Retrospective analyses, improvement of assessment and prediction methods
for haddock stocks.
A.A. Russkikh,
PINRO, Murmansk, Russia K.H. Hauge,
IMR, Bergen, Norway
Murmansk, 2005
Contents
Introduction
Objectives of the work
Statement of the problem
Methods and results
Conclusions
Introduction
NEA haddock annually asessed by AFWG using a catch at age model (VPA) that uses measurement of the number of fish caught in each age group
In common with computing machinery evolution
the software for stock assessment and projection
was changed but calculations were made on the
same principle as previously
Introduction
The standard software and methods used for stock assessment and projection of NEA haddock at AFWG*
Year
Assessment methods Short-term projection catch and biomass
Recruitment projection
1978 -1985 (Survey indices ratio)
1986-1991 IFAP module (VPA) RCRTINX2 (regressions
of yearclasses)
1992-1993
VPA version 3.1(SVPA/XSA),
ADAPT
1994-1998 VPA version 3.11 (SVPA/XSA) 1999-2002
IFAP module (management options
table)
2003-2005
VPA version 3.2 - VPA95 (SVPA/XSA)
MFDP(management options table)
RCT3 (regressions of yearclasses)
* - in brackets - methods
Objectives of the work
Current work will aim at investigating a part of the
uncertainty, i.e. observation errors given a particular model specification using ADAPT algorithm.
Overall objective is to investigate if current assessment and projection procedures can be improved.
It is expected that such improvements may provide for scientific advice. This will help solve main tasks of AFWG.
Materials for investigation were input data for stock
assessment and projection which used in AFWG 2005
Statement of the problem (Unsertainty)
There are several categories of uncertainty in fish science:
¾ natural variation
¾ observation errors in input data
¾ model misspecification
¾ uncertainty in transaction scientific advices into management
¾ imperfect implementation of management strategies
¾ others
Precautionary approach in fishery management which imply care fish stocks on a safe biological limits should (can) been based on total unsertainty estimates in stock assessment.
The AFWG 2004 states that the uncertainty may be underestimated and that difference between Blim and Bpa may be too small therefore it should be investigated
The current uncertainty is reflected in distance between lim and pa points which are adopted by ACFM for this stock as Blim=50 thou. t and Bpa =80 thou. t.
40 90 140 190 240 290
0.10 0.20 0.30 0.40 0.50 0.60
F 2004
SSB 2005
F lim F pa
B lim B pa
2004 AFWG estimates
reference points frame
2005 AFWG estimates
Methods and results
40 90 140 190 240 290
0.1 0.2 0.3 0.4 0.5 0.6
F 2004
SSB 2005
Methods and results
Algorithm of adaptive framework ADAPT was
created in Excel and calculations done with input data from AFWG 2005.
The purpose was estimate part of the assesment uncertainty, i.e. observation errors given a particular model specification, was characterized by using bootstrap methods.
Bootstrap method based on principle random
resampling of the residuals from the observed-
predicted ”tuning” indices.
Methods and results
Model calculated population parameters and recalculated a and b which correspond regressions, what provide smallest differences between observed” and “modeled” population abundance using used nonlinear least-squares minimization procedure
choused the “best” fit to the tuning indices.
0 2 4 6 8 10
1990 1995 2000 2005
Fleet 1 residuals
Methods and results
The residuals of that fit were bootstrapped
0 2 4 6 8 10
1990 1995 2000 2005
Fleet 1 residuals
1000 times and new values of N produced. The
distribution of the associated Fs and SSB provided
an indication of variation and the bias (deviations).
Methods and results
SSB cumulativ e distribution
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
0 100 200 300 400
Spawning stock biomass, thou.m.t.
Cumulative probability SSB 2003
SSB 2004 SSB 2005 SSB 2006 SSB 2007 Blim Bpa Spawning stock biomass
0 50 100 150 200 250 300
1980 1985 1990 1995 2000 2005
Biomass, thou. MT
80 % probability 95 % probability AFWG estimates
Methods and results
F4-7 cumulative distribution
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
0.1 0.2 0.3 0.4 0.5 0.6
Fishing mortality
Cumulative prbability
F4-7 2002 F4-7 2003 F4-7 2004 F4-7 2005 F4-7 2006 Flim Fpa Fishing mortality (age 4-7)
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
1980 1985 1990 1995 2000 2005
Fishing mortality
80% probability 95% probability AFWG estimates
Methods and results
Recruitment (age 3)
0 200 400 600 800 1000 1200 1400 1600 1800 2000
1980 1985 1990 1995 2000 2005
Millions spec.
80% probability 95% probability AFWG estimates
N3 cumulative distribution
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
0 200 400 600
millions spec.
Cumulative probability
N3 2003 N3 2004 N3 2005 N3 2006 N3 2007
Conclusions
Suggested algorithm based on ADAPT framework allow to investigate part of uncertainty in stock assessment and projection procedure and its prototype - program ADAPT can been applied as alternate approach for estimation of population dynamics of NEA haddock
This uncertainty analysis is only the first step in the construction of full analysis of uncertainty in stock assessment.
Additional work to more fully characterizes all important sources of uncertainty in the assessment process can been done by
working group members in relation to estimation more
conservative biological reference points and evaluation of the harvest control rules.
Conclusions
Main weakness of current methods for stock assessment and
projection is in using several partly different, partly similar models
Models takes to account more or less the same input data but receives different estimation of population numbers and fishing mortality, therefore level of uncertainty increasing with each step.
The framework should allow for a procedure stock assessment and short-term projection more simple, objective and robust to
criticism.
Acknowledgments
I would like to thank Kjellrun Hiis Hauge, IMR, Norway for her comments to the manuscript.
I am also grateful to the Dr. Einar Hjorleifsson, MRI, Iceland who provides me strategy and practically helps, and gave me important advices for investigation of uncertainties in stock assessment.
