Remote estimation of nitrogen fixation by Trichodesmium

Abstract

A non-spectral model is described that can be used to calculate N2-fixation rate from remote estimates of Trichodesmium biomass. This model, which is similar to formulations that have been developed for estimating primary production from satellite-derived phytoplankton chlorophyll concentrations, is parameterized using measured Trichodesmium N2-fixation vs. irradiance (I) data and observed subsurface Trichodesmium biomass profiles from the Tropical Atlantic Ocean. These data reveal that the N2-fixation vs. I responses and subsurface distributions of Trichodesmium vary substantially in tropical waters. The calculated rates are sensitive to only one of three forcing variables: the remotely sensed Trichodesmium chlorophyll concentration, BTsat, and two of the model parameters: the maximum N2-fixation rate, PmaxBT, and the depth of the subsurface Trichodesmium biomass maximum, Zm. The model is particularly sensitive to the latter. These results suggest that in order to generate N2-fixation rate estimates with reasonable confidence limits with this model, means must be sought to account for in situ variablity in PmaxBT and Zm. A series of correlation analyses reveal statistically significant correlations between the diffuse attenuation coefficient, Kpar, and PmaxBT, and between wind speed and Zm. These relationships are suggested as potential means of accounting for natural variability in PmaxBT and Zm. An example remote sensing-based rate calculation is made using SeaWiFS-derived Trichodesmium chlorophyll concentration in the South Atlantic Bight described in Subramaniam et al., 2002 (Deep-Sea Research, 2002). Although the optical conditions in the Bight were not all within the range used to derive the model parameters, the model gives rates that are consistent with direct rate measurements in Trichodesmium blooms. Because Trichodesmium biomass can only be detected remotely at relatively high concentrations, efforts to estimate global rates with this model will require the use of both shipboard and satellite data.