Calibration of a coupled biological–physical model for prediction in a coastal inlet

We describe a numerical forecast system designed for prediction of physical and biological dynamics of a coastal inlet. It is based on a coastal ocean observatory that was located at Lunenburg Bay, Nova Scotia, Canada. Biological, chemical, optical, and physical measurements were collected from instrumented moorings, weekly sampling and detailed surveys from 2002 through 2007. Here we present a framework for calibration and evaluation of an ecosystem model using data from the summer of 2007. A three-dimensional hydrodynamic model was coupled to a simple biological (Nutrients–Phytoplankton–Detritus) model; a simple model was used so results could be compared directly to observed biological and chemical variables using skill scores as a first step toward data-assimilation modeling. As a complement to this analysis, variability of model output, e.g., the nutrient limitation term, was examined to understand the modeled biological response to the simulated physical environment. Skill scores based on variances in observed and simulated time-series of biological components were also investigated. Coastal upwelling/downwelling simulated through this model has been found to increase modeled biological activity in the bay. Also model skill in reproducing the observed patterns in nutrients and phytoplankton has been increased due to the restoring conditions for biology set up at the open ocean boundaries of the bay.

► We model biological–physical dynamics in a coastal inlet.
► Observations are analyzed to calibrate and evaluate the three-dimensional model.
► Skill scores based on variances define model performance in reproducing observations.
► Increase in modeled biological activity is due to coastal upwelling/downwelling.
► Defining biology at the open ocean model boundaries increases its performance.