plot_biomass plots a timeseries of landings when provided the example data

[Ovec8hkin]

The plot_biomass(..., module="TIME_SERIES") function, when applied to the example data (e.g., example_data.rda), appears to plot a time series of landings as opposed to a time series of stock biomass (e.g. weight of individuals in the populations, as the function name would seemingly imply). This appears to be related to the fact that "biomass_retained" and "biomass_dead" are also returned, and are what end up being plotted.

If this is intentional, the function name and function documentation are not clear that that is intended behavior.

Additional tests with other modules also did not yield the expected population biomass timeseries.

[Schiano-NOAA]

Thanks for bringing this to our attention @Ovec8hkin ! I am going to discuss some things with our steering committee members. We do want the biomass retained and dead, but we might be missing an additional value to this. There have been some interesting things to deal with while standardizing naming conventions and trying to get comparisons across models.

[Schiano-NOAA]

@sbreitbart-NOAA and I have been working on remedying this issue. We investigated the converted output (thinking this would be the issue), but all seems to be accurate. I think the issue regarding biomass and landings plots being the same lies in what Stock Synthesis calculates biomass when indexed. It does look like there is a slightly higher biomass than landings so they aren't exact, but definitely should not be this close.

@iantaylor-NOAA do you know when producing estimated biomass by fleet, should it be the combination of biomass selected, discard, and retained? or some combination? In the example we used, the landings and biomass plots were almost exactly the same (biomass was using biomass_selected from SS3). However, when Sophie and I tried a combination of these estimates, it still only really resulted in a proportional relationship to the landings plot.

When plotting total biomass, it results in an expected plot and similar to the one found in the assessment:

We looked at it with a sum of biomass selected, biomass discarded, and biomass retained resulting this:

Are any of these methods appropriate? or should biomass be extracted from somewhere else?
Maybe it's landings that's inaccurate?

[iantaylor-NOAA]

@Schiano-NOAA, sorry the columns in TIME_SERIES are not well described anywhere and the table includes a mix of many things.

All of the columns with fleet numbers (e.g. sel(B):_1 dead(B):_1 retain(B):_1) are related to the catch by fleet (selected, dead, and retained, where those numbers will all match each other unless the model is internally estimating discarding). All of the columns without fleet numbers are related to the population as a whole (sometimes subset by sex (e.g. SX:1) or growth pattern (GP:1).

It's not possible to divide the population biomass by fleet, so time series plots that we've made in the past have been for a single population-level quantity, such as female spawning output/biomass (SpawnBio), summary biomass (Bio_smry = biomass of both sexes above the reference age reported above the header of TIME_SERIES (e.g. BioSmry_age:_1), or total biomass (Bio_all which is biomass of both sexes for all ages). The closest thing to a fleet-specific measure of population biomass is the idea of vulnerable biomass, which is independent of catch. Extracting this information from SS3 is discussed in this recent r4ss issue: r4ss/r4ss#1062.

Also note that for (relatively rare) spatial models, TIME_SERIES has separate rows by area, but recent versions of SS3 also include an ANNUAL_TIME_SERIES table which aggregates across areas.

[Schiano-NOAA]

@iantaylor-NOAA okay this clarifies a lot! Thanks so much. I'll make sure to make these changes in the converter and the the biomass plot.

Do you know of one of those rare models that have spatial aspect so I can see an example of it and ensure it functions with the converter as well?

[kellijohnson-NOAA]

@careymcgilliard do you have an example of SS3 with space that @iantaylor-NOAA is talking about that you can share with Sam?

[iantaylor-NOAA]

@careymcgilliard has one of our few multi-area production assessments, but there is also a spatial model available here: https://github.com/nmfs-ost/ss3-test-models/tree/main/models/three_area_nomove (there's no Report.sso file included in the repo, so would need to be run with current SS3 executable to get that).