New approach for predicting nitrification and its fraction of N₂O emissions in global terrestrial ecosystems

Abstract Nitrification is a major pathway of N 2 O production in aerobic soils. Measurements and model simulations nitrification associated emission are challenging. Here we innovatively integrated data mining machine learning to predict rate ( R nit ) the fraction as emissions ${f_{{{\text{N}}_{\text{2}}}{{\text{O}}_{{\text{Nit}}}}}}$?> f N 2 O Nit ). Using our global database on , found that machine-learning based stochastic gradient boosting (SGB) outperformed three widely used process-based models estimating from nitrification. We then applied SGB technique for prediction. The potential was driven by long-term mean annual temperature, soil C/N ratio pH, whereas precipitation, clay content, total N. varied over 200 times (0.006%–1.2%), which challenges common practice using constant value models. This study provides insights into advancing projecting dynamics greenhouse gas approach.