The release of contaminants into the environment can occur from anthropogenic activities, such as oil extraction and transportation, mining, and industrial processes. Remediation associated with reducing contaminant concentrations, and restoration that improves animals and supporting habitat, are often needed to restore ecosystems to their pre-release, baseline condition. We demonstrated the application of Bayesian Decision Networks (BDNs) with two Natural Resource Damage Assessment and Restoration (NRDAR) case studies. We use a stylized case study of riparian restoration following the remediation of a mine-impacted site to evaluate proposed restoration actions aimed at restoring Song Sparrow (Melospiza melodia) populations to baseline conditions. We then use a settled NRDAR case with implemented restoration in the Upper Arkansas River (UAR, Colorado, USA) to demonstrate the application of BDNs to evaluate and forecast restoration effectiveness for Brown Trout (Salmo trutta) (i.e., restoration effectiveness assessment). The riparian restoration model showed differences in the effects of restoration actions on Song Sparrow populations, with the time to reach baseline generally reduced with increased restoration costs, indicating trade-offs between costs and expected recovery. The UAR model showed recovery of Brown Trout populations (i.e., uplift) in response to improved instream habitat restoration, along with forecasted improvements. While the BDNs we developed were specific to two case studies, the structure is adaptable to a diversity of sites, resources, and actions. We suggest that causal network modeling can provide restoration practitioners with a decision advisory tool useful for a wide range of projects.