Machine learning-based estimation and mitigation of nitric oxide emissions from Chinese vegetable fields

Abstract

High fertilizer input and nitric oxide (NO) emissions characterize the intensive vegetable production system. However, the amount, geographic distribution, and effective mitigation strategies of NO emissions over Chinese vegetable fields remain largely uncertain. In this study, we developed a data-driven estimate of NO emissions and their spatial pattern in Chinese vegetable fields based on the Random Forest (RF) model. Additionally, we conducted a field experiment in a subtropical vegetable field to investigate the effect of climate-smart practices on NO emissions. The RF model results showed that soil NO emissions from Chinese vegetable fields were sensitive to nitrogen application amount, soil clay content, and pH. The total NO emission from Chinese vegetable fields in 2018 was estimated to be 75.9 Gg NO-N. The urgency to reduce NO emissions in vegetable fields was higher in northern than in southern China. Our meta-analysis and field experiment results suggested that biochar amendment and replacing chemical fertilizers with bio-organic fertilizers were win-win climate-smart management practices for mitigating NO emissions while improving vegetable production. Overall, our study provided new insights into NO emissions in vegetable soil ecosystems and can facilitate the development of regional NO emission inventories and effective mitigation strategies. These findings highlight the importance of adopting sustainable and climate-smart agricultural practices to reduce NO emissions and mitigate their adverse environmental impacts.

Keywords: Climate change; Climate-smart management; Emission factor; Mitigation potential; Random forest model.