Improved estimation of f-ratio in natural phytoplankton assemblages

Statistical properties in non-linear regression models of f-ratio versus nitrate relationships were investigated using a case study conducted at the European continental margin (Project OMEX). Although the OMEX data fit within the family of empirical models introduced by Platt and Harrison (1985) Nature 318, 55–58), it is shown that the discrepancy between experimental and fitted values is larger than expected, assuming that the data are accurate. Since the ambient ammonium concentration plays a leading role in regulating new production, the present analysis was extended to include explicitly the effect of ammonium on f-ratio versus nitrate plots. Experimental results based on controlled ammonium additions were used to express the f-ratio as a function of both nitrate concentration and ammonium inhibition, i.e. f ∗(NO3, α).I(NH4,β). The estimation behaviour of the data set/model combination was analysed by testing the appropriateness of various model functions for f ∗ and I. The best fit was obtained with a sigmoid curve, for which mean values of the random fluctuation are almost commensurate with the estimated uncertainty of the measurements with natural phytoplankton assemblages. Precision of the f-ratio estimates was further assessed from contour diagrams of constant likelihood. The significance, validity limits of the estimated parameters, and the relevance of the proposed model as a predictive tool are discussed. Overall, this empirically determined model results in improved precision of f-ratio estimations.