Improving Interpretability of Multi-Objective Tradeoff Sets for Environmental Systems

As environmental systems become increasingly constrained from climate change and other factors, managers must balance multiple conflicting objectives. Multi-objective evolutionary algorithms (MOEAs) are able to generate sets of nearly Pareto optimal solutions that quantify varying levels of conflict among planning objectives. These sets often contain hundreds of alternative solutions with many objectives and decision variables. Understanding relationships between these multiple dimensions is both critical and challenging. This presentation summarizes recent research that seeks to aid interpretability of such sets, with the goal of aiding managers' ability to effectively use MOEAs for decision support. We first describe how multivariate regression trees provide coherent groupings of alternative solutions, providing insight on the relationship between decisions and the planning objectives. Subsequently, we introduce an open source software framework that allows users to easily create interactive parallel coordinates plots to visualize tradeoffs. The framework facilitates so-called clutter reduction techniques to improve interpretability even given many solutions. Both of these projects have been carried out in collaboration with groups of relevant stakeholders, enabling further research partnerships for discovering innovative solutions and furthering environmental sustainability.