Artificial neural networks (ANNs) are useful tools for modeling complex ecosystems because they can predict how ecosystems respond to changes in environmental variables (e.g., nutrient inputs). In addition, ANNs can be used to discover relationships among variables, which aids in the understanding of ecosystem function. ANN models were used to predict phytoplankton blooms in three different sites within the same salt marsh estuary located in South Carolina. We (1) compared ANN models with different architectures, (2) applied sensitivity analysis to identify the importance of input variables, and (3) compared the results from ANN modeling to those obtained using linear models.