Gadget Haddock model

1
Gadget Haddock model

• Daniel Howell
• IMR Bergen

2
Introduction

• Brief overview of the Haddock example
• Example freely available on the web
• With documentation
• More Gadget outputs and diagnostics tomorrow
• Several plots presented where they illustrate
  more general features

3
Gadget

• Gadget is a program for running and optimising
  agelength-structured simulation models
• Models growth, recruitment, maturation, mortality
  and fishing
• Can also be multi species and multi area

4
Gadget

• A range of functions available
• Growth, fishing selectivity, maturation, etc.
• Parameters can be fixed or estimated
• Parameters can be set to one value for all years,
  one per year, or in blocks

5
Gadget

• Gadget has a seperation of model and data
• Simulation model independant of data
• Likelihood functions to compare model results to
  the data
• Optimisation of model parameters possible

6
Haddock

• Icelandic haddock
• Agelength structured
• 1978-1999
• Quaterly time step
• Age 1-10, 1cm length classes
• Single-species, single-area model
• Von Bertalanffy mean growth, fitted beta-binomial

7
Haddock

• One commercial fleet, one survey fleet
• Catch in tons input from the data
• Fishing treated as a subtraction, not a mortality
• No stock recruitment function
• Directly estimate recruits each year

8
Data

• Multiple data sources from commercial fleet and
  survey.
• Mean length with variance survey catch
• Length distribution survey catch
• Age-length distribution survey catch
• Survey index abundance by length

9
Data

• Compare modelled catch/survey data with the
  observations
• Different data sets can cover different years
• Single likelihood score as a weighted sum of all
  these
• Optimisation minimizes this overall score

10
Haddock

• 37 parameters
• Initial population (8)
• Annual recruitment (22)
• Growth (3)
• Commercial and survey selectivity (2 each)
• Not all equally significant

11
Haddock Sensitivity analysis
12
Haddock Sensitivity analysis
13
Haddock

• Note that we have reached an optimum
• Most significant parameters are
• Growth
• Fleet selectivity
• Recruitment of large year classes

14
Haddock

• Recruitment in the last years is poorly
  constrained
• No data on these fish available yet
• Potential problem if using the model for
  prediction
• Recruitment in small year classes is relatively
  poorly constrained
• But not very important to the population

15
(No Transcript)
16
Haddock

• Used as an annotated example in the Gadget
  program
• http//www.hafro.is/gadget
• Gadget Example
• gadgethelp_at_hafro.is