Evaluating the impact of watershed development and climate change on stream ecosystems: A Bayesian network modeling approach

• Published in: Water research (Oxford) 2021-10, Vol.205, p.117685-117685, Article 117685
• Main Authors: Qian, Song S., Kennen, Jonathan G., May, Jason, Freeman, Mary C., Cuffney, Thomas F.
• Format: Article
• Language: English
• Subjects: Bayesian network; Climate change; Ecological condition; Regression; Streamflow metrics; Watershed
• ISSN: 0043-1354; 1879-2448
• DOI: 10.1016/j.watres.2021.117685

•The Bayesian Network model developed for this study is integrative, adaptive and updateable.•Stream condition using future scenarios (2025–2065) is predicted to be worse than “good-fair”.•Response in stream condition was more related to changes in precipitation than temperature.•Model outcomes provide actionable science in support of regional water resource management. [Display omitted] A continuous-variable Bayesian network (cBN) model is used to link watershed development and climate change to stream ecosystem indicators. A graphical model, reflecting our understanding of the connections between climate change, weather condition, loss of natural land cover, stream flow characteristics, and stream ecosystem indicators is used as the basis for selecting flow metrics for predicting macroinvertebrate-based indicators. Selected flow metrics were then linked to variables representing watershed development and climate change. We fit the model to data from two river basins in southeast US and the resulting model was used to simulate future stream ecological conditions using projected future climate and development scenarios. The three climate models predicted varying ecological condition trajectories, but similar worst-case ecological conditions. The established modeling approach couples mechanistic understanding with field data to develop predictions of management-relevant variables across a heterogeneous landscape. We discussed the transferability of the modeling approach.