search

cgrandin / csas-latex

Latex style file and template for CSAS Research Documents
0 stars 6 forks source link

help with interpreting PRD retro plots #86

Closed jaclyncleary closed 7 years ago

jaclyncleary commented 7 years ago

I don't understand what's happening with PRD retro plots. Specifically why is the retro-4 (2014). Fig 10b shows a high number of age-2s in 2014, but the survey index barely changes from 2013 to 2014 and the estimated spawning biomass is constant through that time (Fig10a).

PRD AM2.docx

cgrandin commented 7 years ago

I think the high recruitment in the end year (2014) is driving the model. It might be worth running the retrospective back 15 years to 2002 for PRD, where there is a similar large recruitment in the end year and similar survey index pattern to see if the model behaves the same way. You could do it in iscam-gui just to have a look

Shawkshaw commented 7 years ago

I reran the -retro 4 and double checked the input files and have not found anything wrong with this run.
I ran the retro -15 and the sbt does not shoot up to the 40s in the last year like it does for -4 and recdevs do not shoot up either.

I will keep looking for what is making the 2013/2014 years go up so much...

recdevs retro 4_15 prd

sbt prd -15

jaclyncleary commented 7 years ago

Sarah- can you post PRD retro-15 for both AM1 and AM2? I'd like to have it as a slide and I think we'll want to add the retro-15 to the resdoc IN PLACE OF the retro-10 for PRD. Thx.

Shawkshaw commented 7 years ago

I updated these into the Herring repo retrospectives folder

cgrandin commented 7 years ago

I'll have to modify the build() code so that it brings this in. I'll let you know when it's done.

cgrandin commented 7 years ago

Ok, the switch has been made in commit 1214d75b233a2588f415cd86af71c21d361312e0. You need to copy the 15 folder over and then: source("all.r");build(ovwrt.retro=TRUE) and when you build the doc, the PRD plots will have changed.

I struggle to see why the recruitment is so large in the last year. I checked all the parameter outputs and there are no other major differences between retrospectives.

Shawkshaw commented 7 years ago

I added the -retro 11-14 runs for PRD AM1 and AM2 to the Herring repo.
Jaclyn, did you want these added to the plots as well or did you just want to look at them?

jaclyncleary commented 7 years ago

please add them to the plots

cgrandin commented 7 years ago

I can do it.. just working on the prior/post plots first

cgrandin commented 7 years ago

Done in commit 95c1432beaf318b53e6edbe1fb9b174ac87ccd0c.

Once you copy the folders 11-15 in, you'll have to run build(). Don't worry it's fast since it's just the new retros.

I think I'll try to improve the legends so they don't cover up all the action.

cgrandin commented 7 years ago

Legends are much better. I also expanded the y scale so you can see more of what's going on. Commit c14dab812ee936b6901445d9ef58827c76ff2aa1