Modeling the impact of iron and phosphorus limitations on nitrogen fixation in the Atlantic Ocean

Abstract. The overarching goal of this study is to simulate subsurface N* (sensu, Gruber and Sarmiento, 1997; GS97) anomaly patterns in the North Atlantic Ocean and determine the basin wide rates of N₂-fixation that are required to do so. We present results from a new Atlantic implementation of a coupled physical-biogeochemical model that includes an explicit, dynamic representation of N₂-fixation with light, nitrogen, phosphorus and iron limitations, and variable stoichiometric ratios. The model is able to reproduce nitrogen, phosphorus and iron concentration variability to first order. The latter is achieved by incorporating iron deposition directly into the model's detrital iron compartment which allows the model to reproduce sharp near surface gradients in dissolved iron concentration off the west coast of Africa and deep dissolved iron concentrations that have been observed in recent observational studies. The model can reproduce the large scale N* anomaly patterns but requires relatively high rates of surface nitrogen fixation to do so (1.8×10¹² moles N yr⁻¹ from 10° N–30° N, 3.4×10¹² moles N yr⁻¹ from 25° S–65° N). In the model the surface nitrogen fixation rate patterns are not co-located with subsurface gradients in N*. Rather, the fixed nitrogen is advected away from its source prior to generating a subsurface N* anomaly. Changes in the phosphorus remineralization rate (relative to nitrogen) linearly determine the surface nitrogen fixation rate because they change the degree of phosphorus limitation, which is the dominant limitation in the Atlantic in the model. Phosphorus remineralization rate must be increased by about a factor of 2 (relative to nitrogen) in order to generate subsurface N* anomalies that are comparable to the observations. We conclude that N₂-fixation rate estimates for the Atlantic (and globally) may need to be revised upward, which will help resolve imbalances in the global nitrogen budget suggested by Codispoti et al. (2001) and Codispoti (2007).