SEVENTH FRAMEWORK PROGRAMME THEME 7...

SEVENTH FRAMEWORK PROGRAMME THEME 7 Environment

Collaborative project (Large-scale Integrating Project)

Project no: 246 933

Project Acronym: EURO-BASIN

Project title: European Basin-scale Analysis, Synthesis and Integration

Deliverable 5.5 Preliminary progress report on predicting impacts of changing climate and fisheries: on predator-prey spatial distributions and trophic interactions

Contributors: Brian MacKenzie (DTU Aqua) with contributions from CLS, IMR, IFREMER, USTRATH, NMFRI, MRI-HAFRO

Due date of deliverable: Oct 2013 Actual submission date: Oct 2013 Organisation name of the lead contractor of this deliverable: DTU Aqua

Start date of project: 31.12.2010 Duration: 48 months Project Coordinator: Michael St John, DTU Aqua

Project co-funded by the European Commission within the Seventh Framework Programme, Theme 6 Environment

Dissemination Level PU Public X PP Restricted to other programme participants (including the Commission) RE Restricted to a group specified by the consortium (including the Commission) CO Confidential, only for members of the consortium (including the Commission)

EURO-BASIN | D5.5 Report on predicting impacts of changing climate and fisheries, MacKenzie et al., 2013

2

Deliverable 5.5. Preliminary progress report on predicting impacts of changing climate and fisheries: on predator-prey spatial distributions and trophic interactions is a contribution to Task 5.3 (leader: DTU-AQUA) The purpose of this task is to describe and predict the effects of changes in fisheries and climate on trophic pathways and carrying capacity of ecosystems for major pelagic fish stocks. In order to fulfil this task, scenario modelling will be carried out to predict expected changes in trophic relationships due to climate change based on predictions of future habitat and prey availability (WP3 and WP6) and changes in fisheries of the key species. The aim is to determine what the ecosystem wide impacts of such changes might be, when they propagate through the food web, in particular in terms of changes in carrying capacity. Different modelling approaches will be used, ranging from qualitative to quantitative models, while model development itself will be limited to extension and implementation of existing models. T5.3.1 Future projections of Trophic controls: Tuna. Responsible: CLS; Participants: DTU-AQUA Start month 24, end month 47 T5.3.2 Future projections of Trophic controls: Atlanto Scandic herring, blue whiting and mackerel. Responsible: IMR; Participants: IFREMER; USTRATH; MRI-HAFRO; DTU-AQUA Start: Month 24; End Month 47

Contents: Executive Summary: ............................................................................................................................................ 2 Relevance to the project & potential policy impact: ............................................................................................. 3 Access to Data and/or model code: ..................................................................................................................... 3 Section 1: Preliminary scenario impacts of albacore prey fields and albacore distribution ................................. 6 Section 2: Scenario impacts on bluefin tuna distribution and predation impact on prey and food web............... 7 Section 3: Extension of assessment to pre-assessment time ............................................................................. 8 Section 4: A Model of Blue Whiting (Micromesistius Poutassou) Population Dynamics in the NE Atlantic ...... 21

Executive Summary:

This deliverable presents progress towards development of an improved scientific basis for understanding how climate change and fisheries impact the trophic pathways and ecosystems of the North Atlantic. The focus within this task is primarily related to the influence of consumers on their resources (i. e. “top-down effects”), rather than the influence of primary production on higher trophic levels, which is being covered in other work packages. The work in wp5 builds on distributional and abundance data for large and small pelagic species, and the migration behaviour of several of these species being investigated in other work tasks (especially Tasks 5.1 and 5.2). New field data on distributions and diets and models of food webs and migration behaviour are being integrated to enable simulation and scenario studies of how these predator species affect each other and lower trophic levels in north Atlantic food webs. This work is being supported by development of new historical reconstructions of past biomasses (from 1950 forward) of several forage species to assist with model parameterisations and optimisations for future scenario runs. Other input data will be provided by especially WP6 Basin-scale Modelling. Results will underpin activities in WP 6-8.


3

Relevance to the project & potential policy impact:

This deliverable presents progress towards development of an improved scientific basis for understanding how climate change and fisheries impact the trophic pathways and ecosystems of the North Atlantic. The new understanding of how climate change and fisheries affect trophic pathways and ecosystems has potential fishery and ecosystem management policy relevance: fisheries directly impact the abundance and spatial distribution of numerous large and commercially important pelagic fish species in the North Atlantic. Consequently fisheries can affect trophic flows, and the vulnerability of food webs and ecosystems to future perturbations such as climate change and species introductions that could affect their ability to provide goods and services (e.g. fishery yields, employment, biogeochemical cycling) in future. The project may also be able to contribute to the development of new indicators of pelagic ecosystem state, and how these indicators might be influenced by fisheries and climate change.

Access to Data and/or model code:

Input data sets and model codes are available from partner institutes.


4

Report: Work tasks covered by this deliverable according to “Description of Work”: -T5.3.1 Future projections of Trophic controls: Tuna Responsible: CLS; Participants: DTU-AQUA Start month 24, end month 47 - T5.3.2 Future projections of Trophic controls: Atlanto Scandic herring, blue whiting and mackerel Responsible: IMR; Participants: IFREMER; USTRATH; MRI-HAFRO; DTU-AQUA Start: Month 24; End Month 47

Background, progress and current status: Trophic controls depend on an overlap in space and time of predators and prey. As a result, future trophic controls will depend on how individual species change their spatial distributions as both abiotic and biotic conditions change in their habitats. For example, new hydrographic conditions could increase or decrease the time-space windows of overlap between predators and prey, depending on species-specific physiological sensitivities to new conditions. Hence past trophic controls may accentuate, diminish or shift to new locations. This workpackage focuses on how the trophic impacts of the species considered here (herring, blue whiting, mackerel, albacore and bluefin tuna) will change under future exploitaiton-climate scenarios, and in particular how those trophic impacts might affect energy flows at lower trophic levels, including the potential impact on prey species biomasses, and ultimately on carbon sequestration (via integration and collaboration with wp6). The species considered here are some of the commercially and ecologically (in terms of flows of energy and matter) most important species in the region, and consequently those most likely to be directly influenced by human activity in the coming decades. However it is important to realize that the species considered here are only a small subset of those in the north Atlantic, and that many other fish and other species (e. g., squids), could have major influences on trophic flows through food webs. Many of these species are however not (or only lightly) commercially exploited, and their biomasses and food web roles are therefore less directly influenced by fishing activity and / or unknown. Predictions of density distributions of main oceanic predators are key information required for analyzing changes in oceanic foodwebs. Other key informations are the diets of consumers, and how these change in time and space. The work in this sub-task therefore depends on inputs from other sub-tasks in this wp, and on inputs from other wp’s within the project. Within the wp, historical abundances, spatial distributions and dietary information are being used to estimate consumption rates of prey species, and via food web models, impacts on lower trophic levels in the food web. Moreover, new process-based models of the spatial distribution of the key fish species are being developed which take into account factors such as the spatial distribution of prey, abiotic conditions (e. g., temperature, salinity) and the migratory behaviour of the species (see also Deliverables 5.1 and Deliverable 5.3 for details). The new models will be used to conduct simulations of trophic impacts (e. g., on immediate prey species, possible trophic cascades, etc.) by the large and small pelagic fish species under future climate change and exploitation scenarios. The scenarios chosen by the BASIN project for evaluation in wp5-8 were determined at the project coordination meeting, Oct.22-24, Istanbul, Turkey. Since all the IPCC emission scenarios tend to have similar climate change consequences in the period considered by EuroBASIN (2000-2040), it is likely not necessary to use more than one emission scenario. The climate scenario selected was

http://eurobasin.dtuaqua.dk/eurobasin/documents/deliverables/D5.1%20Report%20based%20on%20existing%20and%20knowledge%20assembled%20in%20EURO-BASIN%20online%20version.pdf

http://eurobasin.dtuaqua.dk/eurobasin/documents/deliverables/D5.3%20Report%20on%20top%20down%20control%20on%20key%20pelagic%20species.pdf


5

the IPCC “8.5” scenario to 2040. This scenario was combined with the exploitation levels shown below

Exploitation levels Climate scenario (IPCC terminology)

MSY for top predators (tunas) and forage species 8.5

MSY for top predators (tunas) and 0.5MSY for forage species (e. g., to conserve forage fish for carbon sequestration)

8.5

0.5MSY for top predator and forage fish 8.5

Other technical details of the scenarios are given below. Time scale and horizon: Yearly time steps up to 2040. Spatial scale: the aim is to model the whole North Atlantic basin, which has been accomplished for the biology. Spatial resolution is defined in terms of ICES zones and national fleets. One climate model output for forcing biological and oceanographic models has been received from WP6. This simulation (MEDUSA) provided by the National Oceanography Centre (kindly from A. Yool, NOC, UK) is at intermediate resolution (1°) but with a projection of future environmental conditions over the next century according to the scenario A2 of CO2 release. Activities in the rest of the reporting period will focus on model integrations and developments, and the execution of model runs for selected scenarios. The following sections present details of work activities being conducted within this WorkPackage.


6

Section 1: Preliminary scenario impacts of albacore prey fields and albacore distribution (by Patrick Lehodey, CLS)

A hindcast simulation of north Atlantic albacore has shown that both environmental variability and historical fishing activity interact together with substantial impacts on abundance and distribution of this species (Figure 1.1).

Mean adult albacore distribution in the

1970s

Mean adult albacore distribution in the 1990s

Figure 1.1: Mean predicted density of adult north Atlantic albacore for the decades 1970s and 1990s with observed total catch proportional to circles (from WP5 Task 5.1). A first climate change simulation based on the IPSL CM4 Earth Climate model was made available at the beginning of the project allowing to start the activity concerning tuna modelling with SEAPODYM. The climate projection is the IPCC SRES A2 scenario, i.e., an increase of atmospheric CO2 concentrations reaching 850 ppm in the year 2100. The provided fields included oceanic biogeochemical variables obtained from a off-line simulation with the biogeochemical model PISCES (Aumont and Bopp 2006). This set of variables has been successfully used for a SEAPODYM application to South Pacific albacore population and thus will facilitate the development of similar analyses in the Atlantic Basin. In particular, the long historical forcing was useful to test the sensitivity of the model to its stock - (larval) recruitment relationship, the parameterization of which controls the fate of the stock in the future projection, together with the projected fishing effort (Figure 1.2). This latter is an average of the last five years. Other Earth Climate model runs provided by WG6 should be used to test the variability of response to different model forcings, since the tuna model is also sensitive to some environmental variables (e.g., dissolved oxygen concentration) for which Earth Climate models coupling physics and biogeochemistry still provide divergent results.


7

Figure 1.2: Example of results achieved for the south Pacific albacore tuna stock with SEAPODYM and the IPSL CM4 -A2 climate simulation. Four simulations with different slopes of the stock- (larval) recruitment relationship provided similar fit to catch data in the optimization time window (1980-2000) due to compensation by slight changes of other parameter values (e.g. natural mortality). They are tested from a long historical simulation starting in 1900 allowing to reach an equilibrium state before the beginning of industrial fisheries. Only one (black curve) provides good fit to independent fishing data (1960-1980) not used in the optimization time window, as shown by the comparison of maps of spatial correlation. The projections based on these different solutions start to differ after two decades. The red curve is the estimate from the Tuna Commission (WCPFC) stock assessment (1D) model. The projection used the average fishing effort of the last 5 years.

Section 2: Scenario impacts on bluefin tuna distribution and predation impact on prey and food web (Patrizio Mariani, Brian MacKenzie; DTU Aqua; Patrick Lehodey, CLS).

The scenarios intended for bluefin tuna will be based on the migration, distribution and trophic models being developed in Tasks 5.1 and 5.2 (see also Deliverables 5.1 and Deliverable 5.3 for details). Prey fields provided by SEAPODYM and ERSEM are planned to be used to provide spatial distributions of prey biomass.

http://eurobasin.dtuaqua.dk/eurobasin/documents/deliverables/D5.1%20Report%20based%20on%20existing%20and%20knowledge%20assembled%20in%20EURO-BASIN%20online%20version.pdf

http://eurobasin.dtuaqua.dk/eurobasin/documents/deliverables/D5.3%20Report%20on%20top%20down%20control%20on%20key%20pelagic%20species.pdf


8

Section 3: Extension of assessment to pre-assessment time (Jan Horbowy, NMFRI, Gdynia, Poland)

1. Introduction

Analytical assessment of stock size usually covers period for which reliable catch-at-age data are available. Such period depends on the stock, in most cases it refers to 3 – 4 last decades, and only for a few stocks (e.g. Arctic cod) it extends to end of 1940s. To fulfil main goal of the project: “to advance our understanding on the variability, potential impacts, and feedbacks of global change and anthropogenic forcing on the structure, function and dynamics of the North Atlantic and associated shelf sea ecosystems as well as the key species influencing carbon sequestering and ecosystem functioning ... “ the dynamics of some stocks from north-east Atlantic has to be evaluated for the period when only catch volume is available, while age-structure of the stock/catches is lacking. Thus, an attempt to extend assessment into pre-assessment era was undertaken for the following stocks (in parentheses the period covered by analytical assessment):

o Mackerel (1972-2011) o Horse mackerel (1982-2011) o Blue Whiting (1981-2011) o Icelandic herring (1987-2011) o Anchovies (1987-2011)

The goal was to extend assessment of the above stocks to the period from 1950 onwards, using only total catch volume (other assessment data are not available). The method which could be applied for such an analysis is the one developed by Eero and MacKenzie (2011). The method uses concept of surplus production rate (SPR), which is assumed to be density independent. So, at least theoretically, the method may be used mainly for stocks with low dynamics or density independent SPR. To release these assumptions an attempt was undertaken to develop, test, and apply some other methods for evaluation of stock dynamics in pre assessment era.

2. Data and methods

The catches for considered stocks covering period from early 1950 were provided from ICES data base. Simple method for extension of biomass estimates into pre-assessment time was presented by Eero & MacKenzie (2011). In their method biomass in year y, By, is estimated from the following equation

SPR

CBB

yy

y

1

1, (1)

where C=catch, SPR=surplus production rate. The surplus production rate in given year is estimated from the formulae

y

yyy

yB

CBBSPR

1, (2)


9

using catch data and biomass estimates from most recent analytical assessment. Next, the estimates of surplus production rates are averaged over years and such average rate is used in equation (1) to derive backwards estimates of biomass for years not covered by analytical stock assessment. Eero & MacKenzie (2011) provided some justification that in specific cases the SPR may be only little dependent on time and it may be assumed constant in formulae (1). However, in general the surplus production rate is density dependent as may be seen from the theory of stock - production models. For example, in case of the discrete Schaefer (1954) model the change of biomass within a year y is

, (3)

where E is fishing effort, H, B∞, and q are parameters (intrinsic rate of increase, asymptotic biomass, and catchability, respectively).

Thus, surplus production rate is

, (4a)

and the rate is linearly decreasing with biomass. Similarly, for Fox (1970) and Pella & Tomlinson (1969) models the SPR can be presented, respectively, as

(4b)

and

(4c)

The formulae 4 a-c show that the surplus production rate is biomass dependent if population growth is described by classical stock-production models.

Extension of Eero & MacKenzie (2011) approach (surplus production rate method)

The relationship between “observed” SPR (based on biomass estimates from analytical stock assessment) and biomass may be tested and if such relationship exists it may be used to estimate biomass in pre-assessment era. Thus, in addition to Eero and MacKenzie (2011) method, two other approaches were considered, in which SPR was dependent on biomass linearly or logarithmically

(5a)

and

, (5b)

where a and b are parameters to be determined from the observed SPR and biomasses. Coupling equations (2) and (5) we obtain

(6a)


10

and

(6b)

for two cases of density dependence of SPR. Each of the equations (6) may be solved for By to obtain its estimate when By+1 and Cy are available. The process of solving the equations and estimating the successive By values may be continued backwards until the year with earliest available catch. Solving eq. (6b) for By may be done only numerically, and it may be easily implemented in a spreadsheet. .

Using stock-production models to estimate biomass in pre-assessment time

The Schaefer stock-production model was used as a tool for estimation of biomass in pre-assessment time. First, the model was fitted to current assessment results from ICES, i.e. to the observed catches and estimated biomass. For the fishing effort, the fishing mortalities averaged over ages within the year were taken. The model parameters were estimated by minimizing sum of squared differences between logged observed and estimated catches and biomasses. The biomasses provided from ICES assessment were treated as “observed” biomasses. The Fletcher (1978) parameterization of the Schaefer model was adopted. The relation between By+1 and By in Schaefer model is

(7)

while Cy approximately equals

(8a)

and from (8a) Ft may be replaced by

. (8b)

Thus, having parameters of Schaefer model and the estimates of By+1, and Fy derived from eq. (8b), the equation (7) may be solved for By numerically.

Another production-model based approach used the biomass formulae from production model modified similarly as the stock numbers in Pope’s (1972) cohort analysis. Assuming that catch takes place exactly in the middle of the year, the biomass in half of year is

(9a)


11

and biomass at the end of the year is

(9b)

In backward calculations first By+1/2 is estimated from eq. (9b) and next By is obtained using eq. (9a). Numerical approaches must be used to solve equations (9a, b) for biomasses By+1/2 and By.

For both surplus production rate methods and the stock-production models method the backward calculations for biomass reconstruction, may be replaced by forward estimation procedure. Then, the main parameter in question would be initial biomass, B0, i.e. biomass in first year when catch volume data are available. Equations 6, and 9 may be used to estimate By+1 from values of By, Cy

and parameters. Thus, assuming B0 we can estimate sequence of By and selection of B0 should be such, that biomasses calculated for years when we have analytical assessment would be as close to these analytical biomasses as possible. That could be done by minimization of sum of squared differences between modelled and already available estimates of biomass.

To perform the calculations the excel macros and visual basic programs were developed.

So far, the procedure was applied to the following stocks: Blue Whiting, Mackerel, and Horse mackerel.

First, the procedures were tested for given stock, and the test consisted of the following steps :

- Separation of the data from analytical assessment (here by data we mean catches, biomass and fishing mortality estimates) into two approximately equal time periods: UPPER (covering most recent data), and LOWER (covering earlier data, ALL will be used to denote data from whole analytical assessment)

- Fitting the SPR and stock-production models to the UPPER part of data from analytical assessment,

- Reconstruction of the biomass for the LOWER part of the analytical assessment data, - Comparison the reconstructed biomasses with biomass estimates known from analytical

assessment and drawing the conclusion on how different methods performed in such tests.

Next, taking into account the performance of the methods in testing procedure, the biomass was reconstructed, using ALL assessment data or data from selected period (e.g. LOWER). The final estimate of the reconstructed biomass was the weighted average (inverse of variance weighting) of the estimates derived from applied methods. If some of the methods performed unrealistically or very badly, they were excluded from the average.

3. Results

3.1.Mackerel

ICES assessment of mackerel covers years 1972-2011. The available catch volume data extend backwards to 1950. The task is to reconstruct (estimate) stock biomass for 1950-1971. To test the reconstruction methods on the available data the time series of assessment data (ALL) was separated into two periods: LOWER, 1972-1991 and UPPER, 1992-2011.


12

The surplus production rate methods produce quite similar dependence of SPR on stock size for LOWER and UPPER data (Fig. 3.1.1). The logarithm of the biomass explains slightly more variance of SPR than the biomass itself and much better fit of SPR is obtained for LOWER than for UPPER data (R =0.85 and R=0.45, respectively). The reconstructed biomass using UPPER assessment data and density dependent methods is lower than the analytical biomass but shows similar trend from 1978 onwards. For earlier years the reconstructed biomass declines while analytical biomass goes up (Fig. 3.1.2). The constant density SPR method produce biomass estimates closer to the analytical biomass than the estimates from density dependent SPR but these estimates also decline for years before 1978.

The stock-production model fits quite well to analytical biomass for UPPER, LOWER, and ALL years and parameters estimates are similar for three data periods (Fig. 3.1.3). Only biomass estimate for the initial year (1972) deviates much from the analytical biomass, the estimates for the next years are close to the analytical values. Thus, Schaefer model fitted to ALL data points was selected as a basis for reconstruction of biomass in 1950-1971. The reconstruction of biomass using UPPER assessment data and the stock-production model performed much better than the SPR methods (only the estimate for 1972 deviates largely form analytical biomass) (Fig. 3.1.2).

The results of reconstructions using all methods are presented in Fig. 3.1.4. The density dependent SPR methods show backward increasing trend of biomass estimate, while constant SPR method shows backward estimated biomass declining to zero. The stock-production model method produces reconstructed biomass similar to density dependent method for all years except the earliest. The results of constant SPR method were considered unrealistic and as a final estimate the weighted average of estimates from density dependent SPR methods and stock-production model method were used.

3.2.Horse mackerel

ICES assessment of horse mackerel covers years 1982-2011. The available catch volume data extend backwards to 1950. The task is to reconstruct (estimate) stock biomass for 1950-1981. To test the reconstruction methods on the available data the time series of assessment data (ALL) was separated into two periods: LOWER, 1982-1996 and UPPER, 1997-2011.

The surplus production rate methods produce somewhat different dependence of SPR on stock size for LOWER and UPPER data (Fig. 3.2.1). The logarithm of the biomass explains slightly more variance of SPR than the biomass itself and better fit of SPR is obtained for UPPER than for LOWER data (R =0.42 and R=0.62, respectively). The reconstructed biomass using UPPER assessment data and density dependent methods deviates more from analytical biomass than the reconstructed biomass using constant SPR (Fig. 3.2.2). None of the methods was able to reproduce very high analytical biomass of horse mackerel in second half of 1980s.

The stock-production model fits quite well for UPPER period, while for LOWER and ALL years the problem was to reproduce high stock biomass in second half of 1980s, effect of very strong year-class of 1981 (Fig.3.2.3). The parameter estimates of the model are not very different for the three data periods. Thus, Schaefer model fitted to ALL data points was selected as a basis for reconstruction of biomass in 1950-1981. The reconstruction of biomass using UPPER assessment data and the stock-production model performed very similarly to the density dependent SPR methods but somewhat worse than the constant SPR method (Fig. 3.2.2).

The results of reconstructions using all methods are presented in Fig. 3.2.4. The density dependent SPR methods show slightly increasing backward trend of biomass estimate, while constant SPR method shows backward estimated biomass declining to zero. The production-model method produces reconstructed biomass showing similar trend to the density dependent methods but the values are lower.


13

The results of constant SPR method were considered unrealistic and as a final estimate weighted average of estimates from density dependent SPR methods and stock-production model method were used.

3.3.Blue Whiting

ICES assessment of blue whiting covers years 1981-2011. The available catch volume data extend backwards to 1950 (catches in 1950s were very close to zero). The task is to reconstruct (estimate) stock biomass for 1950-1980. To test the reconstruction methods on available data the time series of assessment data (ALL) was separated into two periods: LOWER, 1981-1995, and UPPER, 1996-2011.

The application of the surplus production rate methods is problematic as the SPR for UPPER and LOWER periods are very different (Fig. 3.3.1), showing clear change in productivity between periods. The attempts to exclude some data points related to higher productivity did not improve the fit of the SPR to biomass for ALL data series. However, the reconstruction of biomass using UPPER assessment data and density dependent methods produced biomass estimates in LOWER period not very different from analytical biomass. The constant SPR method reproduced LOWER period biomass very well (Fig. 3.3.2).

Similar problems were spotted when testing the stock-production model method. The Schaefer model fits relatively well only to the LOWER data points. For the UPPER and ALL data the fit was close to straight horizontal line (Fig. 3.3.3) and unrealistically high maximum production parameter (equivalent to MSY at equilibrium stage) was obtained. Thus, it was not possible to verify stock-production method with the data, while fit to the LOWER period parameters only, produced too low reconstructed values at the end of 1970s (ca. 1.7 mln tons).

The results of reconstructions using all methods are presented in Fig. 3.3.4. Finally, the Schaefer models was fitted to all data except 2002-2007 (these years were considered as years with higher productivity) while the SPR methods were based on LOWER period data. The constant SPR method, which performed very well in reconstruction tests, when applied to all data produced reconstructed biomass declining to zero and it was considered unrealistic. The weighted averages of estimates from density dependent SPR methods and stock-production model method were used as a final reconstructed values of biomass in 1950-1971. They point at rather stable values of about 3 mln tons and the estimates are very insensitive to initial value of B0.

3.4. Icelandic herring and Anchovies

The work on these stocks will start in Reporting Period 3 (Jan-Dec 2014).

Figures and captions


14

Figure 3.1.1. The surplus production rate against stock biomass (10^ tons) for mackerel in 1972-2011.

Figure 3.1.2. The mackerel biomass (10^3 tons) as assessed by ICES and reconstructed biomasses for 1972-1991 using constant SPR, density dependent SPR, and stock-production model methods (the basis for reconstruction were assessment data for 1992-2011).


15

Figure 3.1.3. The biomass (10^3 tons) from ICES assessment and biomass fitted by Schaefer stock-production model to three time periods (1971-1991, 1992-2011, and 1971-2011) for mackerel.

Figure 3.1.4. The mackerel biomass (10^3 tons) from ICES assessment (1971-2011) and biomass for years 1950-1970 reconstructed by SPR methods and stock-production model.


16

Figure 3.2.1. The surplus production rate against stock biomass (10^ tons) for horse mackerel in 1982-2011.

Figure 3.2.2. The horse mackerel biomass (10^3 tons) as assessed by ICES and reconstructed biomasses for 1982-1996 using constant SPR, density dependent SPR, and stock-production model methods (the basis for reconstruction were assessment data for 1997-2011).


17

Figure 3.2.3. The biomass (10^3 tons) from ICES assessment and biomass fitted by Schaefer stock-production model to three time periods (1982-1996, 1997-2011, and 1982-2011) for horse mackerel.

Figure 3.2.4. The horse mackerel biomass (10^3 tons) from ICES assessment (1982-2011) and biomass for years 1950-1981 reconstructed by SPR methods and stock-production model.


18

Figure 3.3.1. The surplus production rate against stock biomass (10^ tons) for blue whiting in 1981-2011.

Figure 3.3.2. The blue whiting biomass (10^3 tons) as assessed by ICES and reconstructed biomasses for 1981-1996 using constant SPR, density dependent SPR, and stock-production model methods (the basis for reconstruction were assessment data for 1997-2011).


19

Figure 3.3.3. The biomass (10^3 tons) from ICES assessment and biomass fitted by Schaefer stock-production model to three time periods (1981-1996, 1997-2011, and 1981-2011) for blue whiting.

Figure 3.3.4. The blue whiting biomass (10^3 tons) from ICES assessment (1981-2011) and biomass for years 1950-1980 reconstructed by SPR methods and stock-production model.


20

References:

Eero, M., MacKenzie, B.R. 2011. Extending time series of fish biomasses using a simple surplus production-based approach. Mar Ecol Prog Ser 440: 191–202.

Fletcher, R.I. 1978. On the restructuring of the Pella-Tomlinson system. Fish. Bull. 76, 515–534.

Fox, W.W. 1970. An exponential surplus yield model for optimizing exploited fish populations. Transac. Am. Fish. Soc. 99, 80–88.

Pella, J.J. and P.K. Tomlinson. 1969. A generalized stock production model. Bull. Inter-Am. Trop. Tuna Comm., 13:419-496.

Pope, J.G. 1972. An investigation of the accuracy of Virtual Population Analysis using Cohort Analysis. Int. Comm. Northwest. Res. Bull. Int. Comm. Northw. Atl. Fish. 9: 65-74.

Schaefer, M.B. 1954. Some aspects of the dynamics of populations important to the management of the commercial marine fisheries.Bull. Inter-Am. Trop. Tuna Comm., 1:25-56.


21

Section 4: A Model of Blue Whiting (Micromesistius Poutassou) Population Dynamics in the Northeast Atlantic. (C. McCaig , M.R. Heath , D.C. Speirs; University of Strathclyde, Scotland)

4.1 Introduction

The blue whiting Micromesistius Poutassou (Bailey 1982) is a wide ranging species in the northeast Atlantic, with a range that stretches from the Iberian peninsula to the Norwegian Sea. Spawning happens mainly to the west of the British Isles before migration to feeding grounds both north and south of these spawning grounds.

The Atlantic fishery for blue whiting has experienced dramatic changes over the past two decades (Payne et al. 2012), with landings rising to very high levels by the late 1990s and falling back to “pre-boom” levels in the second half of the 2000s.

Here we introduce a model to study the population dynamics of blue whiting across this whole domain, which we will use to study hypotheses around the population changes over recent decades. One proposed cause of the dramatic changes is shifts in the sub-polar gyre (Hatun et al. 2009) causing changes in the marine climate around the spawning grounds west of the British Isles.

4.2 Deliverables

The main deliverable from this work is the model of blue whiting in the northeast Atlantic. The sections which follow describe the current state of this model; the driving data used to run the model; and the work required, as well as challenges anticipated, to achieve a completed model of the stock.

4.3 Domain

Our model covers the domain of blue whiting in the northeast Atlantic. Our main reference for determining this was Bailey (1982). The domain we have chosen is from 40○ west to 40○ east and from 40○ to 80○ north. In addition some areas within that region are omitted. There are known to be blue whiting in the Mediterranean, part of which is covered by this region. In adittion our domain also covers some of the Baltic Sea. Any blue whiting that appear in these two areas are treated as separate populations and we do not consider them in our model. The main blue whiting population west of the British Isles do not appear west of the North-Atlantic Ridge and for this reason the region west of 20○

west and south of 55○ north is also omitted from our model. The reason for extending the domain as far west as 40○ west is to cover the seas west of Iceland and south-east of Greenland, where blue whiting have been seen in surveys.

The model divides this domain into cells 2∘ east and 1∘ north. Using the GEBCO 30 arc-second grid bathymetry data (IOC 2003) grid cells that are 50 percent above sea level are considered land and omitted from our domain. In addition cells with their centre point above sea level are also considered land as this point must be in the sea to allow us to perform particle tracking.


22

4.4 Time

The time period of our model was determined by the driving data available for our physical fields. This covers the years 1988-2006. Updates of the adult development and physical transport happen 52 times per year (every 7 days and 27 minutes). Egg development happens with a shorter timestep since the full egg development stage can happen more quickly than the model timestep. The timestep for egg development is set to one day.

4.5 Physical fields

Physical fields from a Medusa model run (Yool et al. 2011) from NOCS are used to produce driving data for our model. These physical fields describe the hydrodynamics, temperature and plankton abundance within our model.

To investigate the hypothesis that changes in the sub-polar gyre as an explanation for the historical changes in blue whiting levels we need to see these changes in the hydrodynamics from the Medusa run. The changes described by Hatun et al. (2009) represented a reduction in the strength of the sub-polar between 1995 and 2003.

In Figs. 4.1–3 we present flow fields from the Medusa model in years before (1990), during (1997) and after (2004) this reduction in the gyre. These flow fields are all taken from the middle of the spawning season (21/3), when changes in the gyre would be expected to have the biggest impact on recruitment. In these we can see that in 1997 the flows to the west of the British Isles are reduced (shorter arrows) compared to 1990, with faster flows in the North Atlantic moved to the west. In 2004 we see that these flows are somewhat increased, though not returned to the levels seen in 1990.


23

Figure 4.1: Hydrodynamic flow during blue whiting spawning season before reduction in gyre (21/3/1990)

4.5.1 Particle tracking

Particle tracking was performed on the Medusa hydrodynamic results with Ichthyop (Lett 2008), with packets of 100 particles released at the centre of each grid cell at 20m depth, and the proportions moving to neighbouring cells by the end of each timestep used to drive the planktonic transport. For some of the cells within our domain Ichthyop failed to run because the 20m depth point was not within the sea. One of these points was over Dogger Bank and reducing the depth to 10m was sufficient to allow Ichthyop to run. The other points where Ichthyop failed were all much closer to shore and reducing the depth to 5m still would not allow Ichthyop to run so these points were considered land. It is not felt that omitting these points will pose a problem for capturing the dynamics of blue whiting as they mostly inhabit areas on or near the shelf edge and we would not expect to find them in areas of less than 20m depth.


24

Figure 4.2: Hydrodynamic flow during blue whiting spawning season during reduction in gyre

(21/3/1997)


25

Figure 4.3: Hydrodynamic flow during blue whiting spawning season after reduction in gyre

(21/3/2004)

4.5.2 Temperature and biological fields

In addition to the flow fields used for the particle tracking the Medusa model results also give us information about the temperature and the biochemistry of the plankton populations across our domain.

In Figs. 4.4 and 4.5 we present examples of maps of sea surface temperature from the spring (21/3/1990) and autumn (12/9/1990). These maps show that the temperature on the spawning grounds (west of the British Isles) during the spawning season is similar to that on the feeding grounds (for instance, off the coast of Norway) later in the year when intensive feeding by blue whiting is observed in these regions. This suggests that using temperature as a driver for the migration of blue whiting offers some hope for producing realistic dynamics.


26

Figure 4.4: Sea surface temperature during spring (21/3/1990)


27

Figure 4.5: Sea surface temperature during autumn (12/9/1990)

In Figs. 4.6 and 4.7 we present examples of maps of mesozooplankton nitrate levels from the spring (21/3/1990) and autumn (12/9/1990). These maps show that the concentrations, and regions of highest concentration, change during the year. In particular the high concentrations of plankton can be said to roughly follow the temperature change discussed above and as such plankton as a food source also offers promise as a driver for our model.


28

Figure 4.6: Mesozooplankton during spring (21/3/1990)


29

Figure 4.7: Mesozooplankton during autumn (12/9/1990)

The Medusa model is on a much finer grid than our model. To calculate values for our grid we take the average of all the points that lie within one of our grid cells, over the depth range we are interested in (currently 100-300m). The Medusa results have a 5 day timestep and to translate this to our timestep we use linear interpolation. These interpolated results for temperature and zooplankton availability act as drivers for our model with fish attracted to areas of high food abundance and along some preferred temperature gradient.

4.6 Growth

We utilise equal width (1cm) length classes and all individuals are assumed to be at the centre of their length class (0.5cm,1.5cm,2.5cm...). Individual growth is governed by a von Bertalanffy (1957) growth curve,

where Δl is the mean increase in length in a timestep of individuals from length class l, l∞ is the asymptotic length (currently 36cm - from Fishbase for west of Scotland - but this will vary with


30

temperature and food availability), γ is the growth rate (currently 0.000767day-1 - from Fishbase for west of Scotland - this will also vary with temperature and food) and Δt is the timestep.

Growth is achieved by use of tent distributions that give the probability of moving from the current length class to each of the other length classes, up to some notional maximum length lmax (with most of the probabilities being 0). [The use of tent distributions to describe migration in discrete space is described by Gurney and Nisbet (1998) and this is the basis for our description of ‘migration’ between discrete length classes.] lmax is currently chosen to be 75. These tents are chosen to have mean as defined by Equation 1 and a variance determined by σ = α + βl for constants α and β. The tents to achieve this growth are estimated by bisection. A tent is accepted when |Δl - ΔlT| < δ × Δl and |σ -σT| < δ ×σ, where ΔlT, σT are the mean and variance of the tent being tested and δ is a proportional error tolerance (currently δ = 0.01).

This approach means that there is a non-zero probability of individuals moving to shorter length classes, and also of individuals growing to longer than l∞. Individuals effectively shrinking in length is somewhat problematic from an ecological perspective but we find that the probabilities of this happening for lengths below l∞ is small, though it becomes larger for lengths close to l∞. For length classes above l∞ the probabilities that individuals become shorter are greater than the probability that they will grow, though there is still a non-zero probability that individuals will grow. We treat the 0cm and lmaxcm length classes (where we assume individuals are on average 0.5cm and lmax + 0.5cm respectively) as terminal and where a tent has a non-zero probability of moving beyond those length classes (shorter than 0.5cm or longer than lmax + 0.5cm) these are added to the probability of moving to the relevant terminal class.

Figure 4.8: Mean length of a cohort over time and von Bertalanffy length over time


31

In Fig. 4.8 we see mean length of a cohort over time gives a nearly perfect match to the von Betalanffy growth curve. There is very slight divergence initially as our model has growth slightly greater than would be expected from a von Bertalanffy curve (since the proportion that would get shorter than 0.5cm cannot) but even at this stage the two are very close. Over time the two representations become closer so that the two lines are indistinguishable.

4.7 Size dependent maturation

At the start of each year, before the start of the spawning season, fish mature, with a length dependent proportion of the immature individuals in each length class moving into the mature class. The proportions that mature in each length class are set by a cumulative normal distribution. This distribution (currently) has mean of 30 (i.e. the length at which annual probability of maturation is 0.5) and standard deviation of 10.

4.8 Fecundity

Using a fecundity constant of 561∕2, based on a maximum of 150,000 eggs (fishbase), l∞ = 36 (fishbase for West of Scotland), length-weight parameters from SFI (a = 0.0082, b = 2.9, which gives a weight of 267.4g), ÷2 assuming 1:1 sex ratio.

4.9 Egg development

Results for egg development times at different temperatures were taken fron the literature (Coombs and Hilby 1979, Seaton and Bailey 1971, Fluchter and Rosenthal 1965 (via Raitt 1968)). These data are presented in Table 1. Turning the egg development times into rates (rate = 1∕time) and performing a linear regression on these gives rate = -0.08378 + 0.2865 ×temp (Fig. 9), which implies a Q10 coefficient of 4.570686.

Table 4.1: Data (temp. and time) used to calculate formula for egg development rate

temp. (∘C) egg dev. time (d) rate (=1/(egg dev. time)) Source

6 8.54 0.117096 Coombs & Hilby 1979

15 2.92 0.3424658 Coombs & Hilby 1979

8 11.5 0.08695652 Fluchter & Rosenthal 1965

10.5 4 0.25 Seaton & Bailey 1971


32

Figure 4.9: Egg development rate against temperature for all 4 data points (fitted line is rate = -0.08378 + 0.2865 × temp)

Looking at Figure 4.9 it seems there may be an outlier in the 8○ data point (from Fluchter and Rosenthal (1965)). Omitting this datum and recalculating gives rate = -0.02641 + 0.02504 × temp (Fig.4.10), which implies a Q10 coefficient of 3.16418. The Q10 value found by omitting this outlier is “better” (closer to the 2 - 3 range we might expect) so here we us this to estimate the rate.


33

Figure 4.10: Egg development rate against temperature omitting “outlier” (fitted line is rate = -0.02641 + 0.02504 × temp)

4.10 Physical transport

Physical transport of individuals employs a method described by Andrews et al. (2006). Transport of individuals between cells is modelled by updating the system at a series of times {U}, seperated by the transport interval Δt. At an update time the state of the system immediately before the update

(e.g. ) to that immediately after the update (e.g. ) according to


34

Here ( )[ ] is the proportion of eggs (immature fish) [adult fish] in cell y at time t -

Δt that are transported to cell x by time t.

The transport of eggs and larvae (currently for all immature fish but should have some upper length above which fish are independently motile) is completely determined by the hydrodynamics coming from the particle tracking. Transport of larger stages will be mostly determined by temperature and food availability (fish move towards some preferred temperature and more abundant food, as well as diffusing), though still partly influenced by the hydrodynamics. At present the proportion of adult fish that move into a destination cell is 10% of the proportion of particles moved to that cell in the particle tracking, with the effects of food, temperature and diffusion not currently included.

4.11 Preliminary Results

In Fig. 4.11 we present results of spawning stock biomass from the end of the model run (December 2006) and in Fig. 4.12 we present results from 9 months before the end (March 2009). The distributions at these two times is very similar even though we know from survey and fishing data that the distribution of the population is quite different at these two times of year. This is likely to be, at least in part, due to the flow fields, which are currently the only determinant of movement of fish in the model. When movement in pursuit of abundant food and desirable temperatures is included we would expect to see more realistic spatial dynamics of the population.

Figure 4.11: Map of blue whiting SSB (t/grid cell) from the end of model run


35

Figure 4.12: Map of blue whiting SSB (g/grid cell) from 9 months before the end of model run

4.12 Future work

4.12.1 Spatial dynamics

At present the spatial dynamics of the adult blue whiting in our model is determined purely determined by the flow fields. As we can see in Figs. 4.11 and 4.12 this does not capture realistic dynamics. Future work will improve upon this by using the food availability and temperature of the ocean as drivers for the independent motion of the fish. Figs. 4.4–7 give us confidence that these can help produce realistic spatial dynamics.

4.12.2 Growth

Currently growth in the model is determined by a set assymptotic length and growth rate. However, these parameters are likely to vary with temperature and access to food. With time series available for these factors (e.g. 4–7) we can use these to drive changes in the growth of fish in the model. One potential problem with this that will need very careful consideration is the treatment of fish that are above the assymptotic length after a temperature/food mediated reduction. In reality larger fish, close to an existing linfty will not shrink just because there is a shortage of food or reduction in temperature.

4.12.3 Forward run

Our other deliverable is a forward run of this model. At present work has not begun on this as we do not have driving data from the forward run of the Medusa hydrodynamic model (Yool et al. 2011).


36

When we have a tuned model and this driving data it is not anticipated that there will be any major difficulties in producing the forward run.

References:

Andrews, J.M., Gurney, W.S.C., Heath, M.R., Gallego, A., O’Brien, C.M., Darby, C., and Tyldesley, G., 2006. Modelling the spatial demography of Atlantic cod (gadus morhua) on the european continental shelf. Can. J. Fish. Aquat. Sci., 63(5):1027–1048.

Bailey, R. S., 1982. The population biology of blue whiting in the North Atlantic. Adv. Mar. Biol., 19: 257355.

Coombs, S. H., Hiby, A. R., 1979. The development of the eggs and early larvae of blue whiting, micromesistius poutassou and the effect of temperature on development. J. Fish Biol., 14(1):111–123.

Fluchter, J., Rosenthal H., 1965. Beobachtungen über das Vorkommen und Laichen des blauen Wittlings (Micromesistius poutassou Risso) in der Deutschen Bucth, Helgolander wiss, Meeresunters, 12:149-55.

Gurney, W.S.C., Nisbet, R.M., 1998. Ecological Dynamics. Oxford University Press.

Hatun, H., Payne, M.R., Jacobsen, J.A., 2009. The North Atlantic subpolar gyre regulates the spawning distribution of blue whiting (Micromesistius poutassou). Can. J. Fish. Aquat. Sci., 66(5):759–770.

IOC, IHO and BODC, 2003. ”Centenary Edition of the GEBCO Digital Atlas”, published on CD-ROM on behalf of the Intergovernmental Oceanographic Commission and the International Hydrographic Organization as part of the General Bathymetric Chart of the Oceans; British Oceanographic Data Centre, Liverpool.

Lett, C., Verley, P., Mullon, C., Parada, C., Brochier, T., Penven, P., Blanke B., 2008. A Lagrangian tool for modelling ichthyoplankton dynamics. Environmental Modelling and Software 23, 9, 1210-1214.

Payne, M.R., Egan, A., Fssler, S.M.M., Htn, H, Holst, J.C., Jacobsen, J.A., Slotte, A., Loeng, H., 2012. The rise and fall of the NE Atlantic blue whiting (Micromesistius poutassou). Marine Biology Research, 8 (5-6), 475–487.

Raitt, D.F.S., 1968. Synopsis of biological data on the blue whiting micromesistius poutassou (Risso 1810). UNFAO Fisheries synopsis No. 34, Rev. 1.

Seaton, D.D., Bailey R.S., 1971. The identification and development of the eggs and larvae of the blue whiting micromesistius poutassou (Risso). J. Cons. int. Explor. Mer I 34 I No. 1 I 76-83.

von Bertalanffy, L., 1957. Quantitative laws in metabolism and growth. Q. Rev. Biol. 32: 217-232.

Yool, A., Popova, E. E., and Anderson, T. R., 2011. Medusa-1.0: a new intermediate complexity plankton ecosystem model for the global domain. Geosci. Model Dev., 4, 381-417.

Date post:	07-Oct-2020
Category:	Documents
Upload:	others
View:	2 times
Download:	0 times

SEVENTH FRAMEWORK PROGRAMME THEME 7...

Documents