To reveal the comprehensive impacts of controlled release urea (CRU) on rice production, nitrogen (N) loss, and greenhouse gas (GHG) emissions, a research based on global meta-analysis and machine learning (ML) was conducted. The results revealed that the CRU application instead of conventional fertilizer can increase rice yield, N use efficiency (NUE), and net benefits by 5.24%, 20.18%, and 9.30%, respectively, under the same amount of N. Furthermore, the emission of N2O and CH4, global warming potential (GWP), the loss of N leaching, and NH3 volatilization were respectively reduced by 25.64%, 18.33%, 21.10%, 14.90%, and 35.88%. The enhancing effects of CRU on rice yield and NUE were greater when the nitrogen application rate was 150 kg N ha-1. Nevertheless, the reducing effects of CRU on GHG emission reduction, nitrogen leaching, and NH3 volatilization was greater at high nitrogen application rate (≥150 kg ha-1). Mitigating effects of CRU on N2O and CH4 emission were significant when soil pH ≥ 6, while CRU posed a measurable effect on reducing nitrogen leaching and NH3 volatilization in paddy fields with soil organic carbon lower than 15 g kg-1 and pH lower than 6. Based on the data collected from meta-analysis, the results of ML demonstrated that it was feasible to use soil data and N application rate to predict N losses in rice fields under CRU. The performance of random forest is better than multilayer perceptron regression in predicting N losses from paddy fields. Thus, it is necessary to promote the application of CRU in paddy fields, especially in coarse soil, in which scenario the environmental pollution would be decreased and the rice yields, NUE, and net benefits would be increased. Meanwhile, machine learning models can be used to predict N losses in CRU paddy fields.
Keywords: Controlled release urea; Machine learning; N2O emission; NH3 volatilization; Paddy fields.
© 2021. The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature.