A hierarchical bayesian model to quantify uncertainty of stream water temperature forecasts

Area of Science:

• Ecology
• Environmental Science
• Hydrology

Background:

• Accurate freshwater temperature modeling is crucial for understanding climate change impacts on aquatic ecosystems.
• Simple linear regression using air temperature can introduce significant bias in freshwater temperature forecasts.
• Existing models struggle to disentangle seasonality and long-term trends, affecting predictive accuracy.

Purpose of the Study:

• To develop and validate a robust modeling approach for reconstructing and forecasting freshwater temperature.
• To compare the performance of a hierarchical Bayesian time series model against simple linear regression.
• To provide reliable temperature forecasts for ecological analyses and climate change impact assessments.

Main Methods:

• Developed a hierarchical Bayesian statistical time series model incorporating seasonal sinusoidal signals and time-varying parameters.
• Utilized water temperature, air temperature, and water discharge as predictors.
• Compared model fitting and forecasting performance with simple linear regression using simulated and real-world stream data.

Main Results:

• The hierarchical Bayesian model demonstrated superior data fitting and forecasting accuracy compared to linear regression.
• The new model effectively addressed long-term trend biases present in linear regression forecasts.
• Forecasts from the Bayesian model provided a more realistic assessment of future warming trends and associated uncertainties.

Conclusions:

• Hierarchical Bayesian time series modeling offers a more accurate and reliable method for forecasting freshwater temperatures.
• This approach improves our ability to predict climate change impacts on freshwater ecosystems.
• The model's outputs are directly applicable to ecological studies and climate change management strategies.