Water quality monitoring and analysis can help researchers predict and learn from natural processes in the environment and determine human impacts on an ecosystem. Stream network models are used to analyze the movement of substances or aquatic life along streams and rivers. In order to model stream network data based on stream distance and water flow, two types of moving average models were developed: tail-up models and tail-down models (Ver Hoef and Peterson, 2010). A stream network could be pure tail-up, pure tail-down, or a mixture of both. In this presentation, a Bayesian hierarchical normal intrinsic conditional autoregressive (HNICAR) model is proposed to approximate stream network models for faster computation. This HNICAR model is a flexible framework for modeling spatially continuous data from stream networks and is applied to a real world example of a stream temperature dataset on the Lewis and Willamette watersheds connected to the Columbia River in the western U.S to help predict the overall thermal suitability for fish in that aquatic ecosystem.