Data-based assessment of environmental controls on global marine nitrogen fixation

Abstract

There are a number of hypotheses concerning the environmental controls on marine nitrogen fixation (NF). Most of these hypotheses have not been assessed against direct measurements on the global scale. In this study, we use ~500 depth-integrated field measurements of NF covering the Pacific and Atlantic oceans to test whether the spatial variance of these measurements can be explained by the commonly hypothesized environmental controls, including measurement-based surface solar radiation, mixed layer depth, average solar radiation in the mixed layer, sea surface temperature, wind speed, surface nitrate and phosphate concentrations, surface excess phosphate (P) concentration and subsurface minimum dissolved oxygen (in upper 500 m), as well as model-based Pconvergence and atmospheric dust deposition. By conducting simple linear regression and stepwise multiple linear regression (MLR) analyses, surface solar radiation (or sea surface temperature) and subsurface minimum dissolved oxygen are identified as the predictors that explain the most spatial variance in the observed NF data, although it is unclear why the observed NF decreases when the level of subsurface minimum dissolved oxygen is higher than ~150 μM. Dust deposition and wind speed do not appear to influence the spatial patterns of NF on global scale. The weak correlation between the observed NF and the Pconvergence and concentrations suggests that the available data currently remain insufficient to fully support the hypothesis that spatial variability in denitrification is the principal control on spatial variability in marine NF. By applying the MLR-derived equation, we estimate the global-integrated NF at 74 (error range 51-110) TgNyr-1 in the open ocean, acknowledging that it could be substantially higher as the 15N2-assimilation method used by most of the field samples underestimates NF. More field NF samples in the Pacific and Indian oceans, particularly in the oxygen minimum zones, are needed to reduce uncertainties in our conclusion. © Author(s) 2014.