The seasonal nitrogen cycle in Wilkinson Basin, Gulf of Maine, as estimated by 1-D biological model optimization

Abstract

The objective of this study was to fit a simple ecosystem model to climatological nitrogen cycle data in the Gulf of Maine, in order to calibrate the biological model for use in future 3-D modelling studies. First depth-dependent monthly climatologies of nitrate, ammonium, chlorophyll, zooplankton, detritus and primary production data from Wilkinson Basin, Gulf of Maine, were created. A 6-box nitrogen-based ecosystem model was objectively fitted to the data through parameter optimization. Optimization was based on weighted least squares with model-data misfits nondimensionalized by assigned uncertainties in the monthly climatological estimates. These uncertainties were estimated as the standard deviations of the raw data from the 6-meter and monthly bin averages. On average the model fits the monthly means almost within their assigned uncertainties.Several statistics are examined to assess model-data misfit. Pattern statistics such as the correlation coefficient lack practical significance when data errors are large relative to the signal, as in this application. Thus Taylor diagrams were not found to be useful. The RMSE and model bias normalized by the data error were found to be the most useful skill metrics as they indicate whether the model fits the data within its estimated error.The optimal simulated nitrogen cycle budgets are presented, as an estimate of the seasonal nitrogen cycle in Wilkinson Basin, and discussed in context of the available data. Wilkinson Basin has spring and fall phytoplankton blooms, and strong summer stratification with a deep chlorophyll maximum near 21 m, just above the nitracline. The mean euphotic zone depth is estimated to be 25 m, relatively constant with season. The model estimates annual primary production as 176 g C m −2 yr −1 , annual new production as 71 g C m −2 yr −1 and sinking PON fluxes of 9.7 and 4.7 g N m −2 yr −1 at 24 and 198 m respectively.Areas for improvement in the biological model, the model optimization method, and significant data gaps are identified.

Optimization of the Spatiotemporal Parameters in a Dynamical Marine Ecosystem Model Based on the Adjoint Assimilation

“…The model performance is highly dependent on those parameters whose values are difficult to be assigned, only a few of which are derived from observational data, and most of which are derived from laboratory studies [6]. Data assimilation, in the parameter estimation context, is a systematic method in a way that is consistent with the available observations [7] and has been the focus of many researches over the past decade [8][9][10][11][12][13][14][15][16][17].…”