Determining & implementing transboundary resource allocation
The recommended allocation option is defined as the one with the greater expected stakeholder satisfaction, which is calculated based on inputs and structure of the Bayesian decision network. Recognizing uncertainties about elicited predictions and satisfaction levels, analysts conduct a sensitivity analysis explore whether the recommended allocation changes depending on the set of inputs used for the analysis. In particular, they run the analysis twice: once using the averaged inputs and then a second time based only the input (from the individual) for each variable that is most favorable for the opposing allocation option (i.e., the option with the lower expected satisfaction under the averaged inputs). If the recommendation changes following the second model run, then the analysts use results from both model runs to calculate the expected value of perfect information. This calculation represents the expected percent increase in satisfaction if the uncertainties about the variables and relationships in the model are fully resolved through further research. This provides a way to check the robustness of the recommended allocation to uncertainty and can lead to recommendations for further research to improve decision-making.
Conducting the sensitivity analysis requires expertise in multi-criteria decision analysis, Bayesian belief networks, and calculating the expected value of perfect information.
Using averaged inputs, expected satisfaction with the optimistic allocation option was 11% greater than the status-quo allocation. Some participants indicated that local farmers and agriculture interests were poorly represented at the workshop. When using only those inputs from the agricultural representative at the workshop, the optimistic allocation remained the preferred option by 10%. The status-quo allocation only became preferred when status-quo favourable inputs were used for at least two of the three ultimate objectives. This indicates that if more evidence becomes available that supports the inputs that favour the status-quo allocation, then this could change the recommendation to following the status-quo. If uncertainty about management effectiveness is completely resolved through additional information, expected satisfaction could increase by up to 5%. This is the maximum expected value of conducting further research to inform the decision model.