Applications of Bayesian modeling to simulate ecosystem metabolism in response to hydrologic alteration and climate change in the Yellow River Estuary, China

The natural flow regime of rivers and their ecosystem health can be substantially altered by various human activities such as damming and reservoir construction. In addition they are also subject to the uncertain effects of climate change. Indeed it is likely that the structures and functions of estuarine ecosystems respond to alterations in river flow regime and climate change simultaneously. Due to the complexity of aquatic ecosystems, functional indicators such as metabolism are required to quantify such ecological responses and establish robust and generalized hydrologic-climatic/ecological response relationships. The Bayesian approach has previously proved to be a flexible and highly valuable approach to model such complex and uncertain environmental systems, especially in data-poor situations. The goal of this study was to develop a Bayesian model to explore how ecosystem metabolism is influenced by alterations to flow and climate change in the Yellow River Estuary of China. The daily metabolism was calculated from measurements of dissolved oxygen taken at 15 minute intervals over a 24 hour period using Odum's classic method. Nine other environmental properties of the water were considered along with two climatic indicators. The samples were taken from the Yellow River Estuary over a 2-year period from April, 2009 to September, 2010. A Bayesian approach was used to simulate the response of the ecosystem metabolism to the significant correlation factors using WinBUGS 1.4 software. Of the 11 variables characterizing freshwater, the freshwater inflow, turbidity, salinity, dissolved oxygen and total daily radiation were identified as significant impact factors for metabolism. The regression coefficient estimation, standard error and P-value of the Bayesian model were compared with the values generated by a multiple linear regression model. The Bayesian model was found to have a narrower confidence interval and higher precision than the multiple linear regression model. The posterior probability distribution of the Bayesian model parameters can be used as the prior probability distribution for subsequent analysis. The Bayesian model proved to be a useful tool to understand ecological responses to the combined action of alterations to river flow and climate change.