UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

An investigation of fisheries yield equations with particular reference to annual state models and seasonal… Wild, Alexander 1981

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


831-UBC_1981_A1 W54.pdf [ 9.94MB ]
JSON: 831-1.0095087.json
JSON-LD: 831-1.0095087-ld.json
RDF/XML (Pretty): 831-1.0095087-rdf.xml
RDF/JSON: 831-1.0095087-rdf.json
Turtle: 831-1.0095087-turtle.txt
N-Triples: 831-1.0095087-rdf-ntriples.txt
Original Record: 831-1.0095087-source.json
Full Text

Full Text

AN INVESTIGATION OF FISHERIES YIELD EQUATIONS WITH PARTICULAR REFERENCE TO ANNUAL STATE MODELS AND SEASONAL PULSE FISHING  by Alexander Wild B . A . S c . , University of B r i t i s h Columbia, 1956 M . S c , University of B r i t i s h Columbia, 1973  A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in the  Department of  Zoology  We accept this thesis as conforming to the required standard  THE UNIVERSITY OF BRITISH COLUMBIA February, (c)  1981  Alexander Wild  In presenting this thesis in partial  fulfilment of the requirements for  an advanced degree at the University of B r i t i s h Columbia, I agree that the Library shall make i t freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the Head of my Department or by his representatives.  It  is understood that copying or publication  of this thesis for financial gain shall not be allowed without my written permission.  Department of The University of B r i t i s h Columbia 2075 Wesbrook Place Vancouver, Canada V6T 1W5  DE-6  BP 75-51 1 E  ii  ABSTRACT  Traditional methods of stock assessment rely on the calculated value of e f f e c t i v e e f f o r t (f)  and the assumed constancy of the c a t c h a b i l i t y  c o e f f i c i e n t (q) to provide estimates of abundance and the effects of f i s h i n g . It  is d i f f i c u l t to account for or to quantify a l l the variables that affect  f and the r e l i a b i l i t y associated with  f  of t h i s term is often questionable.  and  q  The uncertainty  i s r e f l e c t e d in assessments and undermines  confidence in management recommendations.  Under certain conditions the objective of circumventing these d i f f i c u l t i e s i s r e a l i z e d in the development of the annual state models. end of each f i s h i n g year, the f i s h i n g (F) and natural mortality cients and the apparent abundance of each age group (i) without reference to a year c l a s s .  f,  (M) c o e f f i -  are estimated  q , the number of r e c r u i t s , or the entire history of  The data requirements that make this analysis possible are  the catch (C^) and y i e l d in weight (Y of the growth equation parameters. of f i s h (Y^ y'  At the  C.)  w  ^) of each age group and estimates  The derived quantity for the mean weight  i s independent of abundance and provides a deterministic  solution for the total mortality c o e f f i c i e n t ,  .  Consecutive year class  estimates of Z | and Z ^ - j , when coupled.with a r a t i o of catch equations, y i e l d estimates of F^, F . ^ , and M. +  The assumptions of this  particular  model are that growth and mortality are concurrent and continuous during the f i s h i n g p e r i o d , the stock i s closed to immigration and emigration, and M i s constant for a l l age groups.  Alternative models are developed that  provide a simultaneous solution for the stock using equations for mean  iii  length and weight or mean weight alone when .M i s constant or a function of age.  The equation for  is based on a generalized growth model and  integrated by an approximate technique. The effect of seasonal pulse f i s h i n g on equilibrium y i e l d is examined for hypothetical species having twenty d i f f e r e n t distributed growth and natural mortality.  patterns of seasonally  Each pattern is subjected to ten  f i s h i n g strategies that vary in seasonal intensity and annual d i s t r i b u t i o n . The effect on y i e l d of increasing values of M, F and the growth parameter K is also explored.  Relative to .continuous f i s h i n g , the greatest increase in  y i e l d is consistently achieved by concentrating the f i s h i n g a c t i v i t y into a single season.  While the magnitude of this increase varies from three to  t h i r t y per cent, in any p a r t i c u l a r situation the optimal; time and the potential y i e l d improvement is a function of K, M, F and the pattern.  growth-mortality  iv TABLE OF CONTENTS Page TITLE PAGE  i  ABSTRACT  ii  TABLE OF CONTENTS  iv  LIST OF FIGURES  vi  LIST OF TABLES  vii  LIST OF APPENDICES  viii  ACKNOWLEDGEMENTS  ix  INTRODUCTION  1  FISHERY YIELD EQUATIONS  '  8  1.  General Principles  2.  Fishing and Natural Mortality - Catch in Numbers  15  3.  Growth Models  25  3.1  Generalized derivation  25  3.2  A l l o m e t r i c , asymptotic growth parameter estimation  31  4.  5.  Yield Models:  8  Combined Growth and Mortality  4.1  Generalized y i e l d model  4.2  Beverton and Holt constant recruitment,  35 35 equilibrium  y i e l d model  38  4.3  Virtual  43  4.4  An asymptotic, allometric growth y i e l d model  56  4.5  Seasonal pulse f i s h i n g  63  population and cohort analysis models  Regenerative Yield Models  ANNUAL STATE MODELS  69 74  1.  General Remarks  74  2.  Assumptions of the Annual State Models and Bases of Analysis . . .  85  V  Page 3.  Model 1:  Methods of Estimating Growth Parameters and  Mortality Coefficients 3.1  91  I n i t i a l estimates of L , K and Z  94  00 "  3.2  I n i t i a l estimates of W and b •  3.3  Final estimates of parameters, c o e f f i c i e n t s and the critical  4.  96  00  Model 2:  age  97  Methods of Estimating Growth Parameters and  Mortality Coefficients 5.  101  The Dependence of the Mortality Coefficient on t Effect of Variable Natural  Mortality  0  and the 104  DISCUSSION AND CONCLUSIONS  112  LITERATURE CITED  124  APPENDICES  131.  vi LIST OF FIGURES Figure 1  Page A schematic diagram of the s i m p l i f i e d between recruitment,  relationship  growth and mortality for a multi  age group, closed system fishery 2  13  Seasonally intensive periods of growth (g) and natural mortality (m) for a variety of fishes  3  65  (a) Twenty cases of seasonally distributed growth and natural  mortality (m) patterns.  l i f e of an age-group (i)  Each year in the  is divided into  quarters.  (b) Ten, seasonally applied f i s h i n g strategies 4  (g^)  (FS)  66  Schematic y i e l d isopleth diagrams for two d i f f e r e n t multi age group stocks: (a) constant recruitment and (b)  density  dependent recruitment 5  7  Response of 18 conventional and potential  fishery  s t a t i s t i c s in relation to eumetric f i s h i n g curve 6  82  Model 2, residual sums of squares contours for various values of the growth curve intercept,  t , Q  and the total  mortality c o e f f i c i e n t , Z Cl  9  92  (a) Twenty cases of seasonally distributed growth and natural  mortality (m) patterns,  applied f i s h i n g strategies  (FS)  (g-j)  (b) Ten seasonally 191  vi i LIST OF TABLES Table I  Page Age-group catch, mean weight and length f o r two consecutive years of f i s h i n g . . . . .  II  93  Comparison of Model 1 and Model 2 estimates of growth equation parameters and mortality c o e f f i c i e n t s with actual values  III  Expected change in s t a t i s t i c s W.. and L^. due to a change in f i s h i n g mortality of 0.1  IV  100  102  The effect of an error in t  on (a) Z f o r a single year of  f i s h i n g , and (b) the mortality c o e f f i c i e n t s f o r two years of f i s h i n g Al  106  Simulation results and comments.  Typical analysis of a  Schaefer (1957) general production model. All  Results of simulated, experimental  fishing.  Method 4  140  Modified  Beverton and Holt (1957) year-class model containing a Ricker (1958) r e c r u i t function Al 11 Total y i e l d per recruit capture ( t ) c  AIV CI  CII  157/9  (Y-p/R) associated with age of f i r s t  and f i s h i n g mortality ( F ) . .  Predicted total y i e l d and recruitment  166  index  173  Equilibrium y i e l d , rank values and y i e l d gain (+)  relative  to continuous f i s h i n g (FS 10) f o r growth-mortality  Case 1  193  Relative equilibrium y i e l d , seasonal pulse f i s h i n g results for F=1.5  196/ 202  viii LIST OF APPENDICES APPENDIX  Page  A  Analysis of f i s h e r i e s models using a gaming technique  B  Derivation of optimal  stock s i z e , recruitment  131  and f i s h i n g  mortality at MSY in a multi age group, single species  C  fishery  178  Seasonal pulse f i s h i n g results  190  ACKNOWLEDGEMENTS  I am indebted to Dr. P. A. Larkin for his infectious enthusiasm, encouragement and support throughout the preparation this d i s s e r t a t i o n .  The pleasure of working with Dr. J . Schnute  through many mathematically i n s p i r i n g moments is also g r a t e f u l l y acknowledged.  1 INTRODUCTION The early 1960s witnessed the beginning of a rapid escalation in global catches of f i s h and s h e l l f i s h products.  The nominal catch grew from 40.2  m i l l i o n metric tons in 1960 to 70.2 in 1971, and declined s l i g h t l y to 65.7 tons in 1973.  A similar but proportionately greater increase occurred in  the Convention area of the International Commission for the Northwest A t l a n t i c Fisheries (ICNAF) off Canada's east coast.  There, the nominal catch rose from  2.3 to 4.3 m i l l i o n tons in the 1960-71 period.  In ICNAF's experience the  heightened demand for fishery products was met by a substantial increase in f i s h i n g e f f o r t being directed against conventional and non-conventional species.  The effect of increased harvest rates became quickly apparent.  By  the mid 1960s certain f i s h stocks were being over exploited, and i t was recognized that some form of protection was needed short of d i r e c t e f f o r t c o n t r o l . As a r e s u l t , quotas (total allowable catches) were internationally  agreed upon  in 1970 for two halibut stocks, and i t has since been estimated (Regier and McCracken, 1975) tection  that, over a l l , the number of stocks requiring quota pro-  w i l l reach 56 by 1975.  additional  In f a c t , the total proved to be 59 and one  stock was cited for conservation in 1976 (Akenhead, 1976).  The preceding remarks focus attention on one of the chief of f i s h e r i e s management; i t  difficulties  i s , to prepare stock and stock-complex assess-  ments of improved precision at an increasing rate.  Solutions to this problem  are fundamentally important, not only in terms of conservation, but because of their future implications.  Without r e l i a b l e assessments the s c i e n t i f i c  c r e d i b i l i t y of management is placed in doubt and i t becomes increasingly d i f f i c u l t to support regulatory measures when they are most needed.  Without  2  reliable knowledge of the present state of a stock and the fishing mortality inflicted on i t (assessments), yield prediction is also uncertain.  From  this perspective, assessment is a central issue and the problem of improving its speed and reliability of analysis formed the original basis of this thesis. In the past seventy years a substantial body of theory, computational skill and practical information has been developed to carry out stock assessments.  To effect an improvement in the speed and reliability of such anal-  yses i t is possible to review this literature and isolate areas in which a gain in information or the development of a new theory might contribute towards these objectives.  A problem to recognize in such an approach is that  the reviewer is exposed to pre-conceptions that have either a common basis of historical development or are highly specific in their application.  It is  often beneficial therefore to reappraise a long-standing problem by reducing i t to its simplest elements, construct solutions at this level of understanding, and introduce realistic complexity in stages.  This second approach  was used i n i t i a l l y on the assessment problem by adapting a gaming technique to fisheries models.  The essential features of the process are as follows:  An interactive simulation model of a fishery is f i r s t constructed by an outside party and a second individual, known as the "manager," is then allowed to vary the effort level (boats) in successive years of fishing.  Following each application of effort, relevant  information from the fishery is printed out and, on the basis of these data, the manager must (1)  determine the underlying model system,  3  (2)  estimate the coordinates of MSY ( y i e l d , e f f o r t ) , and  (3)  achieve t h i s goal in the shortest possible time by a r e p e t i t i v e sequence of y i e l d prediction and f i s h i n g J  The f i r s t type of f i s h e r y examined here by gaming was a discrete-time version ( L a r k i n , 1974) of Schaefer's (1954, 1957) general production model. The objectives were those given above in the process d e s c r i p t i o n .  The im-  portant conclusion reached was that the time needed to achieve MSY was minimized only when data on c a t c h , e f f o r t and recruitment became a v a i l a b l e . This r e s u l t suggested the importance of e s t a b l i s h i n g - - based on e r r o r - f r e e data - - the minimal time needed to i d e n t i f y the unknown parameters of a y i e l d model, estimate MSY, and s t a b i l i z e the f i s h e r y at that point.  Such a  standard could act as the target in reducing the analysis time of future assessment models.  For t h i s purpose a d e t e r m i n i s t i c , multi age-class  Beverton and Holt (1957) model of a v i r g i n f i s h e r y was developed which included a Ricker (1968) recruitment f u n c t i o n .  The author was only aware  of the model type, and that the instantaneous rates of m o r t a l i t y and the c a t c h a b i l i t y c o e f f i c i e n t were constant f o r a l l age groups. obtained from the f i s h e r y was free of e r r o r .  The information  On the basis of these assump-  t i o n s , complete parameter estimates were obtained a f t e r two years of f i s h i n g and the system was s t a b i l i z e d at the point of MSY w i t h i n ten additional years.  In t h i s case, as in the Schaefer model, data on recruitment were  needed before a s a t i s f a c t o r y s o l u t i o n could be reached. The r e s u l t s and management implications of the two models j u s t described are reported elsewhere (Appendix A ) .  Here, i t - i s more c r i t i c a l to note the  ^See Tyler (1974) for a f i s h e r y simulation model i n v o l v i n g graduated cost penalties f o r time and quantity of information.  4  lessons or p r i n c i p l e s that emerged from the gaming experience to redirect the emphasis of t h i s report.  The remarks that follow are purposely s i m p l i f i e d .  Consider an existing fishery in which the management objective is to maximize the average or long-term y i e l d .  If  this goal is a t t a i n a b l e ,  i t must  f i r s t be assumed that a condition of surplus production w i l l exist and can be generated by a positive response in recruitment or growth, or a decline in natural mortality.  Given that a change in one or a combination of these e l e -  ments occurs, then four interrelated factors must be known to d i r e c t the fishery towards the objective.  F i r s t , management must be able to determine  the present state of the fishable stock at prescribed intervals of time.  The  minimal amount of information needed is an estimate of stock abundance or of year-class strengths.  The description is substantially improved i f  of the rates of growth and natural mortality are also a v a i l a b l e .  estimates Second, a  measure of recruitment or forthcoming additions to the fishable stock must be obtained.  It  i s the  only measurable safeguard available to management to pre-  d i c t a f a i l u r e or temporary decline in the f i s h e r y , even i f the cause cannot be immediately i d e n t i f i e d .  Prior notice allows for the preparation of a response  to the situation and not a reaction a f t e r the event has disturbed the f i s h e r y . This  p r i n c i p l e holds equally well in the case of an improvement in the strength  of r e c r u i t s .  If  values can be assigned to both f a c t o r s , then the instruments  to detect changes in the fishery have been established.  The principal agent  of change, and the t h i r d f a c t o r , i s then readily i d e n t i f i e d as the e f f o r t directed against the fishable stock, and a measure of i t s effectiveness over time is the f i s h i n g mortality  rate.  The reader w i l l recognize that the conditions needed to place the fishing process on an experimental  basis are almost complete.  If management can  establish its present position in a fishery (factor 1), and recognize the changes (factors 1, 2) induced by a disturbance of known magnitude (factor  3),  then it is possible to move towards the objective in a time-consuming but regulated manner.  To complete the experimental analogy, what is missing in  the overall procedure is a method of hypothesis testing. yield prediction, f u l f i l l s this function and more.  The fourth item,  It provides a unifying  model or conceptual framework for all three elements.  Each change in effort  can now be considered a separate experiment to test the predictive value of the model.  Through repeated fishing trials and adjustments confidence in pre-  diction can be improved by degrees. experimental counterpart.  The benefit of each improvement has an  Changes of greater magnitude can be imposed on the  fishery, thereby reducing the number of time steps needed to reach the final objective. The dual potential for reducing assessment time and increasing predictive accuracy appears to rest in the study of yield models.  The essence of such  models, however, is fisheries yield equations, and i f improvements are to be expected in assessments and predictions, the assumptions underlying the components of these equations must f i r s t be understood.  The initial  objective  of this thesis i s , therefore, to reappraise the basic formulae used in yield equations and contrast their assumptions against some of the known realities of a fishery.  Both the practical yield models that use these formulae and the  statistics gathered in the course of a modern fishery form a necessary part of the analysis.  They are examined specifically for their usefulness in diagnos-  ing the present state of a stock. There are two outstanding characteristics of a fishery that negate a simplistic approach.  The f i r s t is that environmental factors predominantly  6  influence the number of recruits and this quantity is effectively a variable. Yield equations which incorporate a deterministic, stock-recruit relationship can therefore lead to inaccurate predictions.  For this reason, recruit  functions are usually excluded from yield equations and estimates are prepared separately.  The remaining elements of fishable-stock abundance and  expressions for growth and both mortalities are then descriptive of only the present state of the stock.  A yield in weight can s t i l l be calculated from  the contracted equations (or catch in numbers i f the growth term is removed), but full predictability is no longer possible. The second characteristic refers to social, economic, or political factors that can constrain management from changing or even controlling-fishing effort at w i l l .  As a result, the experimental use of effort to gain informa-  tion about the system may be substantially reduced, and management must rely heavily on the interpretation of past events in the fishery.  With both char-  acteristics operating simultaneously, i t would therefore be desirable i f a group of statistics could be developed to accomplish the following: (1)  Assess the present state of the fishable stock in terms of year-class abundance and individual growth rates by age group, and supply an average value for the fishing mortality rate.  (2)  Evaluate the fishing mortality independently of the effort used in the fishery.  (3)  Estimate the fishing, natural mortality, and growth rates independently of-variation in year-class strength.  (4)  Allow these entire calculations to be performed at the end of each fishing season, without relying on historical data.  7  With this a b i l i t y ,  estimates of parameters and variables can  be updated seasonally.  The second objective of this thesis is to demonstrate the f e a s i b i l i t y obtaining this complete assessment.  of  Given the assumption of random d i s t r i b u -  tion of e f f o r t or f i s h , the data requirements from the fishery are those of total y i e l d in weight and length, and catch in numbers by age group. independent estimate of the von Bertalanffy t , 0  i s also required.  An  (1938) growth equation constant,  Appropriate y i e l d equations are developed and the  technique applied to a theoretical  model of a multi-age class trawl f i s h e r y .  Growth in length and weight is also assumed to take place according to the von Bertalanffy equations, but the r e s t r i c t i o n of a cubic exponent in the weight-length relationship i s removed.  Practical application of the model is  presently limited because y i e l d in length and weight data in a desirable form are not a v a i l a b l e .  8 FISHERIES YIELD EQUATIONS  1.  General Principles  Ideally, a biomass model of a fishery should include all  time-dependent  factors that contribute to and detract from the fishable mass of the target species.  To achieve this level of understanding it is necessary to know how  the environmental, biological and fishery systems interact to affect:.tne,biomass over the course of time.  A full appreciation of these processes must  also include the self-regulatory response of the population.  The value of  such a comprehensive model is to minimize the residual error associated with yield prediction and stock assessments, and the research efforts to identify the underlying relationships are therefore closely linked with this goal.  The  degree to which the ideal model can be approximated depends partly on the quality and quantity of data that is available from the fishery and from independent research activity.  Collectively, this information must be rendered  into concise mathematical form for two purposes:  1) to establish the working  relationships between the component parts of the model, and the means of estimating the parameters.  2) to provide  This last point is critical to the  solution of the model and is therefore a vital part of the model building process.  If  values  cannot-be assigned to the parameters, the model is  reduced to a theoretical conceptualization that is neither testable nor applicable in a practical sense. The greater source of departure from the ideal case stems from incomplete knowledge of the processes that affect the biomass and hence the yield from a fishery.  As a result, simplifying assumptions are introduced into yield models  9  that lead to an ordered decrease in complexity and an increased departure from reality.  A general description of these models together with the con-  sequences of simplification are given below. The practical requirements of a yield model dictate that each element in its structure be an estimable quantity in either absolute or relative terms.  If the time-dependent factors or process are unknown or cannot be  quantified then their combined effects must be compressed into variables that can be measured or estimated.  Thus, the environmental, biological and  fishery systems are known to interact to affect the biomass, and although the details of their inter-relationships may not be clear, their resultant effects are identifiable in terms of six key variables.  These quantities  are recruitment, growth, mortality due to fishing and all other causes (natural mortality), and fish movement in the form of immigration and emigration.  Together, these quantities represent the first-order simplification  of the idealized model, and each principle component is readily associated with a contribution to or a loss from the existing biomass.  Compared to  the ideal case the complexity of the model, the number of parameters and the amount of information needed to estimate the parameters are substantially reduced. The apparent gain in simplicity, however, is not achieved without cost. By focussing attention on the end results of the system's interactions, each variable now represents a composite of multiple effects, and the sources of variation within each component can no longer be identified.  As a consequence  the reliability of the model is affected in two critical areas.  First, the  derived values of recruitment, growth, mortality and fish movement must each carry a full variance load.  Similarly, by reducing the number of parameters  10  it is less likely that the remainder will each have a constant value.  The  common effect of these features is to increase the degree of uncertainty attached to each principle component and parameter, and ultimately to the predicted yield i t s e l f .  Secondly, since the cause-effect relationships that  influence the components and parameters are not determined, it is extremely difficult to interpret correctly the past and present fishing results or make accurate predictions of yield.  To offset these combined d i f f i c u l t i e s , to keep  pace with events in the fishery and simultaneously offer a sense of direction, management must re-estimate the parameters and monitor the components on a frequent basis. Additional stages of simplification in a yield model may be achieved by reducing the number of key variables through combination or outright elimination.  For instance, it is understandably difficult to estimate the change in  biomass over time that arises from immigration or emigration.  Eliminating  these variables from the model does not deny or ignore the net effect of fish movement. population  It is simply a mathematical convenience that  models a closed-  fishery without specifying how the yield is affected by immi-  gration and emigration.  By not including these variables, however, the unex-  plained variance of the predicted yield is increased beyond the first-order simplification, and this condition holds for the removal of any of the six key components.  The only exception that might occur is i f the fishery oper-  ates over the total surface area of a closed body of water.  In this case the  immigration and emigration terms can be safely ignored and the resulting model is justifiably descriptive of a closed population. A third variable that may or may not be directly incorporated into a yield model is recruitment.  In contrast to the movement terms it is not a  11 question of excluding the recruitment component and allowing the residual error of the yield to increase.  To do so would be to ignore a major factor  contributing to the fishable stock, and the resultant yield predictions would bear l i t t l e relationship to the amounts actually taken by the fishery.  It is  a problem, rather, of choosing between alternative methods of estimating recruitment, three examples of which are: 1)  direct sampling of pre-recruit abundance  2)  assuming constant recruitment as in the simple Beverton and Holt (1957) model;  3)  or  by means of analytical models which rely totally or in part on the assumption of a density-dependent relationship between stock-size and the number of young produced.  This method in-  troduces new parameters into the yield model, however, and the means of estimating these quantities must be considered. The last simplification method that must be mentioned is that of combining two or more of the key components.  Recruitment and growth offer an  immediate choice since these variables represent the principal means of increasing the fishable biomass.  Natural mortality may in turn be combined with  this new variable to achieve a net measure of stock that is subject to fishing. Each of these combinatory steps, however, reduces the breadth of the yield model and its ability to account for changes in the biomass.  The cost of sim-  plification in this case, as in variable elimination, can also be equated to a loss in predictive accuracy.(Watt, 1956).  It is generally to be expected that  as the number-of variables and parameters are reduced the remaining"components will be less able to explain the observed variation in yield.  1 2  Despite the limitations  introduced by s i m p l i f i c a t i o n , the methods just  described provide the most rapid means of gaining information on a f i s h e r y . For management purposes, the emphasis on speed is essential i f the  time-lag  between events in the fishery and suggestions for corrective action is to be minimized.  In general, practical y i e l d models r e f l e c t this purpose by using  a s i m p l i f i e d approach, and either by default  (due to lack of information)  design, the f i s h movement component i s not included.  or-  In the forthcoming  discussion a similar viewpoint is taken.  To begin the a n a l y s i s , i t  is helpful to v i s u a l i z e  relationships between recruitment fishing  (F), and a l l other causes  (R), growth (M).  (Figure 1) the  (G), and mortality due.to  In the diagram these components  appear as independent processes that are linked within a regeneration  cycle.  The f i s h movement terms are not included in order to represent a closed system, multi age-class fishery f o r a single stock. A preliminary condition for the sustained existence of the fishery is the a b i l i t y of the stock to generate a surplus production.  It may occur  by means of an increase in recruitment or growth or a decline in natural mortality,  or a combination of these responses in response to f i s h i n g  pressure.  Under these conditions the opportunity exists to achieve the  greatest y i e l d that i s coincident with maximum, surplus production.  To  assess the progress of the fishery towards this g o a l , a model i s needed to relate the change in fishable stock biomass mortality over a given period of time.  (B)  to the accumulated f i s h i n g  Using Figure 1, the necessary model  can be constructed from a material balance centred on the fishable stock. If  B-i and B  ?  represent the biomass at the beginning and end of the time  13  YIELD  FISHING MORTALITY (F)  FISHABLE STOCK BIOMASS (B)  PRE - RECRUITS  t\  / I  N:  I RECRUITMENT  -3»—GROWTH  >• I  (G)  (R)  N, K GROWTH  I JUVENILES  ADULTS  / NON- FISHING MORTALITY (M)  RECRUITMENT.  Figure 1.  A schematic diagram of the simplified relationship between recruitment, growth and mortality for a multi age-group, closed system fishery. Legend: N = number of fish in age group i ; n = last fishable age group; ••• dotted section indicates that a juvenile age group(s) may or may not be included in the fishery.  14  interval, the change in biomass is represented by the equation  AB = B - B = (R+G) - (F+M), 2  (1.01)  1  and since yield is of primary importance in a fishery, the above expression can also be rearranged in terms of the yield equation F = (R+G) = AB - M  (1.02)  In either case the reader will recognize Russell's (1931) associative model of a closed system fishery in which the biomass increments and decrements are represented by (R+G) and (F+M), respectively.  Since these terms each  refer to a change in biomass they are compiexed variables that include a product of numbers and weight integrated over the time interval.  For ana-  lytical purposes this form of representation is unsuitable, as Russell (1911) himself realized.  Separate and independent expressions are needed for re-  cruitment, growth and mortality to reflect changes in numbers and weights. The problem of representing these variables in a suitable mathematical form, and their amalgamation into yield equations must now be considered. An essential part of this process is to indicate clearly the assumptions used in developing the formulae, and the remainder of this section is devoted to this purpose.  15 2.  Fishing and Natural Mortality - Catch in Numbers  The simplest representation of mortality rests on the assumption that at a given time, t , the rate of change in f i s h numbers, N, is d i r e c t l y proportional to the number present. fisheries l i t e r a t u r e  and i t has been in continuous use since that time.  pressed mathematically,  dN(t) dt  Baranoff (1918) o r i g i n a l l y introduced this idea into Ex-  i t follows that  N(t)  (1.03)  where t and N are recognized as the independent and dependent v a r i a b l e s , respectively.  Since death due to fishing (F) and a l l other causes (M) re-  sults in a decline in numbers, expression (1.03) can be rewritten as an equali t y to represent the rate of change due to each mortality source:  (f)  p  -  - F(t)N,  (1.04)  (1.05)  In these f i r s t - o r d e r equations the proportionality terms F(t) and M(t) are i d e n t i f i e d , respectively, as the instantaneous rates of f i s h i n g and natural mortality.  That their units involve a rate term can be seen by rearranging  the above equations to indicate s p e c i f i c rates of change, i . e . , dN/dt.N, the dimensions of which are "per unit time".  The functional notation (t) w i l l be used on f i r s t appearance of a timedependent variable and thereafter i f needed f o r emphasis.  16 Similarly, i f fishing and natural mortality act concurrently the instantaneous rates may be added to obtain the total rate of decline in fish numbers,  dN  =  .  ( F + M ) N  =  _  z  (  t  )  N  j  ( 1 > 0 6 )  where Z(t) = the total instantaneous mortality rate. It should be added at this point that more complex expressions of mortality can be derived which include second or higher order terms, such as,  = - (aN + b N + c N + ...k 2  N ),  3  n  but there are practical limits to this form of theoretical expansion. The evaluation of the additional parameters (b, c, . ...k) must be considered.  If  they cannot be determined, or i f the residual error in the final yield model is not reduced, there is no practical justification for the increase in complexity.  Since one objective of this thesis is to develop a practical assess-  ment method, on the basis of general acceptance  the treatment of mortality  will be limited to the simplest assumption. The relationship between the rate of capture in numbers, F, and the effort expended in obtaining the catch must now be developed.  The sources of  information used in this task, to which the reader is referred for an enlarged treatment of the subject, are those of Ricker (1940, 1944, 1968, 1977), Widrig (1954a, 1954b), Gulland (1955, 1969), Beverton and Holt (1957), Paloheimo and Dickie (1964) and Cushing (1970).  In their final-form the formulae presented  here-do not differ from those of the above authors. ;  The method of  -  17  presentation, however, i s somewhat d i f f e r e n t i n order to emphasize the assumptions, the meaning of and the units attached to common terms. I n i t i a l l y , i t i s necessary to define two terms: effort.  e f f o r t and standardized  Consider a unit of gear (g) that i s in operation for a short time  interval At.  In a physical sense the gear may be interpreted as the force  or agent of f i s h i n g m o r t a l i t y and At the duration over which i t a c t s .  The  product of these terms constitutes e f f o r t ( f ) , a measure of m o r t a l i t y work that may be expressed in p r a c t i c a l units such as "boat-days".  Over the  i n t e r v a l At the number of gear units in operation i s a time-dependent v a r i able and e f f o r t must therefore be expressed in the generalized form Af(t)  = g(t)At.  (1.07)  The catching power (catch/hour, say) of two d e s c r i p t i v e l y s i m i l a r vessels need not be the same despite an equal expenditure of e f f o r t .  Traw-  l e r s , f o r example, may d i f f e r in horsepower, gear capturing e f f i c i e n c y , net s i z e , operating v e l o c i t y , or experience of the crews.  I f a s i n g l e vessel  and i t s associated gear i s selected as a standard ( g , Af) the catching power of another vessel ( g , A f ) can be expressed in terms of t h i s standard by 1  1  means of a power f a c t o r ( p ) , or r a t i o of catching powers.  If i t i s assumed  that both units operate on the same density of f i s h , then the following expression i s v a l i d : Af = g(At) = p g '  (At) = p A f  To develop a r e l a t i o n s h i p between catch and e f f o r t i t i s usually assumed that the number of f i s h captured (Ac) over a short time i n t e r v a l  (At)  is  18  proportional to the number of encounters between fish and gear.  It is  immaterial at this point whether the gear is stationary and fish move towards i t , or the gear is drawn through the water.  It is easier to visualize the  capturing process in terms of mobile gear and therefore a trawler is used here as an example.  For a unit operation of a standard vessel i t follows that  Ac(t) = q' Af(t) ( t ) ,  (1.08)  ,::  P  where Ac(t) = the difference in accumulated catch in the interval At,  q ' = a proportionality constant, and 1  p(t) = the local density of fish encountered by the net If both sides of equation (1.08) are divided by At, then in the limit as At •->• 0, Ac/At -* dc/dt, and Af/At ->- df/dt, then the rate of capture is expressed as  f = q " ^ P ,  (1.09)  and the rate of capture divided by the effort rate yields  dc/dt _ dc _ , , df7dt~dfn q  ,, ,_  N  p  Equation (1.10) establishes the common relationship that the catch per unit effort is proportional to the local fish density.  Despite the fact that  the expression is given in instantaneous terms, i t is clear that the  19  derivation rests on the "the catch rate per unit effort rate".  The validity  of this expression in terms of comparative rates is preserved in subsequent formulae involving stock density and longer periods of time, such as a fishing year. The above relationship (equation 1.09) operating on the fishing ground.  is valid for any single boat  It bears no relationship to the stock den-  sity (D) as a whole, and does not indicate how the rate of capture is influenced by additional units of gear distributed in time and space.  To make this  transition and establish continuity between the rate of capture and the rate of decline in fish numbers, one of the following assumptions is necessary: (1)  that either the fish or the fishermen are distributed at random over the fishing grounds, or  (2) the fish are highly mobile and capable of re-invading locally depleted areas in order to maintain a random distribution (Ricker, 1940, 1944). Two conditions follow from these assumptions.  First, the proportionality is  established that  P  (t)  = 3 D(t) = 6 H{t)  (1.11)  A  where  g = a dimensionless proportionality constant A  = the area occupied by the total fishable stock, and N(t)  is  specifically redefined as the abundance, or total number of fish that may be subjected to fishing.  Secondly, the catch rate is directly proportional to  the number of units of gear in operation and df(t)/dt is interpreted in a plural sense.  20  Substituting equation (1.11) into (1.09) and (1.10), the relationships follow:  dc  7W= dt  q'  df dt  df A dt  (1.12)  and  dfTA = q' N ,  (1.13)  where q = q"e, and has the units of area/boats x time, and df/A dt = the 1  fishing intensity-measured as total standard effort/unit area/unit time. If the area occupied by the stock is assumed to be constant, then q = q ' / A , and equations (1.12) and (1.13) are reduced to  dc dt  (1.14)  and  f = q N  ,  (i.i5)  where q, the catchability coefficient, takes on the dimensions (boat x time)~\ Under the assumed condition of redistribution the rate at which fish are being captured is now equal to the negative rate of decline in numbers due to fishing.  From equations (1.04), (1.13) and (1.14) i t therefore follows:  21  dN(t) dt  =  dc(t) = F(t) N(t) dt  (1.16)  or, after simplification, and depending on the constancy of A,  F(t) = q  F(t) = q  or  df(t) dt  (1.17)  In these last equations the instantaneous fishing mortality rate is seen to be proportional to either the fishing intensity or the fishing effort rate. The respective catchability coefficients, q' and q, are also recognized to be neither equal dimensionally nor in value. In practice, the instantaneous catch per unit effort or its average value over a short time interval is not likely to be directly proportional to N.  Local concentrations (or depletions) of fish and effort, and cooperation  between fishermen upset the assumptions of random distribution, and may contribute to large variations in catch rates.  It is usual, therefore, to derive  fisheries statistics on a longer time basis over which the fluctuations may be averaged.  Mathematically this simply involves integration of the above for-  mulae, but the realistic interpretation of F(t) [or M(t)] as a function of time represents a problem. the variable F(t), i . e . ,  The difficulty is examined here only in terms of M is assumed to be a constant.  The number of fish at any time t may be calculated by i n i t i a l l y ing equation (1.06) as  d N = d(ln N) - -Z(t) dt N  rearrang-  22  and integrating the left hand side of this equation between the indicated limits of N  /  N(t) No  t t d (In N) = - / F(t)dt - • / M dt o o  ,  to obtain, after taking exponentials and rearranging, t t t - / F(t)dt - / M dt ' - / F(t)dt - Mt N(t) = No e = No e 0  0  (1.18)  0  where No = the initial abundance at time t = o. The integral of F(t) may be treated separately for the moment, and remains unchanged i f multiplied and divided by J*^ dt, i . e . ,  t  A(t)  l W  f  dt =  0  d t  F  /  o  •  dt  t fl dt 0  By definition (Smail, 1949) the integral of the function /gF(t) dt divided by the integral of time over the same limit is equal to the average value of the function, i . e . , F.  Similarly,  dt = t, and substituting these values into  (1.18),  N(t) = No e "  ( F +  M  )  t  (1.19)  The function M(t) can be similarly dealt with to emphasize seasonal mortality, for instance, and to obtain the average values M and Z.  Although the seasonal  and local variation in f, F and M may be appreciable, equation (1.19) usually appears in the literature as  -(F + M)t -Zt N(t) = No e = No e  (1.20)  indicating that the coefficients F, M and Z are assumed to be constant over the time interval.  In the derivations that follow, the latter form of presen-  tation is retained although i t is clear from the foregoing that average values are intended. Substituting the value of N(t) from equation (1.20) into (1.16) the instantaneous increment in accumulated catch is expressed as  dc(t) = F N(t)dt = F No e "  Z t  dt  ,  and the total accumulated catch (C) is arrived at by integrating this equation between the limits c (o,C) for t (o,t):  d c(t) = F  No/J  e"  Z t  dt  (1.21)  and  C(t) = F No  (1.22)  Z  where F / Z = the fraction, by numbers, of total deaths (N -(l-e" ).) ;Zt  0  attributed to fishing.  that is  24  It is more desirable, however, to express the catch in relation to the average abundance N" over the time interval rather than its initial  value, N .  To arrive at this point equation (1.21) may be rewritten as indicated below and divided by  dt, i . e . ,  F  'o  ^o "  N o  e  =  "  '$  d t  f  l  Z t  d  t  d t  and integrating,  C (t) R  = £  (  t  )  = F No  ( 1  ~  e  = F N(t)  Z t ) z  (1.23)  where C = the average catch rate, and R  N" = the average abundance, both take over the total time t. If the time interval is set equal to one year (t = 1), then equations (1.22) and (1.23) are reduced to the common form  C" = C = F No R  Z )  = FN  (1.24)  the catch equation of Baranoff (1918), and on an annual basis the accumulated catch (numbers) and the average catch rate (numbers/year) are numerically equal. Similarly, integrating equation (1.17) over the period of a year leads to the expressions  F =  and  F = qf  ,  (1.25)  25  where F (and incidentally M and Z) is an instantaneous mortality  rate, or  c o e f f i c i e n t , expressed on an annual b a s i s , f/A = the total f i s h i n g i n t e n s i t y , e . g . , total boat-hours/sq. mile/year, and f  = the total e f f o r t  rate;  e . g . , total boat-hours/year  F i n a l l y , by substituting equation (1.25) into (1.24) and rearranging, two interpretive  £  / A  = q' N  formulae are derived:  or  £  = q' | = q D  The conclusions follow:  .  (1.26)  i f A is not constant, the total catch per unit  f i s h i n g intensity is proportional to the average abundance; assumed to be constant, the total catch/total e f f o r t  but i f A i s  (catch per unit e f f o r t )  is proportional to the average stock density.  The principal formulae relating the catch in numbers over time and during the concurrent operation of fishing and natural mortality have been developed. The assumptions that support these relationships have also been s p e c i f i e d . The primary components of growth and recruitment remain to be discussed.  3.  3.1  Growth Models  Generalized derivation  With few exceptions growth models that have been adapted to or evolved for use in f i s h e r i e s have used physiological considerations as a basis of  26  development.  T h u s , the r a t e o f change i n weight has been a t t r i b u t e d to  d i f f e r e n c e i n a n a b o l i c and k a t a b o l i c r a t e s (von B e r t a l a n f f y ,  1938), to  the the  d i f f e r e n c e i n energy c o n t e n t o f r a t i o n s l e s s m e t a b o l i c e x p e n d i t u r e (Paloheimo and D i c k i e , 1 9 6 5 ) , and t o an e m p i r i c a l f u n c t i o n of s i z e t h a t p a r a l l e l s s t a n d a r d i z e d p h y s i o l o g i c a l r a t e s ( P a r k e r and L a r k i n , 1959).  S i m i l a r l y , Taylor  (1962) proposed a g e n e r a l i z e d growth f o r m u l a i n which the parameters are i n tended to have m e t a b o l i c s i g n i f i c a n c e .  Appropriate modification of this  last  model i n c l u d e s the " s e l f - i n h i b i t i n g " , o r a s y m p t o t i c , growth e q u a t i o n o f von B e r t a l a n f f y  ( 1 9 3 8 ) , the p a r a b o l i c e x p r e s s i o n o f P a r k e r and L a r k i n  (1965),  and the e x p o n e n t i a l growth f o r m u l a o f R i c k e r (1968) as s p e c i a l c a s e s . of i t s  g e n e r a l i t y , the T a y l o r (1962) model i s used here to i l l u s t r a t e  v a t i o n o f growth e q u a t i o n s . o f the a u t h o r i s not r e t a i n e d  To a v o i d s y m b o l i c c o n f u s i o n the o r i g i n a l  the d e r i notation  entirely.  Under c o n s t a n t e n v i r o n m e n t a l 1)  Because  conditions, i f  it  i s assumed t h a t  the r a t e o f change i n weight i s equal to the d i f f e r e n c e i n a n a b o l i c and k a t a b o l i c r a t e s , and  2)  these l a t t e r  r a t e s are p r o p o r t i o n a l , r e s p e c t i v e l y , to the a b s o r p t i v e  s u r f a c e area ( s ( t ) ) and weight  ( w ( t ) ) o f the f i s h , the  relationship  can be w r i t t e n  = h s(t)  - k'w(t)  ,  (1.27)  where the c o n s t a n t s h = weight s y n t h e s i z e d per u n i t time per u n i t s u r f a c e a r e a , and k'=  the k a t a b o l i c weight l o s s per u n i t time per u n i t  weight.  27  If i t is further assumed that:  s(t) = A t ) P  ,  and the weightrlength relationship is adequately described by  w(t) = a £ ( t )  ,  b  (1.28)  where p, a = proportionality constants £(t)  = the fish length at time  t  m = an exponent relating length to absorptive surface area, and b  = an exponent relating length to the "metabolically effective" weight (Brody, 1945),  these formulae can then be substituted into (1.27) and rearranged to obtain  dt  =  E£  (m-b+l)_  where E = hp/ba;  K£  _  (  U  2  g  )  and K = k'/b>, a growth coefficient.  Equation (1.29) may be integrated between the limits of Jl (& , £ ( t ) ) and time (0, t) to yield:  £(t)  where i  Q  ( b  -  m )  = E/K - (E/K - £ (b-m) -K(b-m)t )e  = the intercept length at time zero.  In the limit as  t -> °°, £^ ^-> E/K, a constant identified as L«>, the asymptotic length, and i f -ni  28  i  o  =0 when t = t , (the theoretical time when the fish length is zero) the o •  above equation reduces to the generalized form:  a  (b-m)  =  Loo  (b-m)  (1-e  -K(b-m)(t-t )  (1.30)  o  Three well-known growth formulae may be derived from (1.30) depending on the assumptions concerning K and the exponent (b-m). 1.  Asymptotic, isometric growth:  b-m = 3-2  =1.  In addition to a constant environment, i f it is assumed that growth is isometric and the specific gravity of the fish remains constant, then b = 3 and m = 2.  Under these conditions (1.30) is reduced to the familiar von Bertalanffy  (1938) equation  £ ( t ) = L~ (1 - e  (1.31)  and substituting for length in terms of weight from (1.28) the relationship is derived:  w(t)  where W 2.  = W (1 - e - K ( t - t ) 3 0  = a(L°°)  (1.32)  }  , the asymptotic weight.  Exponential growth:  b = m  From equation (1.29), when b = m, the relationship follows that:  dt  = (E-KU  29  and i n t e g r a t i n g between d e f i n i t e l i m i t s f o r  A  ( ) = I T  Q  e  "  ( E  K ) t  = I  Q  e  H t  i  and t ,  ,  (1.33)  and, s u b s t i t u t i n g (1.28) in the above equation,  w(t)  _ o  (e  w  H t  )  b  or w(t) = w e  ,  G t  Q  (1.34)  where H = the instantaneous growth rate .in l e n g t h , and G = bH;  the instantaneous growth rate i n weight, both expressed  on an annual basis (see F, M and Z, page 25). w = the i n i t i a l weight at Q  3.  Parabolic Growth:  m<b;  t = 0 K = 0,  Returning again to equation ( 1 . 2 9 ) , and l e t t i n g K = 0, separating variables and i n t e g r a t i n g between d e f i n i t e l i m i t s , the r e l a t i o n s h i p i s obtained:  In the notation of the o r i g i n a l authors (Parker and L a r k i n , 1959), where a = (b-m)E and z = b-m, s u b s t i t u t i o n y i e l d s  £ (t) Z  = at  +  l 0 Z  30  and letting t = t  when £ = 0, the equations for length and weight (from 1.28)  Q  q  are derived as  *(t) = [a(t-t  )]  1 / Z  and  w(t) = [ a  a (t-t )]  z / b  b / z  Q  The preceding growth models have purposely stressed the physiological interpretation of parameters as intended by Taylor (1962), and particularly by von Bertalanffy (1938) in the case of isometric growth.  The derivation of  Ricker's (1968) exponential formulae and that of parabolic growth, however, are mathematically inconsistent with the implied theory.  In order to derive  the former relationship i t must be assumed that m = b, a clearly untenable proposition unless the "fish", for example, is tube shaped and its net weight never exceeds that of its own intestine.  Similarly, for parabolic growth,  katabolic metabolism cannot be eliminated simply by setting K(= k'/b) equal to zero in order to achieve the correct formulae.  The intention of the original  authors (Parker and Larkin, 1959) to represent growth as a function of size and as the resultant of complex physiological and ecological effects, is clearly summarized in their basic equation  4-git> = w t) x  k  (  where k is a proportionality constant "...and x is a fractional exponent less than unity".  The relationship is in keeping with observed physiological rates  31  and there is no need to invoke a special biological meaning for the parameters. The physiological interpretation of L°°, K, W°° and b in the von Bertalanffy (1938) formulae have similarly been challenged by Parker and Larkin (1959), Ricker (1960), Paulik and Gales (1964) and Knight (1968).  Taylor (1958, 1962)  has also indicated that K is not constant, but a function of temperature. While the search for biological meaning in these growth formulae persists, their immediate practical value is recognized as empirically descriptive models of growth.  In this sense their f l e x i b i l i t y , and particularly that of the  Ricker (1968) model, offers a wide range of choice to meet the observed growth patterns of individual species.  With one possible exception the flexibility  achieved by a suitable choice of parameters becomes a valuable property that can be readily used in assessments and yield predictions.  The exceptional  reference is that of isometric growth, but even here a generalized, allometric weight formula can be derived.  If regression analysis indicates that the best  f i t to length-at-age data is obtained by means of a diminishing growth-rate equation (1.31), and the weight:length relationship is adequately described by w(t) = a £ ( t ) , the appropriate weight formula (Richards, 1959) is b  w(t) = Woo (1 - e ' ^ ^ o V  where  b  ,  (1.35)  can now take on values in the reported range of about 1.4 to 4.0  (Taylor, 1962).  3.2  Allometric, asymptotic growth, parameter estimation An extensive amount of work has been carried out on the techniques used  to estimate the growth parameters K, t , and L°°.  For an introduction to the  32  literature and a critical review of the methods the reader is referred to the computational bulletins of Ricker (1968, 1975), and to Allen (1966) for best least-squares estimates of these parameters when fitting to observed data. Here, only a brief outline of the formula derivations are given to support the fishery statistics developed later in Section  III.  If i t is assumed in general that mortality between age-groups is not size selective, and in particular that growth in length follows the von Bertalanffy (1938) curve, then under constant environmental conditions the length relationship between successive age groups (i)  A. = L j l - e -  K ( l  is described by  '-V)  (1.36)  and  = L j l - e - ^ - V )  ,  (1.37)  and subtracting (1.36) from (1.37) and simplifying,  " *i = L e - ^ - V w  But from (1.36)  e  - 1  ^ ' ^ 1  -  (l-e- ) . K  =  L^-Jl.  .  Substituting this relationship into  the above expression, simplifying and rearranging,  £  i + 1  = Ljl-e"*) + e  _ K  A.  Equation (1.38) is of the linear form 1973) of successive ordinate values of  (1.38)  y = a + bx  and a regression (Ricker,  plotted against those of £^  33  (absissa) yields a line of slope  e~  (the k of Ford, 1933) and y-axis  -K intercept  L^l-e  ).  Values of K and  can therefore be readily calculated.  The intercept of the graph with a 45° diagonal drawn from the origin (0,0) can also be used to estimate L directly. ro  In its original form equation (1.38)  was derived empirically by Ford (1933), later by Walford (1946), and its graphical representation is commonly known as the Walford line. Equation (1.36) can also be expanded in the form  i• =  L- L e  T  00  •e  0  CO  Rearranging this expression and taking natural logarithms gives  In (L, - z.) = (ln  Loo  + Kt ) - Ki.  (1.39)  Q  A regression (Ricker, 1973) of successive values of ln yields a straight line of slope which  t  -K and intercept  (L^-Jo.)  I = lnL  ro  on i , i+1, etc.  + Kt  , from  can be calculated as  to -  -  1  Q  1 n L  "  (1.40)  K  The weight:length relationship w(t) = a I (t) is implicit in the derivation of the weight formula (1.35) given earlier. the intra-age-group values of parameters age, the equation  w  i  =  a  £  i  b  In order to estimate  a and b for integer values of  34  can be expanded as a linear formula  ln(w ) = In a + b ln(&.) i  Again, a regression of ln(w^) on ln(a.) is used to determine the slope and  a  from the intercept  mated directly from  W  m  = a  ,  From a previous estimate of L , W^ is esti-  In a. ( L J *  b  o t  .  5  An alternative method of estimating b, K and W^ is derived as follows: Given that w. = W ( l - e "  o )  K ( i p t  )  ro  and w.  b  = Wjl-e'^^'V )  +1  b  ,  subtracting the former expression from the latter, rearranging and simplifying  1  1  b  w  i + 1  1  b  ,.b - K ( i - t  - w = W e i  v  ro  1  ),,  o (l-e ;  -K*  ).  1 1  But W e ~ ^ o ^ = W - w b  - t  b  , and substituting this expression into the above  b  relationship the following equation can be obtained  1 \  w  1 b  1  = W (l-e" ) + e" w b  +]  K  K  (1.41)  b  Equation (1.41) will assume a linear relationship i f the correct value of b  1  is chosen in a regression of w  b i+l  1  on w  b i  .  To determine this quantity I have written an iterative search program (BSRCH) to minimize the unexplained sums of squares about the line of best f i t . the optimal value of  b  is found, the slope,  provide estimates of K and W , respectively.  -K  e  1  , and intercept, ^b  When ^-j_ -Kj, e  35  4. 4.1  Yield Models - Combined Growth and Mortality  Generalized yield model To inject some reality into a discussion on yield models the following  observations are necessary.  First, there is ample evidence to indicate that  for temperate water fishes recruitment is not constant but a variable quantity, and its magnitude can be measured in terms of year-class strength.  It is  therefore more realistic to develop yield models on the basis of a generalized age group ( i ) ,  of numerical strength N^(t) in the year of fishing, and to  calculate the annual yield as the sum of all contributing ages.  Again, it is  well established that individuals within a given age group differ in size, and in yield equations using length {i^) or weight (w^) these symbols must be interpreted as average quantities.  The conversion of any of the previous growth  equations to that of a particular age group is readily accomplished by adding the integer value of i to the independent variable t, e . g . , from  w(t)  to  w.(t)  = W (l-e" ^ K  t_t  o^)  = w (l-e- < - o>) K  i+t  t  , o < t <«  b  ' °^- ^  b  t  1  (  K  4  2  )  The growth models given earlier imply a continuity or a continuous growth pattern from one age to the next.  As mentioned earlier, to retain this quality  the critical assumption must be made that mortality, in whatever form, is not size selective.  Finally, there is no mathematical or biological reason why a  concept of yield should be limited to that of weight.  Fish grow in length as  well as weight and it is as easy to conceive of yield in terms of "biolength" (Y  L  and B^) as i t is in "biomass" (Y  w  and B ) . w  In fact, as shown above, the  _  For a yield model based on length see Jones (MS, 1974)  36  majority of growth models make use of a weight:length relationship and therefore both types of yield models can be derived simultaneously.  If all  of these points are kept in mind, two basic yield equations can now be developed. In the previous section fishing and natural mortality are represented as independent, concurrent processes.  The catch is calculated on this basis  and supported by assumptions involving the distribution of fish (or fishermen) in space and time. assumed that: and  1)  If these concepts are retained and i t is additionally  gear selection is "knife-edged" (Beverton and Holt, 1957),  2) all fishable age groups are subject to common mortality rates of F  and M, then the basic equations for rates of yield (y) are:  djW*) ^ ^ _ ^  for weight  a n d  l e n g t h  =  d  (  y  ^  i  (  t  )  = F(B^)  i  (t) = F N.:(t)£.(t)  {] A3)  (1.44)  where N^(t) = the fishable age group abundance w^(t) = the average, age group weight, and A.(t)  = the average, age group length, all determined at time  t  (o < t <_ 1).  Equations (1.43) and (1.44) illustrate three major points and they are: 1)  the growth functions w.(t) and ^^(t) are free of any mortality component and are therefore independent processes  ,  2)  mortality, from whatever source, is associated with the rate of decline in numbers, e . g . , F N^(t)  3)  the rates of change of biomass and biolength, and therefore their yields, are dependent on the concurrent and opposing processes of growth and mortality.  To determine the average rates of yield ([Y^ ).j R  , (Y^ )..] R  over a year  both sides of equations (1.43) and (1.44) are multiplied by dt, divided by dt, and integrated between the limits of  , (VPJT  =  < V i  =  F  <Vi =  F  «  t(0, 1) to obtain  T  ( ]  4 5  )  and (Y  L j R  )  i  = (Y ). = F ( B ) . = F (NT). L  (1.46)  L  where (Y..). and (Y.). are, respectively, the annual yields in weight and W  1  L  1  length, and are numerically equal to their average annual rates. (B^).j and (B^)^ represent the average biomass and biolength, and are respectively equal to the average product of numbers x weight (Nw)^ and length (Nft). over the year. The above method can be illustrated using the integrable of weight and length of Ricker (1968).  Substituting (N ) .e" o  n  functions Zt  for N^t)  from equation (1.20) and the exponential formula for weight (1.34) in the general expression (1.43), and transferring dt  38  'l  V  l  <Vi •  d  F  'I  < o>i N  Z t  .(w ) 0  (G.-Z)t  i  <Vi - <v„'i 4 F  e-  G l t i e  n  e  dt  dt ,  •<S>i F B  (G.-Z)t,. e  'J  where (B^) - = the initial age group biomass at t=0 n  .  Following integration,  <Vi - 'Oi F  ^e !!'- '-' 1  1  F  (  Vi  '  (1 47)  i and similarly for length  (V ), - F ( i f ) , L  where G^ and  (e^_  z  -1) . p  ,  are, respectively, the age specific instantaneous rates of  growth in weight and length, expressed on an annual basis. 4.2  Beverton and Holt, constant recruitment, equilibrium yield model In relation to the biomass yield that can be taken from a fishery  Cushing (1972) distinguishes two areas of major concern to management: growth overfishing and recruitment overfishing.  The former refers to the  potential loss in yield i f fish are caught at too early an age and do not achieve maximal growth rates.  Peterson (1894) originally drew atten-  tion to the problem with regard- to North Sea plaice, and subsequently Ricker (1945) defined the point of optimal capture as the "critical size".  It occurs  at the average size at which the net gain in biomass is zero in the absence of  39  ffshing, i . e . , G..-M.. = 0.  Only yield models that separate the growth and  recruit functions, such as the Ricker (1968) model given above, are therefore capable of providing a solution to the growth overfishing problem.  Addition-  a l l y , an outstanding feature of these models is that under equilibrium conditions of constant recruitment, the growth and mortality functions can be used exclusively to calculate the anticipated yield for a given age of f i r s t capture and a range of fishing mortality (F) values.  By combining the projected yields  for all ages and values of F a yield isopleth (Beverton and Holt, 1957, p. 318) diagram showing contours of yield per recruit can then be constructed to estimate the critical age and MSY per recruit, and estimate the eumetric fishing curve. The growth and mortality model that has probably received the greatest attention in fisheries investigations is that popularized by Beverton and Holt (1957).  In its original "simple" form i t has found wide application in trawl  fisheries and for temperate water fishes that exhibit isometric growth according to the von Bertalanffy (1938) equations.  The principal assumptions and  conditions under which the simple model is applicable are as follows: 1.  It pertains to a multi age-class, single species stock in which growth and mortality (F and M) are continuous, concurrent and independent processes.  Additionally, the mortality rates are constant for all  groups in the commercial stock. 2.  Technically the model applies to a closed system.  The equations contain  no immigration or emigration terms. 3.  Environmental conditions are assumed to be constant, or their average effect over time demonstrates no measurable trend.  40  4.  The fish (or fishermen) are randomly distributed at all times. The basic assumptions used to define the mortality functions given earlier therefore apply.  5.  The trawl gear operates with "knife-edge" selection.  6.  A constant number of fish are recruited on the same date each year regardless of the fishing mortality.  7.  Growth in length for the commercial sized stock is asymptotic (von Bertalanffy, 1938), isometric (b = 3), and weight and length are re3 lated by the expression w(t) = a I (t). To develop the simple model let R equal the number of young fish in a year  class that f i r s t become exposed to the gear at an average age a^, but are not actually captured until some later, average age;; a . Conveniently, since time and age are both measured in the same units (years), these ages can be replaced by their time equivalents, t ving to age a  c  and t  respectively.  The number of fish survi-  is then represented by  R' = R e ^ V V  and the number available at any time after capture begins is R'e  -Z(t-t )  -M(t -t ) -Z(t-t ) = Re e , t c  where t' = an integer;  r  c  < t < t'  the age of exit from the fishery.  average weight of a fish at time  t  (1.48) Similarly the  is  -K(t-t ) , w(t) = Wjl-3  0  )  3  (1.49)  The product of equations (1.48) and (1.49) is equal to the biomass B^(t),  41  and the rate of capture is therefore defined as d(y )(t)/dt = F B (t). w  a single year class the total yield (Y ) W  X(= t' - t ) ,  VT  \  over its fishable l i f e span,  is calculated from  c  ( Y (  T  For  w  _ , VT "  t'  (  d(y )(t) = F /  /  0  w  c  Re  -M(t -t ) -Z(t-t ) -K(t-t ) e Wjl-3 ) dt 3  c  r  c  0  Following integration and simplifying the resulting expression  (Y ) =F w  T  W o o  Re  -M(t  c  -t  )  3  -nK(t  Man.e n=o  -t 0  0  )  f-r.M^i/u (^-(F+M+nK)^  (  K  5  Q  )  F+M+nK  where a n = (1, -3, +3, -1) for n = (0, 1, 2, 3). The biolength yield for the year class is also derived as  -M(t  (Y.)-r L  =  L  1  Re  -t c  ) r  [(1-e  _ ZX  Z  z  )  - e  - ^ r ^ J c  0  (1-e  -C7+K)X  ( Z + K ) A  Z+K  )]:  Under conditions of constant, annual recruitment the numbers of fish in successive age groups are serially related by a common mortality term.  The  integrated biomass (biolength) of a year class throughout its fishable l i f e span must therefore also be equal to the biomass (biolength) sum of its separate age groups taken over a period of a year.  The total yield from a year  class must then be equal to both the fishing mortality times the average biomass and the annual yield.  This conclusion was originally arrived at by  Thompson and Bell (1934) in their studies of equilibrium yield in relation to fishing effort.  42  Equation (1.50), its shortened forms (Ricker, 1968;  Gulland, 1969) and  a simplification prepared by Jones (1957) can all be used to determine yield under constant recruitment.  Additionally, i f the basic model is separated  into its constituent age groups the expected yields during the transition period from one equilibrium state to another can also be calculated. tioned earlier, even i f  As men-  R is unknown, the critical size and the maximum  sustained yield per recruit can s t i l l be estimated.  The model is thus best  viewed as having theoretical value, but its practicality is limited by, and the results are dependent upon, isometric growth and constant recruitment and mortality.  It cannot be used to accurately predict yield i f  R  should vary from year to year or be subject to upward or downward trends.  One  method of dealing with these problems is discussed under subsection 4.3. The flexibility of the Beverton and Holt (1957) model can be greatly extended i f the total annual yield in weight or length is calculated as the sum of contributions from each age group (i)  (Walters, 1969).  the identity of the simple model is-essentially .lost.  In the process  Instead, it becomes-a  more generalized age-group model for isometric, asymptotic growth.  The  .  integrated annual .yield formulae may be readily derived as i' (Y  WT }  W  =  1  E  i (  k  V i W  =  1  W  oo  3  1  V i  E  k  1  (  1  V i  i .  VO  0  -nK(i-t )  E  -(F.+M.+nK) O-e  1  )  1  F.+M.+nK  and i' L  1  k  i' L  1  k  -(F.+M.) 1  1  0  1  F +M i  i  -K(i-t )  -(F.+M.+K)  F.+M.+K  -  43  where k is the f i r s t and i' the last age group taken by the fishery, $. = the average, selectivity coefficient of the trawl for age group i ; F. and  0<$ .<!),  are age-group specific mortality rates, and  (N ).j = the starting population size at the beginning of the year. Q  The application of the yield in weight formula, coupled with a Ricker (1968) recruit function is given in Appendix A, part 2. 4.3  Virtual population and cohort analysis models. The virtual population method developed independently by Fry (1949)  had  led to the development of one of the most useful analytical tools available in fisheries work.  In its original form, Fry (1949) recognized that the sum  of the catches taken from a year class throughout its existence in the fishery provided a minimum estimate of the original recruit abundance (in the absence of natural mortality).  In addition, i f data on the entire catch history of a  year class was available, a maximum estimate of  F could be prepared for each  fishing year by interpreting the catch per unit effort as an index of availability rather than one of abundance.  Subsequently, the method has been  adapted and modified by Fry (1957) , Jones (1961, 1964, MS 1967, MS 1974) and 5  See Ricker (1975, p. 184) for references to earlier Russian papers. Presented at the joint ICES, ICNAF, and FAO fisheries conference in Lisbon, 1957. Following this meeting several authors suggested using the Beverton and Holt (1957) method of plotting Z against effective effort (f) to estimate M and q. The linear relationship is based on Z = qf + M, and the virtual population method is used to estimate Z for a range of f values. Subsequentl y , Bishop (1959) and Paloheimo (1958, 1961) drew attention to the bias in individual estimates of M and q. Paloheimo prepared an analytical solution which recognized the positive correlation between Z and f. Bishop showed :i that under conditions of either increasing fishing mortality or effort fluctuations without trend, regression estimates of M were too low and the slope (q) too high. The reverse condition existed i f effort was decreasing. Ricker (1975) indicates that the regression method is no longer used, but the above criticisms are nevertheless valuable.  44  Gull and (1965) to estimate Z and F, independently of effort, for each year that the year class is fished.  Schumacher (1970) has reviewed three varia-  tions of the virtual population technique, and the conditions and assumptions common to all three are as follows: 1.  The basic catch equation of Baranoff (1918) applies, i . e . ,  x n= C  Ax< o>n< - " N  1  e  x Z n  >  x n Z  where x = the year class n = a generalized subscript for the year of fishing C x n = the age group catch of year-class x in year n All previous assumptions concerning the distribution of fish (or fishermen), and the continuous, concurrent operation of both mortalities are therefore in effect. 2.  The mortality rate M is constant for all age groups and must be estimated  independently. 3.  It is assumed that the catch in any year n represents all of the catches  for that age group from year class x. calculated value of  If this assumption is not made, the  F and the back calculated value of the apparent recruitx n . r  ment, 4.  r  R' , will be too low.  A catch model rather than one of yield is assumed and i t is applicable to  a closed system. The virtual population, V, for year-class x is defined mathematically as V =I C x x n  45  where the sum is taken over all age groups throughout the fishable l i f e span of the year class. population,  V A  1  In any given year (n), however, the partial  , may be derived from the catch equation, i . e . ,  11  w • C F (N ) (EN) V _ x n _ x n x on = x on x n = _ j = —j— (1-e x n) xn  ,-, r \ (1.51)  v  where x^n'  =  t  virtual  n  e  c a t c n  i n  from x i f  y  e  a  r  P^  n  us  a  ^  1  future catches to be expected  F and M M remain constant throught the period, xn  It is a partial sum. xn  = the exploitation ratio,  F  x  (F/Z)  n  Similarly,  x 'n+1 V  x n+l  =  C  =  fi „ x^n+l\ (1-e )  x n+l F  x^Vn+1 = x^ Vn+1 E  (1.52)  X^n+l  Equations (1.51) and (1.52) represent the basic formulae used to derive all the methods that follow.  Since the discussion will refer to one year class  only, the subscript x is removed.  Method I Condition:  the year class has been fished to extinction and the virtual population, V, is known. -Z  •If equation (1.52) is divided by (1.51) and (N ) i is replaced by (N ) e Q  n+  Q  n  it follows that V  _i_i  n  E -, e = -fl n  Z n  (1.53)  n  ,  46  If it is now assumed that mortality is constant for consecutive years, then as a f i r s t approximation E ^ = E and (1.53) reduces to +  S  n  = V'  , = e n+1  n  (1.54)  Z f 1  V  n where  = the survival rate in year n.  A process of back calculation is initiated by assigning to V  .j the sum  of catches for the residual age groups that are older than the last age group of x for which catch data are available. known, (S, Z, F)  Since V'  may be calculated directly.  -j, V  and M are then  It is also possible to derive  (S, Z, F) _2 and preceding years by the appropriate substitution of the virtual n  population ratios in ( 1 . 5 4 ) .  The accuracy of the method is impaired by the  assumption that E = E .j. n  Method II Conditions:  the virtual population, V, is known.  An estimate of the  exploitation ratio in the last year of fishing must be prepared or chosen. Jones (1964) recognized that the starting population in year n+1 is related to the catch in the previous year in the following way: Given that -1 (N ) = (N ) e o n+1 on  and  C = E (N ) (1-e n n o'n  '  n  v  v  v  n ;  v  then r  = (N ) n o n+1 n v  y  C  •Z = (N ) e on _ E (N ) ( 1 - e n o n  -Z =  n  v  y  v  v  e  n  z n  )  '  E (1-e n  — n  )  c c  i  (1.55)  47  Since  is a function of F and M, and M is assumed to be constant, a table n  of ratios of r  n  for a given value of M can be prepared for all values of  that may be encountered in a fishery. The table can also be extended to in-Z -Z elude values for the terms e , E and E O - e ) to avoid unnecessary n  n  calculations.  n  From a knowledge of r 3  directly.  n  and Z , F can then be estimated n n  The difficulty at this point is in assigning a value to  second to last year of fishing.  for the  Two relationships provide the necessary  insi ght: V  ,, = E ,, (N ) , -i n+1 n+1 o'n+1 v  and C = V - V' , n n n+1 ;  and rearranging in the form of the ratio r^  r  = (N ) , 1 o'n+1 = -F  V  1  n  i E( ,  x l  n+1  x  n  ^ ^„ . i  c c  \  -  )  (1  56)  If i t is once again assumed that V .| = S^V ^ from equation (1.54), 1  +  expression (1.56) may be replaced by r  =j  n  E  n n+l^W S  ( 1  '  5 7 )  where E -j represents the last exploitation ratio to be applied to the year +  class prior to extinction.  Its value must be assumed or determined in order  to initiate the back-calculation process to evaluate E , E -|, etc. The n  relationship between E and E .j is established by means of equation (1.55) n  and the expression  C  n  =  W n  " W V n + l  =  < o>n N  ( E  n " n+l E  e  ^  48  to yield 7  ' n e  _  -Z  E (l-e n  n  nv  (N ) , ( o n+l C  v  N  y  )  v  )e on  "  Z  n  y  (N.) (E -E e o n n+1  n  v  x l  -Z ) n  '  Simplifying and rearranging the above relationship i t follows that  y -^ 1  +  n+1  E  e  _  Z  n  = n  d'  E  5 8  )  -Z where E (l-e  n  n  ) = the exploitation rate ("u" of Ricker, 1968) and is the  fraction, by number, of fish that will be caught during the year n, and -Z E -|e = the fraction that is caught later. n+  A point that Schumacher (1970) does not make clear is that i f equation (1.58) forms the basis of the back-calculation technique, the assumption must be reiterated that adjacent exploitation ratios are equal.  On this basis E  n  can be set equal to E -j for the preceding year and the working equation for the model is rewritten from (1.58) as  +  E  n1 ^ +  To begin the process r value of E -Z E (l-e n  n +  i. '  n  - n=l  (1.59)  E  is calculated from (1.57) based on the estimated  The tabular value of r  yields the numerical quantities  -Z  ) and e  , and substitution into equation (1.59) determines E^ -j.  This latter value is again substituted into (1.57) to yield r  ^, and the  process is repeated until the catch from the f i r s t age group entering the fishery is encountered.  49  Method Condition:  III  the year class has not been fished to extinction and therefore the partial virtual populations V  1  V ' , V' _-|> e t c . , are n  n  unknown. In this case the catch equation for year n+1, the last year of fishing to date, may be combined with the catch in the previous year to form the ratio  and equating this expression to (1.55)  ^  j  The i n i t i a l ratio r in equation (1.60).  =  n  r  " " ~ z  •  <'.60)  is estimated by means of a starting value of E ^(l-e n+  The quantity of E (l-e n  n  ) , and therefore  are obtained  directly from tables and (1.60) is used as the basis of back calculation  In order to determine the apparent recruitment,  R', the following A  nomenclature (after Pope, 1972) is adopted for year class x.  Let  t  the last age group of a year class for which catch data are available,  equal  50  and let  i  represent the f i r s t age group captured.  The catch equation in  the f i r s t year can then be written as  - ! t ( N  C,  0  O-e"* )  ) ,  1  i and expanding and rearranging, R = (N ). = C i + N..,, x o'l i -pi+l i 1  Z  v  The above expression can be used as a recurrence relationship by substituting for N.j -|, N-j , e t c . , as follows: +  +2  R' = C . i + C . . , i+1 + . . . + C. Z  V  Z  1+1  A h  i  The value of (N )  t-1 + (N ).  t - l r  p  i+l  t  Z  h  t-l  t  refers to age group  (1.61) t  can take on two forms depending on whether the year  class has been fished to extinction or not. C  0  t  If fishing is complete, the catch  and all subsequent residual age groups.  In this  case (N ) Q  t  = C Z /F t  t  (1.62)  t  Similarly, i f (N ) refers to the last age group in an incompletely fished Q  t  year class (N ) 0  t  -  ^ t _ F (l-e t  *)  and substitution in (1.61) yields  (1.64)  X  (1.65)  R" = C. F (l-e  l  t  )  To evaluate  R', calculations begin with the final age group (t) and an x extimate of either F^/Z^ (1.62) or Z^(l-e ) followed by a backward summation Alternate methods of the virtual population technique have also been proposed by Murphy (1965), Pope (1972) and Doubleday (MS 1975).  6  The initial  method involved an iterative procedure of solving for the. F values embodied n  in a sequence of catch-equations ratios between adjacent age groups. Again, the value of M is constant and the initial fishing mortality rate (F..) should be known for ease of computation.  The reverse procedure of estimating F^, as  in the methods described above, introduces the need for laborious and repeated iterations.  In either case the method offers no distinct advantage and has  not been accorded wide usage.  The "cohort analysis" method of Pope (1972),  however, contains several features that are of practical and analytical est.  It is an approximate form of Methods II and III  1.  eliminates the use of tables  2.  provides estimates of F x  n  and (N )  inter-  given above that:  in all fishing years given an  n  estimate of M and the final value F^, and 3.  allows an investigation of the errors introduced into F , x  n  R' and (N ) Q  n  by the arbitrary choice of F^, and sampling errors in catch-at-age data. The approximate formula is derived by rearranging the equation = (N ). e (NJ o i+l 0 1 ;  v  -(F.+M)  into (N ) e MN ) o i+l M  y  Q  - (N ) l-e" ), F i  i  o  i (  See Jones (MS 1974) for a virtual population method based on length.  52  and expressing the last value of (N )^ in terms of catch, Q  < n>l+l oi+l N  e M  =  ( >i o i " i i f d-e N  C  n  ^  F i  ) ( ITjr+M) i ) ] F  (1-e  (1.66)  + M  )  1  The terms inside the square brackets remains unchanged i f multiplied and divided as indicated:  e . (l-e F,/2 F. r L  7  ) . (  ) -, . e -(F.+M) (F.+M)/2 (1-e ) -e I  J  1  1  1  to y i e l d a product of three terms  , V (e  "V) -e  2  2  F. 1  ,F.+M, . ( i ) (F,+M)/2 -(F +M)/2 (e -e ) 1  1  ( i ) e F  .  e  +M  /2  F./2 1  Within the range F. < 1.2 and M <„0.3, the product of the f i r s t two M/2 terms is approximately equal to one. The last term reduces to e  , and  therefore (1.66) can be rearranged and replaced by the following expression of Pope (1972): ( V t  =  (Vi+i  e M  +  c  i  e  M  /  2  For the last year of fishing (1.67) becomes  The author is indebted to Dr. J . Schnute for this insight.  ( - ) 1  67  53  and substituting the appropriate value for  from either equation (1.62) or  (1.64) the process of back calculation can begin. tion of F and (N ) n  n  In a comparative examina-  estimates prepared by virtual population and cohort  analysis, Pope (1972) indicates that the differences do not exceed two percent. Analysis of the cohort method (Pope, 1972) indicates that i f there are no errors in the catch data an underestimate of the true value of  leads to a  positive and negative error, respectively, in the back calculated values of (N )  and F .  The converse is also true, but i t should be added that as the  n  sum of the fishing mortalities (F^ + F ^ + in (N ) Q  n  and F are reduced. n  etc.) increases, the errors  Consequently, the estimates of (N )  n  and F  n  become increasingly more accurate progressing from older to younger age groups. Sampling errors in catch^-at-age data similarly introduce errors in estimates of both age group numbers and their fishing mortality rates.  The magnitude of  these errors, expressed as a percentage of the variance ratio in catch-at-age data, also declines and approach asymptotic values at younger age groups. The errors are persistent, however, and cannot be eliminated unless the catch data i t s e l f is error free. F  n  and (N )  To appreciate the variance associated with estimates of  i t is therefore more critical to know the variance in the catch  data for each age group than to use the true value of F^ as a starting point. For this reason both Pope (1972) and Doubleday (MS 1975) indicate that the latter quantity is likely to be chosen arbitrarily. In the virtual population and cohort analysis methods the number of parameters ( F , (N ) ) to be estimated is greater than the number of observan  tions (C ) and a unique solution is therefore not possible.  To overcome this  difficulty, Doubleday (MS 1975) has applied a least squares method to a matrix  54  of catch-at-age data for a number of consecutive year classes.  An adequate  description of the method is beyond the scope of this report, but the following critical features should be noted: 1.  For age group  i  and fishing year  n,  the fishing mortality rate ^F  n  at any point in the matrix may be considered as a product of -Qf > where n  represents a selection coefficient common to each age, and f  n  is an effective  effort term that is applicable to all age groups fished in year n. and f  may then be expressed logarithmically as ^q = In .Q and f^ = ln f^ to  yield In .F = .q + f . i n n for .C  Both  Similarly, logarithms of the usual catch equation ^  and the ratio of successive catches -jC /-j + -| C -| yield equations that n  n+  are potentially linear and may be subjected to least squares analysis. nential terms in the equations of In  and ln( -Cn / i • C  Expo-  n+1  ) are expanded by  Taylor series approximations to complete the linearization.  By these proce-  i  /  +  dures, and given sufficient data, i t is possible to obtain more than one observation to support each parameter.  In.common with the virtual population  and cohort analysis methods, the least squares technique also requires prior estimates of certain parameters.  If ^q for the last age group and f^ for the  last year of fishing are fixed, the total number of parameters that can be estimated is 2(A+J)-1, where A and J. represent the total number of ages and year classes, respectively.  On this basis, as Doubleday (MS 1975) indicates,  nine ages and eleven years of data are required in order that each parameter be supported by about two and one-half observations. 2.  The principal benefits of the method are that the results lend themselves  to statistical analysis.  Variance estimates (reliability) can be attached to  the parameters, anomalies in the data can be examined, and " . . . the amount of  55  information contained in the catch data about population sizes" (Doubleday, MS 1975) is also indicated. 3.  Unfortunately the model may not offer a unique solution.  Different  starting values of ..q and f^ in the final year may produce different results in terms of ^  and -(N ) n  .  n  Because the catch ratios in successive years are  negatively correlated (the denominator - -|C ^ in one year becomes the numer+  ator in the next) the parameter estimates of .jF opposite directions.  and .j(N ) tend to move in  It is then possible for two or more sets of ^(F, N ) Q  values to generate the same catch observation.  n  The quantity ^q + f^ always  appears as a sum and an increment in ..q coupled with a decrement in f^ by the same amount does not alter their sum.  Finally, for a meaningful analysis of  variance, data for well sampled catches over an extensive period of time are required, but there is l i t t l e assurance that selection criteria will remain stable during the interval. The assumptions and conditions underlying the virtual population, cohort analysis and a least squares method of dealing with catch-at-age data have been described.  Depending on the amount of information available, either technique  can be used to prepare estimates of the fishing mortality and starting population sizes by age group.  It is also clear that these parameter estimates are  conditional upon the accuracy of F^ used as a starting value in the final year of fishing, the assumption of constant M, and the sampling error in the catches. If these limitations are recognized, the parameter values may be used to develop two secondary estimates concerning recruitment and yield prediction.  In the  f i r s t case, i f complete catch data exist for a series of year classes, the apparent recruitment ( R' x  ,  _-|R' , etc.) may also be calculated.  If sufficient  information exists on the adult stock sizes that gave rise to these successive  56  recruitments, then the potential exists to establish the beginnings of a stock-recruit relationship.  At the opposite extreme, i f the fishing mortal-  ity for each age group at the end of the present year of fishing (n+1) can be estimated, the partial age structure of the starting population in year n+2 can be predicted.  Together with an estimate of the apparent recruitment  in the same year, the total age structure may then be combined with: 1)  average net-selection coefficients ($ ),  and 3)  2)  !  average weight-at-age data,  anticipated fishing mortalities ( - j F ^ K to predict the annual yield.  The procedure is currently used by ICNAF (Hodder, 1975) to prepare estimates of the total allowable catches for single and mixed stocks for single species that are capable of being aged. 4.4  An asymptotic, allometric growth, yield model Solutions to the Beverton and Holt (1957) yield equation when growth is  allometric have been proposed by Jones (1957), Paulik and Gales (1964) and Kutty (1968).  In more general terms, the yield model can be established by  substituting equations (1.20) and (1.35) into (Y ) w  i  = F fl  N.(t) w.(t) dt  i.e., (  V i  =  F  o  f  1 [ ( N  _7t -K(i+t-t.) W l - e ) ] dt b  o i )  e  0  b  (1.68)  The u t i l i t y of this equation can be extended to include the allometric -K(i+t-t ) conditions of b = 2 or 4, but i f b f integer the expression (1-e cannot be expanded prior to integration.  b  )  Jones (1957) recognized that i f a  o  For an analytical solution to the enmetric fishing curve for isometric and allometric growth, see Kutty, 1968.  57  suitable substitution is made for this expression, subsequent integration of (1.68) leads to a simplified difference between two incomplete beta functions. On an age group basis his yield formula can be rederived as Z(i-t ) (Y..). = F (N ). W e  where X = e -K(i+l-t ) X = e u  0  ]  3 = the incomplete beta function, and b can have an integer or non-integer value. To evaluate this equation the beta functions associated with the coordinates (X, 7_, b+1) and (X , 1, b+1) can be obtained from tables prepared K K by Pearson (1948) or the more practical edition by Wilimovsky and Wicklund (1963).  The method is simple and direct, but non-tabulated values of b+1,  Z/K and X require interpolation, and i f extensive calculations are necessary the method may be time consuming. Alternative solutions to equation (1.68) are possible and they all depend on the interpretation of the number and weight functions in the general equation  ( V i If w-(t)  =  f  o  N  i  (  t  )w  i  (  t  )  d  t  (1.69)  is treated separately, two solutions that can be integrated directly  are: I.  Approximate the weight function by means of annual arcs of  exponential growth.  The risk involved here is that for each age group of  interest to the fishery growth will be represented by a convex arc, whereas  58  the actual curve may be concave.  The applicable, integrated formula can  be readily derived as (Ricker 1968) (B )i = (N w w  0  o)i  ( e ^ - l )  (1.70)  G.-Z  II.  Approximate the weight function as annual chords of linear growth.  The appropriate weight formula is then w -(t) = ( w ) . + g . t n  Q  , 0<t<l  where g. = the linear growth rate for age group  i.  Substitution into the general biomass equation (1.69) and integrating gives: (B ) w  1  = ( N ) . [(w ) Q  Q  (l-e~ )- ? ! (1-Z e ' - e - ) ] Z 2 Z  i  Z  Z  (1.71)  z  The third and fourth methods rely on approximate integration techniques and deal directly with the product (N.(t) w^(t) as a biomass term.  Attention  is now focussed on the curve generated by plotting the biomass for each age group versus time.  The area under the curve for a given age group taken over  one year may then be approximated as follows: III.  Trapezoidal rule of approximate integration (Smail, 1949). The  method is based on representing a definite integral by an area under a curve and approximating this area by a set of inscribed trapezoids, or replacing the arc of the curve by a set of chords. To generalize the method, let: y = the biomass yg and  = the initial and final biomass, respectively, for a given age  group in the time interval  h.  59  x = N.(t)  w.(t)  a = x evaluated at y  and c = x evaluated at y-j;  Q }  therefore y = f(x)  A = the area under the curve. From these r e l a t i o n s h i p s the area to be approximated i s equal to A = f\  f(x)  dx =  y dx  a  ,  a  and i f the i n t e r v a l  h  i s set equal to one, the above equation i s  :  approximated by A=h(y  0  +  y  i  )  .  S u b s t i t u t i n g the appropriate biomass values for y^ and y^ , and d i v i d i n g both sides by f^ dt (=1), i t follows that vi- + \ 7 "K(l-t ) . -If. ( B „ ) . = (N ). W [(1 - e ) - e 2 0  IV.  b  ( 1  „ 6  -K(i+l-t ) . 0 |Dn > ]  n ( 1  '  7 9 7 2 )  \  Simpson's rule of approximate integration (Smail, 1949).. In t h i s  case the d e f i n i t e i n t e g r a l i s again interpreted as the area under the biomass curve, but the curve i s approximated by a series of parabolic a r c s .  If  the  time i n t e r v a l of one year i s divided into two equal parts (h = 1/2), three biomass values are needed to describe two parabolic a r c s : y g , y-j at the midpoint and y^ at the end of the period.  the i n i t i a l biomass It can then be shown  that the area under the curve i s approximated by  A = h (y  0  +  4 y i  +y ) 2  =  Y' 0 (y  +  4 y  l  + y  2  }  Again, s u b s t i t u t i n g biomass values for y , y-j and y^ and d i v i d i n g both sides Q  by f\  d t ( = l ) , the general biomass formula i s  60  -K(i-t ) L r / \ -K(i + 0.5 - t ) , ) + 4e-°' (l - e ) n  (B ) w  = (N ) W  i  (J  -Z Z  + e (l  - e  1  "  (0^5)  m  K ( i  +  1  [(1-e  0  v 5 ( Z )  b  0  " n b ) ] t  b  )  (1.73)  0  How well equations (1.70), (1.71), (1.72) and (1.73) approximate the actual value of the average biomass can be determined in the following manner. In order to maximize the yield in weight from a multi age-group fishery one criterion of interest to management is the critical size (Ricker, 1968). On a curve of biomass versus time this size is reached when the derivative dB/dt = 0 at the critical time (age) t  ^.  In the absence of fishing mortality the  biomass at time t (0 _< t _< °°) is evaluated by .  M  (B ) (t)  = N(t) w(t) = N e~  w  Mt  Q  Setting t = t  Wjl-e  -K(t-t ) , ) 0  (1.74)  b  differentiating this equation with respect to t  £ r t  and  setting the result equal to zero,  ^W  = 0 = NW  [bKe- (e" M  K ( t c r t  "  t o )  )  (l-e"  K ( t c r t  " ° ) " t  )  b  1  ^rt  d  -  - (l-e" M  M e  and solving for t t t rr  = t  + n  K ( t c r t  "  t o )  )  b  ]  ,  ^. :  1  l n  K  (bK + 1 )  (1.75)  M  For the west Newfoundland cod stock, Wiles and May ( 1 9 6 9 ) estimate the growth parameters as:  K = 0.14, t  = -0.2 and L  OT  = 9 3 cm.  The value of W  OT  may be calculated as 7 2 6 3 gm for combined sexes i f b is conveniently set equal to three.  Choosing an arbitrary value of M = 0 . 2 2 , substituting these para-  meters into ( 1 . 7 5 ) and solving:  t  ^ = 7.4 years.  If ( )y N  Q  = 10,000 fish,  61  the actual average biomass in the critical age interval of seven to eight years is calculated directly from 1 -Zt "K(7+t-t ) (B,,) n = (N ) W f e (1-e T dt = 3997.48 Kg W 7-8 o 7 o 3  L  7  7  0  Z  00  Similarly, by substituting these same parameters into equations (1.70), (1.71), (1.72) and (1.73) the comparative biomass estimates are Method  Type  I II III IV  B~ _  % difference  Kg  compared to actual  7  8  Exponential  3981.49  -0.400  Linear  3997.82  -0.008  Trapezoidal  3982.15  -0.383  Simpson's rule  3997.48  negligible  For all practical purposes the average biomass estimate prepared by Simpson's rule is identical to the actual value.  For this reason it is selec-  ted as the method of approximating the average biomass in the case of both allometric and isometric growth.  It is used exclusively in the development of  the analytical model in Section IV.  Despite the apparent complexity of equa-  tion (1.73) it is readily dealt with by computer, i f required.  The appropriate  annual yield formula for weight, by age group, is therefore:  ( V i *  w -K(i-t ) . F(N ). ^ [ ( 1 - e ) 6 0  0  K  b  +  4e"-  , (1-e  5Z  , -K(i+l-t ) , +e" (l-e )] Z  0  b  -K(i+0.5-t ) . ) 0  b  (1.76)  62  Similarly, i f growth in length occurs according to the von Bertalanffy (1938) equation, the annual yield equation for length, by age group, is  (Y. ) L  = F(NJ. L [(l-e' ) - e " Z  i  o  ro  n  K ( l  "  t p )  (l-e-  z  Z  +  ( Z + K )  )]  (1.77)  K  An interesting feature of equation (1.76) is that i t contains an estimate of the instantaneous age-specific growth rate, G-.;--If the termswithin the square brackets-are rewritten as /  (1-e  -K(i-t ) . „ ' -K(i+0.5-t ) , -K(i+l-t ) . ) [1 4 e " (1-e ) + e" (1-e )] , -K(i-t) -K(i-t ) 1-e 1-e 0  b  5 Z  0  7 Z  b  0  b  +  0  0  the terminal expression, excluding the term e  Z  G  , is equal to e  i  . The  derivation is arrived at by means of the exponential relationship for integer values of age and the allometric growth formula for weight, i . e . ,  w  G. -K(i+l-t ) i+l = e = W (1-e ) i ~K(i-t ) . W (1-e ) 1  0  s D  w  0  1  b  and G. -K(i+l-t) K(i-t) K(i-t ) |/ e = (1-e , , ° )(e ),b = e , -e^yb 1  0  r  0  f  e  - 1  0  e  0  -1  and taking natural logarithms, K(i-t) J. = - b u ln in e re -e G. -e'\  K  0  r  0  -1  i  .  (1.78)  63  Since  = b H. from equation (1.34), the instantaneous age-specific growth  rate in length is therefore  M  1  K  (  i  _  t  °  )  "  K  HL = In re - e i l K(i-t ) J e -1 1  0  4.5  r  Seasonal pulse fishing  Larkin (1977) has described maximum sustained yield (MSY) as an illusionary concept.  For a single stock the fishing mortality needed to main-  tain MSY may erode the genetic variability and reproductive stability that support the original MSY estimate.  Mixed species or stocks of single species  are similarly reduced to the MSY level of their most productive elements.  In  each case the long term value of the sustained yield is less than the hypothetical MSY. In the original analysis of a fishery, however, MSY serves as a theoretical limit beyond which biological productivity of a target species is reduced.  The interpretation  of MSY in this limited sense is common to all the  yield models presented so far.  In each case the basis of estimating MSY is  also restricted to the assumption of continuous fishing, growth and natural mortality.  Mathematically i t is convenient, but i t may not be realistic i f  the species undergoes seasonally intensive growth or mortality.  In fact, the  value of MSY estimated from continuous rates is conditional and may not represent the maximum possible sustained yield.  A fishing strategy adapted to the  actual seasonal growth and mortality pattern of a species may prove superior. The potential improvement in yield is explored here under conditions of seasonal pulse fishing.  64  The seasonally most active period of growth and mortality of several species is illustrated in Figure 2.  For temperate water species the most  important point is that growth does hot take place continuously.  Active  periods may span one or two consecutive quarters of a year, they may be bimodal, or divided seasonally on the basis of maturity or size.  These results  are both sufficient and encouraging in terms of a preliminary analysis of seasonal pulse fishing (SPF).  A greater degree of precision cannot be ex-  pected because the original investigations did not focus on the seasonal nature of growth.  Information on when fish die is practically non-existent,  but there is a suggestion of seasonality for the brown and rainbow troutas well as the walleye and alewife.  To overcome the limitations of these data,  a simulation model is used to calculate equilibrium yield for ent fishing situations.  200  differ-  The details of the model, results and interpretation  are reported in Appendix C.  A brief description of the model and results are  also given here. A hypothetical, multi-age group population is selected as the target species.  Growth and mortality are independent processes that can occur in a  seasonally condensed or continuous mode.  In the seasonally intensive mode  annual growth (mortality) takes place in any one of four quarters.  Moderate  growth (mortality) occurs during any two consecutive quarters, and the extensive phase involves three consecutive periods.  A total of 240 seasonal and  continuous growth-mortality patterns are possible, but only twenty cases are examined by the model (Figure 3a).  Ten fishing strategies (FS) are similarly  arranged into intensive, moderate, extensive and continuous categories (Figure 3b).  Seasonal  SPECIES  Jan  Feb  G r o w t h (g) and N a t u r a l  Mar  Apr  May  Jun  Mortality Jul  (M) P e r i o d s  Aug - - Sep  Oct  Nov  Dec  REFERENCE Swift (1955) Brown (1957)  BROWN TROUT Salmo  trutta  WHITEFISH Coregonus Coregonus Coregonus  _g_  sp. clupeaformis clupeaformis  HERRING Clupea Clupea Clupea Clupea  } non-larval  harengus harengus harengus harengus  _g_ nursery grounds  Pleuronectes  platessa  immature  ARAL SEA BREAM Albramis  brama  bergi,  _g_ small cod  morhua,  Kohler (1964)  large cod  L.  Kelso and Ward (1972)  WALLEYE Stizostedion  vitreum  vitreum  LAKE TROUT Cristovomer  Kennedy (1954)  namaycush  RAINBOW TROUT Salmo  Tody (1964)  gairdneri  1 <2  YELLOWFIN TUNA Thunnus  (speculation)  albacares  ANCHOVETA Cetengraulis  9  mysticetus  Hennemuth (1961) Howard and Landa (1958) Brown (1972)  ALEWIFE Alosa  N i k o l s k y (1963) Nikolsky (1963)  mature  Grieb  ATLANTIC COD Gadus  Nikolsky (1963) Das (1972) Bowers (1952) Jensen (1950)  g}larval  PLAICE  Nikolsky (1963) Healey (1975) Van Oosten and H i l e (1947)  pseudoharengus  Figure 2.  Seasonally intensive periods of growth (g) and natural mortality (m) f o r a variety of fishes. Legend: ..  i  reduced > increased  r  a  t  g  ^ g 3  r o w t n  o  r  mortality.  cn cn  a) CASE  CASE  1  * *  11  m  m i m  12 *  m  ic  * **  * *  g  rk  ic  13  m  * * * *  15  * * * •k  rk  *  rk  *  *  ic  ic  ic  rk  rk  rk  rk  m  rk  rk  rk  16  •k  rk  rk  rk  17  * * * * * * *  m  * * * *.  18  * * * *  19  9i m  m  9i m  20  9, i 7  * *  m b) FS  Intensi ve 1 2 3  4  *  *  Moderate FS  *  *  5 6 7  * *  rk  m  10  * ** * * *  9i  rk  ic  * * *  m*  m  rk  * *  14  * **  ic  Extensive  * *  * * * rk  Continuous FS 10  rk  rk  * * * *  Figure 3. a) Twenty cases of seasonally distributed growth (g.) and natural mortality (m) patterns. Each year in the l i f e of an age group (i) is divided into quarters, b) Ten, seasonally applied fi-shing strategies (FS).  67  I n i t i a l l y , the von Bertalanffy (1938) growth constant is set at K = 0.2, M at 0.1 for all age groups, and for a selected growth-mortality pattern, the critical age ( i  c r t  ) is calculated to the nearest quarter year, using a sea-  sonal Ricker (1968) model. the integer value of i  The age of f i r s t capture (AFC) is set equal to  ^. .  Fishing strategies one to ten are then applied  separately, beginning at AFC and F = 0.1 and the seasonal yields summed (Y ) T  from AFC to 100 years, or until the annual growth rate (G) falls below 0.0002. For each fishing strategy (Y )p is calculated for all odd numbered values of T  F from 0.1 to 1.5 .  At F = 1.5,  (Y )-| g represents the conditional value of T  MSY- that is a function of K, M, FS and the growth-mortality pattern. results calculated to this point represent one data set.  All  Additional sets are  derived for K = 0.4 and 0.6, M = 0.2, 0.3 and 0.4, and the remaining case studies. The final results appear in Table C l i t , Appendix C , and for each data set the  (YJ)-J  g values are expressed as a percentage gain (±) in yield rela-  tive to that obtained from continuous fishing, FS 10.  The conclusions may be  summarized as follows: 1)  In general, as the values of K and M increase, the potential equilibrium yield resulting from seasonal pulse fishing increases relative to continuous fishing.  The degree of benefit,  however, depends on the values of K, M, and more c r i t i c a l l y , on the growth-mortality pattern.  If the entire growth period is free of  seasonally intensive natural mortality, and annual growth is completed before mortality begins, relative yield due to SPF will be greater.  68  2)  Seasonal pulse fishing is least effective when the species' growth and natural mortality are continuous, as in Case 9. In general, as the duration of active growth and mortality is seasonally reduced, the potential for yield improvement increases.  3)  As F increases, the ranked order of yields from the 10 fishing strategies does not remain constant.  Relative to continuous  fishing, the rank of a particular strategy may either increase, decrease, or remain unchanged.  In other words, the yield is a  complex function of the growth-mortality pattern, the fishing strategy and the value of F.  Provided the catchability coef-  ficient remains constant, the interaction of these variables can be used to exploit two opportunities.  The yield from a  seasonally expanded fishing strategy may be maintained or marginally improved by reducing F and increasing the seasonal fishing intensity.  Alternatively, only a portion of the avail-  able fishing fleet can be used to obtain a regulated yield by an appropriate choice of fishing strategy. 4)  Without exception one intensive fishing strategy will provide the greatest yield benefit relative to continuous fishing. A moderate strategy offers a second-best choice and with one exception, extensive fishing is the third alternative.  The magni-  tude of the benefit increases with increasing values of F. Considering all the growth-mortality.cases, the range in relative yield improvement for the intensive strategy is 3.4 to 30.5 percent at F = 1.5.  In an individual case the increase in yield  69  is dependent upon the relationships outlined in item one above. 5)  While one intensive strategy leads to the greatest improvement in y i e l d , another choice in the same mode introduces a nearly equivalent loss.  The risk of an inappropriate choice can only  be eliminated by knowing when fishing should take place. A simple answer cannot be given, but the seasonal relationship between growth and mortality is instructive.  With one exception  (Case 19), i f the end of seasonal growth coincides with or extends beyond the last quarter of mortality, then the optimal choice of intensive strategy is FS 1. age is equal to an integer.  In each case the critical  In all other  growth-mortality  patterns, including Case 19, the critical age contains a fractional value, i . e . , 0.25, 0.50 or 0.75, and the situation becomes more complex.  The optimal strategy then depends on the relative  position, duration and exponential value of g and m and requires a mathematical solution. 5.  Regenerative Yield Models  The yield models described so far include expressions for the principal components of growth and mortality, and as such they can only explain past but not future events.  Unless a recruit function is incorporated in the  model, as in Appendix A, part 2, or separately estimated, predicted yields and the direction of the fishery cannot be estimated. models incorporate recruitment directly.  Regenerative yield  They assume that a stock is capable  of generating a surplus production that can be cropped as y i e l d , and that the  70  rate at which biomass increases is a function of both the present biomass and its departure from an environmentally limiting maximum.  Graham (1935)  and subsequently Schaefer (1954, 1957) formalized the biomass approach based on the Verhulst-Pearl logistic in the following terms:  ^ | = g(B ) - qfBt = r B ^ max " M  - qfB  B  t  t  t  ,  ^max where g(B^) is a density-dependent function of biomass, B^; r is the instantaneous net rate of population increase;  B , is the limiting biomass; max m  v  and the product of the catchability coefficient (q) and effort rate (f) equals the instantaneous fishing mortality rate, F ' , in terms of biomass. F' is not to be confused with F, the mortality rate applied to fish numbers. At equilibrium, the rate of stock increase dB/dt = 0 and the rate of natural increase rB.(B - B.)/B , is balanced by the surplus production, or yield t max t max rate of qfB^ .  A plot of equilibrium yield versus stock size describes a  symmetric parabola with MSY situated at B. = B  m  /2 .  Ricker (1975) indi-  cates that at equilibrium surplus production is also a parabolic function of the fishing rate (F ) and effort, and these relationships provide alternative 1  solutions for locating MSY. Schaefer (1954) perceived that under the usual, non-equilibrium conditions of a fishery, MSY could be empirically derived from a time series of yield and effort data provided that F , and hence q, was separately esti1  mated.  Later, Schaefer (1957) calculated all parameters directly from the  same data stream by an iterative procedure.  The basic equation, requiring  transformation of the data into mean values of yield per unit effort (V) is reproduced in the present notation:  71  T  W W " V-'^L  +  °-  £  7  9  )  where AV . in year j is estimated by linear interpolation to be J AV. = ( V . - V. ,)/2 , and e is a random variable. The initial estimates J J+l J-l 2 of constants r/q and q*B are obtained by dividing the data base in two ^ ^ max v  Ll  3  parts and solving the simultaneous equations.  Substitution of these e s t i -  mates in a summary equation for all the data provides an initial 2  value of 1/q  and hence, second and converging estimates of r/q  Pella and  and q * B  max  .  Tomlinson (1969) state that the iterative procedure leads to a non-unique solution and different values of the constants are obtained depending on where the data is partitioned.  The reliability of the estimates is improved  i f each data sub-set is chosen to include a series of years in which the change in biomass is thought to be large. Recently, Schnute (1977) criticized equation (1.79) on the basis that the linear interpolation method implies that V. -j can be predicted without +  the use or knowledge of the mean effort rate in year j+l.  He proposed instead  the approximation V^ = /V.» Vj ^ to develop the equation for annual events: +  ,„ L ± L - r V.  g '  'V*1+l» 2  _!L_ « max B  ( 7  i * V'' 2  where Vj and V ^ are instantaneous values at the beginning of each year and E is the mean effort rate per year. achieves several important goals.  Subsequently, a model is developed that Parameter estimates are prepared directly  from the data series involving Y, E and V"(=7/E) , the yield in the immediate  72  future is predicted and confidence limits are assigned to both the parameters and predicted yield.  The mathematical development of the Schnute  model is rigorous and comprehensive and supercedes that of Schaefer in being dynamic and stochastic. The use of the logistic growth function dictates a symmetric, parabolic yield curve.  Ricker (1958), Gulland (1961) and Schaefer and Beverton (1963)  indicate that for many fish populations the curve may be positively skewed with MSY occurring at less than B  mgx  /2 .  Gulland (1961) proposed that in an  age-structured population the present yield from the most productive age groups may be influenced by the mean value of their past, fishing effort experience ( T ) .  Ricker (1975) indicates that the relationship between V and  7 is then described by a negative exponential curve.  Fox (1970) derived the  necessary equations for this situation based on a model in which the growth function g(B~.) = rB , t max  _ 1  (InB , - InB". ) , but MSY always occurs at 0.37 B . max t max. m  Pel la and Tomlinson (1969) generalized both the Schaefer and Fox approaches by introducing a four parameter model of the form dB. - ±  -  HB - KB - qfB, m  t  where, in the present notation, H = - r , K = -rB . and m = a constant that max determines the negative or positive skewness of the yield curve. growth is included by the special case of m = 2 .  Logistic  The modified program  GENPROD - 2 (Abramson, 1971) is used to search over a parameter space beginning with a fixed value of m and estimated values of q, r and ( F , V)^y • 1  The goal of the search routine may be to minimize the sums of squares involving actual and predicted yields over a series of years, but the parameters  73  may also be constrained by the experience or knowledge.of the operator. More recently, Mohn (1980), Roff and Fairbairn (1980) and Uhler (1980) have explored the bias and error propagation introduced by catch and effort data when applying different solution methods to logistic-type production models. Hi 1 born (1979) has also compared the merits of different methods of parameter estimation of the Schaefer model by simulation techniques when the fishery is subject to various regulation schemes. The assumptions of the surplus production models are more restrictive than those in which separate functions are used for growth and age structure. The features of a closed population, and non-selective (size) and randomly distributed fishing are maintained.  It is also assumed that each unit of  fishing effort instantaneously removes a fixed percentage of the total biomass.  The effects of fishing on growth rate, recruitment and age structure,  and their interrelated time lags, are essentially ignored.  The existing bio-  mass is the primary determinant of population increase and the response is instantaneous via the parameter r.  Relative to dynamic-pool type models  the information content is reduced by collapsing rates of birth, natural mortality and growth into a single constant, r .  By further assuming that q  is constant throughout the data series, the variance of the parameters is increased and the ability of the model to explain past or future events is impaired.  Silliman (1971) concludes that in the early stages of a fishery  where information on growth, natural mortality and age structure are missing, the simple surplus-production models are useful in preliminary assessment. Ultimately the goal should be to gather data on relevant biological processes and formulate an extended model that is descriptive of the particular species. In what follows attention is therefore limited to a more critical analysis of age-structured models.  74  ANNUAL STATE MODELS  1.  General Remarks  A number of models have now been discussed and each represents a mathematically expressed hypothesis relating yield and biomass of fish.  The  models also suggest the kinds of data that need to be collected to solve for the equation parameters.  A few of these methods have been described and a  more comprehensive account is available in Ricker (1975).  Whatever hypo-  thesis is used, its validity is tested by evaluating how well predicted values of past and future yields agree with reality.  Ultimately, the objective is  to establish the relationship between yield and mean stock biomass or the changes in these qualities over time.  If growth is not included in the model  the analysis can be carried out on the basis of catch and mean abundance. Usually, direct estimates of biomass and abundance are difficult and costly to obtain and the respective indices of V and U are used instead by monitoring y i e l d , catch and effort. Because the parameters are calculated from fishery data and the model is used to determine the f i t to the original data, a high degree of correlation is expected i f the model accurately describes the system.  Yet it is precisely here that difficulties arise.  In a graph of U  versus f for instance, the model assumptions may indicate a linear relationship, but due to the scatter of points curvilinearity cannot be denied.  The  difficulty is aggravated i f the data base does not include a wide range of f values. Historic data may also indicate that several substantially different values of yield or catch have been obtained by the same amount of fishing  75  effort.  These examples illustrate a general deficiency of fisheries models  in that the unexplained variance cannot be assigned with certainty to changes in either q, H or abundance, or to inaccuracies in y i e l d , catch or effort data. The bulk of recent fisheries literature is devoted to error detection in the above quantities, to more sophisticated methods of analysis or to model alterations that reflect the realities of a fishery more closely.  The focus  of these efforts usually begins with the basic catch equation and the implied Constance of q and independence of effort units (DeLury, 1947).  Paloheimo  and Dickie (1964), amongst others, point out that neither of these conditions are satisfied when fish are distributed contagiously and fishermen are able to concentrate their efforts against fish aggregations.  The catchability  coefficient is then neither constant nor as small as that assumed by a random fishing model, and may be further augmented by cooperation amongst fishermen. Fishing success, or U, at the point of capture is then an index of apparent abundance (Marr, 1951) and may not be related to a change in actual abundance over time.  In an extreme case, i f the rate at which fish invade a region of  high aggregation is matched by the catch rate, U is maintained at an artif i c i a l l y high level and will give l i t t l e indication of the decline in abundance until the fishery is near collapse. Despite these criticisms, the illusion of constant q is maintained by assuming that the fishery operates throughout the area! range of the stock and that the same value of q is applicable over a number of years.  Since  catchability is so critical in determining even apparent abundance it is unfortunate that its magnitude is not arrived at independently.  In practice,  q is estimated indirectly during the process of solving equations and, to a  76  large extent, its reliability is dependent on the accurate measurement of effort.  Yet the effort term, or more correctly, the efficiency with which  effort is applied, is subject to three levels of variability that are successively more difficult to quantify.  At the primary level standardization  is necessary to account for the differences in fishing power of individual vessels.  Increased capacity and horsepower reflect the ability to stay at  sea longer, to shorten the time between fishing trips and to reduce searching time.  The net result is an increase in effective effort that has a direct  impact on abundance.  At the fishing site, the interaction between gear type  and fish density, gear selectivity and saturation, and competition between gears introduce the second source of variation in effort efficiency.  The  frequency with which these subjects appear in fisheries literature testifies to their importance, but the effects are usually difficult to quantify.  The  third source of error includes physical and operational components that should properly be grouped under fishing power, but their effects are more subtle than outright changes in capacity or horsepower.  Since each fisherman  is motivated by economic competition i t is to his advantage to seek out and apply technological changes that lead to reduced gear-handling time, both during and between fishing operations, or extend the precision and range of fish detection.  If individual improvements lead.to an increase in fishing  success, the result is inevitably detected by other fishermen and the process of upgrading fleet efficiency is set in motion.  Superimposed on this system  is the intuition, intelligence and experience of individuals that, together with technological change, directs fishing tactics to take advantage of fish distribution.  It is extremely difficult to monitor let alone measure the  effects of these physical and operational contributions to fishing success,  77  and a time lag always exists between the present and historical performance of effort efficiency.  The assumption that the discrepancy can be revealed by  analyzing time-series data implies that either q is constant or fishing conditions have stabilized.  In practice, neither of these assumptions can be  verified and the investigator is led back to the original problem of not knowing whether the change in fishing success is due to a change in q, M, effective effort or abundance. The circular impasse between an immeasurable q and an inadequately measured f can be interrupted by recognizing that the real quantity of interest to management is the product of these variables: of the instantaneous fishing mortality.  the time average value  If this statement needs additional  support, the practical significance of.being able to estimate F directly can be summarized as follows.  During the course of a year, F represents the  mortality inflicted on a stock that is the average of all separate fishing incidents involving an interaction between availability (Marr, 1951), catchability, different gears and their efficiencies.  It reflects the inputs of  intelligence, intuition, experience and technology without the disbenefit of a time lag, and summarizes the effects of environmental vagaries in q and f in a single term.  If F could be calculated directly, there would be no direct  need to estimate f and all the ancillary qualities that are known to modify it but cannot be quantified.  An annually calculated F would also provide the  most recent and reasonable starting value to initiate back-calculations of recruitment via virtual population techniques. However desirable i t may be to. estimate F directly, the fact is that in all the models considered so far, F appears either in the product of two  78  unknowns (FN or F(No).), or in the sum of the mortality terms, Z.  In these  circumstances the numerical value of F cannot be calculated unless one of two methods is used.  Either M is estimated independently and F is obtained  by subtraction from Z, or by the reverse procedure, F is derived from a time series of Z and effort data and M is obtained by subtraction.  In the anal-  ysis to be outlined presently, a third method of estimating M is introduced that relies only on yield (weight and length), catch data for each age group and an independent estimate of t . Q  determine F from these data.  The immediate difficulty then is how to  I n i t i a l l y , a property of a stock or an age  group must be identified that responds directly to changes in total mortality. If this is possible, the characteristic must also be measurable from statistics collected from the fishery, and a model must be developed that utilizes the statistic to estimate Z and solve for the equation parameters.  Each of  these problems is considered below in the context of a regulatory purpose of fisheries management. Figure 4(a) represents the familiar isopleth diagram (Beverton and Holt, 1957) for a multi age-group species at equilibrium under conditions of constant q, M and recruitment.  Each isopleth, or topographical line, indicates  a constant value of total yield per recruit, (Y ) /R. w I W  T  The direction of in-  creasing levels of (Y )j/R is described by the arc of the eumetric fishing w  curve.  The diagram can be easily constructed from a yield model provided M  and a growth function are known.  Growth overfishing and underfishing are  possible and the conditions are identified, respectively, by the regions below and. above the age of f i r s t capture that is coincident with MSY.  If  it  is the purpose of management to regulate such a simplified fishery and guide it towards MSY, the problem is in recognizing where the present fishery  79  F  Figure 4.  —  Schematic yield isopleth diagrams for two different multi agegroup stocks: (a) constant recruitment, and (b) density dependent recruitment. Legend: AFC = age of f i r s t capture; F = fishing mortality rate; GU, GO = growth under- and over-fishing; RU, RO = recruitment under- and over-fishing; E = eumetric fishing curve; 1 to 8 = relative magnitude of y i e l d .  80  exists on the diagram.  Provided that the fish can be aged, a horizontal  co-ordinate can be fixed by selecting the age of first capture.  Since nei-  ther q nor F is known, effort can be temporarily substituted as an index of F, and given sufficient time and a manipulated range of f values, the relationship between the index (Y ) /R and F will eventually emerge. w  T  The process  of analysis may be shortened by sequentially altering AFC and f and monitoring the transitional values of V or U.  In either case, however, the methods are  too time consuming and i t is unrealistic to assume that management can vary f at wi11. In Figure 4(b) the situation is made more complicated and precarious by introducing a density dependent, Ricker (1958) recruit function.  The iso-  pleths below the eumetric fishing curve are collapsed and the possibility of recruitment overfishing is superimposed on that of growth overfishing (Cushing 1972).  In this case, it is not sufficient to select an i n i t i a l value of AFC  and f to begin the analysis of the unknown system.  There is a risk of col-  lapsing the fishery i f AFC is too low and f too high.  To avoid the problem  M should be estimated i n i t i a l l y to determine the critical age of f i r s t capture. If growth is independent of fish density, the critical age is not affected by disequilibrium or transitional which to begin fishing.  states and provides a minimal, safe basis on  It is possible eventually to locate MSY by contin-  ually incrementing effort provided the monitored but time-lagged response in recruitment is positive. The circumstances shown in Figure 4(a,b) are knowingly unrealistic.  The  response of the underlying system to changes in the environment and fishing is unknown and the models are over-simplified, mathematical descriptions.  81  Recruitment may be density dependent, but due to unpredictable environmental conditions the number of recruits entering the fishery is variable even under stable fishing conditions.  As a result, the concept of a fixed value for MSY  is substantially weakened and the best that can be expected is to maximize average yield over a protracted basis of time.  Usually, management intro-  duces regulation many years after fishing has begun and an i n i t i a l analysis is based on data generated by an unregulated fishery.  Additional complications  arise from restrictive model assumptions related to fishing strategy and the distribution of fish.  Under these conditions i t would be beneficial to mini-  mize the time between analysis and a regulative response, and to monitor both recruitment and a biological property that reflects the impact of the fishery. Exactly which property has the desirable attributes is unknown, but the search can be narrowed by considering the statistics that can be generated by a fishery. The response of 18 conventional and potential statistics to changes in either AFC or F are shown schematically in Figure 5.  The model used to gener-  ate the appropriate data is the same as that of Figure 4(b), but the isopleths have been omitted.  For simplicity, i f the response of a group of statistics  is directionally similar they are combined in a single diagram.  The eumetric  fishing curve appears as a solid line i f the rate of change of the statistic is altered by crossing over the curve, and as a dotted line i f the response is unidirectional, or independent of a eumetric curve. arrow points towards increasing values of the statistic.  The direction of an By holding F or AFC  constant the response to changes in either AFC or F are indicated by the vertical or horizontal arrows, respectively.  The basic statistics of yield (Y),  catch (C), and average values of age (A), weight (W) and length (U) are  Figure 5.  Response of 18 conventional and potential fishery statistics in relation to eumetric fishing curve. Legend: AFC = age of first capture; F =• fishing mortality rate; -»- direction_of arrow indicates increasing values; Y = y i e l d ; C = catch; R = recruitment; F = mean weight; L = mean length; A = mean age. Subscripts: T = total stocks; W = weight; L = length; i = age group. Dotted curve in region of arrows indicates unidirectional response. See text for comments on diagrams (a) to (f).  00  ro  83  distinguished by subscripts T (total), i (age group), w (weight) and a (length). The symbol R refers to recruitment in numbers of f i s h . The criterion for judging the suitability of a statistic as an indicator of fishing pressure is that for known values of AFC and the s t a t i s t i c , the corresponding value of F (or Z) should be revealed unambiguously.  Super-  imposed on the equilibrium conditions shown in the diagrams are the effects of variable recruitment, particularly as they may interfere with the determination of F.  In panel (a) for example, data on recruitment, the total catch and  yield in weight or length are insufficient to disclose the location of F. Since the values of the statistics increase and decrease as they cross over the eumetric curve, there are potentially two solutions to the value of F. Moreover, i f recruitment is variable, then (Y ) , (Y ),, R and C are dependent T  T  T  upon the unknown quantities of either fish numbers or population age structure. In panel (b) the unidirectional response of (Y ) /F, (Yr-)./F and C /F in T  I W  T  r  L  I  relation to F is i n i t i a l l y promising as an indicator of fishing pressure, but F itself is unknown and the statistics are further dependent upon either age structure or fish numbers.  In both cases (a) and (b), information in addition  to AFC and the statistics is needed to reveal F.  Similar reasoning applies to  panels (c) and (d). The' only statistic to emerge from this analysis that responds directly to F (or Z) is the average weight of a fish in an age group, The result is not entirely surprising.  (panel e).  It is a direct result of killing fish  before they have had an opportunity to complete their annual expectation of growth.  That the mean weight of the total stock diminishes with increasing  mortality was recognized earlier by Ricker (1968), Gulland (1969) and  84  Ssentongo and Larkin (1973), but the same phenomenon also applies to an individual age group.  As a result, the restriction of constant recruit-  ment that appears in the mean stock size-mortality models of Baranov (1918), Silliman (1945) and Beverton and Holt (1956) is removed because  is inde-  pendent of fish numbers. The concept follows directly from the catch and yield  equations when  expressed annually as (C/F). = N". and ( Y / F ) . = B., and hence ( Y / C ) . = 1  (B/N").. = ¥^W../N. = Wij.  1  w  I  I  W  1  If abundance increases for a given value of F..  the yield in weight and catch also increases, but the yield divided by the catch, or W^, remains unchanged.  This statement implies that:  (1) the  density of fish per unit of fished area must be directly proportional to abundance, and (2) the density relationship is valid whether the fish are distributed at random or contagiously.  The degree by which fish density  is autoregulated by abundance is presently unknown.  To the extent that  i t does occur, however, a change in W. will be a useful indicator of a change in F regardless of fish distribution.  The relationship between  and F..  is maintained even i f the fishermen are. able to direct their efforts against fish concentrations.  In this last case the assumption must be  made that (C/F). is only indicative of the apparent abundance of an age group.  85  The last statistic appearing in Figure 5(f)  is L\, the average length of  a fish in age group i that is influenced by F (or Z).  For all practical pur-  poses the change in U. is negligible for a change in F of 0.1 or less.  On  f i r s t consideration the insensitivity of L. may appear to be a disadvantage, but the usefulness of this property will soon be demonstrated. 2.  Assumptions of the Annual State Models and Bases of Analysis  In this section two models are developed that use data from two immediate past years of fishing to prepare estimates of the growth parameters and the mortality coefficients F and M.  The key statistic that permits this rapid  assessment is W. , and together with a known age of f i r s t capture the state, or position, of the fishery in reference to an eumetric curve can be located. The models that utilize the state property of W^. are termed annual-state models. The assumptions of the models are as follows: 1.  The fishery consists of a single, closed stock in which the fully fishable, multiple age groups are identified by the subscripts i through i + n.  2.  Growth in weight and length are continuous and occur according to the von Bertalanffy-type equations w. = W ( l . e -  K ( i  - o )  K ( i  "V)  o o  and  I.  = L (l-e"  t  )  b  where b may be a non integer. 3.  The mortality rates F and M act continuously throughout the year and  86  are common to a l l age groups.  The models are sensitive to seasonal  changes in growth.and mortality. 4.  The natural mortality rate i s assumed to be constant for any two consecutive years of f i s h i n g .  5.  Fishing i s hot size selective.  6.  The recruitment function i s unknown and the number of f i s h entering the fishery in any year i s subject to environmental variability.  7.  The density of any age group i to i+n i s proportional to apparent abundance, (N ).j , where the subscript "o" indicates the beginQ  ning of the year. 8.  The age of the fish can be determined without d i f f i c u l t y .  9.  Each f i s h that i s included in a sampling program to estimate the annual catch must also be measured f o r length (Model weight (Models 1 and 2).  10.  1) and  The methods used must be consistent.  To retain a s u f f i c i e n t number of decimal places in the calculated mean weights  (W".) and lengths (U^) the catch in each age group  should be no less than T000. 11.  The parameter t  must be independently estimated.  Model 1 makes use of a l l the available data and i t i s used here to describe the general analytical procedure.  To begin the analysis two con-  secutive years of age-group data on catch, y i e l d in weight and length and an annual estimate of t  are needed.  Thereafter, the data of the t h i r d year are  o analyzed in conjunction with the second, and so forth. growth equation parameters  Z and t  By this procedure the  are monitored and updated annually.  addition the analysis always includes the immediate past year of f i s h i n g .  In  The examination of the data is divided into two parts.  In the f i r s t ,  data for each fishing year, j , are treated separately by means of the anal ytical equations developed earlier in Section II,  (4.4).  For convenience  they are repeated here:  ,  b.  , ~ , N W V i , j j o i,J ^  W  (  =  C F  •  M  ( N  }  - K . ( i - t ,) C(l-e °' J  3  4e  J  +  -0.5Z. (lJ  b. -K.(i+0.5-t .) e °' J  J  J  -Z -K (i+l.O-t .) + e (l-e °' ) J  , ( Y  L i }  L  1  i ,J  =  F  i< o i i ~ i J 0 1 , J °°,J N  L  )  [  1  J  "j  -K.(i-t  Z  T — " £j  J  e  J  °'  .)  j ]  (3.01)  -(Z+K). ^ ^ Z+K) J  J  )  ]  ( 3  '  0 2 )  -  Division of expressions (3.01) and (3.02) by (3.03) yields the statistics Z. W . - K . ( i - t .) j 3 - i - ^ [ ( i . e y °.J) b  ff 1 >J  - L.  1-e  +4e  -0.5Z. J  (l-e  -K.(i+0.5-t .) °»J J  O  J  •Z. -K,(i+1.0-t ,) j +e ( l - e °' ) ] b  J  J  J  (3.04)  and Li,j  -Mi-t„  ^ j H - e ^  Z.  -(Z+K)  ^ (-^H ( l ^ J j ] 1-e" j  IZ^Kj  J  (3.05)  88  Each age group generates three equations (3.01 - 3.03) with five parameters (W^, K, b,  t )j Q  and two unknowns (F(N ) ., Z)j . Q  1  Since the  components of the product (F(N-)•),• cannot be estimated separately they are eliminated by restricting the analysis to equations (3.04) and (3.05). any year j , Z is constant and t  For  • is known, a prior estimate having been  made by the method of Beverton (1954), the Walford procedure mentioned earlier or from the known mean size at age of young fish (Fabens, 1965). of these methods L  By either  - and K. are also estimated simultaneously, but to fully  demonstrate the present methodology only the estimate of t  Q  j will be used.  The residual number of unknowns is therefore reduced to five (W , K, b, L and Z) . and the minimum number of equation pairs (W., L^),-. or age groups, needed to arrive at a solution is three. In a multi age-group fishery data are usually available from more than three age groups and the solution must be obtained from an overdetermined system.  At least three programs are available at the University of British  Columbia Computing Centre Library to deal with such a problem. they are:  Specifically,  singular value decomposition (SVD), a solution to a non-linear  system (NONL), and the International Mathematical and Statistical Library program ZXSSQ.  Initial estimates of all unknowns are needed by each program,  but SVD and NONL also require the partial derivatives of the independent variables (W. ,1".) with respect to each unknown.  Program ZXSSQ, however, uses  a modified Levenberg-Marquardt algorithm to find the minimum sums of squares for M functions of N variables and eliminates the need for explicit derivatives (Levenberg, 1944;  Marquardt, 1963).  In the present case ZXSSQ is  favored due to cost and programming simplicity to find the least squares solution such that  89  E (W. - W.)  2  + -2 (L. - L . )  = minimum.  2  (3.06)  The estimates (W., U.). are prepared directly from equations (3.04) and (3.05). Conway (et a l . , 1970) indicates that in the general case of a nonlinear system the Marquardt approach is more efficient relative to either the gradient methods (Spang, 1962) or Taylor series (Glass, 1967).  Whereas the  " . . . Gradient Methods are able to converge on the true parameter values even though initial  values are far removed, . . . this convergence tends to be slow.  On the other hand the Taylor Series Method will converge rapidly providing the vicinity of the correct parameter values has been reached", (Conway, et a l . , 1970).  Program ZXSSQ takes advantage of both tactics by i n i t i a l l y se-  lecting a large value of the internal parameter PARM2 to ensure the steepest descent (Gradient method).  As the correct parameter values are approached  by iteration PARM-2is reduced to provide rapid convergence by means of Taylor series.  One difficulty with this method is that following one appli-  cation of program ZXSSQ the final parameter estimates may be determined at a local rather than a global minimum.  To overcome the problem ZXSSQ can be  applied iteratively such that the parameter estimates from the f i r s t run become the initial  input values of the second run and so forth.  The para-  meter PARM 2 is thereby re-set to its maximum value at the beginning of each run and a local minimum avoided.  The global solution is reached when  equation (3.06) is satisfied. The non-linear algorithm is used to estimate (W^, K, b, L^, Z) . for each of two fishing years.  In the second stage of analysis these results  90  are analyzed by the equation AZ = A F and F. / F . , ?  solve for F._, and F . the c r i t i c a l  = ?  (as on pages 98-99) to  and subsequently for M and the age assodiated with  size.  Before proceeding to the section on i n i t i a l  parameter estimates a  comment is necessary on the unusual requirement of an independent estimate of t .  In contrast to more f a m i l i a r fishery models in which M is derived  independently or by subtraction from Z, an analysis based on the equations for W-  or U. leads to the unpredictable fact that Z and t  1  1  separable sum. the f i r s t  involves equation (3.05).  .  occur as an i n -  Two related examples are chosen to i l l u s t r a t e Since  this point and  is a mean length, the  of a year at which the length of age group i achieves time t  o  fraction  can be equated to  Equation (3.05) can then be rewritten as  "K(i-t ) , -K(Z+.Kh = L. = L [1-e ° • —=—7 • ^rnr ' i+t i °° , -1 Z+K z 1-e {L  L  •K(i+t - t ) z~V)  - L (1-e  (3.07)  where e  "  K t  z  _ _ J L _ " l-e"  and . t z  .  (l-e-(  Z  •Z, _ 1 „„r(l-e" ) - ^ inl - j - L •  Z  +  Z + K  >)  K  Z+K i (Z+K)) 3  (  The l a s t equation demonstrates that for a known value of K, t function of the unknown Z and a constant for a l l  age groups.  3  n  '  Q 0 8  s )  is both a Program ZXSSQ  can be applied to the Lj values derived from a single year's f i s h i n g to  91  establish the difference (t  z  - t ) in equation (3.07), but further  decomposition is impossible unless either t  or Z is known.  The second example draws attention to a theoretical method that can be used to estimate the correct values of t  and Z.  Initially the exact values  of W.j and L~. for a,number of age groups can be calculated directly from chosen values of W , K, b, L , t  and Z.  Program ZXSSQ can then be used to  calculate the residual sums of squares i f W^, U. and the remaining parameters are held constant except for t  and Z.  The results appear in Figure 6.  It  is clear that the least squares solution lies within the elliptical trough formed by the residuals contour 0.1, but the surface of the trough is so flat that the correct values of t  and Z cannot be estimated.  In a more  realistic case the errors associated with the yield in weight and length or catch would be substantially greater than in the example and i t is doubtful that a contour value of even 1.0 could be achieved.  The extent of the  trough also demonstrates the sensitivity and dependence of Z on an accurately determined value of t 3.  and suggests that t  Model 1:  Q  be measured independently.  Methods of Estimating Growth  Parameters and Mortality Coefficients The data presented in Table 1 are used to test the analytical capability of the equations for ¥ . and L~. and to develop the techniques of parameter estimation. (1957).  The model used to generate the data is that of Beverton and Holt  A random multiplier is used to simulate variable recruitment and  the resulting catches from the various age groups.  The age-group yields in  weight and length are not shown but may be calculated from the tabulated values of W. and U. i f required.  92  Table I.  Age-group c a t c h , mean weight and length for two_consecutive years of f i s h i n g , i = age group; C = c a t c h ; W. and L . = mean weight and l e n g t h , r e s p e c t i v e l y ; j = f i s h i n g year.  j=2-  j =l  i  •  C  i  ~  l  W .  T  (mm)  (gm)  C.  W.  L .  i  l  (mm)  (gm)  4  66,040  445.728  804.173  75,936  444.615  798.156  5  71,105  508.994 1,194.647  80,778  508.026  1,187.834  6.  40,023  563.995 1,623.216  86,973  563.154  1,615.950  7  42,202  611.810 2,070.549  48,955  611.079  2,063.120  8  39,023  653.379 2,520.806  51,620  652.744  2,5T;3.443  9  15,309  689.517 2,961.818  47,732  688.965  2,954.692  10  22,342  720.935 3,384.786  18,726  720.454  3,378.015  11  14,724  748.247 3,783.784  27,330  747.830  3,777.444  12  15,525  771.992 4,155.211  18,010  771.629  4,149.345  13  11 ,215  792.634 4,497.267  18,990  792.319  4,491.891  14  9,672  810.580 4,809,504  13,718  810.305  4,804.617  15  6,716  826.181 5,092.443  11,831  825.943  5.088.030  14 E  i=4  C = 347,180 . i  15 E  i=5  C . = 424,663 l  94  3.1  Initial estimates of L , K and Z  A method has been described earlier to estimate the parameters L^and^K by means of a Walford plot based on equation (1.38), i . e . ,  - Ul-e- ) +e-\. . K  A similar relationship can be developed in the general case for successive values of U. ^ +  anc  • z  .  As before, let  _ i _  K t  e  j I".  ' l-e"  (1-e-( ' '  Z  and equation (3.05) for  Z  +  >)  K  can be rewritten as  -K(i-t ) = Ld-e °  h  Z+K  -Kt e- ) Z  -K(i+l-t ) -Kt = L-d-e ° ' e )  a n d  L 1  Z  i+  -K(i-t ) -Kt Subtracting L. from L . , substituting L - L . for L e •e and 0  -  l+l  1  J  o  o  -  |  z  CO  simplifying, L  i + 1  =  (l-e~ ) + e K  Loo  _ K  L,  (3.09)  Consequently, average length values for each age group can be used in a Walford plot to yield initial estimates of L and K that are unaffected by -Kt _ _ the mortality term e . The regression of L.^, . on L. . for each fishi+l.J i ,J ing year in Table 1 leads to the estimates: 00  z  95  L  00  =  , l  K  930.000 mm.  L = °°,2 0  = 0.140  ]  Kg  930.001 mm  =  0.140  Ricker (1975) points out a disadvantage of the above procedure in that each datum appears twice in the calculations except for the first and last. He proposes instead a rearrangement of the basic length equation that leads to the following expression in terms of mean length at age: - L.) = £nL  ln(l * co  ^'  0  - K(t  - t ) - K, . Z  0  (3.10)  1  O  The linearity of this relationship is sensitive to chosen values of L and ro  may be used as the criterion for estimating L  o t  and hence K.. Applying the  method to the data in Table 1 does not alter the initial estimates of L °°»J  and K., but the intercept, i . e . ,  ilnL^ - K(t  z  - t ), Q  for each fishing year  yields the value (t  - t ), = 0.6617 Z  and  (t  0 1  - t ) 0 2 0  Z  The probable sign and magnitude of t difference (t  2  - t )^ . Q  = (l/K)£n(K(l-e* ))  Since  0.6453  .  can be estimated from the  If Z-j is set equal to zero in equation (3.08), a  Taylor series expansion of the term (1-e t  -  )/Z-| reduces to one,  and for K = 0.140 the maximum value of t  K  = 0.4942.  and K in both fishing years are nearly identical it is unlikely  that the growth rate has changed and t for both years.  Substituting t  yields the estimated t  z  is therefore assumed to be the same  = 0.4942 in the difference (t  = -0.168 for Z, = 0. O  - t )^  It follows that t_  I  = 0.4778 )L The magnitude Z  and solving equation (3.08) by trial and error, Z = 0.197. 9  9  96  of Z-j and 1^ are incorrect since  cannot equal zero, but two conclusions  follow from these results that are independent of the actual values of Z-j and 1^ '• ^2 > Z^  and AZ = AF = 0.197 since M is assumed to be equal for  two consecutive years.  If the actual value of t  is found to be -0.200 for o  both fishing years, i t follows that t  -j. = .4617, t  z  2  ~ 0.4453, and by  equation (3.08), Z = 0.391 and Z =0.591. ]  3.2  2  Initial estimates of W and b. A mortality term analagous to t  can be incorporated into an expression  for the mean, age-group weight, e . g . :  W, = V  (1-e  "  K ( i + t  w i- o ' ) t  w  n  )  b  (3.11)  0  In contrast to t , however, i t can be shown that t, • is not constant for a z w,i given value of Z; approaches t  i t is slightly larger than t  as i increases.  for all values of i but  Despite this limitation it is possible to  develop an approximate relationship between W. and  that leads to an esti-  mate of W and b. oo  b If weight is related to length by the equation w = a l weight can be expressed in terms of by the approximation h -K(i+t -t ) b (r.) « L (1-e ) h b  b  z  0  ro  Dividing (3.11) by (3.12) and rearranging,  " mflkO + b & n C L / b  (1-e 1  e  (1-e  -K(i+t w  _  K  (  i  +  '  n  _  t  z  .-t ) °1 }  t  }  0  ]  , the mean  (3.12)  97  If i t is further assumed that t  . = t  for all values of i , the equation z  w,i  ^  above reduces to £n W. S £n(W /L ) + b£n I.  (3.13)  b  "j  OO  CO  •]  *  '  Regression of the tabulated values of Jin W. .. on £n X . . , and substitution of the appropriate value of  ,  1  b^  3.3  .  = 7252.463 gm =  leads to the estimates  w  2.992  » 2  =  b  =  2  7 2 5 2  •  4 1 4  9  m  2.992  Final estimates of parameters, coefficients and the critical age The requirement of program ZXSSQ for preliminary estimates of the  unknowns is now complete.  For the first year of fishing, substitution of  the estimates and a fixed value of t  = -0.200 in ZXSSQ requires one program  run of seven iterations to produce the final results U = 7262.860 gm °°,1  K, = 0.140 1  = 929.995 mm  Z-, = 0.401  3  1  Lco  b  1  =  1  3.000  Convergence is similarly achieved in seven iterations for the second year to yiel d W „ = 7262.545 gm °°,2  K_ = 0.140 2  L  Z„ = 0.602 2  3  0  00  b  2  ,2  = 929.992 mm =  3.000  98  /N  /\  and it follows that Z  /\  - 1  2  /\  = F - F = 0.201 2  ]  The catch equations for two consecutive years of fishing on the same year class may be described by  i,l  C  ' l o»i <'-e"V, F (N  and  -Z C i +  i,2  ^ V i + i  =  ( : ]  ?  )/ 2  e  •  z  - l z  But (N ).,-, 0  = (N ).e  1+1  0  and the ratio between successive catches can  1  therefore be written as i,l  C  F  2  r i+1,2 C  —  •  F  l  Z  2  •  F  2  Z  (1-e  7 l  Z]  _ i  • (  1  _  )  1  Z  '—  -Z  • p  Z  e  2  )  Since F. is assumed to apply equally to all fully fishable age groups within one year, and M is constant for both fishing years, it follows that 14  I  •MnC,' »,» 1  i=4  F  15 i  C = 5  . l  ]  "  Z  FT *  i+l,2  , 2  (1-e  Z7  2  1  -Z, ) ]  -H 2  Z  '  l  e  (- ) 3  14  )  The sums of catches for the appropriate age groups appear at the bottom of Table 1 ,  i.e., 14 J 1 C  4  15 ,1  =  3  4  7  '  1  8  0  a  n  d  ^  i+1 ,2 = 424,663  C 5  99  All other variables in equation (3.14) are known except for F-j/F^ and the solution follows directly that F - | / =  0.499.  Since the difference F^"^  is also known, the final mortality estimates are F  1  = 0.200  ,  F = 0.401  and M -  2  0.201  By applying a catch equation to the data in Table 1, the apparent abundance (N )• for each cohort in each fishing year can be estimated.  If  a sufficient amount of data exists for a number of years the magnitude of recruitment for each year class entering the fishery can also be calculated. The age of first capture that is coincident with the critical size can be estimated by equation (1.75), i  .  crit +  = t  o  + 1  K  i n i ^ r  M  i.e., +  1)  '  .  Since t , b, K and M are equal for both fishing years, substitution leads to the single value of i  c r l  -  = t  7.86 years.  The Model 1 analysis of the data for two consecutive years of fishing is now complete.  A comparison of the initial and final parameter estimates  and the actual values used to generate the data appears in Table 2.  Agree-  ment between the final and actual figures is expectantly good because the fishery's data are without error and meet all the model assumptions.  Under  these conditions the foregoing procedure tests the analytic capability of the approximate equation for w\ in combination with data on mean lengths. The estimates can now be used to construct a yield (weight) isopleth diagram having F and AFC as the x and y axes, respectively.  During the two years of  fishing AFC was held at four years and the successive locations of the  Table II.  Comparison of Model 1 and Model 2 estimates of growth equation parameters and mortality coefficients with actual values. W - maximum weight (gm); - maximum length (mm); b - length weight exponent; K - growth coefficient; F, M, Z - fishing, natural mortality and total mortality coefficients, respectively; j - fishing year one or two; ~ critical age (years); ( ) - assumed Z value. ot  Parameters and coefficients Woo  b  K  Z  F  M Vrt  j  MODEL 1 Initial Final estimates estimates  MODEL 2 Initial Final estimates estimates  Actual values  1 7252.463  7262.860  7260.768  7262.946  7263.000  2  7252.414  7262.545  7260.685  7262.693  1  930,000  929.995  2  930.001  929.992  1  2.992  3.000  3.020  3.000  2.  2.992  3.000  3.030  3.000  1  0.140  0.140  0.140  0.140  2  0.140  0.140  0.140  0.140  1  0.391  0.401  (1.5)  0.405  0.400  2.  0.591  0.602  (1.5)  0.609  0.600  1  -  0.200  -  0.201  0.200  2  -  0.401  -  0.405  0.400  -  0.201  -  0.204  0.200  -  7.86  -  7.79  7.88  930.000  3.000  0.140  -0.200  o o  101  fishery are identified by the coordinates (0.200,4) and (0.401,4).  Since  "I'c'rit is greater than four i t may be argued that the current program of growth overfishing and reduced yield may be improved by simultaneously increasing F and AFC in fishing year three.  This plan, however, is premature.  Despite the reasonable estimates of F^ , F^ and M i t is not known i f the increase in fishing mortality {F^) is due to an increase in the catchability coefficient or effective effort.  A comparison of the monitored effort in  both fishing years can be used to decide the issue and simultaneously e s t i mate the catchability coefficients. 4.  Model 2:  Methods of Estimating Growth Parameters and Mortality Coefficients  In the development of statistics W. and  the statement was made that  L.j is relatively insensitive to changes in Z.  To illustrate this point a  comparison of the changes in W. and I. for a change in F of 0.1 at both a low and high value of Z appears in Table 3.  The parameters used to generate  the data are the actual values listed for the Model 1 analysis.  Several  interesting features are revealed by the table and the first is that for a fish that has the capacity to grow to 930 mm ( L j sponding to AF = 0.1 is less than a millimeter.  the change in U. correThe response of L. for the  older age groups is also less than for the younger fish due to the declining growth rate.  In order that L. be a responsive indicator of changes in  fishing mortality will therefore require a large sample size together with minimal error in counting, ageing and measuring the catch.  The problem can  be circumvented by noting that the mean weight is more sensitive to changes in F and particularly in the region of the critical age (7.88 years).  102  Table HI. Expected change in statistics W. and_ due to a change_ in fishing mortality of 0.1. i - age group; W. - mean weight; L. - mean length; F - fishing mortality coefficient; - instantaneous growth rate.  W. (gm/fish)  i  0. 45  0. 35  0.27  0.22  0.18  0.15  4  5  6  7  8  9  0. 12 10  F=0.3  801. 156  1191. 232  1619.574  2066.825  2517.116  2958.246  3381. 392  F=0 .4  798. 156  1187. 834  1615.950  2063.120  2513.443  2954.692  3378. 015  3. 000  3. 398  3.624  3.705  3.673  3.554  3. 377  F=l .5  766. 973  1152. 477  1578.201  2024.495  2475.140  2917.601  3342. 761  F=l .6  764. 360  1149. 509  1575.029  2021.247  2471.917  2914.479  3339. 792  3.172  3.248  3.223  3.122  2. 969  AW..  AW. i  2. 613  2. 968.  L. (mm/fish)  F=0 .3  445.170  508.509  563.573  611. 444  653.061  689.241  720. 694  F=0 .4  444.615  508.026  563.154  611. 079  652.744  688.965  720. 454  0.555  0.483  0.419  0. 365  0.317  0.276  0. 240  F=l .5  438.811  502.981  558.767  607. 266  649.428  686.083  717. 948  F=l .6  438.321  502.555  558.397  606. 944  649.149  685.840  717. 737  0.490  0.426  0.370  0. 322  0.279  0.243  0. 211  ALT. i  A l  i  103  The increased response of W. to changes in Z or F forms the basis of the second annual state model.  Once again two year's data of age-group  yield in weight, catch and an independent estimate of t the analysis.  are needed to begin  The absence of length information causes the final parameter  values and Z to be derived from the approximate equation for W.. sult the estimates may not be as accurate as in Model 1. quence is that K and the initial estimates of  As a re-  A second conse-  and b cannot be derived  from the usual weight-length relationship and an alternative method must be used. The program BSRCH described earlier in Section 3.2 makes use of the exact relationship between fish weight at time t+1 and t.  Recast in terms  of mean weight, the equation (W  i + 1  )  V b  ^  Wj (l-e- ) / b  k  e- (W.) k  +  (3.15)  1 / b  is an approximation, but its linearity is sensitive to the value of b and this property may be used to establish the least squares solution I ((W  i + 1  )  1 / b  - (ftj- )  1/b  +1  )  2  •  If program BSRCH is applied to the W. data  for j=l , (Table 1), the following results are obtained: n  I  •1  W  "  ,1  K, "1  b, "1  Sums of squares  4 to 15  7261.803  0.140  3.030  185.9 x I O  5 to 15  7260.768  0.140  3.020  16.7 x 10"  6 to 15  7260.531  0.140  3.030  13.1 x l O  -4  4  - 4  104  The greatest reduction in the sums of squares is achieved by age groups 5 to 15 and the underlined values of W , K and b are accepted as the i n i m  tial parameter estimates.  A similar treatment of the second year's data  for age groups 5 to 15 yields the results W  = 7260.685  0  L  A reliable initial estimate of  = 0.140  b_ = 3.030 .  or Z^ cannot be obtained without  data on L. but a reasonable starting value (say, Z = 1.5 for j = 1,2) may be assumed.  The parameter t  is set equal to -0.200 for both fishing years.  From this point the analysis for the final parameter values Z^, Z^ and F is identical to Model 1 except that program ZXSSQ utilizes only the data for W.j.  The results are compared to the actual parameters and mortality coef-  ficients in Table 2. b, K and i  c r l  -  t  The discrepancy between the Model 2 estimates of W^,  and those of Model 1 or the actual values is negligible.  The  principal effect of using the approximate equation for W. is to introduce an error in Z, F and M in both fishing years.  In each case the error is only  two percent or less and indicates that the fishery analysis may be carried out on the basis of W. data alone. l 5.  The Dependence of the Mortality Coefficients on t and the Effect of Variable Natural Mortality ° It is not essential that the growth parameters be estimated by a  Model 1 or 2 analysis.  They can be obtained by alternative sampling methods  and the values substituted into the equations for W. and L. . 1  1  Since t is o  also estimated independently the analyses are reduced to solving for the mortality coefficients, the only remaining unknowns.  Departures from the  105  model assumptions and errors in the independent variables (w\ , L. and t ) can affect Z, F and M, and the r e l i a b i l i t y of the coefficients must be considered.  The most important sources of error include measurement, count-  ing, age determination and sampling as well as gear selection and agedependent fishing and natural mortality.  A complete sensitivity analysis of  the models for each topic is beyond the scope of this thesis.  Only two items  are selected for detailed treatment and they are the dependence of the mortality coefficients on t  and the effect of variable natural  The importance of estimating t  mortality.  correctly cannot be overemphasized.  If  the variables W. and L. contain random errors the effect on Z is linked directly to t  in a sum, and because of the disproportionate relationship, a  small error in t  is compensated by a large change in Z.  To illustrate  this  effect the first stage of a Model 2 analysis of the data (Table 1) for the first year of fishing is re-examined over the range t  = -0.200 t 10 percent.  The resulting percentage change in Z^ (Table 4a) is calculated on the basis of the original solution, Z^ = 0.405.  Atsthis  both b and K is negligible, that of  is < .001 percent and the error in t  is transmitted directly to Z-j .  low value of Z, the change in  In a complete Model 2 analysis, however, in  which all mortality coefficients are estimated over two years of fishing, the effect of an error in t o  is not limited to Z and Z„ . 1 2  All coefficients are  affected, and i f Z^ and Z^ are relatively large some of the parameters may also be influenced. percent error in t  Table 4b demonstrates these results for a minus six and in the situation where Z^ and Z^ are equal to 0.4  and 0.6 (Case 1) and 1.6 and 1.8 (Case 2), respectively. follow from the analyses and they are:  Several conclusions  Table IV. The effect of an error in on a) Z for a single year of fishing, and b) the mortality coefficients for two years of fishing. In both situations the analysis is based on Model 2. In (b) the error in t is minus six percent and the correct values of 1, and Z are either low (Case 1) or high °(Case 2). ?  1  a)  t  0  error Z error  b)  -.220  -.216  -.212  -10  -8  -6  -4  -2  .645  .597  .549  .501  .453  59  47  36  24  12  Case 1  -.196  -.192,  -.188  T.184  -.180  2  4  6  8  10  .357  .309  .261  .213  .165  -35  -47  -59  -11  -23  Parameter and coefficient estimates and errors, Fishing year, j  Results  -.208 ' -.204  W  K  "1  1  7262.786  0.140  3.000  2  7262.239  0.140  3.000  7263.000  0.140  3.000  <-0.02  0  0  1  7241.493  0.140  2.996  1.755  2  7241.100  0.140  2.996  —  7263.000  0.140  3.000  <-0.31  0  Actual values % error  "2  '1  2  0.155  0.549 0.753 0.400  F  0.600  37  26  0.200 22  M 0.394  0.359  0.394  0.400  0.200  -10  97  Case 2 Results  Actual values 7o error  -0.13  0.583  —  1.172  1.946  —  0.774  1.172  1.600  1.800  1.400  1.600  0.200  9  8  -58  -52  —  486  107  1)  Regardless of the error in t , i f M is constant for two consecutive years the difference between  and 1^ (or  and F^) is estimated  correctly even though their invididual values are incorrect. 2)  The difference between the actual and estimated value of a total mortality coefficient is inversely related to the sign and magnitude of error in t  .  The percentage error in Z is also inversely related  to the magnitude of Z. 3)  Relative to the percentage errors in 1^,1^, F-j and F that of M is 2  larger and increases with increasing values of Z^ and Z^ . 4)  For large values of Z^ and Z^ program ZXSSQ attempts to minimize the residual sum of squares by adjusting the parameter estimates.  The  order of increasing sensitivity of the parameters is K, b and  .  No satisfactory procedure has been found to cope with an error in t without increasing the data requirement of the models.  The ability to  correctly estimate the change in F between years (conclusion 1) is only superficially useful.  Unless F-j and F  2  , and particularly M, are correct  the resulting isopleth diagram and the position of the fishery relative to the eumetric curve will be inaccurate. available in addition to't  If an independent estimate of M is  , the information can be combined with the  value of AZ or AF to solve for the remaining mortality coefficients.  For  example, in Case 2, Table 4b, F = F-j + .191, the ratio (data not shown) 2  14 15 V C. . / VC „ = 4.662 and i f M = 0.200 the combined catch equation i=4 ' 5 (3i14) can be expressed on the basis of one unknown, i . e . ,  108  ^ 4.662  =  F  J  {  g  •  l  ^ + .391  ( p  +  }  ( i  •  2 Q 0 )  _ - ( F + .200) -(F. + .391K * g  1  (1-e  The solution for and Z  2  = 1.789  (  )  + .200)  '  1  yields the estimates F -=.1.398, }  Fl  e  F = 1.589, 1^ = 1.589 2  which compare favorably with their actual values (see Table 4b),  An entirely different problem is encountered i f M is a function of age rather than a constant for all age groups.  Since the annual state models are  only capable of estimating a single value for the mortality coefficients, it is reasonable to expect the estimates to approximate the mean values Z. . and ft. i f F. . is constant.  In the special case of variable M., however,  application of the models lead to incorrect estimates of the mortality coefficient means and most of the growth parameters,.  To demonstrate this limita-  tion, consider the situation in which M. assumes the following pattern:  Age group  4  5  M.  .2  .2  6  7  8  9  .23  .26  .30  .35  10 .40  1112 .46  .52  13  14  15  .61  .70  .81  Two consecutive years of fishing at the unknown but constant rates of F. -, 14 15 0.200 and F. = 0.400 yields the catch ratio I C. , / . I C. = 0.978 as 9  9  well as the individual values (not shown) of  for each fishing year.  the methods described earlier are used to prepare the i n i t i a l estimates, and t  If  parameter  is set equal to -0.200, a complete Model 2 analysis pro-  vides the following results:  109  Unknown W a  1  j =2  7239,028  7239.297  Actual Values 7263.000  0.140  0.140  0.140  2.999  2.999  3.000  0.333  0.533  Z  ]  = 0.620; Z = 0.820  0.353  0.553  F  1  = 0.200; F = 0.400  -0.020  -0.020  K  b Z. .(i=4 1  »J  to 15)  M. .(i=4 to 15)  2  2  M. = 0.420 3  Apparently the parameter K, and to a lesser extent b., are insensitive to the condition of variable M. whereas  is more responsive.  The least  desirable situation is also emphasized in that the mortality coefficients contain the largest errors. IAT,  An obvious remedy in this case is to estimate  b and K. independently, substitute their values into program ZXSSQ and  constrain the solution to the only unknown, Z. , in each fishing year.  In  the process a method of distinguishing between the two situations of variable or constant M. is revealed.  If the former condition exists the parameters b  and ]fl estimated i n i t i a l l y by not constraining ZXSSQ will differ from those m  obtained independently. An alternative solution that does not require independent parameter estimates is also possible.  The method is approximate and makes use of the  insensitivity of -K and b and the fact that the difference Z F-|) is estimated correctly whether  is variable or not.  2  - Z-j (or F  2  -  Equation (3.04)  is also collapsed and summed over all fishable age groups such that  no  15 _  Z.  where,  W  A. = (1-e  15  -.bl.  15  - Z . 15  -K(i-t) b -k(i+.5-t ) b -k(i+l-t ) b ) , B. = (1-e ) and D. = (1-e ) . n  0  u  0  Using the fishing data, the original analyses and the given value of t , the following information is either known or may be calculated:  j=l and 2 K - 0.140  b = 2.999  15 \k. = 4.7854 4  15 YB. = 5.1031 4  1  .  2^ + 0.200 15 4  1  1  >  36,824.8 36,750.1  1  15  15 ID. = 5.4180 4  =  1.002  "^1 2 4 •>1  1  C  Substitution of the appropriate values into equation (3.16) for each fishing year leads to two equations involving the sums of the mean weights.  The term  W^/6 is cancelled by dividing one equation by the other, and after rearrangement and simplification, the relationship  Z, - — 1-e^l  -.51, -Z, ,Z, + .20, (1.0 + 4.2656 e '+ 1.1322 e ') = 1 .002 >-(Z-, + ,20) 1-e -.5(Z.+.20) -(Z. + .20K (1.0 + 4.2656 e + 1.1322 e 1  has only one unknown, Z^ .  1  ;  The solution Z-j = 0.590 leads to 1^ = 0.790, and  by means of the combined catch equation,  = 0.197, F = 0.397 and M. = 0.393. 2  m  The results differ from the actual values displayed in the preceding table because of the approximate expression involving  .  The mean 1^ or X,  can now be substituted into equation (3.16) to solve for W^. estimate is combined with F-j or  ar,  d  t n e  If this last  W. values it is also possible to  solve for the individual, NT coefficients associated with each age group and the starting values (N ). • . 3  0  1,]  The analagous situation involving a constant M. but variable F. for each age can be treated similarly to yield the means F^ and F^ as well as the individual, fishing mortality coefficients.  In the more general case in  which M. and F.. are both variable, the solution cannot proceed beyond e s t i mates of Z . , T. and M. .  112  DISCUSSION AND CONCLUSIONS The main topic of this thesis is the development of an assessment method that does not rely on the use of effort data.  Before turning to  this subject, however, a few words about seasonal pulse fishing are necessary. The process by which seasonally applied effort leads to an increase in yield is not difficult to understand.  In contrast to continuous fishing,  which ignores seasonal variation in growth and mortality rates, pulse fishing is adaptive.  By applying effort after the biomass has had an  opportunity to increase, or to counteract the loss from natural the yield will be greater than that of continuous fishing.  mortality,  In the model  used for analysis, fishing always begins with the age group containing the critical size and the difference in yield between intensive and continuous fishing is maximized. based  Clark et a l . (1973) arrived at a similar conclusion  on an economic analysis of the Beverton and Holt (1957) constant  recruitment model.  In succeeding years the yield difference between these  two strategies diminishes and eventually is reversed.  However, the gain  in yield during the f i r s t few years is partially retained due to the reduced growth rate of a declining number of f i s h .  Any species that demonstrates  rapid growth (small K), or high mortality (M), is therefore more susceptible to yield improvement by seasonal fishing than one having opposite characteristics.  Regardless of the growth-mortality  (G-M) pattern or value of M,  an intensive fishing strategy consistently provides the largest yield benefit i f applied at the correct time, or the greatest loss is inappropriately  113  timed.  When to fish is as important as knowing the proper strategy and  this subject is taken up in the next paragraph.  By expanding the fishing  season about the most favorable mode, moderate or extensive strategies also provide an improvement in yield but of diminished magnitude. What is less easily understood is how the remaining variables of fishing mortality (F), the G-M pattern and M interact to affect the potential increase in yield and the optimal time to begin fishing.  The study shows  that the G-M pattern has the greatest influence in one respect.  If growth  and natural mortality are seasonally distinct and concentrated in time, the maximum potential exists for yield improvement and the best time to fish then depends on the value of F.  If F is low, it is necessary to begin  fishing intensively before the active growth period has expired.  At high  levels of F the yield can be increased either by withholding effort until the end of the growth period, or by fishing counteract!'vely to offset the loss due to natural mortality.  An analogy exists between these examples  and a pisciculture situation given limited or abundant resources, respectively, for harvesting the catch.  Case 18 illustrates the gist of these  arguments and represents the most favorable conditions for a large-valued K, F and M as well as a non-overlapping G-M pattern.  In this instance  the increase in yield exceeds that of continuous fishing by approximately thirty per cent.  As the period of natural mortality coincides to a  greater extent with that of growth, the benefits of seasonal fishing must decline regardless of the strategy used.  Continuous growth and  mortality (Case 9) represents the maximum degree of conflict between these  114  processes, and even under the most favorable conditions of K, M and fishing strategy the yield improvement is less than five per cent. Because of the complexity of interactions that exist between the variables, the theoretical benefit of seasonal fishing can only be e s t i mated by analysis.  To do so requires detailed knowledge of the G-M  pattern, M and the anticipated value of F; practical difficulties arise.  yet it is here that the f i r s t  If temperate species undergo a period of  accelerated growth during the year, the magnitude and duration of the increase, as well as the general growth pattern, can usually be found by sampling the catch.  By comparison, i t is substantially more d i f f i c u l t to  obtain a reliable estimate of M and particularly its seasonal variation. The more realistic situation of age-group variation in mortality adds further uncertainty to the choice of optimal fishing strategy.  It has also  been pointed out that the decision of when to fish is dictated by the magnitude of F.  Since total allowable catches, quotas and regulated  open seasons are subject to change, they can affect the intensity, duration and time of fishing.  The structure for applying seasonal fishing may .  already exist in the form of regulations, but the influence of these additional variables should be considered in any future appraisal of its benefits. A substantial part of this thesis is concerned with the assumptions and derivation of three basic types of fisheries models. are:  Specifically, these  the regenerative, or stock production models, the dynamic pool model,  and the virtual population and cohort analysis techniques.  The purpose in  115  analyzing them in such detail is to discover which factors contribute to or limit their assessment capability and to use this information in developing the annual state models.  The following paragraphs summarize how each model  type differs, by choice or capacity, in its approach to the assessment problem. The u t i l i t y of the regenerative models depends on the relationship established between catch and apparent abundance or yield and apparent biomass when a method of aging the stock is unknown.  Of the two components of  an assessment, age-group abundance and the effects of fishing, only the latter has the potential of being estimated.  Although i t is usually not stated  explicitly, an assumption of the models.is that stock structure-be reasonably stable.  Changes in catch or yield are therefore the result of fishing  activity and the displacement of the abundance or biomass from their respective environmental carrying capacities.  If there are environmentally  induced changes in recruitment, however, the abundance and biomass are a l - . tered independently of the effects of fishing or the natural growth rate of the stock.  If the catchability (q) is also variable because of behavioral  changes, weather or accessibility, the consequence of both activities is to introduce uncertainty (or variance) in the parameter and q estimates.  More-  over, since q is not determined each year, but only through analysis of a long time-series of data, the observed changes in fishing success (U or V) cannot be attributed with confidence to fluctuations in either the stock or q.  Collectively, these problems are further aggravated by the difficulty in  obtaining a reliable yearly estimate of fishing effort. y i e l d , the effort term cannot be measured directly;  Unlike catch or  i t must be assembled  from and weighed according to a variety of influences that exist in a fishery.  116  Since the distribution of fish and fishermen is usually non-random, what is required is an estimate of the effective fishing effort (f).  It is the  quantity of effort that would have been expended to obtain the catch or yield i f fishing had been conducted at random and continuously during the year.  Provided f can be calculated U and V will then serve as indices of  apparent abundance and biomass, respectively.  The derivation of f, however,  involves more than just the standardization of vessels and gear.  The co-  operation of the fishermen is essential in revealing where and when they fished, the quantities and species of fish caught, and the time spent searching, fishing and handling gear.  For competitive reasons fishermen may be  reluctant to divulge all or part of this information, despite management's assurances of confidentiality, and on occasion the data may be intentionally misleading.  Taken together, these factors emphasize the difficulty in ob-  taining a reliable estimate of f and the corollary product F.  In contrast  to the problems associated with recruitment, q and f, the regenerative models offer a distinct advantage in not requiring an estimate of M. If it is possible to determine the age of a fish the potential exists for carrying out a complete assessment.  Any doubt concerning the appropriate-  ness of the logistic equation to relate yield to biomass, as in the above models, is also removed.  Growth equations in length or weight supply the  multiplicand in calculating the appropriate biomass product and the remaining unknown is reduced to fish numbers.  All of these benefits are reflected in  the simple and more complex forms of the dynamic pool model, but it is equally clear that the difficulties of an assessment are compounded.  In  addition to the fishing mortality coefficient the growth parameters, apparent abundance and two new variables, the total and natural mortality coefficients  117  must be estimated for the stock or each age group.  Of these quantities the  growth parameters are obtained most easily, and since the abundance estimates depend on the coefficients, the confidence in an assessment rests on the accuracy of the mortality rates.  The methods used to calculate these rates,  however, rely extensively on the historical and current values of f, the assumed constancy of q over variable periods of time and, less frequently, on a reasonably stable recruitment and application of fishing effort.  Similar  or identical assumptions and dependencies occur in connection with the regenerative models and their affect on the reliability of an assessment is not repeated here. The virtual population and cohort analysis techniques reconstruct the experience of a year class during its fishable lifespan by interlocking a series of annual catch equations.  An estimate of the apparent recruitment  is obtained by back-calculation once the year class has l e f t , or nearly l e f t , the fishable stock. required.  Furthermore, data on growth in weight are not  These simplifications, together with information on age, create  the potential for a complete assessment.  Calculations can begin with any  age group in a year class, the requirement being a known or assumed constant value of M and a starting value of F.  Both Jones (1961) and Pope (1972)  indicate that the accuracy of back-calculated estimates of (N ).j and F Q  for the earliest ages improve with the number of age groups included in the analysis, but the improvement is relative to the chosen value of M. Although this feature is important for historical reasons, management is more concerned with the accuracy of a stock assessment for the immediate past year of fishing.  In this respect the virtual population and cohort analysis techniques  offer no advantage over the dynamic pool model, and they are equally dependent  118  on f and q to prepare the initial estimates of the mortality coefficients. The position taken in the above summaries is that variable  recruitment,  the reliability of f and the assumed constancy of q restrict the confidence and timeliness of stock assessment procedures.  Under certain conditions the  annual state models circumvent these difficulties and permit a complete assessment of the stock or an individual age group at the end of each fishing year.  The potential implications for management are that the dependence on  fishing effort and historical data is reduced, while growth parameters and the elements of an assessment are updated in time to apply regulations to the fishery. The capacity of the state models to provide this rapid assessment is attributed to three modifications of the dynamic pool model and the use of the age-group statistic for mean weight, W"..  I n i t i a l l y , i t is a reasonable  assumption that past events in the l i f e of a year class are summarized in the abundance and growth characteristics of each age group currently being exploited.  By reducing the yield model for the stock to that of an age  group, the analytical complications of variable recruitment and the historical effects of fishing are avoided.  The second modification, that of sub-  stituting the generalized Richards (1959) growth equation -k(t-t ) (w = Wjl-e t  u  b  )) u  for the von Bertalanffy (1938) model, removes the  cubic restriction on the parameter b.  The purpose of this change is to  allow a more accurate description of a species' growth.  The final adjust-  ment involves integration of the yield model by an approximate method without a serious loss in accuracy.  These alterations, together with informa-  tion on yield in weight (Y^ .) and catch (C^.) lead directly to an expression  119  for W. that is independent of abundance. for W.j  More importantly, the equation  provides a deterministic solution for  since the only other un-  knowns, the growth parameters, can be readily obtained by sampling.  If  these methods are applied to the same year class for two consecutive years, the difference between 1-,,  and Z. • estimates the change in annual  fishing mortality rates assuming that M^ = M.^ .  The ratio  j/F^+i  »  obtained from consecutive catch equations, can then be used to solve for the individual values of F.. and F ^  , and simultaneously, for M.. ^  and the  apparent abundances. The above method is one of several model variations that can be used to carry out an assessment.  In Section III  alternative solutions are presented  that deal with either the total stock (Models 1 and 2) or a single age group when M is constant or a function of age.  Although these alternative methods  are analytically interesting, in some cases their practical application is limited.  If a species grows rapidly in length a model based on the mean  length of fish (L^) may be adequate.  For slower growing species, the length  of fish sampled from the annual catch probably cannot be measured with sufficient accuracy to estimate Z^ from Lj .  The quantity  in this respect and is the logical indicator to use.  is more sensitive  If-L/ is not included  in the calculations (Model 2) it is neither practical nor advisable to solve simultaneously for the growth parameters and the total mortality coefficient. The surface over which program ZXSSQ must search for a solution is flatter than that of Model 1 and there is greater chance of being trapped in a local minimum. K, b and t  To avoid this problem the independently determined values of W , ro  can be substituted directly into the Model 2 equations.  The  120  final transition from a stock model to that of an age group is readily accomplished and is described in the preceding paragraph. In all of the above models the key variable that determines the r e l i a b i l i t y of an assessment is the total mortality coefficient.  The  accuracy of Z.., however, depends on the uncertainty associated with all of the parameters (W^, K, b,  , t ) and variables (w\ , U ) .  sense the limited exploration of errors in t insufficient.  In this  that affect Z^ (p. 105) is  A complete sensitivity analysis is required to indicate  whether the uncertainties in the parameters and variables are acceptable in determining Z^, and how their uncertainties contribute relatively to that of Z.j (Majkowski, 1981, MS). of  In practice, to improve the accuracy  i t may be necessary to abandon the traditional  use of the von  Bertalanffy (1938) and Richards (1959) functions as composite descriptions of growth for all age groups, particularly i f growth has age-group variability between years, or is a function of sex or stage of maturity.  In  these cases the sampling program that runs concurrently with fishing operations to establish W. or  can be expanded to obtain the necessary  information. An assumption that is common to all of the above models is that the stock is closed to immigration and emigration.  Although neither condition  can be prevented, their occurrence can affect the results of an assessment. For example, i f an immigrant source of fish enters the fishing grounds, the a r t i f i c i a l l y high values of increase that of abundance.  and  will reduce the estimate of 1. and  Throughout the development of the state models  i t is also assumed that growth and mortality act concurrently and continuously during the fishing year.  If there are seasonal changes in either  121  growth or the rate of fishing, W. in particular will be affected and lead to an erroneous estimate of Z . .  In this case the models can be partitioned  into shorter periods of time in which the original assumption is maintained. The equation for  W.  is then modified to:  Z. —j-  =  1-e  •  1  W —  -0.5vZ. -vZ. v(A + 4Be + De ') 1  6  where v = the fraction of a year;  i = the age at the beginning of the  interval v , and A, B and D are equal, respectively, to ( _ - ( k  1  i +  e  5 v  - o)) , t  b  ( _ -k(i v-t )jb  (1 - e " ^ " V ) , b  +  a n d  1  e  0  >  I n  t  h  e  i  r  p  r  e  s  e  n  t  f o r m  t  h  e  models apply only to the fully fishable age groups and are limited by the assumption that within-age size-selective mortality does not occur.  If the  smaller members of an age group are not harvested, to\ will increase and lead to an underestimate of I. and an overestimate of (N )^. Q  The opposite  situation is expected to occur i f the larger fish in an age group are able to avoid capture. =  M-j+j j+l  o r  The last important assumption to consider is that M. •  ' ™ ff e  e c t  ' that M is constant for all age groups.  The  state models for two consecutive age groups, as well as the cohort analysis technique (Pope 1972), are particularly sensitive to this assumption. is invalid, the ratio F.. j/F-j+j  If  it  determined by the catch equations will  s t i l l be correct, but the difference between the total mortality coefficients will not estimate the change in fishing mortality alone.  In this case it is  not possible to solve for the individual values of F.. .. and F^ .  directly  and two optional methods of analysis must be considered. In the f i r s t procedure i t is immaterial whether M is constant or a function of age, but it must be assumed the fishing mortality rate of all age groups in a given year is  122  constant.  If in addition F. f F . , , then it is possible to solve for the +  mean values  Z. , Z .  + 1  and M. . , , and eventually for F., F. , and the  individual values of M. .  +  If both fishing and natural mortality are age  dependent, the second procedure is restricted to a stock analysis in consecutive years and cannot proceed beyond the mean values of the mortality coefficients. With the exception of seasonal pulse fishing the preceding remarks are summarized in the following conclusions: 1.  The potential of the annual state models to provide a complete assessment of a stock is demonstrated.  2.  The fishing and natural mortality coefficients and the apparent abundance of each age group is estimated without reference to effective effort, the catchability coefficient, the number of recruits or the entire history of the year class.  3.  The data requirements of the models are an age group sample of: 1)  the annual catch and yield in length and weight, or 2)  annual catch and yield in weight, and 3) of the growth curve intercept.  the  an independent estimate  This information is used to cal-  culate the mean length (L^) and weight (W\) of each age group, to develop suitable expressions for each of these quantities and to maintain an accurate description of growth. 4.  Equations for yield in weight and mean weight incorporate a generalized growth model by means of an approximate integration technique.  Potentially, the same method can be applied to other  123  empirical growth equations to improve the accuracy of the state models and extend their range of application. The assessment capability of the models is also flexible.  Depending  on the assumed constancy of the natural or fishing mortality coefficients, the analysis can be carried out on a stock or age group basis by combined (T. and W^) or individual The  (W..) equations.  major assumptions of the models are that:  closed to immigration and emigration;  (1) the stock is  (2) growth and mortality are  concurrent and continuous during the period of analysis;  (3) fishing  is not size selective within age groups;  and (4) the natural mortality  rate of the fishable stock is constant.  If this last assumption is  invalid, alternative solutions are presented for constant fishing mortality in each year of fishing or for age-dependent fishing and natural mortality.  124  LITERATURE CITED Abramson, N.J. 1971. Computer programs for fish stock assessment. FAO Fish. Tech. Paper, 101:11-154. Akenhead, S. 1976. Personal communication. Comm. Northwest At!. Fish. (ICNAF).  Former biostatistician,  Int.  Allen, K.R. 1966. A method of fitting growth curves of the vonBertalanffy type to observed data. J . Fish. Res. Bd. Canada, 23:163-179. Baranov, A. 1918. On the question of the biological basis of fisheries. Nauchn. Issled. Ikthiol. lust. Izv. 1:18-128 (in Russian). By_: On the dynamics of exploited fish populations. R.J.H. Beverton and S. Holt. 1957. Fish. Invest..Series 2, 19. London. 533pp. Beverton, R.J.H. 1954. Notes on the use of theoretical models in the study of the dynamics of exploited fish populations. U.S. Fish. Lab., Beaufort, N . C , Misc. Contrib. 2. 159p. Beverton, R.J.H. and S . J . Holt. 1956. A review of methods for estimating mortality rates in fish populations with special reference to sources of bias in catch sampling. Rapp. Conseil Expl. Mer, 140(1):67-83. . . 1957. On the dynamics of exploited fish populations. Invest. Series 2, 19. London. 533pp. von Bertalanffy, L. 1938. Biol. 10:181-213. Bishop, Yvonne M.M. 1959. virtual populations.  A quantitative theory of organic growth.  Fish. Hum.  Errors in estimates of mortality obtained from J . Fish. Res. Bd. Canada, 16:73-90.  Brody, S. 1945. Bioenergetics and Growth. New York: Reinhold. 1023p. By: Taylor, C.C. 1962. Growth equations with metabolic parameters. J . Cons. int. Explor. Mer, 27:270-286. Brown, M.E.  1957.  The physiology of fishes.  New York:  Academic Press. 447p.  Brown, E.H. 1972. Population biology of alewives, Alosa pseudoharengus, in Lake Michigan, 1949-70. J . Fish. Res. Bd. Canada, 29:477-500. Bowers, A.B. 1952. Studies on herring (Clupea harengus L.) in Manx waters: The autumn spawning, and the larval and post-larval stages. Proc. Trans. L'pool. Biol. Soc. 58:47-74. Clark, C , G.E. Edwards and M. Friedlander. 1973. Beverton and Holt model of a commercial fishery: optimal dynamics. J . Fish. Res. Bd. Canada, 30:1629-1640. Conway, G.R., N.R. Glass and J . C . Wilcox. 1970. Fitting nonlinear models to biological data by Marquardt's algorithm. Ecology 51:503-507.  125  Cushing, D.H. 1970. Fisheries Biology. A study in population dynamics. Univ. Wisconsin Press. 200p. . 1972. A history of some of the International Fisheries Commissions. Proc. R. Soc. Edinb. Sec. B, Biol. 73:361-390. J_n: A link between science and management in fisheries. D.H. Cushing, 1974. Fish. Bull. 72:859-864. Das, N. 1972. Growth of larval herring (Clupea harengus) in the Bay of Fundy and Gulf of Maine area. J.Fish. Res. Bd. Canada 29:573-575. DeLury, D.B. 1947. 3(4):145-167.  On the estimation of biological populations.  Biometrics  Doubleday, W.G. MS 1975. A least squares approach to analyzing catch at age data. ICNAF Res. Doc. 75/35. 33p. Fabens, A . J . 1965. Properties and fitting of the von Bertalanffy growth curve. Growth, 29:265-289. Ford, E. 1933. An account of herring investigations conducted at Plymouth during the years 1924-1933. J . Mar. Biol. Assoc. U.K. 19:305-384. Fox, W.W. 1970. An exponential surplus-yield model for optimizing exploited fish populations. Trans. Am. Fish. Soc. 99(l):80-88. Fry, F . E . J . 27-67.  1949.  Statistics of a lake trout fishery.  Biometrics  5(1):  1957. Assessment of mortalities by use of the virtual population. Paper P-15. Lisbon meeting sponsored by ICES, FAO and ICNAF. J_nj Errors . in estimates of mortality obtained from virtual populations. Yvonne M.M. Bishop, 1959. J . Fish. Res. Bd. Canada 16:73-90. Glass, N.R. 1967. A technique for fitting non-linear models to biological data. Ecology, 48:1010-1013. Graham, M. 1935. Modern theory of exploiting a fishery and application to North Sea trawling. J . Cons. int. Explor. Mer, 10:264-274. Gulland, J.A. 1955. Estimation of growth and mortality in commercial fish populations. U.K. Min. Agric. and Fish., Fish. Invest. Ser. 2, 18(9). 46p. . 1961. Fishing and the stocks of fish at Iceland. Agric. and F i s h . , Fish. Invest. Ser. 2, 23(4). 52p. . 1965. Estimation of mortality rates. working group report, ICES, CM. (3). Mimeo.  U.K. Min.  Annex to Arctic fisheries  . 1969. Manual for methods for fish stock assessment. Part 1. Fish population analysis. FAO, No. 4. 154p. Halliday, R.G. 1972. A yield assessment of the eastern Scotian shelf cod stock complex. ICNAF Res. Bull. 9:117-124.  126  Healey, M.C. 1975. Dynamics of exploited whitefish populations and their management with special reference to the Northwest Territories. J . Fish. Res. Bd. Canada 32:427-448. Hilborn, R. 1979. Comparison of fisheries control systems that utilize catch and effort. J . Fish. Res. Bd. Canada 36:1477-1489. Hennemuth, R.C. 1961. Size and year class composition of catch, age and growth of yellowfin tuna in the eastern tropical Pacific Ocean for the years 1954-58. Inter-Am. Trop. Tuna Comm. Bull. 5(1):1-112. Hodder, V.M. 1975. Personal communication. Int. Comm. Northwest A t l . Fish.  Assistant Executive Secretary,  Howard, G.V. and A. Landa. 1958. A study of the age, growth, sexual maturity and spawning of the anchoveta (Cetengraulis mysticetus) in the Gulf of Panama. Inter-Am. Trop. Tuna Comm. Bull. 7(9):391-457. Jensen, A.J.C. 1950. Amount of growth of herring fry in Danish waters. Rep. Dan. Biol. Sta. 51:20-48. Jones, R. 1957. A much simplified version of the fish yield equation. Doc. 21, presented at the Lisbon joint meeting of ICES, FAO and ICNAF. 8p. . 1961. The assessment of the longterm effects of changes in gear selectivity and fishing effort. Mar. Res. (Scotland) 2:1-19. . 1964. Estimating population size from commercial statistics when fishing mortality varies with age. Rapp. P.-V. Reun. Cons. Perm. int. Explor. Mer, 155:210-214. . MS 1967. Appendix to the report of the northwestern working group. ICES. C M . (1968) Document No. 20(mimeo). Jj^: Estimation of fish mortality in the stock of cod off west Greenland. A. Schumacher (1970). Proc. Ger. S c i . Comm. Oceanic Res. 21(4):248-259. . MS 1974. Assessing the longterm effects of changes in fishing effort and meshsize from length compositon data. Int. Counc. Explor. Sea; Demersal fish (northern) comm. C M . 1974/F:33. 13p. Kelso, J.R.M. and F . J . Ward. 1972. Vital statistics, biomass and seasonal production of an unexploited walleye (Stizostedion vitreum vitreum) population in West Blue Lake, Manitoba. J.Fish. Res. Bd. Canada, 29: 1043-1052. Kennedy, W.A. 1954. Growth, maturity and mortality in the relatively unexploited lake trout, Cristovomer namaycush, of Great Slave Lake. J . Fish. Res. Bd. Canada, 11:827-852. Knight, W. 1968. mathematics.  Asymptotic growth: an example of nonsense disguised as J . Fish. Res. Bd. Canada 25:1303-1307.  Kohler, A.C. 1964. Variations in the growth of Atlantic cod (Gaelus morhua L.). J . Fish. Res. Bd. Canada 21:57-100.  127  Kutty, K. 1968. Estimation of the age of exploitation at a given fishing mortality. J . Fish. Res. Bd. Canada 25:1291-1294. Larkin, P.A. 1974. Personal communication. Professor, Institute of Animal Resource Ecology, University of British Columbia. . 1977. An epitaph for the concept of maximum sustained y i e l d . Trans. Am. Fish. Soc. 106(1):1-11. Majkowski, J . MS 1981. Sensitivity analysis and its extensions—applicability and usefulness in a multi-species approach for designing a research program and validating the outcome of this program. P.A..Larkin (ed.). Paper submitted for ICLARM/CSIRO workshop on the theory and management of tropical multi-species stocks. Cronulla, N.S.W. 1981. Marr, J . C . 1951. On the use of terms abundance, availability abundance in fishery biology. Copeia 2:163-169.  and apparent  Mohn, R.K. 1980. Bias and error propagation in logistic production models. Can. J . Fish. Aquat. Sci. 37:1276-1283. Murphy, G'.I. 1965. A solution of the catch equation. Canada 22:191-202. Nikolskii, G.V.  1963.  The ecology of fishes.  J . Fish. Res. Bd.  New York:  Academic Press. 352p.  Paloheimo, J . E . 1958. A method of estimating natural and fishing mortalities. J . Fish. Res. Bd. Canada 15:749-758. . 1961. Studies of estimation of mortalities. I. Comparison of a method described by Beverton and Holt and a new linear formula. J . Fish. Res. Bd. Canada 18:645-662. and L.M. Dickie. 1964. Abundance and fishing success. Reun. Cons. Perm. int. Explor. Mer, 155:153-163.  Rapp. P.-V.  . 1965. Food and growth in fishes. I. A growth curve derived from experimental data. J . Fish. Res. Bd. Canada 22:521-542. Parker, R.R. and P.A. Larkin. 1959. Res. Bd. Canada 16:721-745.  A concept of growth in fishes.  J . Fish.  Paulik, G.J. and L.E. Gates. 1964. Allometric growth and the Beverton and Holt yield equation. Trans. Am. Fish. Soc. 93:369-381. Pearson, K. 1948. Tables of the incomplete beta function. Press, Cambridge. 494p.  The University  Pella, J . J . and P.K. Tomlinson. 1969. A generalized stock production model. Inter-Am. Trop. Tuna Comm. Bull. 13(3):421-496. Petersen, C.G.J. 1894. On the biology of our flat-fishes and on the decrease of f l a t - f i s h fisheries. Rep. Dan. Biol. Stn. 4; 146p. By_: A link between science and management in fisheries. D.H. Cushing 1974. Fish. Bull. 72(4):859-864.  128  Pope, J.G. 1972. An investigation on the accuracy of virtual population analysis using cohort analysis. ICNAF Res. Bull. 9:65-74. Regier, H.A. and F.C. McCracken. 1975. Science for Canada's Shelf Seas Fisheries. Fish. Res. Bd. Canada Report No. 3. Ottawa. 46pp. Richards, F . J . 1959. A flexible growth function for empirical use. Biol. 10(29):290-300.  J . Fish.  Ricker, W.E. 1940. Relation of "catch per unit effort" to abundance and rate of exploitation. J . Fish. Res. Bd. Canada 5:43-70. . 1944. 1:23-44.  Further notes on fishing mortality and effort.  Copeia,  . 1945. A method of estimating minimum size limits for obtaining maximum y i e l d . Copeia, 2:84-94. . 1948. Methods of estimating vital statistics of fish populations. Indiana Univ. Publ. S c i . Ser. 15:101p. . 1960. Ciba Foundation colloquia on aging. Vol. 5. "The l i f e span of animals." London: J . and A. Churchill Ltd. 1959. Reviewed in J . Cons, internat. Explor. Mer 26:125-129. . 1968. Handbook of computations for biological statistics of fish populations. Fish. Res. Bd. Canada, Bull. 19. Ottawa. 300p. . 1975. Computations and interpretation of biological statistics of fish populations. Fish. Res. Bd. Canada, Bull. 191, Ottawa. 382p. . 1977. The historical development. J_n: Fish population dynamics. J.A. Gulland (ed.). London: John Wiley & Sons. 372p. Roff, D.A. and D.J. Fairbairn. 1980. An evaluation of Gulland's method for fitting the Schaefer model. Can. J . Fish. Aquat. S c i . 37:1229-1235. Russell, E.S. 1931. Some theoretical consideration on the "overfishing" problem. J . du Conseil, VI(l):3-20. Schaefer, M.B. 1954. Some aspects of the dynamics of populations important to the management of the commercial marine fisheries. Inter-Am. Trop. Tuna Comm. Bull. l(2):25-56. . 1957. A study of the dynamics of the fishery for yellowfin tuna in the eastern tropical Pacific Ocean. Inter-Am. Trop. Tuna Comm. Bull. 2(6):245-285. .and R.J.H. Beverton. 1963. Fishery dynamics--their analysis and interpretation. In: The Sea. N.M. Hill (ed.). New York: Wiley Interscience Publ. Vol. 2. 554p. Schnute, J . 1977. Improved estimates from Schaefer production model: Theoretical considerations. J . Fish. Res. Bd. Canada 34:583-603.  129  Schumacher, A. 1970. Estimation of fish mortality in the stock of cod off west Greenland. Proc. of Ger. Sci. Comm. for Oceanic Res., 21(1-4):248-259. (Translated from German by Fish. Res. Board Can. Transl. Ser. No. 1690, 1971). Silliman, R.P. 1971. Advantages and limitations of "simple" fishery models in the light of laboratory experiments. J . Fish. Res. Board Can., 28:1211-1214. Smail, L.L.  1949.  Spang, H.A. 1962. functions.  Calculus.  Appleton-Century-Crofts, L t d . , N.Y.  592p.  A review of minimization techniques for nonlinear SIAM Review, 4:343-365.  Ssentongo,„G.W. and P.A. Larkin. 1973. Some simple methods of estimating mortality rates of exploited fish populations. J . Fish. Res. Board Can., 30:695-698. Swift, D.R. 1955. Seasonal variation in the growth rate, thyroid gland activity and food reserves of brown trout (Salmo trutta Linn.) J . Exp. B i o l . , 32(4):751-764. Taylor, C.C. 1958. 23:366-370.  Cod growth and temperature.  Cons. Int.  . 1962. Growth equations with metabolic parameters. Explor. Mer, 27:270-286.  Explor. Mer, J . Cons.  Int.  Thompson, W.F. and F.H. Bell. 1934. Biological statistics of the Pacific halibut fishery. (2) Effect of changes in intensity upon total yield and yield per unit of gear. Rept. Int. Fish. (Pacific Halibut) Comm., 8. 44p. Tody, W.H. 1964. An investigation of the success of rainbow trout populations in ten lakes relative to limiting environmental factors. MS Ph.D. thesis. Michigan State Univ., Agric. Forestry and Wildlife. Tyler, A.V. 1974. Users manual for PISCES. A general fish population simulation and fisheries program. Fish, and Mar. Ser., Can., Tech. Rept. No. 480. 30p. Uhler, R.S. 1980. Least squares regression estimates of the Schaefer production model: some Monte Carlo simulation results. Can. J . Fish. Aquat. Sci. 37:1284-1294. Van Oosten, J . and R. Hile. 1947. Age and growth of the lake whitefish. (Coregonus clupeaformis, Mitchi11), in Lake Erie. Trans. Am. Fish. S o c , 77:178-249. Walford, L.A. 1946. A new graphic method of describing the growth of animals. Biol. B u l l . , 90(2):141-147.  130  Watt, K.E.F. 1956. The choice and solution of mathematical models for predicting and maximizing the yield of a fishery. J . Fish. Res. Board Can., 13:613-645. Widrig, T.M. 1954a. Method of estimating fish populations with application to Pacific sardine. U. S. Fish and Wildlife Ser. Bull. 94, 56:141-166. . 1954b. Definitions and derivations of various common measures of mortality rates relevant to population dynamics of fishes. Copeia, 1:29-32. Wiles, M. and A.W. May. 1969. Biology and fishery of the west Newfoundland cod stock. Fish. Res. Board Can. Studies, Part 1, No. 1318:487-525. Wilimovsky, N.J. and E.C. Wicklund. 1963. Tables of the incomplete beta function. Inst, of F i s h . , Univ. of British Columbia. 291p.  APPENDIX A Analysis of Fisheries Models Using a Gaming Technique  The results, conclusions and management implications arrived at by gaming with two fisheries models are reported in this section.  In each  case the general objective was to identify the minimal amount of information needed from the fishery to solve a given problem in the least amount of time.  A gaming technique is readily adapted to such questions  and in addition it is a valuable, educational tool.  In the models that  follow the reader will note that the investigator is constantly weighing the information content of the fishery data against the desired objective.  Because of the minimal time limit, information that is mis-  leading or ambiguous is quickly identified and rejected.  By a process  of elimination the "manager" is then led to ask the following question: What piece or pieces of information are at present not being collected from the fishery that would help solve the problem more directly?  If  such statistics can be identified, a second question must immediately be answered:  How can the information be collected from the fishery,  or in the present case, is the simulation model programmed to reveal this information?  If not, the model can be altered to disclose increas-  ing amounts of information in keeping with a greater, understanding of the problem. The equivalent situation in a real fishery is recognizable in that data collection should be oriented towards both economy and high information content.  132  The method of presentation and specific objective is different for each of the models.  Details of the Schaefer (.1954, 1957) simulation are  summarized while those for the modified Beverton-Holt (J957) model are reported in ful1.  1.  The Schaefer general production model.  Objective:  to determine the catch and effort co-ordinates of.maximum  sustained yield (MSY) using the least number of separate applications of fishing effort (f = boats).  The only information available to the  investigator is that the maximum number of boats cannot exceed 1000, a convenience reflecting the scaling of the computer programme.  The  operator is not aware of the Schaefer model parameters or those of the stochastic processes. The simulation is based on a discrete time version (Larkin, personal communication) of Schaefer's production model.  Variability in  recruitment (R) and catchability (q) is achieved by multiplying each :  quantity by a normally distributed, random number with mean equal to one and a standard deviation of 0.5 and 0.25, respectively.  Given the  basic Schaefer model, in time step j , dN  W  j  = r N. - r N . K 1  J  9 8  ,  CA 1)  J  where, r = 1; net rate of population increase in numbers K = 1000; carrying capacity dN./dt = rate of change of stock size N.; 3  3  ment AR.  equal to absolute recruit-  A flow chart for a single time step may be diagrammed as follows; Start  AR = l.N. -0.001 N . ' 1  J  9 8  J  R = AR.v N.  1  = N. + R  J  J  CB = f . q . n F = CB/960, C = F. N . ' -»• Print out catch (c) Res = N.' - c Set  ff. i = Res and, Return to Start +  where, v = recruitment random number multiplier N-' = fishable stock J n = catchability coefficient (q) random multiplier f = number of boats applied by investigator;  o<CB<960  F = effective fishing mortality Res = residual reproductive stock Solution Methods and Discussion 1.  It is tempting at f i r s t to begin a cautious analysis of the problem  by the successive application of low effort levels.  A knowledge of the  resulting catch and effort, and hence the catch per unit effort CU), may be interpreted as an index of average, fishable stock abundance,  A record  of catch or effort versus U should then reveal the proximity of MSY. Unfortunately, the above method is extremely wasteful of time, and information on catch or  U can be both misleading and ambiguous.  For  a given percentage change in effort the resulting change in catch may  134  be relatively greater of smaller, and there may be no apparent relationship between effort and stock-size index.  The investigator is therefore  left in doubt as to the cause or causes that induces the change in catch. It may be due to the manipulation of effort, but equally a change in availability or vulnerability can produce similar results.  In addition  to the loss in time the manager cannot formulate a rational basis of future effort changes based on past events in the fishery. 2.  A bolder approach would be to increase effort by a factor of two or  three in consecutive years.  Hopefully, the variation induced by envir-  onmental effects will be overshadowed by the comparatively larger changes in effort, and a positive sense of direction may be revealed. The decision to follow this course in a fishery is potentially dangerous to the stock and may also be wasteful of effort.  If the i n i t i a l  stock size is unknown, and particularly i f the productive capacity is both low and unknown, a rapid rise in effort may drive the population to extinction.  In addition to this risk the information on catch and U  from the exploratory fishing period can be misleading.  Repeated trials  indicate that i f effort is increased in three successive time steps i t is possible for the catch to show a consecutive decline while the U simultaneously increases.  The random number multipliers, v and n, can  produce such an effect and the investigator is denied a logical interpretation of the data.  The rationale for subsequently changing effort  is missing and therefore the procedure cannot lead to the objective in minimal time.  135  3.  If both small and large changes in effort can produce equivocal  results a more informative approach might be to maintain the same effort level for two or more time steps.  The variability within the  system would then be partially revealed and similar information could be gathered at different effort levels. The above method is the least productive of information over time. By maintaining a constant effort an associated distribution of catch and  U may be generated provided the system is at equilibrium.  The  random multipliers, however, introduce stochasticity and equilibrium cannot be achieved.  There is a greater liklihood that the f i r s t choice  of effort will introduce oscillations in the system that can be vaguely monitored by a catch (or  U) record.  In time, the variability may become  damped, but the manager will only achieve information relative to a single effort value.  The balance of the system will remain unknown.  The cause of variation in  U or catch will be similarly confused and  unassignable to either a change in fishing efficiency, availability or recruitment. Of the three methods just described the cautious manipulation of effort (methods 1 and 3) appears to involve low risk, and given suffir cient data, the co-ordinatres of MSY could be estimated.  The cost of  using either approach, however, can be equated to a loss in analysis time and a low rate of return of information for the effort involved. The second procedure, involving large changes in effort, increases the breadth of information taken from the fishery and the perturbations are confined to a short time interval.  The exploratory potential of the  136  method i s appealing, but the benefit i s associated with a high r i s k . Future catches may be jeopardized and be accompanied by severe s o c i a l and economic d i s l o c a t i o n i f the experiment  should f a i l .  Nor can a  large amount of e f f o r t be committed to or removed from the f i s h e r y at w i l l .  To reduce the r i s k there i s a need to i d e n t i f y a p a r t i c u l a r  action with i t s consequences and the time lag between these events should be minimized to allow for remedial action i f necessary. measure of recruitment can f u l f i l l both requirements.  A  In a d d i t i o n i t  i s the only s t a t i s t i c that can act as a warning device with the l e a s t ambiguity, a property that i s not shared by e i t h e r catch or alone. 4.  U  data  A fourth s o l u t i o n method i s therefore p o s s i b l e .  The simulation model can be modified to p r i n t out an estimate of  recruitment following step two i n the flow chart.  Since the informa-  t i o n appears p r i o r to f i s h i n g i t i s assumed that the r e c r u i t s form part of the f i s h a b l e stock and the f o l l o w i n g a n a l y t i c a l model i s useful as a f i r s t  approximation:  Let, = number of r e c r u i t s entering f i s h e r y p r i o r to f i r s t a p p l i c a t i o n of e f f o r t N  Q  = residual stock before f i s h i n g begins  .N-j = f i r s t f i s h a b l e stock s i z e C-j = catch taken from N-j Res-j = residual stock a f t e r removal of C-j g = symbol to i n d i c a t e a function = underlined symbol known from f i s h e r y data  137  Then, given a starting value of R , the analytical sequence is estabQ  1ished: Start  N, = N + R 1 o —o Assume a value for N (= N ) o o Apply effort R e s  l = l " h N  R  = \  +  ^  ~ l C  }  = q(Re ) = g (N - C )  }  Sl  ]  1  Return to start and substitute:  R-, for R ; — l — o  and N for N o o  The algorithm above is used recursively to establish a relationship between R_^ and (N + R^ - C^).  Notice that in the last expression the  Q  initial and correct value of N is never determined. Q  It is an internally  generated number to which an arbitrary, large value is f i r s t assigned by the investigator and up-dated following each return to start.  Using  the i n i t i a l value of N it is then possible to estimate Res-j and estabQ  lish a functional relationship between this quantity and the number of recruits (R-j) derived from i t . of  Operationally, the successive values  R become the focus of attention for the region of MSY is associat-  ed with the highest average value of R.  In this connection the manager  will discover that the original objective, to find the co-ordinates of MSY, is inconsistent with its determination, " . . . i n the least number of separate applications of fishing effort".  A low number of iterations  is partly fortuitous because a new value of N is used at the beginning Q  of each trial run and cannot be duplicated. component is attached to both  Similarly, since a variance  R and effort, only the approximate region  of MSY, followed by a repetition of this procedure.  A typical Simula-  tion, with appropriate comments is recorded in Table A l , The solution method used in the example depends c r i t i c a l l y on the variance (stochasticity) assigned to recruitment and catchability.  If  the stochasticity is zero a production curve can be estimated using a minimum of three levels of effort.  Alternatively, i f the variance of  R and q is high relative to the change in catch with effort, a predictive relationship cannot be established over a short time period. In a real fishery, historical data may offer some guidance, but the length of time over which the data must be collected depends on the stochasticity inherent in the system and the accuracy with which effort can be measured. An analytical solution to the present problem is possible using equation (Al) below.  The investigator would, of course, be unaware of  the parameter values and the details are offered here for the sake of completeness. The rate of change of stock size is related to the stock by the expression: dN _ N - 0.001 N ' dt 1  (Al)  9 8  At the point of equilibrium (e), where the catch is maximized, the rate of change of d N /dt is zero, and the value of N is obtained by taking g  g  the second derivative of equation (Al): d  e 2 dt^ 2  N  =0=1-  (0.001) (1.98) N  0 , 9 8  139  and solving,  N = 573 fish g  The maximum catch associated with N is derived by substitution in e  equation (Al): C = ^ e = 1(573) - 0.001 (-573) ' dt 1  = 284 fish  98  e  The number of boats, f , needed to obtain the equilibrium catch is obg  tained from the relationship: C =_e__. N 960 f  e  e  or f  e  = C . 960/N , e  e  and substituting the values of C and N g  f  = 284 (960)/573 = 476 .boats  g  TABLE  AI.  Time Step  Recruits R  1  Simulation results and comments.  Boats 15  Catch C  50  44  Typical analysis of a Schaefer (1957) general production model.  C/BEstimate:Estimate: or Fishable Residual U Stock, FS Stock, Res  35  2035 0.88  1991  Method 4.  Comments Estimate s t a r t i n g stock ( N „ ) = 2000; FS = 2000 + R = 2035 Enter w i t h low e f f o r t to determine system r e s p o n s e . Res = FS-C = 1991. R = g ( R e s i ) , but f u n c t i o n unknown. 2  2  143  2143 150  3  270  4  228  6 7  253  0.84  450  693  1.54  450  423  0.94  450  317  0.70  199  9  82  10  128  12  299  13  219  14  322  15  229  1346  R must have been u n u s u a l l y h i g h . Repeat e f f o r t Res appears to be s t a b i l i z i n g , but  88  190  276  Repeat  Increase  effort.  effort.  5  to reduce Res.  r e c r u i t response begins to d e c l i n e . FS f a l l i n g and u n s e t t l e d . Repeat e f f o r t to see i f r e c r u i t m e n t c o n t i n u e s to d e c l i n e .  0.74  1211  0.28  1220  0.29  0.63  F S , Res and R show d e c l i n e . May have exceeded MSY. Upper bound MSY i s u n c e r t a i n but lower l e v e l o f 300 boats appears to generate a favorable recruitment. Reduce e f f o r t to 300, a l l o w stock to r e c o v e r , repeat e f f o r t i n c r e a s e to l o c a t e upper l i m i t MSY. U lowest on r e c o r d , but h a s n ' t been s a t i s f a c t o r y index in p a s t . Both R and FS showing nominal  •  increase.  Repeat  effort.  1260  1387 1686  450  6  i n time step 1.  6  1577  300  Strong r e c r u i t r e s p o n s e . FS beginning to i n c r e a s e . e f f o r t to see i f r e s u l t due to chance or p o p u l a t i o n b u i l d i n g up. Res beginning to respond p o s i t i v e l y .  Repeat actually  R e c r u i t response h o l d i n g , FS recovered to time . t e p Begin plan to i n c r e a s e e f f o r t to overshoot MSY.  7 level.  1410  0.61 1629  450  354  0.79  1275  Repeat e f f o r t t o determine  1323  Small change in R and F S , but Res beginning to d e c l i n e s h a r p l y . If MSY in excess of 450 boats must determine now w h i l e Res s t i l l has the c a p a c i t y to respond. Increase e f f o r t to 500 b o a t s .  1302  Repeat e f f o r t to determine  1597  500  274  0.55  if  decline  in R continues.  1552 500  226  R e c r u i t response may be e x c e p t i o n a l . Res = Res . Near MSY?  1348 300  317  83  B to reduce F S .  300.  1444  1303  300  Hold e f f o r t a t  R improving. FS no worse than Unexpected high c a t c h .  1545 334  Increase  1386  1663  450  11  F S , R and Res d e c l i n i n g .  1867  219  B.  1805 2079  481  Increase  1830 2058  300  Both FS and R I n c r e a s e d .  Both FS and R c o n t i n u i n g to i n c r e a s e . Note: U seems to be h i g h l y v a r i a b l e .  1.21  274  8  16  362  + 143 = 2143.  1922  1.47 2192  300  5  221  FS = 1191  250  0.50 1528  recruit  response.  REMARKS: Repeated e f f o r t of 500 boats f o r two a d d i t i o n a l time steps caused a d e c l i n e i n R. Conclude t h a t MSY i n r e g i o n o f 400 to 500 b o a t s . At time step 19, r e t u r n t o B=400, a l l o w s t o c k to r e c o v e r and repeat incremental p r o c e s s .  141  2.  A modified Beverton and Holt (1957) model containing a Ricker (.1968) recruit function  A critical requirement in the application of the Schaefer (.1957) model is that the fish species have an early age of reproduction and entry to the fishable stock.  Under these conditions i t is usually as-  sumed that the rate of change of stock size at any time is a function of the stock abundance or biomass at the same point in time.  Conse-  quently no time lag exists between a change in stock and the net response in recruitment.  However, for a large number of commercial species these  assumptions are inappropriate, and a model is called for that can simultaneously accomodate time lags and be descriptive of a multi age class fishery.  The "dynamic pool" model of Beverton and Holt (1957) is a  suitable choice, but its simplified form (ibid, p. -36) is restricted to the analysis of an equilibrium condition with constant recruitment.  A  considerable improvement in realism is possible i f the usual yield and catch equations are made to apply to individual age groups.  The total  yield from the fishery over (say) one year will then equal the sum of yields from each cohort in the fishable stock over the same time interval.  The modification also allows for variation in year-class strength  and the inclusion of a deterministic recruit function.  In this final  form the modified age-group model is suitable for analysis by a gaming procedure. In the pages that follow several empirical relationships and formulae are used without the support of mathematical derivation.  At the  142  time of their development the author was in search of rapid and approximate methods of analysis in which these objectives figured more prominently than mathematical rigour.  Subsequently, the necessary proofs  and derivations were carried out in f u l l .  They appear in the main body  of the report and appropriate references are given in this appendix. The gaming rules are modified in the present case to achieve particular objectives.  The investigator is forewarned that the following  assumptions apply: 1.  The fisheries model is deterministic and contains a Ricker 0968). type of recruit function.  2.  Information from the fishery is perfect or error free.  3.  The distribution of the target species, cod, is random.  Together  with assumption (2), all fishery and research samples are therefore fully representative of the stock age distribution. 4.  Fishing and natural mortaltiy occur simultaneously and act continuously throughout a fishing year.  5.  The trawl gear operates with "knife-edge" selection, and the average environmental effects over time is zero.  6.  The natural mortality rate, M, is constant for all age groups.  7.  The investigator is free to change the age of first capture [t ). and effort (f = boats) each year.  All fishable age groups are  equally affected by the effort, and therefore by the fishing mortality rate, F.  The catchability coefficient (q) is constant.  143  8.  In any year that fishing takes place a nominal number of research vessels ( f  = 10 boats) continuously sample all age groups less  than t . c 9.  Growth is density independent, constant for each age group, and f o l lows the von Bertralanfly (1938) equations for length and weight.  10.  In the year before fishing begins the single stock is at equilibrium and in a virginal state.  11.  Effort is restricted to the range: o<tVI000 boats.  Given these assumptions the objectives are expressed as follows: 1.  Determine the minimum number of years needed to estimate the parameter values in the yield and recruit equations.  Specify the i n -  formation needed to do this and its source, i . e . , fisheries or research data. 2.  Estimate the y i e l d , effort and t  3.  Equilibrate the fishery at MSY by means of a repetitive process of  c  co-ordinates of MSY.  yield prediction and experimental fishing. 4.  To avoid social and economic distress the boats assigned to the fishery by the investigator are "permanently" committed.  This is  an unusual condition, but i t imposes an additional constraint on the investigator. The interactive, simulation model is only responsive to the number of boats (f) and the value of t  £  specified by the manager.  Internally,  the program converts f into a fishing mortality rate that is used in computing the annual yield and catch from age groups equal to and older  144  than t c  Exit from the fishery occurs at age t' = 13.  At the end  of each fishing year reproduction is calculated on the basis of the residual sum of adult fish four^ years old and older.  Fecundity at  age is constant for all adults and i t is further assumed that larval mortality occurs in one instant in time.  The recruit function cal-  culates the.residual progeny which enter the fishery in the next year as age-group ' 0 ' . is three.  The exponent in the length-weight relationship  The von Bertalanfly constants are taken from Halliday (1972)  and apply to the Scotian shelf cod stock complex. of events for each fishing year (j)  Finally, the sequence  and the internal formulae are out-  1ined below: Sequence  Model input and calculations  -  (External operator:  1'(Start)  pre-select M and t  m  Specify f. and t J  2  c  F. = q f. and F' = q f , where f  = 10  Z. = F. + M and Z' = F' + M 3 -nK(i-t) -(.Z, +nK) Y. . = F (N ). . W» z %_e °_ (1-e ) n=o Z. + nK J  3  J  J  1 , J  4  C  i,j  " J o»i.J (l-e"^)/Z. F  ( N  (A2) (A3)  t'-l  t'-l i=  5  s  I  -z. G N  t i,j }  e  J  The age of f i r s t maturity (t ) is a pre-selected variable; l<t  <10. m  range:  145  Sequence  Model input and calculations a[l-S./S ] R  7  Print out requested information  8  Advance all age groups by one year, -Z.(orZ') 1  9  =cS.e  CA4)  6  j+1  - -  <Vi i, i  e  +  Set R j  10  J  +1  e  - (Vi,j  J+  e  J  equal to age group ' 0 ' , or (.N ) 0  0>  j  + 1  Return to start for the next fishing year.  The notation is standard except for the following: i = age group subscript j = fishing year subscript; j=o for virginal state S. = sum of reproducing stock at the end of the year 3  (N ). . ; o 1 ,J  (N ). . = number of fish in age group i ,j  i  at the beginning  and at the end, respectively, of the year (N )  =  Q  0  j+1  P 9 y calculated from S^ which will enter popula-  =  r o  e n  tion as age-group 'o' in the following year. S = equilibrium adult stock size, 1106 units, in the absence of fishing. g  Based on equilibrium, age-group 'o' starting value of 1191 units (R )e  (The starting population is calculated internally from -M  the recursive relationship, i.e.,  (N ) Q  1  . = (N ) Q  o  . e"  N^ ^ = +  M  e  = 1191 e " "  , where M = 0.28; = 900 units, etc.  28  Since reproduction takes place at the end of the year, at equilibrium (j = o) the number of adults i s ,  S = P e  1106 units) c = R / S = 1191/1106 = 1.0769 e  e  i=4  CN.). e~^ = z  1  Additional constants: W~ = 11,410 gm  K = 0.14  t  a = 1.895  = 0.07 yr. M = 0.28;  a pre-selected variable; range: 0,2 <^ M<£ 0.3  Method of Analysis:  Year One The f i r s t decision the investigator must make is to assign values to f and t .  A low, i n i t i a l effort is desirable because the reproduc-  tive capacity of the stock is unknown, and there is a social and economic obligation not to retract any boat committed to the fishery. choice of t  is also c r i t i c a l .  The  The f i r s t few years of fishing must be  viewed as an exploratory period during which important information on the system is obtained.  A high valued choice of t  least productive of information.  is cautious and  Alternatively, a low value of t  can  produce four, desirable results: 1.  provide data on a broad range of age groups,  2.  permit research to establish the age-of f i r s t maturity Ct 1, and estimate the age-group instantaneous growth rates, G., from fishing and research samples,  3.  initiate the "fishing-up" process for the surplus stock, and  4.  produce a strong recruit response in the second year of fishing provided t  < t . m  147  There are at least two risks that must also be considered and both argue in favor of a low, i n i t i a l effort. 3  The value of t is unknown m  and therefore the time lag between recruitment in the second year and the point at which these fish enter the reproducing stock is also unknown. In the interval their numbers will decline due to natural mortality and possibly fishing i f they are removed as juveniles.  The second point is  also related to the time lag, but concerns the decline in age-groups 1, 2, 3, . . . t  immediately following the first year of fishing.  The quant-  ity of fish in these age groups cannot exceed their i n i t i a l , equilibrium numbers.  They too will decline due to M and possibly F, and therefore  the fishable stock will be substantially reduced for an unknown number of years.  Recovery may begin when the f i r s t year recruits become mature,  but the degree of recovery will depend on the variable choices assigned to t  in the intervening years.  The manager must anticipate and guard  against these risks to both conserve the fish stock and protect the investment of fishing boats. In consideration of the above arguments, the levels of f and t for the first year are set to: f  1  = 50 boats  t  £  = 2 years  The only information needed from the fishable stock is yield in weight and catch in numbers by age group. less than ance (oy  t  Research data on age groups  are also available and are reported as indices of abund-  = C./f)  and biomass (V.'  = y./f* }.  The yield and catch data,  however, are used in several repetitive calculations and the model is programmed to carry out these operations directly.  The techniques used  148  in preparing the derived statistics are given below, (i)  W. The statistic W. is obtained by dividing the yield (Y.) by the  catch (C^). age group  It represents the average weight, by number, of fish of i  taken during the fishing year.  For further details on the  derivation and meaning of w\ , the reader is referred to Section IV, p. 37.  The values of W. for the fishable age groups can be used in an iterative, regression technique to provide an estimate of the length-weight exponent,  b.  The details appear in Section II  (3.2)  p34.  In the pres-  ent analysis, the above method was unknown and the f i r s t estimate of  b  was taken from the literature on cod (Kohler, et al_., 1970) and set equal to three. K, t , W and L ' o °° The length-weight relationship w(t) = a £ ( t ) can be rearranged 00  b  into an approximate expression involving  :  1 (W\)  b  ~=  (A5)  QL  where Q is a proportionality constant and L. is a statistic analogous to ¥ . , but refers to the average length, by number, of fish caught in age-group i.  From the known values of M^, W ^ , . . . ! ^ and b, the corres-  ponding products Q L ^ J QL^• - • QL-j2 are determined and used in a Walford (Walford, 1946;  Ford, 1933) line regression of QU. .| on QL.. +  The  149  slope provides an estimate of the von Bertalanfly growth constant K, A  and the intercept produces an estimate of QL  OT  in equation (A5).  s i m i l a r Walford regression of ln (:QLs> - QL^) on age (i + 0.5) an estimate of t .  It  A  yields  i s necessary to increment each age by the con-  stant 0.5 to represent half a year, the approximate time at which the weight and length of a f i s h reach t h e i r average values of W. and * (see Section IV, p.  , for an improved estimation technique),  h. The instantaneous growth rates are calculated d i r e c t l y from the equation:  G  - ~ i  =  b  l  • ^ " V n  (^  s  e  K(i-t v  -K (A6).  ) . o - l  The derivation of this equation i s reported in Section II  (4-3), p  The results of the above c a l c u l a t i o n s , together with the research and fishery values of IK and V^, are reported under the heading FISHING YEAR in Table A II (ii)  at the end of this appendix.  Estimation of the natural mortality r a t e , M. The assumption of an equilibrium condition for the virginal stock  permits an estimate of M to be obtained in the f i r s t year of f i s h i n g . Let R equal the number of progeny entering age-group ' o' in year the year p r i o r to f i s h i n g . '3'  j=o,  The number of f i s h in age-groups 2 ' and  at the start of the fishing year (j  l  = 1} w i l l then be:  150  CN )  = R e"  2 M  (N )_ = R e "  2 M  0  ?  , and  e"  M  (A7)  = R e'  ,  3 M  (A8)  o  since the number of fish in adjacent age groups are related by the ex-M pression ( N ) - | = (N )^ e Q  at equilibrium.  i+  Equations (A7) and (A8) can be substituted in the catch equation  (A3)  to y i e l d i n d i c i e s of abundance:  U  2  = q (N ) Q  = q R e"  2  2 M  (1-e  Z l  )/Z  (A9)  1  and U 3  q R e"  (1-e  3 M  ^/Z-,  where, Zj- is the total mortality  ,  (A10) rate in year j = 1.  Dividing equation (A9) by (A10), simplifying and taking natural  logar-  ithms: In U - In U = M 2  3  ,  or more generally, for the fishable stock where  i  takes on values  ( 2, :  12) In U. - In U . Equation (All) 0  + 1  = M  (All)  indicates that i f successive values of In  are plotted against age (i  + 0.05;  resulting slope w i l l equal -M.  (ordinate)  abscissa) in a catch curve, the  Again, the ages must be incremented by  0.5 because each U\ term i s an index of average stock size over the year.  The method is approximate and therefore the derived value of M  is only an estimate.  A straight-line relationship is obtained from the  data in Table A II:  ln U.  In U.  Age, i +0.5  Age, i + 0.5  ln (1132) = 7.0317  2.5  5.3519  8.5  6.7511  3.5  5.0689  9.5  6.4708  4.5  4.7875  10.5  6.1924  6.5  4.5109  11.5  5.9108  7.5  4.2341  12.5  5.6312  8.5 Slope  (iii)  0.2800 = -  , and M_  Estimate of the optimal age of f i r s t capture,  28  t  ^  In the absence of fishing the critical size (Ricker, 1968) is obtained at the point where the difference between the instantaneous growth rate (.G^) and M is zero. if  t  G.  values in Table A II,  The yield will be less than optimal  is less than or more than the critical size.  A = G 5  A  6  A  7  5  M  A search of the  for j = 1, indicates the following:  = 0.371 - 0.280 = + 0.091 = 0.290 - 0.280 = + 0.010 = 0.231 - 0.280 = - 0.049  152  Therefore, the f i r s t estimate of the optimal age of capture is t  Q p t  = 6 years.  Year Two An estimate of q is needed to convert effort into fishing mortality rate and allow the development of a predictive equation for yield. If the f i r s t year levels of . t .c  and  f  are maintained for a second  year it is possible to achieve both goals.  The cost of obtaining this  information can be equated to a temporary, future loss in catch from all age groups greater than 1 ' . 1  It is more important, however, to  be able to recognize the present position of the fishery and predict a future direction;  and i f need be, the depleted age groups can be a l -  lowed to recover by setting  t  c  = t  in year three.  The ability to  advance the age of f i r s t capture reduces the experimental risk and, with this assurance, the second year values are set to: boats, (i)  t  = 2 years.  =  50  The fishing results appear in Table A II,  Estimation of the catchability coefficient, q. The following relationship is valid i f the fishing effort is held  constant for two, consecutive years:  l n  if^ i+1, U  J+1  "  J \ J+1  (A12)  Equation (.A12) is analogous to (7\11) in that the logarithm of the ratio of two indices is equal to a mortality rate.  A catch curve based on  expression (A12) would be incorrect, however, because the indices refer  to different sampling years.  Calculation of the geometric mean (G.M.)  from individual ratios is indicated (Ricker, 1968), and the details appear below. Let the notation i , j/i+1, j+l Age-group  be abbreviated to i , i+1. In  ratio  i_ U  In Z,  ?  1 , d i + 1  - K i  2.3  M  = 0.3802  -0.9671  3.4  0.3795  -0.9689  4.5  0.3795  -0^9689  5.6  0.3812  -0.9644  6.7  0.3774  -0,9824  7.8  0.3789  -0.9705  8.9  0.3820  -0.9623  9.10  0.3776  -0.9739  10.11  0.3808  -0,9655  11.12  0.3837  -0.9579  154  Number of samples of Z-j , n = 10 2  1 11,12 G.M. = antilog z In Z, , = 0.3798 n 1,2 2  and Z  1  2  j  3  = 0.38  An estimate of q and F is obtained directly:  Z  1,2  = F  1,2  +  A  =  q  f  = q (50) + 0.28 and solving, and  (ii)  l , 2  +  A  ,  q = 0.002  F.j  2  = 0.10  , for f = 50 boats  A predictive equation for yield. The average weight of a fish at time t can be approximated by the  exponential growth formula (Ricker, 1968): w  (t)  i  = (w )  G.t e 1  i  ,  for o<t<l ,  (A13)  and the decline in age-group numbers by, -(F+M)t  N (t) = (N ) e i  0  i  J  = CN ) e Q i  -Z.t J  ,  (A14)  where, (w )^ = the i n i t i a l , average weight at the beginning of the time interval (t = o). Over a year of continuous fishing, the age-group yield is obtained by integrating the general expression for biomass:  155  Y  i,j  =  F  j <l i , j B  C  t  )  d  t  = F. f] N. Ct) w. Ct) dt  ,  .  . t(o,l)  (A15)  Substituting equations (A13) and (A14), into (A15), integrating and rearranging: Y  ,  i,j  = F  (G. -  j <o>1 <"o>1 o N  7  (G. = F. ( B J . . (e b  e  Z.)t  '  1  dt  - Z.) 1  i  Z  A. - 1) = F, ( B J . , (e  J  j  - 1)  1  A  (A16)  i  or, to  \  (BJ-i  '  J  -i  Y. 1  .-. A. i / Ai  'J F. J  =  Ce  1  =  V. .  ,J  i  q  n  - 1)  (e  A.  A. i  /  >  A  1  7  x  (Al7)  - 1)  1  where, (B ). . = the i n i t i a l biomass at t = o, and o i ,j V.  • = an index of biomass (Y.  The biomass at the end of year j , (B ).  ./f.). ., is derived from equation  + ' >J (A17):  ( »i,r ;B  ( B  +  But (B ) +  o'i,J  j = (B ) i  i  0  i +  j  e A i =  ^i  —  —  r  <  —  > and the yield in year j+1 can be predicted  : + 1  by integrating the following equation: (G  -i  Y  i  +  l  j  j  +  1  = F.  + 1  (B ) o  i + l j j + 1  /Je  Substituting equation (Al8) for CB )^ i 0  and integrating:  A18  +  i + 1  1 + 1  J  + 1  - z. ^  At  dt  in the above expression  >  Y 1  i + 1  +  1  ,  0  \l  = F  i + 1  1  J  A  i  (e  q  l  The total y i e l d , (-Y J  ^sr  , from a l l  T  biomass of pre-recruits in year j+l.  j  - 1).  1 + 1  (A19)  "AT——  fishable age groups must include the that w i l l enter the fishery in year  The generalized, predictive y i e l d formula i s ,  (v )  T j+1  -  FJ  +1  V^i  6 (  t -u-M)  G t ( e  c  (1-e  c"  )  c  therefore:  Z  j  +  1  -  n  c  11 +  i+l=t +1 c  Equation (A20)  Y  i  +  1  '  j  +  ( ^  1  A20  can be suitably modified to allow for a change in t  the discretion of the investigator, i . e ,  t  . , f C,J  program i s contained in the subroutine PREDY. y i e l d in year j + l , Table A II  T  I  t  ..  at  The entire  C, J  To predict the total  the necessary values of (G. and V . ) . are taken from  and combined with the previous estimates of q and M.  At the end of two years of f i s h i n g i t i s necessary to assume that the analysis of fishing and research samples have led to an accurate measure of t , the time of spawning and the correct values of K, W^, b, t  o  and G . . i  In Table A II  the reader w i l l note that estimates of these  parameters are prepared by the simulation program in year three and thereafter.  The recorded values are now known to be inaccurate, but  they are included in the table to i l l u s t r a t e real values as e f f o r t i s increased.  In a l l  the departure from the . . future calculations and  predictive subroutines the following quantities w i l l apply:  TABLE  All.  R e s u l t s of s i m u l a t e d , experimental f i s h i n g . M o d i f i e d Beverton and H o l t (1957) y e a r - c l a s s model- c o n t a i n i n g a R i c k e r (1968) r e c r u i t f u n c t i o n . Legend: i = age-group s u b s c r i p t ; j = f i s h i n g y e a r . s u b s c r i p t ; V . , = y i e l d per u n i t e f f o r t ( g m / f ) ; U . = catch per u n i t e f f o r t ; W. = average weight by numbers (gm); ' G, = instantaneous growth r a t e ; f = effort (boats); t = age f i r s t c a p t u r e ; K, W^, t = estimated von B e r t a l a n f f y (1938) growth equation c o n s t a n t s ; R = '0' age group progeny. Note: Values o f W . . G . , K, W and t o n l y change w i t h a change i n e f f o r t ( f ) . i  J  1  c  1  j =1  1  0  AGE GROUP  R 0  oo  1  2  3  no  306  534  1 132  .855  10  11  12  5  6  7  8  9  719  833  873  854  795  712  619  526  .646  .489  .369  .279  .211  .159  .120  .091  .069 6375  .4  50 2  f v  i,j  b.i w  G  2.058  1 555  438  i,j  2.230  i,3  !il.J • i  2.375  3062  3771  4469  5142  5780  1 052  .687  .371  .290  .231  .188  .154  .128  .107  108  300  483  651  753  790  773  719  644  560  476  397  f  1 524  1 109  .774  .585  .442  .334  .253  .191  .144  .109  .082  .062  fc  117  308  494  613  578  608  596  555  498  433  368  307  1 652  1 129  .788  .550  .343  .259 .  .196  .148  .112  .085  .064  273  627  1116  1687  2346  3043  3751  4450  5124  5762 .108  W  G  CO  i.j  U  i.j  2.634  V  i,j  2.656  2  .048  K  0.137  6359  W  11,574  00  .683  .490  .371  .289  .231  .188  .155  124  334  525  651  567  392  384  358  321  279  237  198  1.760  1.224  .837  .584  .336  .167  .126  .096  .072  .055  .041  .031  138  355  568  692  602  384  248  231  207  180  153  127  1 .951  1.303  .907  .620  .357  .164  .082  .062  .047  .035  .027  .020  j =5  c  219 5  1 042  j = .4  50  f  .129  i  V  0.061  2365  j =3 V  v  1704  .492  j =2  U  W  1113  i  i,3  0.138 11,512  624  271  i  V  K  fc  0.039  f  219  o  5  f  219 5  TABLE  A l l (contirfued) R 0  AGE GROUP 6 7  1  2  139  394  606  750  639  408  243  149  133  116  99  82  1.968  1.445  .966  .672  .379  .174  .080  .040  .030  .023  .017  .013  K  .0.137  273  627  1116  1687  2346  3043  3751  4450  5124  5762  6359  W  11,574  •1.042  .683  .490  .371  .289  .231  .188  .155  .129  .108  3  4  5  8  9  10  11  12  j =6 i.j  V  U  i,j  *i i  G  f 2 648 - • -  -  219 5  CO  -  *<>  0.039  j =7 _  i.j  V  i.j  U  139  398  672  799  693  433  258  146  86  75  63  •53  1.962  1.458  1.071  .715  .411  .185  .085  .039  .019  .015  .011  .008  139  396  677  886  694  442  258  146  79  45  38  32  2. 657  1.964  1.453  1.080  .793  .413  .189  .085  .039  .018  .009  .007  -  -  273  627  1116  1680  2339  3035  3743  4442  5117  5756  -  -  1.040  .683  .490  .371  .290  .232  .188  .155  .129  .108  -  139  397  675  893  758  406  241  134  73  38  21  2. 652  3 =8 i,j  V  i,j  U  W  i  G  i  i,j  U  i,j  W  i  G  i  219  tc f  -  5 285 5  .005  t K  0.137  6353  W»  11,594  c  -  0.034 f  j =9 V  f  18  302  t K  c  5  2. 657  1.968  1.455  1.077  .800  .452  .174  .080  .036  .016  .008  .004  .003  -  -  273  627  1116  1678  2337  3033  3741  4441  5115  5754  6352  -  1.039  .682  .490  .371  .290  .232  .188  .155  .129  .108  -  139  398  676  890  764  436  218  123  65  34  17  9  1 969  1.458  1.078  .798  .455  .187  .072  .033  .015  .007  .003  .002  139  398  677  892  760  438  233  110  60  31  16  8  1.080  .799  .453  .188  .077  .030  .014  .006  .003  .001  K  0.137  W»  11,599  0.137 11,598  -  0.033  j = 10 V  U  i.j i,j  2. 657  i,j  U  i,j  W  i  G  i  302 5  f  j = 11 V  f  2. 657  1 969  1.458 '  -  -  273  627  1116  1678  2336  3033  3741  4440  5115  5754  6351  -  -  1.039  .682  .490  .371  .290  .232  .188  .155  .129  .108  -  t  304 c  *0  5  0.033  TABLE A l l (continued) R 0  1  2  3  . 4  5  AGE GROUP 6 7  8  9  10  11  12  j = 12 V  U  i,j i.j  2.657  139  398  678  893  761  435  234  118  54  28  14  7  1.969  1.458  1.080  .800  .454  .187  .077  .032  .012  .006  .003  .001  f  304 5  j = 13 V  U  i.j i,j  2.657  398  678  893  763  436  232  118  57  25  13  6  1 .969  139.  1.458  1.080  .800  .455  .187  .077  .032  .013  .005  .002  .001  139  398  678  893  763  437  233  118  58  27  11  6  1 969  1 458  1 080  .800  .455  .187  .077  .032  .013  .005  .002  . .001  f  304 5  j = 14 V  U  i,j i,j  2.657  f  304 5  160  (iii)  K = 0.14  W*, = 11,410 gm  t  b = 3.0,  and from equati on : (A 6 ) ,  = 0.07 years  Q  i  G. i  ,  1  .1.9873  7  0.2301  2  1.0543  8  0.1867  3  0.6877  9  0.1533  4  0.4917  10  0.1272  5  0.3704  11  0.1064  6  0.2886  12  0.0.895.  G. i  i  Estimation of the r e c r u i t function parameters There are two methods of estimating the parameters R and S g  the Ricker (1968) recruit R  Vl7V 5  a  ^  g  in  function,  ~  (A21)  e  The f i r s t procedure makes use of q to express S  g  in absolute terms, and  the second leads to a calculation in which the parameter S variables S. and R are given as functions of q.  g  and the  I n i t i a l l y , both tech-  niques are used interchangeably, but in developing the f i n a l the prediction equation the l a t t e r method i s preferred.  form of  The fishery  information, measured as abundance i n d i c i e s , can then be used d i r e c t l y . A size index of the reproducing population at equilibrium (S ) may be calculated from the knowledge that t place at the end of the year.  = 4, and that spawning takes  Expressed mathematically,  it  follows  161  that:  o r , replacing the age-group numbers by t h e i r respective indices after the f i r s t year of f i s h i n g ,  S  Z1  = p  1  (  - Z  h  (1-e  2  . + U  E U.  5  n  i o,  .)  (A22)  i  q  The only unknown in equation (A22)  is the index  An estimate of  y  this quantity can be prepared from the s e r i a l relationship that exists between adjacent age groups, namely:  U  U  i,l  =  CA23)  r  i 1, 1 +  The geometric mean value of 1  r  U  1  s  G.M.(r) = a n t i l o g l  ln , i , l  1-5  n  is calculated as follows:  <U.+1,1  = antilog 1 In ,0.489 7 0.369 l  and, from equation (A23)  Uio I J >  x  0.369 0.279  0.09K 0.069  =  ,  ,  i = 12, 1 = 0.069 = 0.052 1.323 U  1  Substituting known values in equation (A22), where Z^ = 0.38, (I  U  l f l  '  ;  +  U  1  3  f  l  l  = 1.839,  and  162  S  or,  =  0-38  (1.839) = 2.210/q ,  S = 2.210/.002 e  =  1105 f i s h  S i m i l a r l y , R can be expressed as a function of .q: g  R  VLL  = e  Z' _2 - .  q  and  = 2.058 (0.30)  (1-e" ')  4  7  Q  n  = 2.382 P  _ *-- \ 30  ( 1 - e - )  R / S =1.078 e e  An estimate of parameter  a  (equation A21) can be prepared from the  general p r e d i c t i o n r e l a t i o n s h i p : a(l - Sj/2.210) R , , , = o,j+1  J  n  1  = 1.078 S , e  V  U  J  -7'  (1 - e ) Z ' / ( l - e " ' ) = 0.30/(1 - e " - ) = 1.157 q  where,  Z  Z  and,  ,  3 0  a ( U  ,  q  l  - Sj/2.210)  o , j + l = 0,931 S , e q  (A24)  q  J  The variables S . can occur in two forms depending on the r e l a t i o n s h i p between t  and t  c  .  m  1 2  S. = J  and, f o r t  -  t  For t  < t^ : c — fm  II . = E i,j t q m  Z  1 2  E (,N,). 1  m  u  ,  J  e  j Z. j _  ,  (A25)  1  > t c m  1  (, ^  —fr-  +  }  *Li  ^  —  )  (A26)  163  Using the data from Table A II  for the f i r s t year of fishing, and  equation (-A25) , S, = (0.646 + 0.489 + . . . + 0.069)  (0.38)  Ce" - 1) 38  1 = 2.000 , q  .  q  and U  o,2= 2.230/q q  Substituting these values into equation (7\24), rearranging and solving, a = 1.898  The general, predictive equation for the index of recruitment is therefore given by: 1.898(1 - S ./2.21) U  o J + ]  = 0.931 q S. e  q  ,  (A27)  where, S. can take either of the two forms indicated by equations (J\25\ 3 and (A26).  The variable  q  in the above expression cancels..when S. is  given as a sum of indicies divided by  (iv)  q.  A refined estimate of the age of first capture From the estimates of G., -M and  q  i t is possible to construct  a constant-recruit model to determine the optimal age of f i r s t capture. The procedure is contained in the subroutine AFC0PT, which will be described shortly. age-group effort (Vj)  '1'  For a given starting value of pre-recruits entering , the model generates data on the total yield per unit  and total yield per pre-recruit (-Y/R)j for t  val (2, 12), and for F values in the range (-0.10, 1,50),  in the interFrom the  total amount of printed information i t is easy to select the best age [t  .) that will maximize V or (Y/R),, opt i i T  A change in the number of pre  recruits only introduces proportional changes in Vj or CY/Rl ; T  the  optimal age of capture remains unaffected and for this reason the calculations are only carried out once.  The precaution is taken to re-  place the natural mortality rate MI by V (.= 0.30) for all age groups less than t .  During the gaming process these same age groups are  continuously sampled by research vessels and effectively produce an added mortality rate of 0.02. In Table A II the index of pre-recruits in age-group ' V during 2 the second year of fishing (U^) ' i s 1,555. mortality the starting biomass for i  In the absence of fishing  in the interval {2, 12). is read  i l y calculated from the relationships: N = U,/q  B  n  l  =  N  l  W  l e  G -Z n  and  , (B ) = (B ) Q  2  Q  e  1  =B -tG,-Z')  (1 - e < o3 B  }  =  < o>2 B  6  - Z  < o>12 B  =  (Vn  'The subscript  6  j  is temporarily removed for clarity.  - 1  165  For a given value of t (B ). 0z  and F, the appropriate starting biomass,  , is selected from the above listing and the following yield  c  calculations are performed: Sequence 1. Start  Comments and Calculations Select t  = 2, F = 0.1,  and (B )  2  from l i s t  2.  Calculate R = uyq = 1.555/.002 = 900 units  3.  Set AZ = G - (F + M)  4.  Calculate individual yields for fishable age-groups, t  2  2  using (B )  2  as the equilibrium starting value:  2  Y = F (B ) Q  2  (e  ^-1)/AZ  AZ AZ, Y = F (B ) e *(e 9  Q  :  2  :  =  ii X AZ  12  Y  5.  =  o2  f{B  ]  2  - D/AZ^  J  3  (2, 12),  q 2  (  e  i  A  z  1 2  '  1 ) / A Z  Calculate total yield (Y ). T  1  12  12 F and Y /R from Y = I Y. T  T  V  t  6.  Print results from sequence (5.).  7.  Return to Start, increment F by 0.1 and repeat.  1  c  return to Start, reset F = 0.1, increment t  When F = 1.5,  by one and sel-  ect appropriate starting value of (B ) . Stop when calculations for t = 12 are completed. +  8.  A synthesis of the Y^/R output from the program AFC0PT appears in Table A III.  Only the data for age groups in the region of t  ^ is given.  Age group '6* appears to offer the greatest potential Y /R , but its T  TABLE A I I I .  Total y i e l d per r e c r u i t (Yj/R) associated with age o f f i r s t capture ( t ) and f i s h i n g m o r t a l i t y rate ( F ) . A Ricker (1968) constant recruitment model. Maxima are underlined. c  Total Y i e l d Per Recruit (gm/R) Age o f F i r s t Capture ( t ) c  F  4  5  6  7  0.1  263  240  210  175  0.2  380  361  326  281  0.3  431  424  393  346  0.4  453  456  431  387  0.5  462  474  455  414  0.6  463  484  470  432  0.7  462  489  480  444  0.8  460  492  486  453  0.9  456  493  491  460  1.0  453  493  494  465  1.1  450  493  496  469  1.2  447  493  498  472  1.3  444  493  499  474  1.4  441  492  500  477  1.5  438  491  501  478  167  maximum value, as well as MSY, will not occur until  F approaches F°°,  This feature is a characteristic of constant recruitment models having a fixed natural mortality rate.  To avoid this difficulty and introduce  a practical compromise fishing should begin one year earlier, and accordingly t  j. is set at age five.  The maximum Y^/R in this age group  demonstrates a broad plateau at 493 gm/R.  It is sufficiently close to  the relative value of MSY (.501), and the mortality rate is also held in the realistic range of F = 0.9 to 1.3.  (v)  Estimation of optimal stock size and effort at maximum recruitment In Appendix B, a method is developed to calculate the stock size,  recruitment and fishing mortality to yield MSY for the particular fishery described here.  The exact results indicate that at MSY the ex-  pected yield is 508 Kg, recruitment is 1532 f i s h , and the optimal effort is 371 boats.  Because of the high value of  a  in the recruit function  the yield at MSY is only about one percent higher than that obtained at maximum recruitment, i . e . , effort needed at R max  R = 1538 and Y = 503 Kg., but the max i is 22 percent lower (303 boats). In this case m=v  T  it is not economically justified to stabilize the fishery at MSY.  For  this reason R is chosen as the target value of recruitment in the max current problem. If the age of f i r s t capture is theoretically set equal to the age of f i r s t maturity (4) then equation (A25) is descriptive of the reproducing stock size S.. Let U . . . be the index of the average, mature j mat,j stock size. Then i t follows that the predictive recruitment equation  168  (A27) can be rewritten as: 1-898 (1 - ^mat.j U  o,j l  °'  =  +  9 3 1  U  mat,j  e  ^ j  l  -  e  ^ j  ) ,^  Z  -  2  Q  )  ~"  Ce^-11  - 1  Maximum recruitment occurs where the slope of the stock-recruit curve is zero. 2  Therefore, differentiating equation (A28) with respect to (-IL. . mat,j  Z./e  - 1), setting the result equal to zero and simplifying:  J  Z  j  2  "  e  '  2 1  °  (e^ - 1)  C  2  '  2  1  0  "  '  J  Z  j  1  In the above equation the term in the square brackets must equal zero, and solving for li . .: mat,j 3  U  m a t > j  = 1-164  e ^ J .  WZ91  3 i.  Substituting this value into equation (A28), the (Z./e  J  - 1) terms can-  eel and i t follows that: U  O  J  +  -  1  2.661  ,  and, the maximum number of recruits is then equal to: R  =  U  o,j+l q  Z'  ( 1  _ -V e  Y  = 2,661 .002  (0.3) = 1539. fish ^ ^  0 > 3  If the fishery is removing the surplus stock at a rate sufficient to produce R  max  in successive years, then the reproducing stock is at equil-  ibrium and at its optimal value ( S ^ ) . -  The derivative (dR  max  /d S p ) Q  t  .  3  169  is therefore equal to zero.  Substituting the values for R / S Q  g  , and S  g  (1105 fish) in equation (A21), i t follows after differentiating with respect to S  that;  Q p t  S „. = e = 1105 = 583 fish T 17898 S  o p t  With the theoretical age of first capture s t i l l set at four the terminal stock size S ^ must be equal to the sum of the residual, fishable stock, i . e . ,  But at maximum recruitment ,  <  N +  )  (N ) +  = 4  (  N 0  >4  ^  6  =  max  R  e  =  4  6  4  ^  -I. -21. = (N ) e = 464 e J  5  ^  +  J  4  ! = : -91. ( N ) = 464 e , J  +  12  and summing to obtain S  S  AF.A nopt n t == 464 (e(a  opt '  -Z, J  -91. + e ) = 583  -2Z,  + e  J  +...  j  The value of Zj that satisfies the above equation could be determined by Newton's method of solving transcendental equations (Smail, 1949).  How-  ever, in the present case, trial and error methods are equally quick, . and the solution is Z. = 0.583  ( f = 152 boats). :  Substituting this value  of Z. in equation (A29), the catch per unit effort index of the mature J  population that will produce R  max  in the following year is obtained  directly: IJ m a t  ..= " 3  1.164 ( e ' - 1) = 1.580 0.583 5 8 3  Year Three The sum of the catch per unit effort figures for the mature stock (U  j.) at the end of fishing year one is 2.434 and in year two i t is  2.203.  (Table A IV).  Both figures exceed the optimal value of 1.164  needed to produce R . max r  The age structure of the population in year -  two, as revealed by the IK figures, indicates both the source of the high U . value and the potential means of correcting the situation. mat 3  In its present state the population is unbalanced.  There are too many  adult fish aged five and over, and not enough in the age-group range of two to four due to the previous fishing strategy.  In addition, the  increased index of recruitment in year two ( U Q ^ = 2.230). will not begin to benefit the reproducing population until fishing year six.  The  effort in the fishery must be curtailed until this event occurs. The effort to be used in year three is based on the level needed to ensure maximum recruitment in year four, provided all mature age groups are theoretically fished.  If this same effort is actually ap-  plied to age groups five and over in the simulation model, three purposes are accomplished. year of recovery;  First, age group '4' is allowed an additional  second, the risk of having to remove excess boats  from the fishery at some future time is reduced; abundant fish in age groups '5+' are removed.  and third, the over  171  The number of fish entering the adult population at the beginning of year three i s : 12 E (Nj. 4 0  1  ,  J  11 , = z (N ). 3 ^  7 ^2 Z (e - 1)  li E i,2 3 q  1 1  =  9  u  1198 fish  9  2  For maximum recruitment the desired number of fish at the end of year three should not exceed S „. therefore it follows that: opt 12 12 -Z„ -Z, E (NJ. - = E (NJ. e = 1198 e = 583. J  4  '  4  u i  5  J  o  and solving: Z = - ln (583/1198) = 0.72 3  and f  3  = (0.72 - 0.28)/.002 =  Substituting the values of V.  219 boats. (Table A II)  9  for  i  in the interval  (1, 12) in subroutine PREDY, the predicted yield is 865 kg for t From this point on the value of t  = 5.  will not change and the predictive  recruitment index in year j+1 can be rewritten and abbreviated as follows : From equation (A26), S  Ii 1 q  =  u (  4  j  , ' z  +  q t *>3 7 7 7 r  e  •  E  1 2 u  i  j  Z  5 ' 1  J ) Z  J  e  1  J  - 1  12 = 1 ,0.8575 U, . + E u. . y j , q ' 5 ' Z. e - 1 ?  z  l  1  4  J  J  ;  J  \  The numerical value of  q  is not needed in this equation.  It cancels  when S./q is substituted into £A27) to calculate U . , , . j o,j+l Similarly, the predictive formulae for the total mortality and effort in year j+l 1.  to produce R ^ in j+2 may be abbreviated to:  11 = - In (408 - 317.62 IL .)/(428.74 U, . + _ J _ S U. ,  7 Z  r  e and,  f\  +1  = (Z  j + 1  j  )-, - 1  J  - 0.28)/0.002  The actual and predicted total yields and indices of recruitment for fishing years three to fourteen are summarized in Table A IV.  The  formulae given above are used in each case and the detailed calculations are not reported.  Years Four to Seven The effort required in years four to seven to produce R  mav  in the  following year are: f  4  = 194 boats  f  5  = 140 -  Each value is less than the committed level of 219 boats, and the adult population from age group '5' onward is purposely overfished during the interval.  At the same time the recruitment index Ui . = 4 to 7) obtains o,j  a maximum value of 2.657;  (the difference is negligible and is due to  rounding error in the computer).  The reason why the expected recruit-  ment is not lower can be attributed to one major cause:  at the maximum  effort difference of 219 less 140 boats the reproducing stock is driven  173  TABLE A IV.  Predicted t o t a l y i e l d and recruitment Legend:  Y^ = t o t a l y i e l d (Kg); U  Fishing Year J  T  = catch/effort,  recruitment;  . = c a t c h / e f f o r t , mature stock.  Values Actual Predicted y y 'T  index.  Values Actual Predicted U U U  R  U  R  Effort f (Boats)  Age First Capture  Total/. Mortality  t  Zj  c  u  mat (optimum 1,580 )  1  361  -  2.058  -  50  2  0.100  2..434  2  328  -  2.230  -  50  2  0.100  2. .203  3  864  865  2.375  2..361  219  5  0.718  1,.805  4  600  605  2.634  2,.637  219  5  0.718  1..509  5  468  471  2.656  2,.660  219  5  0.718  1..413  6  410  412  2.648  2..653  219  5  0.718  1,.428  7  397  397  2.652  2,.656  219  5  0.718  1,.489  8  496  499  2.657  2..660  285  5  0.850  1.,559  9  511  514  2.657  2..661  302  5  0.884  1..572  10  504  506  2.657  2..661  302  5  0.884  1..571  11  505  508  2.657  2,.661  304  5  0.888  1..570  12  503  506  2.657  304  5  0.888  1 .571  13  503  505  2.657  -304  5  0.888  1 .572  14  503  505  2.657  2. .661 2.661 2.661  304  5  0.888  1 .572  -  174  to the left of optimum on the stock-recruit curve, but i t is s t i l l within the region of S  t  -  Years Eight to Twelve The strategy of advancing the age of f i r s t capture to five initiates, in year four, a continuing recovery of the first reproducing age group. By fishing year eight the increase in the terminal number of four-year olds is sufficient to allow the second, major increase in effort to 285 boats.  In subsequent years the effort is incremented to insure maximum  recruitment and the final effort of 304 boats is introduced in year eleven.  A second estimate of the maximum effort is calculable at the  end of fishing-year nine for at that time maximum recruitment has stabilized all age groups from '0' to '4'. From Table A II,  the equilibrium catch per unit effort of four  year olds is 0.800 (U^ g).  The terminal number of fish at this age is  therefore: (N ), Q = 0.800 q ^ +  Je  = 0.800 -002 - 1  (0.3) = 343 fish 0.3 , e - 1  V 1  i  At equilibrium, the optimal reproducing stock is: 12 S  opt  =  3  4  3  +  z  l 5  (  N  } +  i,9  -Z 343 + 343(,e  0  =  5  . p  t  8  3  -2Z + e  and solving by trial and error,  o p t  ,  -8Z + ... + e  o p t  . )  = 583,  175  1. = 0.888, opt  and  f  . = 304 boats, opt  The data in Table A II  indicates that for practical purposes the  introduction of the above effort stabilizes the fishery at the economic level of MSY CMSY } by the end of year twelve. Conclusions: The objectives and assumptions made available to the investigator are detailed at the beginning of this appendix, part 2, tions the one  Of these assump-  concerning perfect information from the fishery is prob-  ably the least tenable, but i t is included for a specific reason:  under  ideal conditions the minimum, standard time needed to analyze the modal can then be determined.  The information is useful as a reference point  in determining the conditions that prevent an equivalent analysis in more realistic models. At the completion of this exercise the following conclusions may be drawn: (1)  Given the i n i t i a l assumptions, estimates of all the unknown para-  meters can be prepared within two fishing years.  Armed with this in-  formation the necessary steps to reach MSY in a controlled manner g  can then be undertaken.  In this sense, what appeared i n i t i a l l y to be  a problem in fishery population dynamics, is recast as an issue involving parameter estimation. formation are necessary:  To prepare the estimates two sources of inthe yield in weight and catch from both the  fishable and pre-recruit age groups.  Whereas the former data is help-  ful in assessing the present state of the stock, yield prediction and  176  parameter estimates for the recruit function cannot be prepared unless pre-recruit information is available.  Ultimately, the parameters of  the von Bertalanfly (.1938) growth equation must be obtained from research sample analysis. (.2)  If the age of first capture is set equal to six, the critical size,  the fishery must remove all fish of this age and over to maintain maximum recruitment.  Provided there is no pre-recruit, sampling mortality the  real value of MSY will equal the biomass of fish reaching this age, or 584 kg.  With sampling mortality of pre-recruits the potential MSY is  reduced to 518 kg.  Both conditions require the application of an in«v  finite fishing mortality rate.  For practical reasons the age of first  capture is therefore reduced to five, and under this condition, the maximum sustainable yield (MSY ) is limited to 503 kg.  The fishery .  can be stabilized at this production level in twelve years with a maximum committment of 304 boats (-F = 0.89). (3)  The parameters of the growth equation and instantaneous growth  rates can be estimated in each fishing year.  The necessary information  is limited to age-group values of W. and a separate estimate of the length-weight exponent.  At the stage of development shown in this ex-  ercise, however, the accuracy of the estimation technique is restricted to low fishing mortality rates. The statistic -W-. also shows strong potential as a 'state' indicator. The values of W^ associated with each age group differ substantially, but they are also individually response to changes in the total mortality rate.  Potentially, this rate could be estimated from W. statistics i f  the analysis is limited to the fully fishable age groups.  In addition,  177  since  represents an average weight by numbers, i t s value i s independ-  ent, of year-class strength and could therefore be used as a diagnotic aid under non-equilibrium conditions. (4)  The a n a l y t i c a l method provides f o r m u l t i p l e estimates of  obtained during the development of the f i s h e r y .  q  to be  A l l that i s required i s  that the f i s h i n g e f f o r t be held constant f o r two consecutive years.  By  t h i s procedure the parameter may be updated at the d i s c r e t i o n of the i n vestigator.  In the present a n a l y s i s the o b j e c t i v e of s t a b i l i z i n g the  f i s h e r y i n the shortest time only allowed a s i n g l e estimate of be prepared.  q  to  178  APPENDIX B Derivation of optimal stock size, recruitment and fishing mortality at MSY in a multi age-group, single species fishery.  To calculate the absolute value of MSY in a multi age-group, single species fishery, two questions must f i r s t be answered.  What is the  best, average size at which to capture the f i s h , and how many of them should be taken in order to generate the maximum yield on a sustained basis?  The critical size (Ricker, 1968) is the solution to the f i r s t  question;  the average size at which the gain and loss in biomass due  to growth and mortality, respectively, are in balance.  A solution to  the second problem also involves a balance, but the conflict is between the number of above critically-sized fish that are harvested and those that must be left behind to maintain reproduction.  At MSY, this balance  is achieved at the optimal, reproductive stock size (-S^) support the highest, equilibrium yield.  that will  The fishing mortality rate at  MSY is in turn a function of S j., the stock-regeneration function and the time during the year when reproduction takes place.  Given the f o l -  lowing assumptions about the biology of the species and the conduct of the fishery, the objective is to determine S (R p ) and fishing mortality rate Q  t  1.  optimal recruitment  ^.)., and the value of MSY.  Natural mortality (M) is constant for all age groups and F  affects all ages equally.  The fish Cor fishermen) are distributed at  random at all times, and growth and mortality are continuous and concurrent. 2.  Growth and natural mortality are density independent. Reproduction takes place at the end of the year (as for the  179  and according to the Ricker 0 9 6 8 ) . r e l a -  cod stock in Appendix A(2) tionship  R  =c  j + 1  S j  a(i . e  -sys } (,B1)  and a(l  R  - S./S )  j+1 ,= c e d S.  J  d  where Rj -|  =  +  e  0  - a S./SJ J  (,B2)  e  the number of fish entering age-group '0' of year  at the beginning  j+1,  c = a constant,  R e  /S J E  where R  g  and S  g  are, respectively,  equilibrium progeny and stock size in the absence of  the fish-  ing S. = the reproducing stock size at the end of year  j.  All  adults  are assumed to be equally fecund. 3.  The c r i t i c a l  s i z e , or age in this case, has been determined, •'<  but f i s h i n g actually begins at some e a r l i e r t i c e f f o r t rate.  The age of e x i t from the fishery is t ' ,  of age and time are the same, i . e . , i s "knife-edged". onmental effect  age (-.t ) to maintain a r e a l i s -  years.  and the units  Selection of the trawl gear  The fishery i s a closed system and the average  envir-  over time is constant.  Under these conditions the annual, equilibrium y i e l d (Y ) may be w represented by the general equation Y  = F  But N (t)  Hit) w It) dt. c -Z(t - t ) = R' e ,  (,B3)  A'  c  t  1 t < t"  ,  or  180  -Mt N(t)  -Z(t - t )  = Re  e  where R' = the number of recruits entering the fishery at time t , and c  i t is also equal to the equilibrium progeny,  R,  times the  accumulated mortality rate up to time t c  Substituting these relationships into (C3): -Mt , -Z(t Y = Fe R / J e C  w  t) c  w (t)  dt  or -Mt  Y = e  RQ  c  w  where  (B4)  Q = F times the integral. At MSY, the rate of change of yield with respect to fishing mortality,  d Y / d F , must equal zero, and differentiating w ^w = e dF  c  (B4),  ,RdQ+QdR.dSx=0 dF dS dF '  (-B5)  K  -Mt  Replacing R by (Bl), dR/dS by (-B2) and canceling the term e  c  , equation  (B5) reduces to S dQ_ + Q (1 - a S dS = 0 dF S dF e Q  and dividing by SQ, dQ_ + J - a \ dS = 0 Q dF S S dF l  (B6)  j  e  The above equation can be used to test whether the choice of are such that the yield will be equal to MSY. a residual (y) that will differ from zero.  S and F  If not, (B6) will provide  181  The differential  dS/dF can be expressed in three different ways  depending on the relationship between the age of f i r s t capture it \ and the age of f i r s t maturity C.t ). Each case is developed separately, as fol1ows: t „c  tm  >  m  The reproducing stock size at the end of the year is equal to the terminal number, (N ).> of adult fish in each age group +  i.  Expressed mathematically i t follows that:  S =  C  (N  Z  -  M  (  ).  T  ™  *  +  (N ).  I  "  -M  S = Re  (1 + e  -Mt + Re  c  7  {e~  M  +e  ? 7  + e  L  d  +  L  -a m  + ... + e  -"tVtj. - i) c  m  -Z(t'-t) ... + e ) c  (B7)  -M  The f i r s t summation of terms involving e  may be collapsed into a simpl-  er form as follows: Let x = e  and . T = 1 + x + x^ + . . . + x  m  ]  then  (t -t ) - xT, = 1 - x  and T ]  = 1- x 1-x  c  (t -t ) c  m  =  m  1- e ]  _  E  -M(t -t ) c  m  -M  Similarly, the second summation may be abbreviated to  182  T  -Z(.f Q - e  7 Z  2  = e"  - t ).  1 - e" If  and T  2  -ZCt'-t )  ^L) = 1 - e  c  e  Z  Z  - 1  are substituted into (B7), and  R is removed as a  common term, then (B7) gives S = R (T  -M(t e  ]  or, replacing  )  -Mt + T e )  x l m + l  (B8)  c  2  R by equation (Bl),  a(l - S / S J -MU S = c Se (T-, e  .)  m+1 m + 1  Cancelling  S,  + T  ?  e  -Mt c  ).  (B9)  taking natural logarithms, rearranging and solving  for S, S = je In c + S + In a a  e  m  1  + T e  c  2  )  Only the last term in the above equation is a function of differentiating with respect to  F,  F.  Therefore,  the f i r s t and second terms equal  zero, and with the appropriate substitutions for T-j and T , it can be 2  shown that Hq  s  -"IVV'I  d_S = _e e^ dF a Z_ r  L  ( e  1 }  2  x  -ZCt'-t ) (t'-t ) e (e c  c  7 l  - 1)-  7  e  Z  -z(.t'-t ) (1 - e )  "M^c-tJ -M(-t - t - 1 ) (1-e ) + e (1-e c  1 - e"  m  M  c  m  c  -ZCt'-tJ ) c  e  Z  - 1  ]  ( B 9  -  1 }  133  2  -  K =*  m  c  m  In this case, i t can be similarly shown that t'-l -Mt -Z(t'-t). S = z (N+). = R e (-1 - e ) m  m  m  e  -  1  and following the above procedures, dS  ^  =  (  f  3 e  - t  J  m  Z(t'-t ) _ ,  j  1-e  m  (  Bg>2  )  t <t c m  3.  t'-l S = z  " ^ V ^ "^V^ (NJ, = R e e m  c  m  +  ]  )  ( 1 - e  c  •  Z ( t  '- m )  1- e "  \  t  )  m  Z  and  f f(^T7T-- -z 7T- »- ^ ) = !  1  e  ( t  t  1 )  <B,3,  Notice that equations (B9.1, .2 and .3) all include the term S / a , g  and therefore when substituted into (B6), the latter formula with the residual y may be rewritten as  Q  f  +  ( | f - i ) ijs-y  ,  (BIO)  where dS'/dF = only the bracketed portions of equations (B9.1, .2 and .3)'.  184  The expression f o r Q as well as the d i f f e r e n t i a l dQ/dF may be problematical.  Before d i f f e r e n t i a t i o n can take place,  Q  must f i r s t be  i n t e g r a t e d , and the ease with which t h i s can be done depends on the weight function w(t).  A general s o l u t i o n to the problem i s therefore  not presented here, but the method i s i l l u s t r a t e d below using as an example the cubic, von B e r t a l a n f l y (1938) growth equation.  Example In appendix A ( 2 ) , the exact values of c e r t a i n v a r i a b l e s and parameters for a hypothetical cod population are as f o l l o w s : t  = 4  a =  1.895  t  = 5  S  V  = 13  ' c = R /S =1191/1106 = 1.0769  = 0.07  S /a  m  t  Q  1106 e  e  K = 0.14  M =  b = 3.0 = 11.41  =  V Kg  '  e  = 583 0.28  = M + 0.02  =  0.03  q = 0.002  where Z' = the natural m o r t a l i t y rate plus a small component due to research sampling of age groups less than t c  The value of Z' must  therefore replace M i n a l l formulae developed p r i o r to the example. The o b j e c t i v e i s to determine S ^ , ables that govern the p o s i t i o n of  R . and opJ  the three v a r i -  MSY.  To begin the a n a l y s i s a s t a r t i n g value of R must f i r s t be  determined.  A reasonable choice i s R ,„, the maximum recruitment at which the slope max of the stock-regeneration function (B2) i s zero. At R the appropriate max stock s i z e , and the f i r s t estimate of S, i s m  185  S, = S /a = 583. 1 e and substituting this value of S in equation (Bl) R, = R =1538 1 max At equilibrium, since t  > t , the sum of the reproducing age groups  £  may be determined by equation (68)..  The component terms in the formula  are  T, =1 - e " ' i1 - e Z  y  2  =  e"  Z  ( t c  (1 - e "  "  Z ( t  1 - e" and since t  c  - l - e " - ^ i 1 - e- 0 . 3  t m )  0  '"  t c )  ) = e"  Z  3  (1 - e ' ) 82  1 - e"  Z  = 1 - e"  Z  9Z  1 - e"  Z  = t +1, m  -z'(yi)  -z-t  ,  c  e  5Z  = e  =e  It follows that equation (B8) may be reduced to S = R e" " 5Z  (1 - e" ) 9Z  1 - e"  ,  and substituting the values of 1 - e~  9Z  1 -e"  Z  (Bll)  Z  -5Z' , R, and e , and rearranging  = 1.701  Solving this equation by t r i a l and error..for Z,  186  Z  = 0.886  ]  and t h e r e f o r e ^  = 0.886 - 0,28 = 0.606.  Substituting equation (Bl) for  in dBll), taking natural logarithms  and solving for S-j  h  ^  -  " % " ' !& [InO - e" ) - ln(l - e" )]  1 nc + S e  e  5Z  d S, . S d  F  9Z  a  a  and  +  l  a  • {  a  ,  0  e  Z  - 1  9 Z  e  - 1  Z  Therefore, replacing Z by 0.886 in the above expression  l d F d  S  ^e ( - 0.6985) or l.= -0.6985 a dF d  = ]  S  ]  Since the value of Z is reasonably large and t' - t  = 8, the product  of these two terms, when used as a negative exponent of yield a small number. Q  e  , should  On this basis, the shortened form of the integral  can be used, i . e . , Q = FW  OT  t  f  or Q =  1 IT -  -Z(t - t ) e  (1 - e  c  L  F  .  x  c  - (V o K  3 e  t  "  )  +  3  e  Z+K  K L  ^  -K(t - t ) . 0  W  Z + ZK  T dt  3  -  e  K  ^V o t  Z + 3K  )  \  '  Similarly, -K(t -t ) dQ = M - 3:e (M + K) c  d F  Z  tZ + K)  2  + 3 e  0  2  Z k  (t-t ) (M + 2K) c  0  ( Z + 2K)  2  -3K(t„-tJ - e (M (M+3K) +3K c  0  (Z f e!<)  2  (B13)  187  Substituting the values of  M,  and  in equations (B12) and  (B13), it follows that Q = 0.1290  and  dQ = 0.0154 l  d F  All the values needed to determine the f i r s t residual in equation (BIO) have now been calculated, (BIO)  Y, = 1  dQ + Q dF  ,h K  1  a  i.e.,  - K > dF  1  and substituting, Y, = 0.0154 0.1290 1  + (583 583  - 1) (-0.6985) = 0.1194  Equation (BIO) can also be used to solve for the value of will in fact make the residual zero.  S that  On this basis, and using the prev-  iously derived values, 0 = 0.0154 0.1290  + (583 - 1) (-0.6985) S  and solving, the second estimate of S  2  S is  = 498.  The entire procedure can now be repeated by using S to prepare a 2  second estimate of R from (Bl), solving for Z and 2  2  error, and calculating a second residual y 2  by t r i a l and  If i t is not zero, a  third estimate, S^* is calculated and the procedure is repeated. results of five iterations are given below.  The  188  Estimate F  Z  1538  .606  .886  +0.1194  498  1520  .863  1.143  -0.0569  3  556  1536  .678  .958  +0.0483  4  518  1528  .793  1.073  -0.0276  5  544  1534  .713  .993  +0.0209  6  534  1532  .742  1.022  +0.0014  No.  S  1  583  2  R  Y  The absolute values of the residuals from estimates 4 and 5 are almost equal and are sufficiently small to allow an estimate of S ^ to be prepared graphically.  By plotting  S on the abscissa,  y on the  ordinate, and using the co-ordinates (S^, Y^) and (S^, Yg), a line joining these points intersects the S-axis at 534 (-= S values of (R, F and Z)  Q p t  ).  The remaining  ^. are then readily determined.  The yield may now be calculated using the exact expansion of the Beverton and Holt (1957) model at equilibrium.  To complete the com-  parison the yield at maximum surplus production (msp) is also included. R  S .  F  Z  f (boats)  Yield (.Kg)  1532  534  0. 742  1. 022  371  508.2 (MSY)  R = 1538 max  583  0. 606  0. 886  304  502.7  = 1456  408  1. 300  1. 580  650  485.0  R  opt  R .  msp  =  For the particular case defined in Appendix A(2), the effort (boats) needed at R  opt  is 22 percent higher than at R  the gain in yield is only about one percent. yield obtained at R  max  , where as  Economically, the •  is therefore a more justifiable goal.  190  APPENDIX C Seasonal Pulse Fishing Results The effect of seasonally applied fishing mortality on the value of equilibrium yield is examined by means of a hypothetical, constant recruitment model.  Annual, instantaneous growth rates (G^ for successive age groups (i)  are calculated from equation (1.78) for-t  G  =  3  = 0, b = 3.0, i . e . ,  *n <e -e ) (e -l) l K  using  K values of 0.2, 0.4 and 0.6.  Annual, instantaneous natural mortality  rates of 0.1, 0.2, 0.3 and 0.4 are applied separately for each K value and are constant for each age group.  Seasonal growth (g^) and mortality (m) rates are  arrived at by appropriate division for each growth-mortality combination . reported in Figure Cl(a).  The processes are independent of each other, and  their annual rates may be condensed into a single quarter, or occur consecutively for half, three-quarters, or a full year. continuous growth and natural mortality.  The last situation represents  For a given case the growth-  mortality pattern is constant throughout the l i f e span of the cohort. I n i t i a l l y , the critical age is determined to the nearest quarter year by means of the recursive equation  B. • = B. . . e 1 - J ( g  where  m )  j = the quarter index and B. , . , and B. . are the initial and final  biomass respectively.  Calculations begin by arbitrarily setting the f i r s t  year-old stock (N )-j at 1000 fish and Q  = 100.  Thus, the biomass of one-  year olds at the beginning of the f i r s t quarter is calculated from  191  ( a )  M  S  CASE  E  m 2  * *  9 i  m  1  5  * * * *  1 3  „. m * * * * 9i * * * m * * m  * **  m  * * **  Qi * * *  m 7  12  * * * *  g- * * * * m * *  1 6  9i * * * *  1 7  91 * * * * ID  8  * * *  *  *  q,- * * * *  m 9 10  (b)  *  g. * * * * m * * * *  *  9-j  g. * * m * *  20  m  *  *  g i  m  INTENSIVE  MODERATE  FS 1 * 2 * 3 4  FS 5 * * 8 * * * 5 * * g * * * 7 * * CONTINUOUS FS 10 * * * *  Figure C l .  *  *  EXTENSIVE  (a) Twenty cases of seasonally distributed growth (g-j) and natural mortality (m) patterns. Each year in the l i f e of an age group (i) is divided into quarters, (b) Ten seasonally applied fishing strategies (FS).  192  B  l,0  =  ( N  o l "J " "*) 1  }  6  3  =  1  x  1 0 5  d-e" ) . K  3  Once the critical age is determined, the age of f i r s t capture (AFC) is set equal to the integer value of the critical age without regard to the fraction.  The f i r s t seasonal fishing strategy (FS 1) shown in Figure Cl(b)  is then applied to AFC and successive ages for either 100 years or until the annual growth rate (G^) falls below 0.0002.  Quarterly yields are summed to  give the equilibrium yield (Y ) by means of the equation i = 100 or G.<0.0002 Y-r  T  =  :  E  .  E  i=AFC  f,B. 'j i,j-l  (e  D  j=l  (g f-m). r  1  - 1.0), (g-f-mjj J  where f. = the seasonal value of the annual instantaneous fishing mortality rate, F, that is constant for all age groups.  For each fishing strategy Y^  is calculated separately for odd numbered values of F from 0.1 to 0.5 inclusive.  Intensive fishing describes the situation in which the annual  rate is condensed into a single quarter, moderate and extensive i f F is distributed over two or three consecutive quarters, respectively.  Strategy  FS 10 represents continuous fishing. The amount of data generated by this program is voluminous, but they need not be reported in full to grasp the essentials of seasonal pulse fishing.  In Table CI several general principles are illustrated which apply  to all growth-mortality combinations.  The data refer to the relative equi-  librium y i e l d , yield rank and percentage gain (+) in yield relative to continuous fishing.  Strategy FS 10 is chosen as the basis of comparison  Table CI  Relative equilibrium y i e l d , rank values and yield gain (+) relative to continuous fishing (FS 10) for growth- mortality Case 1. K =0.4; M = 0.3; F = 0.1 to 1.5; c r i t i c a l age = 3.5 y r , ; age of f i r s t capture = 3.0 y r . Range = percentage sum of absolute maximum gain and loss relative to FS 10. CASE 1  G |* |* | M |  FISHING STRATEGY (FS)  |  |  |  1*1*1  Relative Equilibrium Yield (Yj) .1  .3  .5  .7  .9  1.1  1.3  1.5  Yield Rank F .1 .3 .5 .7 .9 1.1 1.3 1.5  Percentage Gain (+) Relative to FS 10 .1  .3  .5  .7  .9  1.1  1.3  1.5  -1.9 -4.8 -6.7 -8.2 -9.2 -9.9 -10.5 -10.9  1  9.53 16.76 19.12 20.01 20.35 20.46 20.47 20.42  8  9  9 10  10  10  10  10  2  10.44 18.85 21.91 23.25 23.90 24.22 24.37 24.44  1  1  3  3  3  3  3  5  7.5  7.1  6.9  6.7  6.6  6.6  6.5  6.6  3  10.18 18.70 22.05 23.69 24.60 25.16 25.54 25.81  3  3  1  1  1  1  1  1  4.8  6.3  7.5  8.7  9.8 10.7  11.7  12.6  4  8.76 16.09 18.98 20.39 21.17 21 .66 21.98 22.22  10 10 10  9  9  9  9  9  -9.8 -8.6 -7.4 -6.4 -5.5 -4.7  -3.9  -3.1  5  9.97 17.75 20.40 21.44 21.86 22.00 22.00 21.94  5  6  8  8  8  8  8  9  2.6  0.8 -0.5 -1.6 -2.5 -3.2  -3.8  -4.3  6  10.31 18.77 21.97 23.45 24.20 24.62 24.85 24.98  2  2  2  2  2  2  2  2  6.1  6.7  7.2  7.6  8.0  8.3  8.6  9.0  7  9.48 17.46 20.64 22.22 23.14 23.72 24.14 24.44  9 8  6  5  5  5  5  4  -2.4 -0.8  0.7  2.0  3.2  4.4  5.5  6.6  8  10.03 18.04 20.88 22.08 22.60 22.82 22.88 22.86  4  4  5  6  6  6  6  7  3.2  2.5  1.9  1.3  0.8  0.4  0.0  -0.3  9  9.80 17.95 21.11 22.63 23.46 23.95 24.27 24.48  6  5  4  4  4  4  4  3  0.8  2.0  3.0  3.8  4.6  5.4  6.1  6.8  10  9.71 17.60 20.50 21.79 22.41 22.72 22.87 22.93  7  7  7  7  7  7  7  6 17.3 15.7 14.9 16.9 19.0 20.6  22.2  23.5  RANGE  194  since i t is the usual method of reporting yield.  The principles that emerge  are as follows: 1.  As F increases from 0.1 to 1.5 the yield for each fishing strategy approaches an asymptotic value.  The rate of increase is not the  same for each strategy, however, and leads to a changing pattern of relative rank yields. 2.  To be able to benefit from seasonal pulse fishing i t must be applied on a species basis for which are known.  , M, K and the growth-mortality  pattern  From item 1 above the annual fishing mortality rate and  the degree to which i t can be seasonally concentrated must also be estimated.  Once these factors are known, a suitable strategy can  be selected to improve or maximize equilibrium yield. 3.  Alternatively, only a portion of the available fishing fleet can be used to obtain a regulated yield by an appropriate choice of fishing ° strategy.  4.  The opportunity to conserve fishing effort (at constant catchability) and obtain an annual yield equivalent to that of a greater effort can be demonstrated.  For instance, i f current practice is to fish inten-  sively (FS 4) at F = 1.5, the same yield can be obtained with less than half the effort (F = 0.7) by means of the extensive strategy FS 7. Similarly, a change from continuous fishing to a moderate strategy (FS 6), coupled with a reduction in F from 1.5 to 0.7, can lead to a small  (2.3  per cent ) but positive increase in yield. The ten fishing strategies are divided into four modes: moderate, extensive and continuous.  intensive,  Within these categories the strategy  195  options are reduced, respectively, from four (FS 1-4), to three (FS 5-7), to two (FS 8-9) and one for continuous fishing.  From the rank values in  Table CI it is clear that for each fishing mortality rate there is a best and worst choice of fishing strategy within the f i r s t three modes as well as on an overall basis.  As F increases, the rank of a particular strategy can  adopt only one of three courses: unchanged.  i t can improve, deteriorate, or remain  At low values of F the rate of change of improvement or deteriora-  tion is rapid but tends to stabilize as F assumes higher values.  At F = 1.5  the difference between strategies leading to yield improvement and loss is therefore magnified, and the absolute value of the largest difference, or range, also takes on an extreme value.  If data on only the best fishing  strategies to be used is extracted from the percentage gain column of F = 1.5, then Table CI can be condensed into a single line.  This feature forms the  basis of Table CII, but additional information on all K and M values and the 20 growth-mortality cases can now be included. The usefulness of Table CII can be judged in relation to two important management questions: 1.  Are there general guidelines that can be used to indicate whether seasonal pulse fishing will benefit the yield from a fishery?  2.  If the gain is sufficiently attractive, how and when should fishing take place?  To answer the f i r s t question, the following generalizations emerge from a study of Table CII. exceptions exist. of a target species.  They are intended to act as guidelines only; a few  Ultimately, there is no substitute for a detailed analysis  196  TABLE  R e l a t i v e e q u i l i b r i u m y i e l d , seasonal pulse f i s h i n g r e s u l t s f o r F = 1.5.  CII  The a s t e r i s k s (*) i n d i c a t e the quarter during which a f r a c t i o n of the annual instantaneous growth (G) and natural m o r t a l i t y (M) rates apply. Strategy r e f e r s to the f i s h i n g s t r a t e g i e s of Figure CI(b) and the r e s u l t s are reported as +_ gain i n y i e l d r e l a t i v e to continuous f i s h i n g (FS 10). % Range = sum of f i r s t capture (years) AFC = age of f i r s t capture (years) AFC + f r a c . = the c r i t i c a l age S t r a t e g i e s reported under I n t e n s i v e , Moderate or Extensive r e f e r to the best choice in each f i s h i n g mode.  CASE 1:  G ! M !  *  j  *  I  i *  i  Worst Strategy  i i  *  % Gain  INTENSIVE Strategy % Gain  MODERATE Strategy % Gain  EXTENSIVE Strategy % Gain  0.2  0.1 0.2 0.3 0.4  - 2.9 - 6.2 - 7.8 -10.8  3 3 3 3  3.4 7.2 9.5 13.2  2.6 5.5 7.5 10.3  0.4  0.1 0.2 0.3 0.4  - 4.4 - 6.5 -10.9 -20.2  3 3 3 3  4.8 7.5 12.6 23.0  6 6 6  3.3 5.6 9.0 14.6  9 9 9 9  0.6  0.1 0.2 0.3 0.4  - 3.6 - 6.8 -15.0 -14.1  3 3 3 3  4.0 7.8 16.7 16.5  6 6 6 6  2.9 5.7 10.8 11.7  9 9 9 9  CASE 2:  G ! M !  *  j  *  i *  % Range  A FC Frac.=.50  1.8 3.9 5.0 6.9  6.3 13.4 17.3 24.0  9 6 5 4  2.7 4.0 6.8 12.4  9.2 14.0 23.5 43.2  5 4 3 2  7. 14. 29. 30.6  4 3 2 2  CTi  I  i *  i *  i  Frac.=.50  0.2  0.1 0.2 0.3 0.4  • • • •  1.8 4.1 5.3 7.1  2 2 2 2  I. 9 3.9 5.9 7.5  6 6 6 6  1.8 3.9 5.5 7.0  9 9 9 9  1.1 2.5 3.2 4.1  3.7 8.0 11.2 14.6  6 5 4  0.4  0.1 0.2. 0.3 0.4  - 3.4 - 4.4 - 7.9 -16.3  3 3 3 3  3.1 4.0 7.3 15.7  6 6 6 6  2.5 4.0 6.6 11.4  9 9 9 9  2.0 2.6 4.7 9.6  6.5 8.4 15.2 32.0  5 4 3 2  0.6  0.1 0.2 0.3 0.4  - 2.5 - 4.8 -12.0 -10.1  3 3 3 3  2.3 4.4 II. 4 9.4  6 6 6 6  2.1 4.1 8.4 8.5  9 9 9 9  1.5 2.8 7.1 6.0  4.8 9.2 23.4 19.5  4 3 2 2  197  CASE 3:  G ! M!  Worst Strategy  % Gain  INTENSIVE Strategy % Gain  MODERATE Strategy % Gain  EXTENSIVE Strategy % Gain  % Range  AFC Frac.=.50  0.2  0.1 0.2 0.3 0.4  -  1.4 3.3 3.5 5.1  3 3 3 3  1 3 4 6  7 8 3 3  6 6 6 6  1.3 2.9 3.6 5.0  9 9 9 9  0.9 2.1 2.3 3.3  3 7 7 11  1 1 8 4  9 6 5 4  0.4  0.1 0.2 0.3 0.4  - 3.0 - 3.6 - 6.7 -14.9  3 3 3 3  3 4 7 15  1 1 4 9  7 6 6 7  2.1 3.0 5.0 11.7  9 9 9 9  1.8 2.2 4.1 8.9  6 7 14 30  1 7 1 8  5 4 3 2  0.6  0.1 0.2 0.3 0.4  - 2.1 - 4.0 -11.0 - 8.6  3 3 3 3  2 3 4 4 11 .5 9 .6  6 6 7 6  1.6 3.1 8.4 6.5  9 9 9 9  1.3 2.4 6.6 5.2  4 4. 8 4 22 .5 18 .2  CASE 4:  G ! * ! M ! !  4 3 2 2  I * i *  Frac.=.50 2 5 7 10  6 6 4 2  6 6 6 6  1.7 3.6 5.0 6.7  9 9 9 9  1 2 3 5  3 9 7 1  4 10 13 18  8 3 2 3  9 6 5 4  3 3 3 3  2 5 9 9  3 8 8 4  6 7 7 6  1.6 3.7 5.7 .6.5  9 9 9 9  1 3 5 4  1 0 1 6  4 10 18 16  1 8 3 7  6 4 3 3  3 3 3 3  3 6 5 12  1 1 6 8  7 7 6 7  2.1 4.0 4.4 8.7  9 9 9 9  1 7 3 2 2 .6 6 .6  5 9 11 4 10 2 23 .7  4 3 3 2  0.2  0.1 0.2 0.3 0.4  • • -  2.2 4.7 5.8 8.1  3 3 3 3  0.4  0.1 0.2 0.3 0.4  -  1.8 5.0 8.5 7.3  0.6  0.1 0.2 0.3 0.4  - 2.8 - 5.3 - 4.6 -10.9  CASE 5:  G ! * ! M ! !  i i  j * i  Frac.=.75 2=3 3 2 2  0 2 2 3  9 2 8 7  6 6 6 6  0.9 2.0 2.6 3.6  9 9 9 9  0 1 1 2  6 5 5 3  2 4 5 7  0 9 5 7  9 6 5 4  - 2.4 -'2.9 - 5.4 -12.3  3 3 3 3  2 2 4 10  1 4 6 7  7 6 7 7  1.9 2.0 3.7 10.7  9 9 9 9  .1 1 3 6  4 6 0 8  4 5 10 23  5 3 0 0  5 4 3 2  -  3 3 3 3  1.1 2.1 7.7 4.5  9 9 9 9  1 1 5 3  0 8 0 8  3 5 16 12  1 9 8 7  4 3 2 2  0.2  0.1 0.2 0.3 0.4  1 1 1 1  -  0.4  0.1 0.2 0.3 0.4  1 1 1 1  0.6  0.1 0.2 0.3 0.4  1 1 1 1  1.1 2.7 2.7 4.0  1.7 3.2 9.0 6.9  1 4 2 .7 7 8 . 5 .8  7 6 7 7 •  198  Table CII  (continued) i * i  G ! M!  CASE 6:  1*1*1  Worst Strategy 0.2  0.4  0.6  INTENSIVE Strategy % Gain  % Gain  MODERATE Strategy t Gain  EXTENSIVE Strategy % Gain  AFC Frac.=.75  Range  0.1 0.2 0.3 0.4  0.7 1.9 1.5 • 2.4  3 3 3 3  0.9 2.3 2.3 3.4  7 7 7 7  0.8 2.0 1.7 2.7  9 9 9 9  0.4 1.1 1.0 1.5  1.6 4.2 3.8 5.8  9 6 5 4  0.1 0.2 0.3 0.4  • 2.0 - 2.1 - 4.3 -10.9  4 3 3 4  2.4 2.4 4.7 14.4  7 7 7 7  2.2 2.3 4.6 12.2  9 9 9 9  1.2 1.2 2.5 6.2  4.4 4.5 9.0 25.3  5 4 3 2  0.1 0.2 0.3 0.4  - 1.3 - 2.4 - 7.9 - 5.4  3 3 4 3  1.4 2.7 10.2 5.9  7 7 7 7  1.4 2.6 8.8 5.9  9 9 9 9  0.8 1.4 4.5 3.1  2.7 5.1 18.1 11.3  4 3 2 2  i * [ i * i  G ! M!  CASE 7:  Frac.=.50  3 3 3 3  1.7 3.6 4.7 6.5  6 6 6 6  1.3 2.7 3.9 5.2  9 9 9 9  0.9 2.0 2.5 3.4  3.1 6.8 8.5 11.8  9 6 5 4  0.2  0.1 0.2 0.3 0.4  1 1 1 1  - 1.4 - 3.2 - 3.8 - 5.3  0.4  0.1 0.2 0.3 0.4  1 1 1 4  -  1.1 3.4 2.1 4.7  3 3 2 3  1.4 3.7 3.4 5.9  6 6 6 6  1.2 2.7 3.4 5.0  9 9 9 9  0.7 2.1 1.5 3.0  2.5 7.1 5.5 10.6  4 3  0.1 0.2 0.3 0.4  7 1 1 1  -  0.4 3.6 2.3 7.5  2 3 3 3  1.2 3.9 3.5 8.2  6 6 6 6  1.1 2.8 3.5 5.6  9 9 9 9  0.4 2.2 1.6 4.6  1.6 7.5 5.8 15.7  5 3 3 2  0.6  CASE 8:  G ! * ! * • * • * • M ! 1 * ! * ! * !  Frac.=.25 1.8 5.1 5.8 7.3  10 7 5 4  8 8 8 8  0.2 0.8 1.1 0.8  2.3 6.3 8.9 8.5  6 5 4 3  8 8 8 8  0.5 1.0 1.0 2.0  2.5 7.1 8.3 14.5  5 4 3 3  0.1 0.2 0.3 0.4  7 4 4 4  -  0.9 3.3 3.2 3.9  2 2 2 2  0.9 1.8 2.6 3.4  5 5 6 6  0.5 0.9 1.5 2.0  CO OO CO CO  0.3 0.6 0.5 0.6  0.2  0.4  0.1 0.2 0.3 0.4  4 4 4 4  - 1.4 - 4.4 - 6.1 - 5.0  2 2 2 2  0.9 1.9 2.8 3.5  6 5 5 6  0.5 1.3 1.7 1.8  0.1 0.2 0.3 0.4  7 4 4 4  - 1.5 - 5.4 - 5.6 -10.7  2 2 2 2  1.0 1.9 2.7 3.8  6 5 5 5  0.8 1.7 1.5 3.4  0.6  CO  199  Table CII  (continued) G ! * ! * ! * i * i  CASE 9:  Worst Strategy  % Gain  INTENSIVE Strategy % Gain  MODERATE Strategy % Gain  10 7 6 5  0.2 1.5 1.5 0.7  0.1 0.8 0.8 0.4  0.8 4.6 5.9 2.8  6 5 4 3  0.7 1.6 1.3 3.2  0.4 0.8 0.7 1.6  2.8 6.0 5.2 12.4  5 4 3 3  - 0.9 - 1.7 - 3.9 - 5.2  0.6 1.1 2.5 3.4  5 5 5 5  0.4 0.7 1.7 2.3  0.4  0.1 0.2 0.3 0.4  4 4 4 4  - 0.5 - 2.5 - 3.6 - 1.7  0.3 2.1 2.3 1.1  5 5 5 5  0.6  0.1 0.2 0.3 0.4  4 4 4 4  -  1.1 2.3 2.0 4.8  5 5 5 5  CASE 10::  0.2  0.4  0.6  i  G M  * ! *  Frac.=.0 8 8 8 8  0.2 0.3 0.7 1.0  6.6 13.1 20.2 27.0  10 7 6 5  5 5 5 5  8 8 8 8  0.4 0.5 1.5 1.6  7.0 13.5 21.7 28.2  7 5 5 4  5 5 5 5  8 8  0.6 0.7 1.6 1.5  7.5 13.9 22.1 28.0  6 4 4 3  3 3 3 3  - 3.4 - 6.7 -10.1 -13.4  3.2 6.4 10.1 13.6  5 5 5 5  0.1 0.2 0.3 0.4  3 3 3 3  - 3.5 - 6.8 -10.5 -13.7  3.5 6.7 11.2 14.5  0.1 0.2 0.3 0.4  3 3 3 3  - 3.6 - 6.9 -10.6 -13.7  3.9 7.0 11.5 14.3  i G . M * ! *  8 8 8  * !*! ! !  0.1 0.2 0.3 0.4  CASE 11::  1.3 2.5 4.2 5.7  * !* ! * ! !  Frac.= .0 2.5 5.0 8.0 10.7  5 5 5 5  1.2 2.4 4.0 5.4  0.2  0.1 0.2 0.3 0.4  3 3 3 3  - 2.7 - 5.3 - 8.0 -10.6  0.4  0.1 0.2 0.3 0.4  3 3 3 3  - 2.8 - 5.4 - 8.4 -11.0  2.8 5.3 9.0 11.5  5 5 5 5  1.5 2.7 5.0 6.2  4 3 4 3  - 3.3 - 5.5 - 7.5 -10.9  3.1 5.6 10.7 11.4  5 5 5 5  1.9 3.0 6.7 6.1  0.6  0.1 0.2 0.3 0.4  .  AFC Frac.=.0  1.5 2.8 6.4 8.6  4 4 4 4  1.7 3.7 3.2 7.6  Range  0.2 0.4 0.8 1.1  0.1 0.2 0.3 0.4  0.2  EXTENSIVE Strategy % Gain  8 8  0.3 0.5 1.1 1.6  5.2 10.3 16.0 21.3  10 7 6 5  8 8 8 8  0.5 0.8 1.8 2.1  5.6 10.7 17.4 22.5  7 5 5 4  0.7 1.0 3.4 2.0  6.4 11.1 18.2 22.3  6 4 4 3  200  Table CII  (continued)  I *  G ! M !  CASE 12:  i  Worst Strategy  *  % Gain  INTENSIVE Strategy % Gain  MODERATE Strategy % Gain  EXTENSIVE Strategy % Gain  % Range  AFC Frac.=.0  K  M  0.2  0.1 0.2 0.3 0.4  3 3 3 3  - 1.7 - 3.4 - 5.3 - 7.0  1.8 3.5 5.7 7.6  5 5 5 5  0.8 1.5 2.7 3.7  8 8 8 8  0.2 0.3 0.8 1.1  3.5 6.9 11.0 14.6  10 7 6 5  0.4  0.1 0.2 0.3 0.4  3 3 7 3  - 1.9 - 3.6 - 5.9 - 7.4  2.0 3.8 6.7 8.4  5 5 5 5  1.1 1.9 3.8 4.5  8 8 8 8  0.4 0.6 1.5 1.6  3.9 7.4 12.6 15.8  7 5 5 4  0.6  0.1 0.2 0.3 0.4  4 3 4 3  -  2.4 4.1 7.0 8.3  5 5 5 5  1.4 2.1 4.0 4.4  8 8 8 8  0.6 0.9 1.7 1.5  5.2 7.8 14.3 15.7  6 4 4 3  2.8 3.7 7.3 7.4  G ! ! * ! * !*! i I M l * ! *  CASE 13:  Frac.=.0 0.2  0.1 0.2 0.3 0.4  3 3 3 3  0.4  0.1 0.2 0.3 0.4  3 3 3 3  0.6  0.1 0.2 0.3 0.4  3 3 3 3  CASE 14:  8 8 8 8  0.1 0.1 0.5 0.8  4.5 8.8 14.7 19.5  10 7 6 5  5 5 5 5  8 8 8 8  0.3 0.4 0.5 1.4  5.4 9.9 14.3 22.4  7 5 4 4  5 5 5 5  8 8 8 8  0.3 0.6 1.5 1.3  5.3 10.8 19.0 22.0  5 4 4 3  2.4 4.7 7.9 10.6  5 5 5 5  - 2. - 4. - 6. -10  2.9 5.3 7.7 12.1  - 2.5 - 5.0 - 8.8 -10.1  2.8 5.8 10.2 11.9  2.1 4.1 6.8 8.9  0.8 1.6 3.1 4.1  4.9 5.1  G ! ! * ! * ! * ! M ! * ! * ! * ! ! Frac.=.0  0.2  0.1 0.2 0.3 0.4  3 3 3 3  0.4  0.1 0.2 0.3 • 0.4  4 3 3 4  0.6  0.1 0.2 0.3 0.4  7 4 7 4  1.3 2.6 4.6 6.0  1.7 3.3 5.7 7.7  5 5 5 5'  0.8 1.5 2.8 3.8  8 8 8 8  0.2 0.4 0.9 1.3  3.0 5.9 10.3 13.7  10 7 6 5  - 1.9  - 3.1 - 4.5 - 8.2  2.2 3.9 5.5 9.2  5 5 5 5  1.2 1.9 2.7 5.0  8 8 8 8  0.4 0.7 0.9 1.9  4.1 7.0 10.0 17.4  7 5 4 4  1.7 - 3.8 - 7.4 - 7.7  2.1 4.4 8.0 8.9  5 5 5 5  1.1 2.3 4.6 4.8  8 8 8 8  0.4 0.9 1.9 1.8  3.8 8.2 15.4 16.7  5 4 4 3  -  _  ro o o  201  Table CII  (continued)  CASE 15:  G ! ! * ! * ! * ! M i * j * j * j * i  Worst Strategy  INTENSIVE Strategy % Gain  % Gain  MODERATE Strategy I Gain  EXTENSIVE Strategy % Gain  % ' Range  AFC Frac.=.0  K  M  0.2  0.1 0.2 0.3 0.4  2 2 4 4  -  0.6 1.2 2.2 2.9  1 1 1 1  0.9 1.8 3.4 4.6  5 5 5 5  0.3 0.6 1.5 2.1  8 8 8 8  0.1 0.2 0.6 0.8  1.5 3.0 5.6 7.5  10 7 6 5  0.4  0.1 0.2 0.3 0.4  4 4 4 4  -  1.5 1.7 1.9 6.3  1 1 1 1  1.4 2.4 3.2 6.1  5 5 5 5  0.8 T.l 1.4 3.2  8 8 8 8  0.3 0.4 0.5 1.4  2.9 4.1 5.1 12.4  7 5 4 4  0.6  0.1 0.2 0.3 0.4  4 4 4 4  -  1.2 2.8 7.2 5.8  1 1 1 1  1.3 2.8 5.7 5.8  5 5 5 5  0.7 1.5 3.3 3.0  8 8 8 8  0.3 0.6 1.6 1.3  2.5 5.6 12.9 11.6  5 4 4 3  CASE 16:  G ! * ! * ! * ! * ! M ! * ! * ! ! ! Frac.=.0  0.2  0.1 0.2 0.3 0.4  3 3 3 3  -  2.1 4.0 6.7 8.9  1 1 1 1  2.1 4.0 6.9 9.4  5 5 5 5  0.9 1.7 3.2 4.3  8 8 8 8  0.2 0.3 0.8 1.1  4 8 13 18  2 0 6 3  10 7 6 5  0.4  0.1 0.2 0.3 0.4  3 3 3 3  - 2.5 - 4.6 - 6.6 -10.3  1 1 1 1  2.7 4.7 6.7 11.1  5 5 5 5  1.3 2.2 3.0 5.4  8 8 8 8  0.4 0.5 0.7 1.6  5 9 13 21  2 3 3 4  7 5 4 4  0.6  0.1 0.2 0.3 0.4  3 3 3 3  - 2.5 - 5.0 - 8.7 -10.1  1 1 1 1  2.5 5.3 9.7 10.8  5 5 5 5  1.2 2.6 4.9 5.2  8 8 8 8  0.3 0.7 1.7 1.5  5 10 18 20  0 3 4 . 9  5 4 4 3  CASE 17:  G ! * ! * ! * ! * ! M ! * ! * ! * ! ! Frac.=.0  0.2  0.4  0.6  0.1 0.2 0.3 0.4 0.1 0.2 0.3 0.4  4 4 4 4 -  0.1 < 0.2 0.3 0.4  4 4 4 4 4 4 4 4  1.4  - 2.6 - 5.4 - 7.1 2.4  - 3.8 - 5.1 - 9.9 _  2.2  - 4.7 - 9.6 - 9.5  •  1  1.3 2.6 4.8 6.4  5 5 5 5  0.8 1.6 3.0 4.0  8 8 8 8  0.3 0.6 1.2 1.6  2.7 5.2 10.2 13.5  10 7 6 5  1 1 1  1  2.0 3.3 4.6 8.1  5 5 5 5  1.2 2.0 2.8 5.1  8 8 8 8  0.5 0.8 1.1 2.2  4.4 7.1 9.7 18.0  7 5 4 4  1 1 1 1  1.8 3.8 7.4 7.8  5 5 5 5  1.1 2.4 4.7 4.9  8 8 8 8  0.5 1.0 2.0 2.0  4.0 8.5 17.0 17.3  5 4 4 3  1 1 1  202  Table CII  CASE 18:  (continued)  G ! M!  Worst Strategy  K  M  0.2  0.1 0.2 0.3 0.4  1=4 1  0.1 0.2 0.3 0.4  1 1 1  0.1 0.2 0.3 0.4  1 1 1 1  0.4  0.6  CASE 19:  1  1  INTENSIVE Strategy I Gain  % Gain  MODERATE % Gain Strategy  EXTENSIVE Strategy % Gain  AF Frac.  Range  - 2.3 - 5.2 - 7.9 - 8.9  3 3 3 3  4.0 8.8 11.3 15.9  6 6 6 6  2.5 5.3 7.4 10.1  9 9 9 9  1.7 3.8 4.7 6.7  6.3 14.0 19.2 24.8  9 6 5 4  - 4.0 - 5.5 - 9.7 -19.4  3 3 3 3  6.1 9.1 15.8 30.5  7 6 6 7  3.6 5.3 8.6 19.1  9 9 9 9  2.7 3.9 6.8 13.0  10.1 14.6 25.5 49.9  5 4 3 2  - 3.1 - 5.9 -14.2 -12.5  3 3 3 3  4.9 9.6 22.0 20.8  6 6 6 6  2.8 5.5 10.3 11.4  9 9 9 9  2.2 4.1 9.5 8.8  8.0 15.5 36.2 33.3  4 3 2 2  c .50  G ! M! Frac.=.75  0.2  0.1 0.2 0.3 0.4  3 3 3 3  - 3.8 - 7.3 -11.3 -14.7  1 1 1 1  2.3 4.3 7.2 9.2  5 5 5 5  0.9 1.5 3.0 3.8  10 10 10 10  0.0 0.0 0.0 0.0  6 11 18 23  1 6 5 9  9 6 5 4  0.4  0.1 0.2 0.3 0.4  3 3 3 3  - 3.2 - 7.3 -10.2 -11.2  4 4 4 4  5.3 6.0 11.7 29.4  5 5 5 7  0.7 1.4 1.2 • 4.1  10 10 10 10  0.0 0.0 0.0 0.0  8 13 21 40  5 3 9 6  5 4 3 2  0.6  0.1 0.2 0.3 0.4  3 3 3 3  - 3.6 - 7.1 - 8.7 -13.6  4 4 4 4  3.6 6.7 20.8 14.8  5 5 7 5  0.5 1.2 2.5 1.8  10 10 10 10  0.0 0.0 0.0 0.0  7 13 29 28  2 8 5 4  4 3 2 2  CASE 20:  G ! M! Frac.=.0  0.2  0.4  0.6  0.1 0.2 0.3 0.4  4 4 4 4  -  1.3 1.3 5.6 7.4  0.1 0.2 0.3 0.4  4 4 4 4  -  0.7 4.0 5.2 2.4  0.1 0.2 0.3 0.4  4 4 4 4  - 2.5 - 5.3 - 4.6 -10.7  .  1 1 1 1  0.7 1.3 2.9 4.0  5 5 5 5  0.6 1.0 2.3 3.1  8 8 8 8  0.3 0.5 1.1 1.5  2.0 2.6 8.5 11.4  10 7 6 5  1  0.4 2.1 2.7 1.3  5 5 5 5  0.3 1.6 2.1 1.0  8 8 8 8  0.2 0.8 1.0 0.5  1.1 6.1 7.9 3.7  6 5 4 3  1.3 2.7 2.4 5.6  5 5 5 5  1.0 2.2 1.9 4.4  8 8 8 8  0.5 1.0 0.9 2.1  3.8 8.0 7.0 16.3  5 4 3 3  1 1 1 1 1 1 1  ro O  203  1.  As  m increases for a given  value of to the  K value, or as  K  increases for a given  M, the benefits from seasonal pulse f i s h i n g increase r e l a t i v e y i e l d from continuous f i s h i n g .  In t u r n , the percentage y i e l d  increase depends on the degree of overlap between the seasonal growth and mortality patterns.  If  the growth period is e n t i r e l y free of any  natural mortality component, the expected y i e l d improvement is enhanced.  2.  The c r i t i c a l age depends on the values mortality pattern.  of K and  M and the growth-  In the model used here, i f the c r i t i c a l age includes  an annual f r a c t i o n , the age of f i r s t capture (AFC) i s reduced to the integer value. 9.50,  For instance, i f the c r i t i c a l age is calculated as 9.75,  9.25 or 9.00,  then AFC i s set equal to 9.00.  further AFC departs from the c r i t i c a l seasonal f i s h i n g is increased.  Apparently the  age, the y i e l d improvement due to  In p r a c t i c e , commercial f i s h e r i e s  usually exploit age groups younger than the c r i t i c a l age, and therefore seasonal pulse f i s h i n g may o f f e r a d i s t i n c t advantage over the continuous mode.  The question of how to f i s h on a seasonal basis can be answered without ambiguity.  Seasonally intensive f i s h i n g offers a superior y i e l d compared to  a continuous strategy, and a moderate program provides a second best choice. With only one exception (Case 19), small but positive y i e l d benefit. guidelines given e a r l i e r . are seasonally separate,  the extensive category also contains a These comments must be tempered by the  For example, in Case 1, where growth and mortality K=0.2 and M=0.4, an intensive strategy (FS 1)  generate a 13.2 per cent increase in y i e l d over continuous f i s h i n g .  can  In  Case 9, however, where growth and mortality are continuous for the same K and M values, the y i e l d gain from FS 1 i s only 3.2 per cent.  The decision to  implement an intensive program w i l l therefore depend on the species charac-  204  teristics and whether concentrating the fishing activity is considered worthwhile. While one intensive strategy leads to the maximum, relative increase in yield one of the three remaining strategies in that mode usually represents the opposite extreme. nearly equal.  The magnitude of the potential gain and loss are  The same situation applies in the remaining modes and there-  fore it is as important to know not only how to fish but when.  The answer :  is not clear cut, but the relationship between the growth and mortality patterns is instructive.  If the critical age is equal to an integer value,  the best strategies in the intensive, moderate and extensive modes are consistently FS 1, FS 5 and FS 8, respectively.  In intensive fishing the  worst strategy is also removed from the best choice by two or three quarters. The difficulty in assigning the best time to fish arises when the critical age contains a fraction.  If the growth period is free of natural mortality,  the optimum intensive strategy usually falls in the season immediately after growth is completed.  Where growth and mortality overlap, the decision is  influenced by the seasonal intensity of growth and mortality.  For example,  in Case 2 one third of the annual mortality rate coincides with the terminal but moderate growth period.  For values of K = .2 and M = 0.1 to 0.4,  intensive fishing in the second quarter is advantageous compared with either FS 1 or FS 3.  As K increases, however, intensive fishing can be delayed  until FS 3 to achieve the most favorable yield.  In Case 4, growth is  distributed over three seasons, mortality confined to two, and the relative intensity of these patterns is the reverse of Case 2.  Here, because of the  exponentially diminished value of g^, intensive fishing must begin in the third quarter (FS 3) to offset the increased loss from natural mortality.  In summary, the question of when to fish cannot be answered on the basis intuition or the appearance of the growth-mortality pattern. action between K, g^ and mathematical solution.  The inter-  m is too complex, and the decision requires a  


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items