J~THE ICLARM!KISR CONFERENCE

ON THE THEORY AND APPLICATION OFLENGTH-BASED STOCK ASSESSM~NTS

by

Daniel PaulyICLARM

MCC P.O. Box 1501

Makati, Metro Manila

Phili ppines

The international conference on the

"T he 0ry and Application of Length-basedSt0ck Assessments" took place from 11-16february, 1985 at the Instituto di Tech-nologia dells Pesca e del Pescato (ITPP)at Mazara del Vallo, Sicily, Italy, asann 0unced in the previous issue of fish-byte. We had great hopes for this con-ference. All of them were met, and in

many ways, the conference went beyondwhat we had hoped it would achieve.

Specific aims of the conferencewere:

to identify the basic features ofsampling schemes for representativelength-frequency data,to review length-based methodo-logies for fish stock assessments,with particular reference to theirprecision and accuracy,to examine the interrelationshipsbe twee n length-based and age-basedmethods, and possible approachesfor combining the two, and finally,to review and test computer pro-

grams implementing length-basedmethods.The conference, the planning of

which was initially in the hands of Dr.J.L. Munro, ICLARM, was organized bythis author, acting as conference secre-tary and Dr. G. Morgan of the Kuwait Ins-titute of Scientific Research. It hadthe active support of the Division ofMarine Resources, fAD and of ~heConsiglo Nazionale delle Ricerche,through ITPP, the host institution.

Both the host, ITPP Director Dr.Dino Levi and the conference ChairmanDr. J.A. Gulland contributed to thesmooth running of the conference, whichbegan with a two-day presentation of the22 contributed papers (see Appendix fortheir title, authors and abstracts).

March1985

The conference participants and sixo b s e r v e r.s (t h r e e fro m K I S R an d t h r e e

from ITPP) then split into four workinggroups, each of which looked at one ofthe following areas:

. sampling of length-frequency data,review of existing length-basedmodels,sensitivity, precision and bias ofvarious models,review and testing of relevant com-puter programs.A report for inclusion in the confe-

rence proceedings was prepared by eachof the groups, to which will be added areport, presently being written by Dr.J.A. Gulland, on the conference as awhole.

Although it is Dr. Gulland's reportwhich will contain the finer points ofthe conference, three areas of concensusemerged clearly enough for them to bementioned here:

i) Although length-frequency data canbe obtained relatively cheaply,they must be obtained through well-designed sampling schemes if theyare to form the backbone of stock

assessments. Particularly, atten-tion must be given to appropriatesampling of young stages,

ii) There are a vast number of methods

for estimation of growth, mortalityand recruitment patterns and forderiving other information fromlength-frequency data. It is there-fore possible to select the methodbe st suited to the structure of the

data set. However, when applyingsuch methods, care must be taken toassess the likely errors involved,and to take remedial measures whereappropriate.

5

iii) The precision and accuracy oflength-based methods generallyimproves considerably when growthinformation (e.g. as obtained byreading ages from a few otoliths)is incorporated into an analysisbased predominantly on length-composition data. This suggeststhat the most cost-effective

met hods for stock assessments, bothin developed temperate and in tropi-cal developing countries should be"hybrid methods", combining lots oflength-frequency data with some ageinformation from hard parts (e.g.'daily rings; see Fishbyte ~(3) ).

The proceedings of the conference,will be edited by this author and Dr.

Gary Morgan and will be published in twovolumes in the ICLARM Conference Proceed-

ings Series. The first volume will con-tain the 22 contributions 'listed below,

the four group reports and the Chair-man's report. The second volume willinclude user's instruction, listings,test runs and test data for a number of

the computer and calculator programs pre-sented and tested at the conference

(excluding the ELEFAN programs, whichwill be published separately). Theseprogram's will not be standardised to acommon language, however, and thus willinclude listings in FORTRAN, various dia-lects of BASIC, as well as routines inHP's Reverse Polish Notation.

The E LEF AN program whose review anddetailed examination made up about 1/4of the conference contributions have"survived" the very critical tests towhich they were put, in the sense thatthe conditions could be defined underwhich they can be expected to providereasonable answers.

It will thus be possible, when the

ELEFAN programs are prepared for publica-tion to state these conditions expli-

citly - as well as to incorporate thesuggestions for improvements that weremade during the conference.

Altogether, the conference was aresounding success. This was best summa-rised in a letter just received from oneof the participants which concluded with"...good science, good people, goodfood. What more could one ask for a con-ference?"

6

Abstracts of papers presented at theconference

METHODS Of SIZE fREQUENCY ANALYSISAND THEIR INCORPORATION IN PROGRAMMES

FOR FISH. STOCK ASSESSMENTIN DEVELOPING COUNTRIES:

FAO INTERESTIN RECEIVINGADVICE

J. Csirke,FAa,

J.F. Caddy* and S. Garcia*Fisheries Department

Rome, I tal y

Abstract: FAD's task requires theadaptation of conventional methodologiesdeveloped in high latitudes .for use intropical situations and the developmentand further refinement of length-basedmethodologies. However, a number of pro-blems need to be resolved before these

methods can be applied on a grand scale.Questions for which FAD requiresdetailed answers from this conference

rei ate among other things to the follow-ing items: "quick and dirty"'y'! elabo-rate approaches, problems with cohortseparation, bias due to selectivity andheterogenous fishing effort, departuresfrom steady state, sampling problems inmultispecies, multigear fisheries, pro-blems with very short-lived animals,appropriate trawl survey stratificationfor sampling length frequency data,effects of sampling errors and biases onmortality and growth estimates from dif-ferent methods, use of mean size statis-tics and translation of results from

length based models into managementadvice.

[Answers to most of these questionswill b e f 0 u n din the 21 other con t r i -butions presented at the conference.D.~.]

A REVIEW OF THE ELEFAN SYSTEM FOR

ANALYSIS OF LENGTH-fREQUENCY DATA INfISH AND AQUATIC INVERTEBRATES

D. PaulyICLARM,P.O. Box 1501

Makati, Metro ManilaPhilippines

Abstract: Some aspects of the historyand present role of length-based methods

for the study of the growth, mortalityand recruitment of fish and invertebrate

stocks are reviewed. The recently deve-

FISHBYTE

lop e dEL E FAN ( ,Ll e c t r 0 n i c .b£n g t h

Lrequency ~alysis) system of BASICmicrocomputer programs is presented andsome of its key features are discussedin detail.

Examples of the use of the ELEFAN

system are provided; these examples areus e d to illustrate the need for furtherresearch on length-based systems in gene-ral and to emphasize the need for ade-

quate sampling.

ESTIMATING GROWTH AND SURVIVALPARAMETERSBY NONLINEAR REGRESSION

USING AVERAGE SIZE IN CATCHES

T. A. Ebert

Department of BiologySan Diego State University

San Diego, CA 98182, U.S.A.

Abstract: When recruitme-nt is period-

ic, average size in samples is smallestin the month of recruitment and largestin the month just before recruitment.The change in average size during a yearis a function of growth and survival

par a meters, which can be estimated by aniterative nonlinear regression technique

using average size as the dependentvariable and time since recruitment as

the independent variable. The modelthat is presented uses a constant rateof survival and the Richards function to

model growth. Four parameters can beestimated: Z, the mortality constant; K,

the growth rate constant; SCD, asymp-totic size; and, n, the shape para-meters. Two examples are presented toillustrate applications of the method.

SOME MODIFICATIONS OF EBERT'S METHODTO CALCULATE GROWTH AND MORTALITY

PARAMETERS FROM AVERAGE LENGTHSIN A POPULATION

U. Damm

Abteilung FischereibiologieInstitute fur Meereskunde

Dusterbrooker Weg 2023 Kiel, FRG

Abstract: Ebert's method for the

analysis of average sizes in a popula-tion is discussed and its similarity tocertain methods for fitting a von Berta-

lanffy growth curve is pointed out. A

March 1985

simple fitting method is proposect. Itis s how n h ow this method can be appli edto the special case .of two recruitmentsper year; the robustness of thisapproach is discussed. A BASIC programis given whiph can fit the relevantfunction to a given set of data.

ESTIMATING GROWTHAND MORTALITYIN STEADY STATE FISH STOCKS

FROM LENGTH-FREQUENCY DATA

J. A. Wetherall*, J. J.Polovina*and S. Ralston*

National Marine Fisheries Service, NOAAP.O. Box 3830, Honolulu, Hawaii 96812

U.S.A.

Abstract: Methods of using length-frequency statistics to estimate LCDand the ratio Z/K in steady state fishstocks with von Bertalanffy growth andexponential mortality are studied. Seve-ral standard procedures applicable whenmortality is constant are reviewed, andnew methods are introduced which have

superior performance. The variousmethods are evaluated using Monte Carlotechniques. Problems of systematic biasare discussed, and remedial measures aresuggested. Finally, a method is deve-loped (but not yet evaluated) to esti-mate length-specific mortality ratesunder the steady state model.

A METHOD FOR ESTIMATION OF GROWTH,MORTALITY AND GEAR SELECTION/RECRUITMENT

PARAMETERS FROM MULTIPLE

C.P.U.E. LENGTH FREQUENCY DATA.

P. SparreDanish Institute for Fisheries

and Marine Research2920 Charlottenlund

Denmark

Abstract: This paper describes amethod and a program for the estimationof von Bertalanffy growth parameters,total mortality rates and gearselection/recruitment parameters frommultiple length frequency samples in theform of C.P.U.E. (e.g. numbers caught

per hour by length group each quarter ofthe year). The length frequencies areseparated into normally distributed com-

ponents, each of which represents abrood.

7

The method is constructed to allow

for continuous (although seasonallyvariable) recruitment all year round andis intended to be used espec;ially for

tropical fish stocks with such a recruit-ment pattern.

The program is implemanted inFORTRAN 77 for VAX 11/750.

TWO METHODS fOR SIMULTANEOUSLY

ESTIMATING GROWTH, MORTALITY ANDCOHORT SIZE PARAMETERS fROM TIME SERIES

Of CATCH AT LENGTH DATA fROM

RESEARCH VESSEL SURVEYS.

J. G. PopeMinistry of Agriculture,

Fisheries and Food

Fisheries LaboratoryLowestoft, Suffolk, NR33 OHT

United Kingdom

Abstract: Two methods for estimating

growth, mortality and cohort size para-meters from time series of g-roundfish

survey catch at length data are des-cribed. Examples of their use are givenbased on North Sea cod data from English

groundfish surveys.

A WEAKLY PARAMETRIC METHOD fOR THE

ANALYSIS OF LENGTH COMPOSITION DATA

J.G. Shepherd

Ministry of Agriculture,Fisheries and Food

Fisheries LaboratoryLowestoft, Suffolk NR33 OHT

United Kingdom

Abstract: A new and robust method for

the analysis of length compositions interms of the parameters of a specifiedgrowth curve (e.g. von Bertalanffy) wasdeveloped. It does not require thenumber or width of modes to be speci-

fied, and is insensitive to samplingnoise of the larger length groups. Agoodness-of-fit measure is mapped as afunction of growth parameters, whichavoids problems caused by multiplemaxima, and enables the size and shapeof approximate confidence regions to beestimated. Once growth parameters havebeen selected, these may be used to'slice' the length composition into age

groups. However, it is probable thatfurther analysis of length data directly

8

is preferable to methods relying on sucha decomposition.

ESTIMATION Of GROWTH AND MORTALITY

PARAMETERS fOR USE IN LENGTH-STRUCTUREDSTOCK PRODUCTION MODELS

Minnesota

500 Cedar

J. M. HoenigDept. of Natural ResourcesAvenue, St. Paul, MN 55146

U.S.A.

Ab s t r act: The s u g g est ion by Va u gh a nand Kanciruk that linear methods of fit-

ting von Bertalanffy growth curvesshould be abandoned in favour of non-

1 i n ear est i mat ion a p pea r s p r ema t u r e ,because the direction of the bias is

ex.plainable by the use of an ordinarypredictive regression line in their simu-1 a t ion s t u d Y ins tea d 0 f a fun ct i 0 na 1regression. An estimator for the total

mortality, Z, based on mean length isderived for use when reproduction occursperiodically instead of continuously.Production models employing mortalifyrat e or I/K as the independent variable,as suggested by Csirke and Caddy, may.appear tilted to the left if natural mor-tality rate is compensatory. If dataare limited to a small portion of theproduction curve, attempts at curve fit-ting may give unreasonable results. Inthis case, one can constrain a parametertoe qual an independent estimate so thatthe other estimates become more reason-able.

ESTIMATING MORTALITIES fROM LENGTH

(OR AGE) SPECIFIC RATIOS

M.J. SandersFAO/UNDP Regional Project RAB/S3/023

Cairo, Egypt

Abstract: Methods are presented for

estimating F's and M's from the propor-tions by length (or age) of each sex in

the catch; and knowledge of the totalmortality coefficient and age at firstcapture for each sex. When dealing withseasonal fisheries, the additional infor-mation required are the ages at the endof the first fishing season, and thefishing season duration. The methods are

supported by the provision of programmessuitable for Hewlett Packard 67/97 calcu-lators and worked examples.

FISHBYTE