Use of life cycle assessment and modeling techniques for prediction of energy-environmental indicators in different wheat production systems

Abstract

This study was conducted to analyze and model the energy consumption and environmental indicators for different wheat production systems in Qazvin province. Information needed for this study, related to 2021–2022, data was collected by monitoring machinery, and inputs during wheat production were interviewed by farmers and filled out questionnaires. The purpose of this study includes: (1) estimating the fuel consumption and energy of inputs in different wheat production systems, (2) estimating the environmental impact associated with input consumption using life cycle assessment (LCA), and (3) providing suggestions for optimization. The results showed wheat production in irrigated conservation tillage system has energy ratio (2.12) and energy productivity (0.18 kg MJ−1) suitable compared to conventional irrigated and rainfed tillage, and conservation rainfed tillage. Nitrogen fertilizer (55%–68%) and fuel consumption (15%–31%) of affect the energy consumption of four wheat cultivation systems. Three categories of damage to human health, ecosystem and resources in LCA were using the ReCiPe2016 method. Resource damage category for conventional tillage irrigation (76.05 USD2013) has significant pollution. The share of seed emissions, On-Farm emissions and nitrogen emissions affect the categories of damage to human health, ecosystems and resources, respectively. The results of the Cobb-Douglas model showed that T3 (fertilization and weed control) was significant at the level of 1% among other inputs. T1 (chisel packer roller), T2 (planting) and T4 (harvest and straw packaging) were significant at 5% level. The results of artificial neural network (ANN) of conventional rainfed cultivation showed that 6 input layers, 7 hidden layers with 10 hidden neurons and 1 output parameter were selected as the best structure. R2 and RMSE for fuel consumption were 0.95 and 43.65, respectively. As a result, the ANN results show higher accuracy than the Cobb-Douglas model.