Modelling of CO2 emission on dairy farms using ANN “Case study in Canterbury Province, New Zealand”

Abstract

This study was carried out across 35,300 hectares of pastoral ( PDF) and barn (BDF) dairy farming systems in the Canterbury region of New Zealand. In pastoral systems, the primary contributors to CO₂ emissions were fertilizers (27%) and machinery and equipment (25%). In contrast, barn systems showed feed supplements (30%) and machinery and equipment (24%) as the dominant sources of emissions. The estimated CO₂ emissions from milk production were 2,857 kgCO₂ per hectare (1.920 kgCO₂ per kg of milk solids) in PDFs and 3,379 kgCO₂ per hectare (2.129 kgCO₂ per kg of milk solids) in BDFs. To predict carbon emissions in dairy milk production, a Multi-Layer Perceptron (MLP) artificial neural network (ANN) model was developed using several direct and indirect factors. The final model, which achieved a high R² value of 0.95, incorporates variables such as farm characteristics (number of cows, stocking rate), socio-demographic factors (farmer's age), and energy-related inputs (electricity use). This model offers a reliable tool for estimating emissions under varying farm conditions and management practices