Modelling uncertainty in greenhouse gas emissions from UK agriculture at the farm level

We outline a framework for considering uncertainty in estimates of emissions of greenhouse gases (GHGs) where the objective is to determine least cost (profit foregone) adaptations to reduce emissions at the farm level. Three sources of uncertainty were identified as being associated with GHG emissions: lack of understanding of biological systems, poor validation of results and weather-induced variability. Output from existing models and other information on emissions were included in a farm-level optimisation model, Farm-adapt, set up to represent a dairy and beef farm in north west England. Monte Carlo simulation was used to examine the effect of uncertainty on total GHG emissions and the most cost-effective adaptations for reducing these emissions to 60% of the baseline level. We assumed a triangular distribution for all parameters and sampled 1000 times from these distributions. Farm-adapt results showed that cost-effective adaptations were to: (i) eliminate intensive beef production, (ii) reduce stored manures and increase frequency of manure spreading, (iii) substitute concentrate feed for grass and conserved grass in milk production and (iv) apply less mineral nitrogen to grassland. Monte Carlo simulation showed that there was a large degree of uncertainty in the level of absolute emissions associated with the optimal adaptations and the mean of the simulation output (9265 kg ha−1 year−1 CO2 equivalent) was greater than the Farm-adapt results using default parameters (7787 kg ha−1 year−1). However, there was a high degree of certainty in the adaptations required to reduce farm-level emissions, indicating that the cost-effective adaptations were robust to uncertainty in the GHG emission data.