Quantitative assessment and mitigation strategies of greenhouse gas emissions from rice fields in China: A data-driven approach based on machine learning and statistical modeling

Abstract

The accurate prediction of greenhouse gas (GHG) emissions from paddy fields is critical for developing mitigation strategies to reduce emissions, while realizing the large-scale prediction of GHG emissions from paddy fields remains to be a challenge. Here, we established machine learning models to predict the GHG emissions from Chinese paddy systems using a dataset including 782 CH4 and 679 N2O emission observations based on 118 published studies across China. Our results identified XGBoost was the most suitable model with the outstanding efficiency and accuracy for predicting both CH4 (R2 = 0.754, RMSE = 0.485 kg ha−1) and N2O emissions (R2 = 0.762, RMSE = 0.423 kg ha−1) from rice fields in China. We found mineral and organic fertilizer rate, irrigation mode, straw returned proportion and tillage depth were key factors in regulating GHG emissions. Specifically, CH4 emissions trended to increase first and then decrease with increasing mineral nitrogen fertilizer rate, with the inflection point delayed under the application of organic fertilizer. On the other hand, N2O emissions continued to increase until the N fertilizer rate reached approximately 150 kg ha−1. The use of organic fertilizer, tillage, straw return in half and full quantity increased global warming potential (GWP) by 80.3 %, 33.8 %, 25.2 % and 111.6 %, respectively. Frequent drainage (FD) was identified as the most promising water management mode, with a higher potential for GHG emission mitigation of 39.5 % compared to continuous flooding, followed by mid-season drainage at 18.4 %. We found the combination of a mineral nitrogen fertilizer rate of 128 kg ha−1, FD water management, without straw, tillage, and organic fertilizer could achieve the most effective GHG emission mitigation, with a GWP of 3.13 Mg CO2 equivalent ha−1. Our findings provided a new insight for predicting GHG emissions from rice fields on a large scale, and offered guidance for mitigating GHG emissions from rice production in China.