These pages document initial testing of the atlantisom package in development at https://github.com/r4atlantis/atlantisom using three different Atlantis output datasets. Development of atlantisom began at the 2015 Atlantis Summit in Honolulu, Hawaii, USA.
Draft atlantisom landing page for review, June 9, 2020
Need to diagnose the annage outputs; they look strange
Starting cod catch at annual age diagnosis, Sept 30, 2021
Multispecies dataset from NOBA sacc_30 with climate and fishery selectivity, Sept 29, 2021
Code converting these outputs to multispecies production model and length structured model inputs is in development as of spring 2021 and is on the ms-keyrun repository. See also Initial simulated data, May 24, 2021
Apply survey specifications to diet outputs, December 7, 2020
Compare multispecies survey specifications, November 9, 2020
Explore NOBA diet comps and multi-survey multispecies wrappers, October 21, 2020
Including detailed diet exploration, multiple surveys, survey function bug fixes for species with < 10 cohorts.
Apply and extend wrappers to get full age structure, updated June 5, 2020
Interpolate to get average weight at true age, June 5, 2020
Also bug fixed config files for multispecies dataset. Interpolate wtage includes data generation for NOBA cod (max age 20), redfish (max age 40), and capelin (max age 5).
Different
packages to visualize atlantisom functions, updated June
15, 2020
Sardines in the CalCurrent, all true data, May 22, 2019, updated May 23, 2019
Sardines in the CalCurrent, input to SS functions: May 24, 2019 - now generalized to run for any species in the cal current with a unique species config file
May 24: Sardines have a full dataset, but still working on converting to true ages for species with multiple ages per age class. See below under True age compositions.
The below generates a readable dat file for input into SS3 for sardines. The control file and biological parameters still need hand modification, but SS3 now runs.
So we can start some comparisons of true and model estimated attributes.
Compare Atlantis truth with SS3 estimated B, R, and F, October 1, 2019
Big picture and new functions needed - May 10, 2019, updated May 16, 2019
May 11: Changes to atlantisom::sample_fish and new
atlantisom::aggregateDensityData functions seem to work.
See below under Length
compositions.
May 15: Changes to atlantisom::calc_age2length to allow
user specified max length bin seem to work. Also listed under Length compositions.
May 20-22: Fisheries, corrections to nums at age for Cal Current and decision to read in the catch.txt file for total catch in weight
Installation and initial tests - April 26, 2019, updated April 30, 2019
Survey census visualization for NEUS - April 30, 2019
Initial survey tests - April 30, 2019, updated May 1, 2019
Standard survey visualization for NEUS - May 1, 2019
Initial age comp tests - May 2, 2019, updated May 6, 2019
CCA outputs had some 0 length at age, requiring changes to
atlantisom::sample_fish
True and sample fish census comparison NEUS - May 6, 2019
True and sample fish census comparison CCA - May 6, 2019
True and sample fish census comparison NOBA - May 6, 2019
Standard survey with composition sampling: impact of selectivity,
sample_fish
Compare true cohort comp with survey + sampling - May 6, 2019
True cohort comp vs survey comparison NEUS - May 6, 2019
Difficult when you don’t have a good idea of the outputs you are using!
Defeated by length compositions - May 8-9, 2019
Better when you do:
Victory: length compositions - May 11, 2019
CCA length composition examples - May 11, 2019
But don’t try running length comps at the subannual level for all species.
NOBA length composition examples - May 13, 2019
NEUS length composition examples - May 13, 2019
Standard survey with length sampling: impact of selectivity,
sample_fish
Compare survey + sampled lengths with true - May 14, 2019
CCA compare survey + sampled lengths with true - May 14, 2019
NOBA compare survey + sampled lengths with true - May 15, 2019
NEUS compare survey + sampled lengths with true - May 15, 2019
Test new user specified max length bin for a large species:
Default 150 cm, try different max - May 15, 2019
Do we see ever small fish for large species? Is this normal, or a problem with length estimation?
Are length comps OK on the small end? - May 15, 2019
May 16: Isaac says CCA fish look reasonable. NOBA Greenland halibut growth parameterization was reportedly “a nightmare” with conflicting science and datasets, so maybe we don’t worry too much about them right now.
How much variability in weight at stage? - May 16, 2019
Work in progress on interpolating weight at age in InterpolateWeightAtAge.Rmd
Fishery catch problem, older codebase has a bug in output. Test with NOBA…
Suggests a workaround needed for catch in weight - May 21, 2019
Test Z calc and compare with true annage NOBA - May 24, 2019
May 24: Either calc_Z or calc_stage2age or
both are incorrect if we expect them to exactly match atlantis annual
age ouput where it is available. We match in timestep 0, then diverge.
May 27 update: also true for Bocaccio in CCA, the interim numbers at age
drop to 0.
June 7: calc_Z seems incorrect when compared to
outputMort.txt; stepping through code needed
June 11: fixed matching of recruitment to timestep, still seems off
June 13: I think we have the best calc_Z we are going to
get. After discussion with Beth, we should not expect to match output of
Mort.txt because M gets rescaled so much. However, the comparison looks
reasonable now for sardine and even for NOBA cod.
Dec 11: Our calculated Z is less than the input F. Does Atlantis rescale F?
June 14: fixed a bug and calc_stage2age works better now
that it is using Z, and now that Zs are in the right ballpark. Still not
a perfect match to true age output.
Test stage to age calculations and compare true age output NOBA - June 14, 2019
June 15: while the age splitting is functioning better, applying it to a species with more than 2 ages per age class (CC Bocaccio rockfish) shows that our application of an annual Z estimate may not be the best approach. The functions themselves could use further tweaking.
Overall it is advisable to use true age output from more recent Atlantis codebases where possible.
Redo Bocaccio test with CCA F scenario, June 15, 2019
Further experiments with M, F, and Z across models and species confirm that mortality needs a lot more work:
In which I broke the very first thing that worked, then fixed it - June 4, 2019