Nitrous oxide(N_(2)O)is a potent greenhouse gas with about 60%of its emissions are attributed to agricultural activities.Its fluxes are influenced by a range of crop-specific factors,such as nitrogenous fertilizer inp...Nitrous oxide(N_(2)O)is a potent greenhouse gas with about 60%of its emissions are attributed to agricultural activities.Its fluxes are influenced by a range of crop-specific factors,such as nitrogenous fertilizer inputs,soil N availability,tillage practices,temperature,pH and soil moisture.These factors interact in complex,nonlinear ways,creating the need for predictive modeling of N_(2)O emissions to both improve understanding and estimation and identify mitigating strategies.This proposes proposes data-driven machine learning techniques,particularly multilayer perceptron and random forest(RF)algorithms,for estimating soil N_(2)O fluxes in a sugarcane plantation under different irrigation regimes and to contrast machine learning results with conventional analytical methods.The findings indicate that RF modeling achieved a coefficient of determination of 87.4%for N_(2)O emission prediction,and identified ammonium,nitrogen nitrate,soil temperature,and water-filled pore space as the most influential predictors,in that order.The results open new possibilities for integrating machine learning to study N_(2)O fluxes in sugarcane and other major crops.All data and code used in this study are provided openly to support further research.展开更多
基金funded by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior,Brasil(Finance Code 001)the partial support provided by EMBRAPA(Brazilian Agricultural Research Corporation)and the University of Brasília.
文摘Nitrous oxide(N_(2)O)is a potent greenhouse gas with about 60%of its emissions are attributed to agricultural activities.Its fluxes are influenced by a range of crop-specific factors,such as nitrogenous fertilizer inputs,soil N availability,tillage practices,temperature,pH and soil moisture.These factors interact in complex,nonlinear ways,creating the need for predictive modeling of N_(2)O emissions to both improve understanding and estimation and identify mitigating strategies.This proposes proposes data-driven machine learning techniques,particularly multilayer perceptron and random forest(RF)algorithms,for estimating soil N_(2)O fluxes in a sugarcane plantation under different irrigation regimes and to contrast machine learning results with conventional analytical methods.The findings indicate that RF modeling achieved a coefficient of determination of 87.4%for N_(2)O emission prediction,and identified ammonium,nitrogen nitrate,soil temperature,and water-filled pore space as the most influential predictors,in that order.The results open new possibilities for integrating machine learning to study N_(2)O fluxes in sugarcane and other major crops.All data and code used in this study are provided openly to support further research.
