possible bug in readFLSss3 with new version of SS (3.30.19)

[hgerritsen]

Hi
I upgraded to SS v 3.30.19.
When i used readFLSss3 i found that the catch.wt was incorrect in the FLStock object. I changed the argument wtatage to TRUE, which fixed the catch.wt problem but introduced a new problem with the m slot. Neither of them give a warning or an error. This problem did not occur with the previous version of SS.
My SS model has 1 season, 1 area, 2 commercial fleets and 2 sexes; I can provide the model files if it helps.

[iagomosqueira]

Yes, getting the files would help. What version of r4ss is installed?

[kellijohnson-NOAA]

SS version 3.30.17.01 changed the parameter label from "NatM_p_1_Fem_GP_1" to "NatM_uniform_Fem_GP_1" which might be causing problems.

[iagomosqueira]

Has r4ss changed to reflect this? ss3om takes it from SS_output$endgrowth.

[kellijohnson-NOAA]

Looks like endgrowth takes realized natural mortality, i.e., "Natural_Mortality", from the table of natural mortality values rather than the "parameter" so I think everything should be fine with respect to endgrowth.

[hgerritsen]

Apologies, the problem is with maturity, not natural mortality. My bad. Will send my ss files tomorrow.

[hgerritsen]

CheckreadLFSss3.zip
Attached are two sets of SS files, same model run in 30.3.18 and 30.3.19
Also a short script to compare the two.
I am using:
R version 4.1.1 (2021-08-10)
Platform: x86_64-w64-mingw32/x64 (64-bit)
Running under: Windows Server 2012 R2 x64 (build 9600)
ss3om_0.5.0 r4ss_1.43.0 FLCore_2.6.18

[hgerritsen]

I think the wtatage argument might be a red herring. It does change the maturity slot but it does this consistently between versions of SS.
The immediate issue is the difference in catch.wt, landings.wt and discard.wt between SS versions when wtatage=FALSE (default value)

[iagomosqueira]

Using current r4ss from github, the Mat_F_Natage column of the endgrowth data.frame comes up as character with some strange values when loading from version 19

s19<-r4ss::SS_output('SSv19/')
s19$endgrowth[, 'Mat_F_Natage']

shows

 [1] "9.14165e-050.5"    "0.0004234360.5"    "0.00284120.500309" "0.01628550.500413"

while it is loaded correctly for version 18.

All columns after this one are incorrect. Could it be some separator not being written properly?

I noticed the version 19 data.frame has an extra column, sex ratio, could this be causing it? Also, Is this column containing the sex ratio at birth?

[iagomosqueira]

There is no separation from row 2 between Mat_F_natage and sexratio, so rather than

9.14165e-05 0.5

it has

9.14165e-050.5

@kellijohnson-NOAA are you able to check if this is a problem in the writing section of the TPL?

Still looking if this is the reason for the ss3om error.

[kellijohnson-NOAA]

@iagomosqueira it very well could be a bug in SS3 I am compiling a new version now, which I will then use to run SS3 using the provided files to see if it fixes the output.

I also recommend reading the change log. Issues #136 and #283 might be relevant.

[iagomosqueira]

Thanks, will check.

The error in catch.wt comes from the difference in endgrowth, the sel_wt columns are wrong in the version 19 run.

The difference in SSB comes from maturity when loading from wtatage. readFLSss3 substitutes maturity from Report.sso with the result of dividing fleet -2 in wtatage.ss with fleet -1. This is to be done only when a wtatage.ss file has been used, as determined by looking at SS_output$wtatage_switch. This might not be correct in this case when we force it to read it by calling readFLSss3 with wtatage=TRUE.

[iagomosqueira]

Fixing the issue with sex ratio column separator, as in Report.sso.zip, makes the SS 19 FLStock equal to the SS 18, if loaded using wtatage=FALSE.

The difference in maturity when using wtwtage=TRUE comes from the disparity between what is taken from endgrowth when Maturity_option equals 1

data.table(ss19$endgrowth)[Sex == 1, `Mat*Fecund` / Wt_Beg]

and what is computed from wtatage.ss_new

wtage <- data.table(SS_readwtatage('SSv19/wtatage.ss_new'))

wtage[Sex == 1 & Yr == 1950 & Fleet==-2, 7:22] /
  wtage[Sex == 1 & Yr == 1950 & Fleet==-1, 7:22]

Happy to hear which one of the two is really correct. Should readFLSss3 always use wtatage.ss_new if available?