search

gavinfay / fluke-mafmc-recdisc

Code for runnning analyses for the 2021 MAFMC Summer Flounder MSE focused on recreational discards
3 stars 3 forks source link

update fleet selectivity at length #2

Closed sgaichas closed 2 years ago

sgaichas commented 2 years ago

@gavinfay am I correct that this is what I will be replacing after converting Mark's selectivity at age from the assessment?

https://github.com/gavinfay/fluke-mafmc-recdisc/blob/d1f188c4ce78b9f28aba7232a359dbe4ba7d0ddf/mse/Sinatra.CTL#L108-L112

gavinfay commented 2 years ago

@sgaichas yes that's right. Thanks!

sgaichas commented 2 years ago

@gavinfay is it better that these starting values represent a smooth curve or stay pretty close to what the assessment would have predicted? The assessment selectivities are not smooth so they result in something that will deviate from the assessment if I fit a function to it. I could just do an interpolation between points to get the length-bin specific values if we want to stay closer. Also, the highest mean length is far from the max bin of 92 cm currently in Sinatra, so some extrapolation will be needed either way:

image

gavinfay commented 2 years ago

@sgaichas Thanks for looking at this. In the past to condition these OMs based on purely age-structured assessments I have predicted selectivity at age given a functional form of selectivity at length and growth curve parameters, rather than fit a curve of selectivity age in terms of mean length.
Perhaps I made this more complicated at our meeting earlier this week.
Depending on the growth curve parameters, there will be some large fish in the population even if the mean length of the plus group is 60cm.
I think it makes sense to make all the calculations that rely on growth internally consistent.
I added scratch.xlsx to the repo, which is the spreadsheet I used for setting up the initial version of the model for the MAFMC fluke work Jason McNamee and I did.
It's currently set up to estimate a logistic maturity at length given the input maturity at age and growth curve parameters. We could change this to a double logistic for the selectivity to represent the doming - indeed I might have an example of this for another case study stock.
I updated the maturity at age in here from the most recent fluke assessment, but the other pieces (growth curve parameters, W-L relationship) are based on the 2018 assessment. We should check that these are consistent with the new assessment. I am wondering if we should increase the age of the plus group so that the size distribution of fish is properly accounted for (only makes a real difference if fish haven't gotten close to L infinity by the time they reach the first age of the plus group).

sgaichas commented 2 years ago

@gavinfay I have selectivities for the landings fleets and working out strange dome fits for the discards

What order do you want the fleets in the Sinatra.CTL file? I'll ensure the correct order and send a pull request with this update plus updates to length-weight equation parameters and weights at age.

gavinfay commented 2 years ago

Thanks! Fleet order is:

  1. Commercial landings
  2. Commercial discards
  3. Recreational landings
  4. Recreational discards
sgaichas commented 2 years ago

@gavinfay so the discard selectivities as fitted by double logistic are a bit strange. I'm comparing fitted curves (curves 0 and 1 which just makes full selectivity =1) with some forced down to no selection (curve 1d) for the largest sizes here--an extreme assumption about selectivity, but at least it bounds the issues.

I am unsure how we want to represent discards in the commercial and recreational fisheries. If we believe no legal sized fish are discarded, forcing the curve to 0 for larger sizes may be ok, but if we think there is some relatively constant discard above a certain size then keeping the fitted curve may be ok for both fleets. However I think the fitting is extrapolating from that last point which as Mark explained has many reasons to be a bit off.

image

It matters less for recreational discards since they will be updated but we probably want commercial discards in the right ballpark.

Thoughts on which curve to use? I can also continue to fiddle with the descending limb of the double logistic until we get something between if that is more reasonable.

gavinfay commented 2 years ago

Thanks! As we discussed I think it makes sense to go with '1' and see what this means for the population model output derived quantities then adjust if it seems to be giving something wacky during iteration with the WGs.

sgaichas commented 2 years ago

OK, sounds good. I'm sending a pull request with this and several other items updated to try to align with the ASAP inputs.

gavinfay commented 2 years ago

Fixed with PR #4 f6f056f.