Freshwater conservation planning is often confronted with limited data and uncertainty about how populations will respond to different environmental conditions. Despite these challenges, conservation decisions need to be made using the best available information. This challenge is particularly acute for freshwater mussels, many of which are both highly imperiled and poorly known. Conservation decision-making for such taxa can benefit from the use of Bayesian Network models– flexible, extensible tools that are designed to support decision-making under uncertainty, and that can be readily updated as understanding improves. Bayesian Network models are probabilistic models composed of multiple nodes (variables) that contain non-overlapping categorical states. Linkages between each node are described by conditional probability tables that can be parameterized based on empirical information, where available, and/or expert knowledge of the study system. I will discuss how Bayesian Network models can be used support the conservation of imperiled freshwater taxa. I use a case study that applied this modeling framework to estimate the biological response of six imperiled freshwater mussel species to proposed water conservation management actions in the Lower Flint River Basin, Georgia, USA. This model linked stream flow, air temperature, and physical stream characteristics (e.g., stream temperature, dissolved oxygen, physiographic district) to the percent population loss of mussels during an extreme low flow event at the stream reach scale. The relationships between variables were parameterized using a combination of empirical data from the Lower Flint River Basin, data from the literature, flow models, habitat availability models, and expert knowledge. Moving forward, the development of modeling approaches that maximize what can be learned from data limited systems will improve our ability to respond to aquatic conservation challenges and help prioritize information gathering in an often funding limited environment.