Improving Net Ecosystem Production Estimates Using Weighted Regression and Bayesian Statistical Methods for Apalachicola Bay, FL

Abstract:

Since 1956 the Odum model has been used to estimate net ecosystem production (NEP) in streams, lakes, and the open ocean. Over time, adaptations of the Odum model have estimated NEP using varying statistical methods, such as Bayesian, but have not successfully differentiated between physical (i.e., lateral advection by tides) and biological processes, producing poorly constrained errors that influence our interpretation of NEP. The goal of this research is to improve the quantification and understanding of NEP by removing the lateral advection signal and then estimate NEP using two statistical methods, weighted regression and a Bayesian framework. WtRegDO, an open-source R^© package, uses a weighted regression method to remove the contribution of physical processes that influence NEP from dissolved oxygen (DO) time series, thus isolating the biological signal, to then estimate rates of NEP. BASEmetab, another open source R^© package, uses a Bayesian framework to estimate NEP based on DO time series. Here, we will demonstrate the effects of removing the physical process signal from DO time series, referred to as detiding, and then compare outputs of NEP from the two methods, WtRegDO and BASEmetab. To evaluate these methods, we use data from the Apalachicola National Estuarine Research Reserve, FL, which has continuously measured water quality and meteorological parameters in Apalachicola Bay since the mid-1990s. This 540 km² bay is fed by the Apalachicola River and is mostly separated by the Gulf of Mexico by barrier islands. Using detided DO, preliminary results for both methods show that the bay was net heterotrophic during the summer of 2018, but BASEmetab estimates a wider range of average daily NEP. For example, BASEmetab estimates that respiration rates in July were up to 14 mg O₂/L/day greater than estimated by WtRegDO. Improving estimates of modern and historical NEP trends in estuaries after detiding the DO time series helps us identify climatological impacts on NEP, ultimately aiding our understanding of processes that affect economically viable and diverse ecosystems. Future research will include using a controlled volume approach to quantify lateral advection in the bay to verify detided DO time series, as well as air-sea gas exchange measurements to further improve NEP estimations.