Comparison of aquatic system models using outdoor mesocosm data for ecological risk assessment, part II: results and...

Mechanistic effect models are important research and decision-support tools in ecology and ecological risk assessment (ERA). A way to foster model utilization and increase confidence in their results in ERA is to compare different models developed to represent the same experimental system. Here, we present the results of such a ring study conducted to examine the feasibility and capability of four aquatic system models (ASMs) to represent mesocosm data and evaluate whether such models could be used as an extension of mesocosm experiments in support of ERA. The ring study methodology is described in a companion article published alongside this article. The four models we tested were AQUATOX, CASM, StoLaM+, and Streambugs. These ASMs were calibrated to represent the biological communities in mesocosms with (treatment) and without (control) the application of azoxystrobin. Mesocosm data were also available to perform model control and treatment validation. We show that the models, especially two of them, performed very well in reproducing the control dynamics according to the calibration criteria we set up, and the control validation simulations confirmed those scores. Models also reproduced the more evident effects (including indirect ones). We could identify similarities among the models' challenges, crucial data needs, and common knowledge gaps about mesocosm communities. The ring study also highlighted some weaknesses in the defined model performance criteria, which did not account for model differences and resulted in favoring some models over others. Despite the detailed model comparison, we could not quantify to what extent different performances were due to model structures, implemented processes, or calibration techniques. This ring study established a basis for adjusting models to mesocosm studies and proposed possible solutions towards advancing capabilities in using ASMs to inform ERA.