Earth System Models underestimate natural terrestrial nitrogen fixation by up to 18%

The research team, led by Professor Shushi Peng from Peking University, has delivered an observation-driven isotope-based estimate of global biological nitrogen fixation (BNF) for natural land ecosystems, revealing substantial biases in widely used Earth System Models (ESMs). Their findings indicated that the BNF in current ESMs may need to be better represented, and the parameterization may need to be constrained by many observations.

Grounded in nitrogen isotope (δ¹⁵N) mass-balance theory of a plant–soil system, the authors derived a theoretical negative relationship between the fraction of plant external N demand met by symbiotic fixation (f_BNFs) and the isotope fractionation of plant N uptake (e_U). Combining this framework with machine-learned global maps built from thousands of plant and soil δ¹⁵N observations, they estimated the spatial pattern of BNF and quantified key environmental controls.

Mean annual temperature (MAT) was identified as the primary predictor of symbiotic fixation, increasing monotonically from cold to warm regions and explaining ~29% of the spatial variability. Natural abundance of ectomycorrhizal fungi (ECM) explained ~14% of the variability, and MAT and ECM could interact to affect the fixation. 

Among the 11 ESMs participating the CMIP6 ensemble runs, only MPI-ESM-1-2-HAM can produce spatial pattern of BNF close to the isotope-based map, while the other models have either too large (ACCESS-ESM1-5) or too flat (CESM2, EC-Earth3-Veg, NorESM2-MM, MIROC-ES2L, UKESM1-0-LL, and TaiESM1) latitudinal gradients.

The team estimated global natural terrestrial BNF at 83.0 (78.2–89.8) Tg N yr^-1, indicating ~18% underestimation in CMIP6 models with a multi-model mean of ~67.7 Tg N yr^-1. To improve the model performance in the next phase of CMIP, the team suggest adding temperature and ECM into the representation of BNF, and also to constrain the parameters with N isotope measurements.

“By embedding real-world isotope constraints into a Bayesian framework, we reveal a larger natural nitrogen inputs than many models assume,” said Prof. Shushi Peng. “Incorporating temperature, abundance of mycorrhizal fungi and isotope constraints could improve model performance substantially.”

 

---

---