Meeting Mintues March 7-8

Forage Fish Meeting Minutes

Date: March 7th – March 8th 2016

Venue: CSIRO, Hobart

1. Attendance

Name

Membership

Ray Hilborn

University of Washington

Beth Fulton

CSIRO

Carl Walters

University of British Columbia (Emeritus)

Tony Smith

CSIRO

Tim Essington

University of Washington

Olaf Jensen

Rutgers University

Eva Plaganyi

CSIRO

Ana Parma

CONICET (Argentina)

Cathy Bulman

CSIRO

Robert Leaf

University of South Mississippi

Ricardo Amoroso

University of Washington

Planned on attending but had to cancel

Will remain group members

Cody Szuwalski

University of California Santa Barbara

Keith Sainsbury

Private Consultant Hobart Tasmania

Andre Punt

University of Washington

2. Background

Two large efforts have been conducted to assess the impacts of harvesting forage fish species on the higher trophic level predators (the Lenfest Forage Fish Report and the Marine Stewardship Council Report on Developing best practice management for low trophic level fisheries – evaluation of harvest strategies). Both initiatives used existing ecosystem models for different regions around the world to assess the tradeoffs between forage fish harvesting and the effects on the broader ecosystem, in particular the abundance of their predators. The general conclusion of these reports was that marine predators and other parts of the food web can, in some instances, be strongly influenced by harvesting of key forage fish at historical exploitation rates. Additional evidence in the Lenfest study came from empirical case studies and reference to meta-analysis showing relationship between seabird reproductive success and forage fish availability. However, both group’s work emphasized analysis of ecosystem models. The MSC group found that the number of indirectly affected groups was generally small at moderate harvest rates (50% of Fmsy, corresponding to about 75% of Cmsy) but increased sharply thereafter in some ecosystems (Smith et al 2011). The MSC study was used to develop new

guidelines for assessing low trophic level fisheries under the MSC standard. The Lenfest group used a management strategy evaluation routine, and found that hockey-stick type control rules led to the best performance when there is estimation error in forage fish stock size used in setting annual harvest levels. The Lenfest report made specific recommendations about the minimum biomass levels used in harvest control rules and maximum exploitation rates and how those could be adjusted based on the amount of information available about forage fish population dynamics and predator dependencies.

In subsequent work, Essington and Plaganyi evaluated the suitability of existing food web models for evaluating consequences of forage fish fisheries on ecosystems. Most models were not built with the explicit intention of evaluating forage fish fisheries, so unsurprisingly many models did not include features of forage fish population biology or food web structure that are relevant for evaluating all fishery impacts. Among the most common features that were absent in most models are: 1) natural variability of forage fish stocks, 2) the extent of size overlap between fisheries and predators, 3) taxonomic resolution of both forage fish groups and predators was often coarse, and (4) very little explicit representation of spatial structure, including localized depletion or natural variation in spatial ranges.

The MSC work also revealed that different ecosystem models for the same region can lead to different results in terms of the impacts of low trophic level fisheries on predators and other parts of the food chain. Consequently, these models are best used to identify high level predictions (e.g. the number of groups potentially affected), but it is commonly difficult to identify which particular species will be positively or negatively impacted.

The goal of the meeting was to review and discuss the main issues regarding the assessment of forage fish and the impacts of fishing them on their predators, and to agree on a work plan to address the issues that are identified as missing from the earlier analyses.

3. Recruitment variability of forage fish stocks

Recruitment variability is a key feature of forage fish populations, although its characterization has proved challenging. A common approach to investigate the presence of regime shifts involves the use of a change point detection technique to identify breakpoints in the recruitment time series and estimate the mean and standard deviation of each regime. In particular, the STARS method (a sequential t-test method) has been applied to identify changes in marine fish productivity and recruitment in synthesis papers (Vert- Pre et al., 2013; Szuwalski et al. 2014). A limitation of the STARS method is that it does not explore the entire segmentation space and results are very sensitive to user input parameters. During the meeting alternative algorithms for change-point analysis were presented and applied to a data set of 55 forage fish stock recruitment time series. In

particular the results of applying the PELT algorithm (Prune Exact Linear Time) and Bayesian change point methods to the data were discussed. Preliminary results indicate that when an AIC approach is followed to select between a biomass driven recruitment and a regime shift hypothesis 47% of the stocks support the regime shift hypothesis. The workshop discussed the need to include a mixed model hypothesis, where a stock recruitment model is fit for each regime.

In the context of evaluating the performance of different harvest strategies the characterization of regime shifts is not nearly as relevant as being able to identify if the observed changes in recruitment are due to variations in productivity or carrying capacity. The group discussed the inability of change point analysis to assess this problem and identified potential methods, algorithms and alternative sources of information that could improve our understanding of the main drivers of recruitment variability. The main alternatives were: (i) the use of Kalman filters, (ii) the implementation of hidden Markov algorithm and (iii) the analysis of the spatial expansion-contraction dynamics of spawning aggregations.

Britten et al., (2016) used a Kalman filter to detect temporal changes in productivity and capacity in a global meta-analysis. However it has been pointed out that simulation exercises showed that Kalman filters do not perform well when abrupt changes occurs (as in regime shift situations) due to the Gaussian nature of the process. It was suggested that an alternative probability distribution with fatter tails (horse shoe) could help to improve the ability to reconstruct regime-shift like trajectories. Additionally, it was suggested that the use of particles filtering methods should be explored.

Cunningham (2015) used a hidden Markov algorithm to assess changes in productivity and carrying capacity on salmon stocks. The limitation of this method is that the number of regimes should be set a priori.

A new approach based on a risk sensitive movement of fish coupled with a range expansion-contraction dynamic was suggested. This would require compiling spatial data of spawning areas and range of diverse forage fish stocks. It was suggested that the main mechanism causing regime shifts may be related to range changes. Under this hypothesis an amplification of environmental variability should be observed as range contracts.

During the workshop an analysis was conducted to identify patterns in productivity and capacity changes across 55 forage fish species: under a Ricker stock-recruitment model either the productivity or capacity were kept fixed and the other was calculated for each point in the time period. Graphs of the temporal changes in the calculated productivity and capacity were produced and visually inspected. The results suggested that some common patterns arose, and more effort should be expended to classify those patterns and use some well-known study cases to assess their validity.

4- Modeling approaches

Ecosystem models have been the preferred tools to analyze the impacts of fishing on predators. During the workshop a simple prey-predator model was presented to illustrate how model structure can affect our understanding of complex interactions and system dynamics. The model used classic predator-prey theory (biomass dynamic, logistic prey growth) assuming that predation rate is a function of prey density (Type 2 functional response). This presentation illustrated the implications of different structural assumptions on how fishing pressure on forage fish affected the predator biomass and how natural mortality (due to predation) and fishing pressure interact. The group agreed that such simple models should be explored further to identify key uncertainties in the structural assumptions, and that empirical data should be used to distinguish between alterative hypotheses and model predictions.

There are several ecosystem model frameworks that can be used to assess the impacts of forage fish fisheries on top predators – e.g. EwE (Ecopath with Ecosim), Atlantis, MICE (Models of Intermediate Complexity in Ecosystems), OSMOSE (individual based models) and a suite of size spectrum models. The group agreed that to further advance the understanding of the tradeoffs between fishing and environment it will be necessary to maintain a diversity of methods and to further explore key mechanisms that are not always well represented, such as natural variability of forage fish and the pattern of size overlap between top predators and fisheries.

It was suggested that given the time-frame of the project and the capabilities of the team a potential strategy would be to (i) modify the available EwE models to further disaggregate the forage fish species into size groups and force the models with recruitment variability and (ii) make use of existing MICE models where forage fish are a significant model component (e.g. of the Benguela, California Current and the Coral Sea).

Topic

EWE

Atlantis

OSMOSE

Size Spectrum

MICE

Structural rigidity

Depends greatly on which elements are included and parameterized

Depends greatly on which elements are included and parameterized

Depends greatly on which elements are included and parameterize d

Only based on size, which can make it quite rigid unless modified using trait- based approaches.

Totally flexible

Recruitment variation

Can force recruitment in multi-stanza (not done for LENFEST or MSC)

Either driven by correlate or directly fed with recruit time series.

Driven with correlates

Can drive with correlates

Totally flexible

Model fitting

Easy but not often done

Very hard

Hard but not impossible

Easy

Easy

Spatial Representation

Implicit in foraging arena equations, foodweb specifications and explicit in Ecospace

Explicitly 3d

Explicitly 2d

Can be done but typically not

Flexible

Size dependent predation

Depends on use of stanzas

Standard, though specifics depend on how groups are defined (groups as age structured vs biomass pools)

Explicitly age and size structured

Yes

Flexible

Specialists vs generalists

Initial diet matrix (effective search and switching)

Prey switching primarily through size- based refugia, and spatial structure

IBM with emergent functional responses (type II can emerge but not always)

Default is generalist, but trait based versions means specialists can be defined

Flexible

Stock Recruit for forage fish and predators

Can be explicit on first stanza or if no stanza then surplus production

Explicit

Emergent

Many have explicit relationships, often force bottom sizes

Yes flexible

Harvest control rules

Either inputs or outputs. Hockey-stick and constant catch were used in the Lenfest Report

All fisheries management rules can be represented.

Typically F based by space

Not clear some quotas some Fs

Yes

Table 1: Comparison between ecosystem modeling frameworks.

5- Trophic interactions

The size structure of both predators and prey is critical to understanding the dynamics of ecosystems. When analyzing the effects of fisheries on predators the overlap in prey size eaten by predators and fisheries selectivity should be considered. During the meeting preliminary results on the changes in food supply for predators and fishing pressure when natural variability is considered were presented. Nine forage fish species from the California Current, the East Coast of the US, and the Gulf of Mexico were used as study cases. For the analysis data on forage fish predators, time series of predator abundance, diet composition and prey size eaten were compiled from the literature. A simulation model for each forage fish

was run to assess the tradeoff between fishing mortality and changes in food available for predators in different size ranges. The model used was age-structured and parameters for each forage fish were compiled from the most recent stock assessment. In order to include natural variability two procedures were followed (i) the model was forced with past recruitment observations and (ii) recruitment was calculated using a stock-recruitment relationship with an error (which was sampled from the residuals from the best stock-recruitment fit). Preliminary results indicated that: (i) there is an important proportion of predators whose diet mostly comprises small fish, which are not targeted by the fishery, and (ii) the changes in food availability for most size-ranges due to fishing are relatively small compared to changes induced by natural variability.

The fluctuating nature of forage fish stocks suggests that top predators have developed strategies to cope with changes in food availability. During the meeting the group discussed some results showing changes in predator abundance/survival and prey abundance. For the nine forage fish species in the US for which the group has collected data on predator abundance time-series, the per capita rate of change for predators was plotted against prey abundance. For almost all cases for which the group have data no relationship was observed. Results on the effects of food availability and a penguin population in South Africa were also shown (Robinson et al. 2015). Although anchovies are the main component of penguin diet, no relationship was observed between penguin reproductive success and prey abundance. However, it was observed that as sardine shifted their spatial distribution and became less abundant in the vicinity of the colonies, penguin adult survival decreased. This case illustrated that the spatial distribution of prey abundance can have major impacts on predator dynamics, and that the impact can be primarily on adult survival rather than breeding success. It also revealed that overall diet proportion alone provides misleading indications of predator responses to reduced food availability. Rather, diet proportions during critical time periods need to be known to identify predator dependencies and sensitivities to reduced forage fish abundance.

Another important issue discussed during the meeting was the diversity of forage fish in the ecosystem and the ability of predators to switch their diet in response to changes in abundance. One of the most striking examples of changes in prey availability is the multi-decadal cycle between sardines and anchovies in some upwelling ecosystems. However during the meeting results from a retrospective study analyzing the covariance and synchrony between sardines and anchovies abundance cycles were presented, and the analysis clearly indicates that sardines and anchovies are not replaceable in terms of biomass. Additionally, the expected large negative correlations between these forage fish species when they fluctuate under a cyclic fluctuation hypothesis were not observed, but instead there exist some negative and some positive correlations. It was pointed out that instead of assessing the relationship between predator’s survival/abundance and the biomass of a single forage fish species, the entire portfolio of prey species should be

analyzed. However, in practice there exist several limitations for an empirical analysis since data from many forage fish species are often unavailable.

6- Harvest control rules

During the meeting the topic of optimal harvest control rules in the presence of high natural variability was addressed. A summary of Essington et al. (2015) that analyzes the role of fishing in the collapse of forage fish stocks was presented and possible improvements in the analysis were suggested. In their paper authors used a simple harvest control rule that involved the closure of the fishery when the biomass was below a threshold. This harvest control rule seemed to perform well in reducing the probability of collapse. However, during the meeting it was discussed that the performance of a utility function using log catch should be considered, since fisheries closures often have severe socio economic impacts. There was an agreement that methods similar to those used in Essington et al. (2015) should be expanded, also considering the age structure of the population and alternative harvest control rules.

It was also shown with an example how dynamic programming could help to provide input on assessing the performance of a harvest control rule in the presence of highly auto correlated recruitment.

Work plan:

The group discussed priority research action that should be completed:

  1. Evaluate a range of alternative methods to characterize the recruitment variability, including Kalman-Filters and Hidden Markov approaches. Produce a summary paper of recruitment in forage fish fisheries. Lead responsibility: Amoroso, Szuwalski, Walters, Hilborn.
  2. Gather information on the spatial changes in the distribution of forage fish. Responsibility entire group
  3. Evaluate spatial changes in distribution of forage fish for implications regarding spawner recruit relationships and variation in availability of forage fish to predators. Lead responsibility: Walters for recruitment, Amoroso for availability to predators.
  4. Expand the data base on prey size and when possible try to get the raw data in order to calculate the mean size of prey eaten and the standard deviation. Lead responsibility: Amoroso, Hilborn
  5. Construct age or size structured predator prey models of major forage fish ecosystems that incorporate the recruitment analysis from research item (a) above, and evaluate harvest strategies. Lead responsibility: Amoroso, Hilborn
  6. Convene a workshop of developers of existing EwE models for scientists to evaluate their model in the context of more realistic

Timeline

scenarios (predator size preference and recruitment variability). Lead responsibility: Fulton

Individuals will begin their analysis and the next step is a conference call likely in June or July to assess progress.

A further group meeting, perhaps in a year’s time would be desirable if funding can be arranged.

References

Essington, T. E. and E. E. Plaganyi. 2013. Pitfalls and guidelines for “recycling” models for ecosystem-based fisheries management: evaluating model suitability for forage fish fisheries. Ices Journal of Marine Science.

Pikitch, E., P. D. Boersma, I. Boyd, D. Conover, P. Cury, T. Essington, S. Heppell, E. Houde, M. Mangel, and D. Pauly. 2012. Little fish, big impact: managing a crucial link in ocean food webs. Lenfest Ocean Program, Washington, DC 108.

Robinson, W. M., D. S. Butterworth, and É. E. Plagányi. 2015. Quantifying the projected impact of the South African sardine fishery on the Robben Island penguin colony. ICES Journal of Marine Science: Journal du Conseil:fsv035.

Smith, A. D., C. J. Brown, C. M. Bulman, E. A. Fulton, P. Johnson, I. C. Kaplan, H. Lozano-Montes, S. Mackinson, M. Marzloff, and L. J. Shannon. 2011. Impacts of fishing low–trophic level species on marine ecosystems. Science 333:1147-1150.

Appendix I: Agenda

March 7
am First Session Recruitment dynamics

Forage fish recruitment patterns: randomness, regimes and what changes, carrying capacity or productivity: Carl, Tim, Ricky, Ray

patterns of total forage variability and covariance patterns among forage species in the US Atlantic and Gulf Coasts, Olaf, Robert

am Second Session Ecosystem Models
How important are between model differences: Tony, Beth … ?

Model predictions of impacts of fishing forage fish on predators: Carl

pm First Session Impacts on predators
Size overlap between predators diet and what the fishery takes … Ricky, Ray Predator population dynamics: Is it only food? Ray

pm Second Session Harvest strategies
Harvest strategies: constant F vs minimum biomass threshold — Carl and Tim A Study Case: penguins and forage fish in South Africa (Eva)

March 8
AM follow up discussion of first day topics

PM First session: design of follow up analysis and data collection
PM Second session: who else to involve, funding, future meetings
The importance of small scale spatial structure and area based foragers. Ray Robert and I can present some preliminary results on

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s