Quantifying uncertainties associated with the measurement of primary production

Abstract

Depth-integrated primary production was estimated in the Celtic Sea (European Shelf) by 2 independent methods. Daily production rates were derived from photosynthetic parameters determined by photosynthesis/irradiance (P/E) experiments on samples incubated with 14 C-bicarbonate ( 14 C method) and from in situ profiles with the fast repetitive rate fluorometer (FRRF) method. Production was estimated to be between 332 and 910 mg C m -2 d -1 by the 14 C method and from 351 to 1278 mg C m -2 d -1 by the FRRF method. The uncertainty, or error, of these estimates was determined by calculating 95% confidence intervals using Taylor series and Monte Carlo modelling approaches. Both approaches revealed significant uncertainty in estimates. The Taylor series analysis of the 14 C method indicated that 95% confidence interval widths were between 4 and 20% of the estimated production. The smallest error in this study, calculated using a Taylor series analysis, gave a 95% confidence interval of (578, 622) mg C m -2 d -1 for an estimated production rate of 600 mg C m -2 d -1 . The largest error was calculated using a Monte Carlo analysis and found that a production rate of 377 mg C m -2 d -1 had 95% confidence limits of (300, 453) mg C m -2 d -1 . Larger uncertainties were associated with the FRRF method, but reflect only the increased complexity of equations used to derive photosynthetic parameters and not a known greater degree of error in the approach. The precision of primary production measurements is a neglected but important factor that should be considered in primary production experiments and in testing satellite remote sensing algorithms.