Machine Learning Predicts Biochar Aging Effects on Nitrous Oxide Emissions from Agricultural Soils

Shujun Wang,Jie Li,Xiangzhou Yuan,Sachini Supunsala Senadheera,Scott X. Chang,Xiaonan Wang,Yong Sik Ok

DOI: https://doi.org/10.1021/acsagscitech.4c00114

2024-01-01

ACS Agricultural Science & Technology

Abstract:Biochar effects on agricultural soils change over time as biochar ages. To better understand the long-term impacts of biochar application on climate change mitigation, the effect of biochar aging on nitrous oxide (N2O) emissions has been widely investigated in field experiments. However, the underlying relationship of N2O emissions with biochar properties, fertilization practices, soil properties, and weather conditions is poorly understood. We collected data from 30 peer-reviewed publications with 279 observations and used machine learning (ML) to model and explore critical factors affecting daily N2O fluxes. We established and compared models constructed using neural networks (NN), support vector regression (SVR), random forest (RF), and extreme gradient boosting (XGB). We found that the gradient boosting regression (GBR) model was the optimal algorithm for predicting daily N2O fluxes (R2 > 0.90). The importance of factors driving daily N2O fluxes is as follows: fertilization practices (44%) > weather conditions (30%) > soil properties (21%) > biochar properties (5%). In addition, the aging time of biochar, potassium application rate, soil clay fraction, and mean air temperature were critical factors affecting the daily N2O fluxes. When biochar is initially applied, it can reduce N2O emissions; however, it has no long-term effects in reducing N2O emissions. The accurate prediction and insights from the ML model benefit the assessment of the long-term effects of biochar aging on N2O emissions from agricultural soils.