Simulating algal dynamics within a Bayesian framework to evaluate controls on estuary productivity

Abstract

The Neuse River Estuary (North Carolina, USA) is a valuable ecosystem that has been affected by the expansion of agricultural and urban watershed activities over the last several decades. Eutrophication, as a consequence of enhanced anthropogenic nutrient loadings, has promoted high phytoplankton biomass, hypoxia, and fish kills. This study compares and contrasts three models to better understand how nutrient loading and other environmental factors control phytoplankton biomass, as chl-a, over time. The first model is purely statistical, while the second model mechanistically simulates both chl-a and nitrogen dynamics, and the third additionally simulates phosphorus. The models are calibrated to a multi-decadal dataset (1997–2018) within a Bayesian framework, which systematically incorporates prior information and accounts for uncertainties. All three models explain over one third of log-transformed chl-a variability, with the mechanistic models additionally explaining the majority of the variability in bioavailable nutrients (R2 > 0.5). By disentangling the influences of riverine nutrient concentrations, flows, and loadings on estuary productivity we find that concentration reductions, rather than total loading reductions, are the key to controlling estuary chl-a levels. The third model indicates that the estuary, even in its upstream portion, is rarely phosphorus limited, and will continue to be mostly nitrogen limited even under a 30% phosphorus reduction scenario. This model also predicts that a 10% change in nitrogen loading (flow held constant) will produce an approximate 4.3% change in estuary chl-a concentration, while the statistical model suggests a larger (10%) effect. Overall, by including a more detailed representation of environmental factors controlling algal growth, the mechanistic models generate chl-a forecasts with less uncertainty across a range of nutrient loading scenarios. Methodologically, this study advances the use of Bayesian methods for modeling the eutrophication dynamics of an estuarine system over a multi-decadal period.