The properties and functions of complex systems apply across a variety of domains of science and are at the heart of the solutions to many global crises facing the world today. As such, understanding of complex systems has been increasingly recognized as a core goal of science education. Learning about complexity-related concepts and phenomena is persistently considered difficult for many students--even at the undergraduate level, but traditional pedagogical approaches have been unsuccessful in teaching complex systems effectively. Evidence indicates that agent-based participatory simulations can be promising for this purpose. Using mixed methods, cross-case analysis, we examined how undergraduates experience changes in their understanding of the watershed as a complex ecosystem with the use of a particular agent-based participatory simulation called the UVA Bay Game. While one of the cases yielded evidence of nonsignificant quantitative change between pre- and post-simulation concept maps, this study observed an overall positive increase of complex system understanding through both concept map analysis and narrative reflections on learning. Our findings extend the current understanding of the role of participatory agent-based simulations in teaching and learning about complex systems in classrooms. Implications and limitations are discussed.