Machine Learning\nPredicts Biochar Aging Effects on\nNitrous Oxide Emissions from Agricultural Soils

Biochar\neffects on agricultural soils change over time as biochar\nages. To better understand the long-term impacts of biochar application\non climate change mitigation, the effect of biochar aging on nitrous\noxide (N<sub>2</sub>O) emissions has been widely investigated in field\nexperiments. However, the underlying relationship of N<sub>2</sub>O emissions with biochar properties, fertilization practices, soil\nproperties, and weather conditions is poorly understood. We collected\ndata from 30 peer-reviewed publications with 279 observations and\nused machine learning (ML) to model and explore critical factors affecting\ndaily N<sub>2</sub>O fluxes. We established and compared models constructed\nusing neural networks (NN), support vector regression (SVR), random\nforest (RF), and extreme gradient boosting (XGB). We found that the\ngradient boosting regression (GBR) model was the optimal algorithm\nfor predicting daily N<sub>2</sub>O fluxes (<i>R</i><sup>2</sup> > 0.90). The importance of factors driving daily N<sub>2</sub>O fluxes is as follows: fertilization practices (44%) >\nweather conditions\n(30%) > soil properties (21%) > biochar properties (5%). In\naddition,\nthe aging time of biochar, potassium application rate, soil clay fraction,\nand mean air temperature were critical factors affecting the daily\nN<sub>2</sub>O fluxes. When biochar is initially applied, it can reduce\nN<sub>2</sub>O emissions; however, it has no long-term effects in\nreducing N<sub>2</sub>O emissions. The accurate prediction and insights\nfrom the ML model benefit the assessment of the long-term effects\nof biochar aging on N<sub>2</sub>O emissions from agricultural soils.