On the interest of using field primary production data to calibrate phytoplankton rate processes in ecosystem models

In many ecosystem models based on empirical formulations, parameters generally are calibrated in order to achieve the best fit between measured and simulated chlorophyll a standing stocks. An accurate calibration of rate processes as primary production rarely is taken into account. In this paper, we test the usefulness of calibration of phytoplankton photosynthetic processes in an ecosystem model using field primary production data. We used 18 months of photosynthetic process data from the Baie des Veys ecosystem (Normandy, France). Five empirical formulations of photosynthesis–irradiance curve models amongst the most widely used were tested. In each formulation, the variability of photosynthetic parameters (i.e. the light-saturated rate of photosynthesis (P_max^B) and the initial slope of the photosynthesis–light curve (α^B)) was considered depending on environmental factors (temperature and nutrient availability). The fit of the five equations as well as the calibration of parameters on field measurements (i.e. the light-saturated rate of photosynthesis (P_ref^B), the initial slope of the photosynthesis–light curve (α_ref^B), the half-saturation constant for nitrogen (K_N) and silicates uptake (K_Si), and the coefficient in the exponential thermal effect (K_T)) was performed using the whole available data set of P vs. E curves (n = 143, P vs. E curves). Then, the Smith formulation allowing the best simulation of the Baie des Veys primary production and corresponding parameters were introduced in an ecosystem box model. This formulation led directly to a satisfactory representation of the Baie des Veys phytoplankton dynamics without additional calibration. Results obtained were compared with a more classical approach in which ecosystem models were calibrated using published values of parameters. This comparison showed that for the two years studied, annual primary production estimated through the ecosystem model was 13% and 26% higher with our approach than with the more classical approach. This work emphasizes the importance of accurately representing rate processes in ecosystem models in order to adequately simulate production as well as standing stocks.