Internal meeting of WP4 May 2016, Öregrund
GOHERR Expert round held at SLU 13.05.2016
Purpose of the 1st expert round was to map potential management alternatives related to SLUs work in GOHERR and plan what materials could be provided for the integrative modelling (herring – salmon – human dioxin intake system’s analysis) during 2016. First experts were interviewed one by one, but in the end they could comment on each other and discuss when writing common notes.
Participants and their roles:
Magnus Huss, domain expert (PhD, Docent, Fish ecology, SLU)
Philip Jacobson, domain expert (PhD student, Fish ecology, SLU)
Anna Gårdmark, domain expert (Professor, Fish ecology, SLU)
Andreas Bryhn, facilitator (PhD, Docent, Environmental analysis, SLU)
Annukka Lehikoinen, modeler / system’s analyst (PhD, Environmental sciences, UH)
Content:
The group listed management options (MOs) / environmental changes that would potentially affect the dioxins concentrations in Baltic Sea herring and further on (via herring) in salmon. AB and AL interviewed each of the experts to map their thoughts about the mechanisms through which the experts saw the regulatory potential of these MOs. Potential ecological side effects of the MOs were discussed too. Available materials and models, as well as the knowledge gaps were listed.
Listed / considered management options were:
1. Size-selective herring fishing, where small individuals are targeted (< 17 cm). Smaller / younger fish include less dioxin. Increased mortality could also increase herring growth, which in turn decreases their dioxin uptake (also in salmon, but only if the salmon eats herring).
2. Decreasing herring density by intensified herring fisheries. This would lead to higher growth and thus decreased dioxin uptake.
3. Increasing cod density. This would potentially lead to higher predation on herring (which has the same effect as intensified herring fisheries in alternative 2).
4. Changing food recommendations (area specifically?). Interesting is, how the formulation of the recommendations affects to dioxin intake (number of fish vs. size of fish). Also, as the food-web structure differs between the Baltic Sea basins, area-specific recommendations might be justified. a. Increase: i) Amount of fish in diet, ii) Size recommendation in diet; b. Decrease: i) Amount of fish in diet, ii) Size recommendation in diet.
5. Decrease dioxin intake by developing dioxin-free fish products. This way people could get the health effects of eating fish but without dioxins.
6. Decrease eutrophication. This would lead to smaller total biomass in the system and thus decreased biological dilution of dioxins, adding dioxin concentration in fish. Thus this management action should maybe be taken into account when planning fisheries-related management or agreeing on the dietary recommendations.
The group is going to continue their work by organizing an internal two-day workshop in the end of August 2016 in Öregrund, where the knowledge of the expert group is going to be combined to a system’s analytic Bayesian Network (BN) model. In practice, this is a preliminary sub-model of what happens between the input and output variables marked for the SLU in the current governance model, and where in this system different MOs would affect and how. This BN should integrateable to the GOHERR governance model - an aspect that will be kept in mind when building the model. Another issue is that the results by PJ should be relatively easy to read in the sub-model when they are ready (in the very end of the project). The planned main result of the 2-day workshop is the defined structure of the model and a plan on how each of the conditional probability tables are going to be populated (filled) + schedule for that. For some parts of the model we have quantitative estimates (based on modeling) with varying confidence levels, but in some parts of this complex entity only rough qualitative estimates on the direction and type of the dependencies between some variables can be provided. Anyway the model will be based on the best available knowledge. BN as method is suitable for integrating quantitative results with qualitative estimation, still producing numerical output. As all the knowledge is handled in probabilistic form, varying confidence levels related to different elements of the system can be included. Some parts of the model will be improved / updated in the end of the project based on the results by PJ.