add more info on reasons for choosing discard approach?

[iantaylor-NOAA]

In two independent settings today I was asked about approaches to modeling discards. The section in our handbook on discards, which @chantelwetzel-noaa recently expanded, has lots of really valuable detail about the different ways to model them: https://pfmc-assessments.github.io/pfmc_assessment_handbook/01-data-sources.html#notes-and-best-practices-for-observer-data-and-discards.

However, I think it would be useful to provide a bit more guidance on why you would choose to estimate them within the model vs outside. Here's a rough representation of my thinking on this subject. Missing from this is the nuances of recreational data where we have things like death-by-depth estimates applied by the states.

Primary reason for discarding	What to do if discards are high or length comps differ	What to do if discards are low or length comps very similar
Catch limits or lack of market for all sizes (e.g. dogfish)	Model separate discard fleet (challenge is extrapolation to period without discard data)	Combine discard and retained catch and ignore discard comps or combine retained + discard comps
Size limits or lack of market for small fish (e.g. lingcod)	Model retention within the assessment (probably needs to be time-varying)	Combine discard and retained catch and ignore discard comps or combine retained + discard comps

Does this reasoning match how other people are thinking about it? Would it be useful to add?

Tagging @31ingrid and @brianlangseth-NOAA.

[chantelwetzel-noaa]

I would add an additional row for species with low observation rates within the WCGOP data (e.g., nearshore species). In this instance you would may not want to model retention but would want to input catches (landings + discards) into the model. In this instance you also would want to look at the GEMM or get input from the states around discarding practices across time to determine the catches, since the rates in WCGOP may be highly variable and not representative of discarding practices on average given low sample sizes.

[brianlangseth-NOAA]

@iantaylor-NOAA This is a great start. I would be curious to know the metric for whether discard is high or not. Species with catch limits contributing to the reason for discards may have high discards relative to catch (because of restrictive limits). Does that mean its high? Im not sure

I think for me, the bigger issue is having details of how to go about modeling discards. linking to github files for a few species could help. if Im planning to combine catch and discards together, having code to show how others have done it would be helpful.

Something more detailed like, for combining discards with catches you can
a) Pull the gemm report, and sum together discard estimates by sectors that match your fleet definitions. Add these discards by year to your landings estimates from PacFIN.
b) If your fleets are state based, because the GEMM does not provide state-specific commercial discard estimates, use the landings AND discards from WCGOP data for the desired fleets, sum these across states to come up with state proportions and multiple these by the coastwide discard estimates for that sector from the GEMM. This was explored for copper rockfish and quillback rockfish in 2021.

For using discard rates
a) be wary of pulling rates from WCGOP data as these data are a subset of the total discard data that is used by the GEMM.

I still dont know what 'rates' to use, should I want to include discards as discard rates. Are they rates from WCGOP, rates from the GEMM, or rates calculated from the GEMM's discard estimates and landings from PacFIN? Im not sure, and Im sure many different ways have been used. Alltogether, clarifying the details of what to do would I hope reduce the amount of questions each cycle.

[chantelwetzel-noaa]

The challenge in my opinion is that for determining how to handle discards, either inputting catches or estimating discarding in the model, varies by species and the sectors catching that species.

Species caught by the catch share fleet can often be easier, especially if the catch share fishery is the primary source of mortality. In these case we have relatively good observer information prior to 2011 and then full observer coverage starting in the 2011. In these instances the most common approach is to estimate discard via retention within the model. Additionally, depending upon the information you could estimate retention and discarding within the model for select periods while inputting catches (discard + landings) for other periods by using time-blocks on retention. For example, the earliest catch reconstruction years for Dover sole represented catch while later years only had landings for the same fleet. This was handled via time-blocks on the retention curve.

If a species was not a quota species or not a species often landed (Pacific spiny dogfish, big skate), the landings data will be only a small portion of the catch. Additionally, there may be limited information on discarding practices, particularly before the start of the WCGOP program. In these instances, assessors have looked at other factors (catch of other species) to estimate catch outside the stock assessment model.

For species where removals will be comprised from a non-fully observed sector, the discard rates from WCGOP observation may be used to estimate discarding in the model but more careful consideration is needed. If you species is commonly observed by WCGOP the discard rates may be reasonably representative of discarding. However, if the observations of the species being assessed is more limited the discard rates based on only observed trips may be more variable. In these instance, it is recommend looking at the discard rates calculated based on the observations and comparing those rates to the estimated discards within the GEMM. The GEMM calculates expansions of landings and discards based on finer resolutions (areas) and sectors using the observed data. If the rates from the observations provided by @chantelwetzel-noaa are not generally similar to the calculated overall discard rates from the GEMM, then you may want to apply a discard approach that aligns more closely with the GEMM estimates.

Also, remember within a single model, one can have a mix of fleets where the input removals are landings and/or catches since retention is fleet specific. Regardless of how catches are determined, estimated within the model and calculated externally, it is critical that there is comparison between those values and the values within the GEMM. While these values would not be expected to match, they should be very similar. In the instance where discard mortality is being estimated within the model and those estimate do not have reasonable coherence with the GEMM values, one may want to re-evaluate either the model fit to the discard data or the representativeness of these data.

[Rick-Methot-NOAA]

Important to get SEFSC perspective where much discarding is due to bycatch in fisheries targeting other species. Do they have same selectivity for the subject species as fisheries that are targeting the subject species?

[iantaylor-NOAA]

@Rick-Methot-NOAA, the table was definitely west-coast focused and incomplete even for this region.
Folks from SEFSC provided lots more good info during the FIMS meeting.

[iantaylor-NOAA]

I will attempt to aggregation the information from this issue into the handbook for people to look over and edit in the future as needed.