Machine learning-assisted life cycle environmental sustainability assessment of crop straw biochar carbon sequestration

Crop straw (CS) biochar carbon sequestration has gained widespread attention as a negative emission technology for mitigating climate change. However, the diversity of feedstock types and the complexity of the pyrolysis process present significant challenges to conducting a comprehensive assessment of environmental sustainability. This study employs a life cycle assessment (LCA) approach, using 1 ton of CS as the functional unit, to evaluate the life cycle environmental sustainability of biochar carbon sequestration of three common agricultural CS (rice straw, corn straw, and wheat straw) produced at different temperatures (300–700 °C). The biochar sequestration life cycle stages were categorized into feedstock collection, pretreatment and pyrolysis, and product application. The product application considers both biochar carbon sequestration and by-products. In addition, we developed a machine learning (ML) model that employs random forest algorithms to predict the f influencing the global warming potential (GWP) of CS biochar throughout its entire life cycle and to identify key influencing factors. The results indicate when pyrolysis temperatures are maintained between 600 and 700 °C, CS biochar achieves optimal carbon sequestration while minimizing impacts on ecosystems and human health. The production characteristics of biochar (bio-oil yield, syngas yield, temperature, and biochar yield) are the main influencing factors of GWP, contributing 85 %, followed by the properties of biochar (oxygen, hydrogen, carbon, and ash content), which account for 14 %. In contrast, the characteristics of the feedstock have the least influence on GWP, contributing only 1 %.