgifter. För att läsa mer om hur detta går till, klicka här
E-mailing SLU will result in SLU processing your personal data. For more information on how this is done, click here

[Ian Taylor - NOAA Federal]

Chantel,

Thank you for posting this tricky issue to the forum!

I agree with Max on 1. The growth parameters still exist in the model and I think expected values for length distributions could still be calculated, but they are internally inconsistent with the weight at age data so you would be applying one set of assumptions for the discard lengths that is different from the assumed growth for the retained fish. There may be additional problems with this approach and I think the executable may stop if you try to add length comps to a weight-at-age model.

Adding to Max's suggestion, I think there are two related issues for creating a discard fleet in a weight-at-age model in your case. The first is getting the age distribution of your discards. The second is providing weight-at-age for that fleet. If you used an age-based retention function in a weight-at-age model as max suggested is possible, you would only need the age comps for the discards, not separate weight-at-age.

If you switch your parametric growth model from using a retention function to having a separate discard fleet you would first be able to test the impact on the model results of this alternative structure. You could then use the discard fleet model to get information to help set up a weight-at-age model. First, the wtatage.ss_new file would contain the internally estimated weight-at-age for that discard fleet which could be used as an input to a weight-at-age model. Those weight-at-age values would be the result of applying the estimated selectivity at length to the growth curve to get expected selected lengths at each age and then converting them to weights.  Second, if you create placeholder observations for conditional age-at-length data for the discard fleet (with lambda = 0), it would get you the expected age distribution associated with each length bin. That is essentially the age-length key you would need to convert your lengths to ages (by multiplying a length comp vector by the matrix of age-at-length). Note that this is different from the distribution of lengths at each age used internally in SS3 and reported as $ALK in r4ss::SS_output().

Using a parametric growth model to create inputs for a weight-at-age model isn't a sustainable long-term solution to provide inputs, but would help you see a few things

1. How similar are the estimated weights-at-age for the discard fleet vs retained fish or survey catch? If not very different then you could just use a single matrix for all fleets as is typical for these models. 

2. How similar is the result of using the internally estimated age-length key to something you computed outside the model (either via an empirical age-length key or just converting lengths to ages via a growth curve).

All that adds up to a lot of work, but I think it would tell us a lot about the similarities and differences among these approaches.

-Ian

To view this discussion visit https://groups.google.com/d/msgid/ss3-forum/347b8a06b3404b958ebfb810b1880399%40slu.se.

[Richard Methot - NOAA Federal]

Here is a bit of additional information.  SS3 never needs to know discard wt-at-age because the expected value for discard amount comes from total_catch - retained_catch.  This may seem paradoxical because it can produce  output of discard wt-at-age, but it is true.  You can even get output of discard weight-at-age by including it in this data section with partition code = 2 and a negative ageerr code:

1 #_Use_MeanSize-at-Age_obs (0/1)
# sex codes:  0=combined; 1=use female only; 2=use male only; 3=use both as joint sexxlength distribution
# partition codes:  (0=combined; 1=discard; 2=retained
# ageerr codes:  positive means mean length-at-age; negative means mean bodywt_at_age

When the empirical wtatage feature was added to SS3, the features supporting parametric growth and length comp were not removed.  Those features still exist in an EWAA model, but when the model needs to get the wtatage for a fleet, it uses the empirical input and bypasses the values calculated parametrically.  Ian gives some good tips on how to check for consistency between them

Rick

[chantell.wetzel]

Max and Ian,

Thank you both.  Based upon your comments, I have some thoughts on how to move forward.

Chantel

On Monday, March 24, 2025 at 10:35:32 AM UTC-7 Ian Taylor wrote:

[Mark Maunder]

Hi Everyone,

 

We have a 3 area model and we are estimating initial recruitment deviates to define the initial age-structure. We have time varying recruitment partitioning among the three areas and want this to go back before the start of the model to influence the initial age-structure for each area, but it does not appear to be doing that. Has anyone done this before and do you have an example you can send us.

 

Thanks,

Mark