Understanding complex scientific data visualizations in 2D can be challenging. Virtual Reality (VR) provides an alternative, combining realistic 3D representations with intuitive, natural interactions with data through embodied experiences. However, realistic 3D representations and associated immersive experiences are prone to misrepresentations as they are selectively representative and often leave little room for abstraction. This is particularly challenging for topics such as modeling natural environments where users value realism. We discuss the causes and categories of potential misrepresentations in VR with a particular focus on scientific visualization. We contextualize our discussion by presenting an application prototype that translates ecological model output data into a high-fidelity VR experience that allows users to walk through forests of the future. We also designed and implemented two methods to display uncertainties in high-fidelity VR environments: A multi-scenarios approach to provide users access to alternative scenarios, and a slide-and-show approach to view the environment within the confidence interval.