Different representations of biological nitrogen fixation cause major variation in projected terrestrial biosphere responses to elevated levels of atmospheric CO2

Abstract:

Including a land nitrogen (N) cycle in current Earth system models has led to substantial attenuation of predicted land-climate feedbacks, but the magnitude of this N effect remains highly uncertain. The current magnitude and global change responses of major land N cycle processes are still not well understood. Biological nitrogen fixation (BNF) is one particularly important process, being the largest natural land input of N. However, global terrestrial BNF rates are highly uncertain and models lack observations on which to base their predictions. The current variety of terrestrial biosphere models use a wide array of differing, largely untested BNF representations. We tested the six most widely used formulations within the O-CN model and examined the resulting differences in model predictions both under current atmospheric [CO₂], as well as under future scenarios of elevated atmospheric [CO₂]: a prescribed global map of static BNF rates, two simple empirical relationships between BNF and other ecosystem variables (net primary production and evapotranspiration), two process-based formulations based on plant N status, and an approach following a basic form of optimality of plant N acquisition. We found that the predicted global BNF rates for current conditions were fairly comparable, ranging from 93 to 134 Tg N yr^-1 (median 118 Tg N yr^-1). However, at 587 ppm atmospheric [CO₂], model responses in BNF rates ranged from -5 Tg N yr^-1 (-4 %) to 113 Tg N yr^-1 (+88 %) (median 14 Tg N yr^-1 (+15 %)). As a consequence, future projections of global net primary productivity and carbon storage (increases of different magnitudes), as well as N₂O emission (negative responses or unchanged) differed significantly across the different model formulations. Our results emphasize the importance of better understanding the nature and magnitude of BNF responses to change induced perturbations; particularly through new empirical perturbation experiments.