Structural model sensitivity alters predictions and affects management decision-making ability

• Structural sensitivity of models used to manage the Murray-Darling Basin is high. • Sensitivity appears at many steps in a complex modeling process. • Scenario comparison reduces but does not eliminate sensitivity. • Assessing model sensitivity identifies management risks and opportunities. • Better matching model structure to ecological processes would reduce uncertainty. Ecological modeling is increasingly used to predict the consequences of environmental management under global change, but such models are difficult to develop and test. Poor models can lead to real-world environmental and societal harm. Model structure (i.e., how the system is represented) can contribute significant uncertainty. Here, we analyze structural choices in a widely-used model linking hydrological indicators to environmental outcomes in the Murray-Darling Basin, Australia. We assess the sensitivity of modeled outcomes to the shape of relationships (e.g., linear vs. nonlinear) and choices of aggregation methods in space, time, and ecological groupings. Modeled outcomes within a single ‘historical’ scenario ranged from very low to very high ecological condition based solely on a simple set of modeling choices. Relationship shape, aggregation functions, when aggregation functions were applied, and the length of time over which assessments were made were all highly influential. Comparing 12 climate and management scenarios, sensitivity depended on the amount of water and the distribution of values being aggregated. Common default choices (e.g., linear relationships and arithmetic means) were not conservative and may inflate risk. Common approaches to reduce structural uncertainty, such as an ensemble of various model choices or comparing scenarios, were only partially successful. We recommend grouping by the type of ecological response (e.g., threshold vs . linear) to balance parsimony and realism, as well as reporting several competing structures. Uncritical use of a single structure represents the highest risk of poor decision-making. Explicit testing and reporting of structural uncertainty is essential for robust, science-based environmental management.