Jump to Index of Abundance

Jump to Length Based Indicators

Jump to Relative Harvest Rate

Jump to Classification Skill

Jump to More Information

Jump to References

Use of Observation Error Model in Management Strategy Evaluation

The Observation Error Model is a key component in Management Strategy Evaluation (MSE). Its role is to generating data for use in simulation when testing and validating various management strategies. The Observation Error Model is a vital component of MSE, and simulates the types of data that can realistically be collected, and used either in a stock assessment or in an empirical harvest control rule.

The roles of the Observation Error Model are to

Simulate Realistic Data: Generates pseudo-data that mimic real-world fisheries data, such as catch numbers, relative abundance indices, and length frequency data.
Incorporate Uncertainty: Accounts for uncertainties in data collection, including sampling errors and biases
Evaluate Robustness: Uses simulated data to evaluate the performance of different indicators and management strategies under scenarios that represent plausible alternative hypotheses about the resource and its fisheries.
Compare Data Collection Schemes Allows the benefits of alternative monitoring and data collection schemes to be evaluated. For example, to evaluate the skill of different data sets to monitor the state of the stock, using an index of abundance or length data.
Validate of Stock Assessment Methods: Uses the simulated data in a stock assessment allowing the estimates to be compared with the ‘true’ values from the Operating Model.

Packages

The packages required are

FLR

library(FLCore)  
library(FLBRP)
library(FLasher)
library(FLife)
library(ggplotFL)

library(FLCandy)

FishLife

library(rfishbase) 
library(SPMpriors)
library(FishLife)

Plotting and Data wrangling

library(plyr)
library(dplyr)
library(reshape)

library(ggplot2) 
theme_set(theme_bw())

library(ggpubr)
library(ggcorrplot)
library(GGally)

Operating Model Conditioning

Choose parameters, and create an FLBRP object

par=lhPar(FLPar(linf=30))
eq=lhEql(par)

Based on the equilibrium FLBRP object create an FLStock object to simulate time series.

om=as(eq,"FLStock")

ftarget=m(om)[4,-1]

om=window(fwd(om,f=ftarget,sr=eq),start=19)

[1] "maxfbar has been changed to accomodate new plusgroup"

dimnames(om)$year =an(dimnames(om)$year)+1980
range(om)["plusgroup"]=range(om)["maxage"]

Project using fwd

om=fwd(om,fbar=window(fbar(om),start=2021)%*%FLQuant(seq(1,10,length.out=61)),sr=eq)
om=window(om,start=2010,end=2080)

Add uncertainty in recruitment

nits   =101
ftarget=rlnorm(nits,log(fbar(om)),0.1)[,-1]
recDevs=rlnorm(nits,rec(om)%=%0,0.3)

om     =propagate(om,nits)

om=fwd(om,fbar=ftarget,sr=eq,residuals=recDevs)

plot(om, iter=5)

Figure 1 Time series.

Indicators

Index of Abundance

Creat an index of abundance

om   =window(om,start=2010,2030)

index=rlnorm(nits,log(vb(om)),0.2)
 
plot(index,iter=12)

Length Based Indicators

lbi=FLQuants(indicators.len(om, 
                   indicators='lmean', 
                   params=FLPar(apply(par,1,median)),
                   metric='catch.n'))
 
plot(lbi,iter=17)

Relative Harvest Rate

HR=catch(om)%/%index
 
plot(HR,iter=17)

Length data

ak =invALK(iter(par,1),cv=0.1,age=an(dimnames(om)$age),bin=1)
lfd=lenSamples(catch.n(om),ak,n=5000)
 
units(lfd)="cm"

plotLengths(group(lfd, sum, year=year-year%%10))

Classification Skill

dat  =model.frame(FLQuants("Biomass"=vb(om),
                           "F"      =fbar(om),
                           "Index"  =index, 
                           "HR"     =HR,
                           "LBI"    =lbi[[1]]))
dat=subset(dat,year%in%2021:2030)

ggplot(aes(Biomass,Index),data=dat)+  
  geom_point()+ 
  geom_smooth(col="blue",method=lm,se=FALSE)+
  geom_vline(aes(xintercept=mean(dat$Biomass)),col="red",linetype=2)+
  geom_hline(aes(yintercept=mean(dat$Index)),  col="red",linetype=2)+
  theme_bw(16)

True Skill Score

OM =fbar(om)>mean(fbar(om))
OEM=HR>mean(HR)  

dat=model.frame(FLQuants(OM=FLQuant(OM),OEM=FLQuant(OEM)))
with(dat,table(OM,OEM))

   OEM
OM     0    1
  0 1168  208
  1  106  639

Receiver Operator Characteristic

Figure 2 ROC curves

More Information

For more information on the FLR Project for Quantitative Fisheries Science in R, visit the FLR webpage ¹.

Author information

Laurence Kell. laurie@seaplusplus.co.uk

Acknowledgements

Software Versions

R version 4.2.1 (2022-06-23)

knitr: 1.43
FLCore: 2.6.19.9105
FLBRP: 2.5.9.9022
FLasher: 0.7.1.9221
FLife: 3.4.0
ggplotFL: 2.7.0.9132
FLCandy: 0.1.0
rfishbase: 3.1.9.99
SPMpriors: 1.0.4
FishLife: 3.0.0
plyr: 1.8.9
dplyr: 1.1.4
reshape: 0.8.9
ggplot2: 3.4.4
ggpubr: 0.6.0
ggcorrplot: 0.1.3
GGally: 2.1.2

Compiled: Sun Jan 21 12:36:51 2024

References

http://flr-project.org ↩︎

02: Observation Error Model

FAO: Small Pelagics

L Kell

21 January, 2024