A data-driven modeling approach for simulating algal blooms in the tidal freshwater of James River in response to riverine nutrient loading

Abstract

Algal blooms often occur in the tidal freshwater (TF) of the James River estuary, a tributary of the Chesapeake Bay. The timing of algal blooms correlates highly to a summer low-flow period when residence time is long and nutrients are available. Because of complex interactions between physical transport and algal dynamics, it is challenging to predict interannual variations of bloom correctly using a complex eutrophication model without having a high-resolution model grid to resolve complex geometry and an accurate estimate of nutrient loading to drive the model. In this study, an approach using long-term observational data (from 1990 to 2013) and the Support vector machine (LS-SVM) for simulating algal blooms was applied. The Empirical Orthogonal Function was used to reduce the data dimension that enables the algal bloom dynamics for the entire TF to be modeled by one model. The model results indicate that the data-driven model is capable of simulating interannual algal blooms with good predictive skills and is capable of forecasting algal blooms responding to the change of nutrient loadings and environmental conditions. This study provides a link between a conceptual model and a dynamic model, and demonstrates that the data-driven model is a good approach for simulating algal blooms in this complex environment of the James River. The method is very efficient and can be applied to other estuaries as well.