Estimating energy consumption and GHG emissions in crop production: A machine learning approach

It is necessary to observe the relationship between the energy inputs and outputs of agricultural production in the context of long-term strategies to determine optimal sustainable solutions and identify the most susceptible system components. Comprehensive research to assess energy performance data over a relatively long-term is required to predict crop production and greenhouse gas (GHG) emissions based on energy inputs. This study collects and analyzes energy inputs and GHG emissions from 17 crops of Iran in five major categories (cereal, pulse, oilseed, fiber, and tubers) during 1970-2019 to quantify the long-term energy efficiency of major crops. Three machine learning (ML) algorithms are developed to quantify the relationship between the resources consumed per unit of energy output and GHG emission models. These models are evaluated through three quantitative statistics. The analysis considers all physical and chemical inputs used to produce the major crops. These energy variables include human labor, machinery, fossil fuels, electricity, irrigation water, fertilizers (phosphate, nitrogen, and potassium), pesticides and seed rate. The results revealed that the input and output energy values increased from 39.47 to 190.18 MJ ha-1 in 1970 to 253.43 and 654 MJ ha(-1) in 2019. The GHG emissions from electricity (47.1%) were followed by those of nitrogen fertilizer (25.75%). Reducing the consumption of electricity and nitrogen fertilizers is the most efficient options to improve the energy efficiency of the crops studied.