Fine-scale agricultural fertilizer management promotes global NH3 emission reduction

Abstract

Mitigating NH3 emission from cropland soil is crucial for further improving air quality. Despite crop-specific emission factors (EFs) being provided by several studies, the lack of high-resolution EFs under different human management practices limits the further exploration of global cropland NH3 mitigation potential. Agricultural NH3 mitigation potential varies widely depending on where emissions are released. Thus, comprehensively and systematically assessing impacts on a fine scale is useful when developing strategies to efficiently mitigate the effects of NH3 emission. In this study, we performed several machine learning models on a global NH3 emission factors response dataset to find the relationship between EFs and climate conditions, soil properties, and human management. The random forest (RF) model with an R2 of 0.78 and an RMSE of 0.88 showed the best estimation ability. Our data-driven approach indicated that the EFs were mainly affected by temperature, water input, N placement, crop type, and fertilizer type. The combine-effect of N application rate and temperature need to be further studied since these two variables interacted most in our RF model. Using the RF model, we provided five-arcminute high-resolution NH3 emission factor maps under different management practices. The results showed that under proper management, the global NH3 emission of rice, wheat, and maize production has a reduction potential of 44%, 34%, and 37%, respectively. The effects of different human management practices vary everywhere due to the interaction of environmental conditions and management. Our management-specific EFs can provide insights for fine-scale NH3 emission control.