An R package for providing an interface between the SAM stock assessment (SAM, https://github.com/fishfollower/SAM) and the Fisheries Library in R (FLR, https://flr-project.org/).
The package provides methods for running SAM with FLR objects
(FLR_SAM) and functions useful for conducting a Management Strategy
Evaluation (MSE). This includes methods for turning the output of a SAM
model fit into FLR’s FLStock (SAM2FLStock) and for creating
simulation replicates based on SAM’s variance-covariance matrix
(SAM_uncertainty).
Before installing FLfse, the following non-CRAN R packages should be installed:
- stockassessment (available from https://github.com/fishfollower/SAM)
- FLR packages FLCore and ggplotFL (see http://www.flr-project.org/#install)
FLfse can be installed with
remotes::install_github("shfischer/FLfse/FLfse")Load package
library(FLfse)The package includes example data for some stocks, see
data(package = "FLfse")Among others, this includes the 2018 assessments for North Sea cod, North Sea haddock and North Sea whiting and the 2019 assessment for North Sea herring.
Fit SAM to North Sea cod data from ICES WGNSSK 2018, with default parameterisations:
fit <- FLR_SAM(stk = cod4_stk, idx = cod4_idx)use the SAM assessment configuration as used by ICES WGNSSK 2018:
### fit SAM model
fit <- FLR_SAM(stk = cod4_stk, idx = cod4_idx, conf = cod4_conf_sam)
### check convergence
fit$opt$convergence## [1] 0
### The result of FLR_SAM() is an object of class "sam".
### All methods defined in the stockassessment package can be used on it.
### summary table
tail(summary(fit))## R(age 1) Low High SSB Low High Fbar(2-4) Low High
## 2013 226194 173181 295435 99494 81167 121960 0.442 0.375 0.521
## 2014 317568 242921 415153 105714 86598 129050 0.445 0.380 0.520
## 2015 155316 118799 203058 119893 97172 147926 0.433 0.371 0.506
## 2016 109912 82134 147083 119699 97197 147411 0.423 0.359 0.498
## 2017 385593 259272 573460 113502 90267 142718 0.444 0.373 0.528
## 2018 97383 40347 235049 118387 90333 155154 0.446 0.353 0.563
### plot model
plot(fit)
The resulting
“sam” object can then be converted into an FLStock object:
cod4 <- SAM2FLStock(fit, cod4_stk)
### plot with ggplotFL
library(ggplotFL)
plot(cod4)
The 2019 ICES HAWG assessment for North Sea herring can be replicated
with FLfse. To get this assessment running, a different branch
(“components”) of the stockassessment package needs to be installed.
The following code snippet shows how to do this without interfering with
the global packages (please restart R if another version of
stockassessment is currently loaded):
### 1st: add a temporary R package library
.libPaths(c(tempdir(), .libPaths()))
### 2nd: install stockassessment branch "components"
devtools::install_github("fishfollower/SAM/stockassessment", ref = "components")
### now, the assessment can be run:
library(FLfse)
library(stockassessment)
### replicate HAWG assessment
fit <- FLR_SAM(stk = her4_stk, idx = her4_idx, conf = her4_conf_sam,
conf_full = TRUE, NA_rm = FALSE)
### check results
fit
plot(fit)
summary(fit)
### convert into FLStock
stk <- SAM2FLStock(fit, her4_stk)We can use SAM’s variance-covariance matrix to create simulation replicates (e.g. for MSE).
### fit SAM to North Sea cod
fit <- FLR_SAM(stk = cod4_stk, idx = cod4_idx, conf = cod4_conf_sam)
### create simulation replicates
set.seed(1)
unc <- SAM_uncertainty(fit = fit, n = 10)
### turn into FLStock
stk <- SAM2FLStock(fit, cod4_stk)
### insert replicates for stock numbers, catch numbers and fishing mortality
stk <- propagate(stk, 10)
stock.n(stk) <- unc$stock.n ### insert stock numbers
stock <- computeStock(stk) ### update stock biomass
catch.n(stk)[, ac(1963:2017)] <- unc$catch.n ### insert catch numbers
catch(stk) <- computeCatch(stk) ### update total catch
harvest(stk) <- unc$harvest ### insert fishing mortality at age
units(harvest(stk)) <- "f"
### plot stock and replicates
plot(stk, iter = 1:10)