Streamflow is a complex physical phenomenon closely related to flooding. Flooding is among the most costly natural disasters so accurate streamflow prediction has the potential to bring substantial social benefits. Streamflow is driven by several factors including stream connectivity, precipitation, snow melt, and the activity of dams therefore, an accurate streamflow prediction system should incorporate all these factors to maximize predictive accuracy. However, due to the time constraints of this summer project, we chose to focus our work on the influence of stream connectivity. The influence of stream connectivity is an interesting subject for our initial exploration because, to the best of our knowledge, prior work on flood and streamflow prediction has not extensively studied the impact of stream connectivity. Incorporating streamflow connectivity into predictions requires a way to model the spatial relationships between stream gauges. We accomplish this by imposing a graph structure on the data and utilizing a graph network to make predictions.