Abstract: The European Union Effort Sharing Regulation (ESR) will require a 30% reduction in greenhouse gas (GHG) emissions by 2030 compared with 2005 from the sectors not included in the European Emissions Trading Scheme, including agriculture. This will require the estimation of current and future emissions from agriculture, including dairy cattle production systems. Using a farm-scale model as part of a Tier 3 method for farm to national scales provides a more holistic and informative approach than IPCC (2006) Tier 2 but requires independent quality control. Comparing the results of using models to simulate a range of scenarios that explore an appropriate range of biophysical and management situations can support this process by providing a framework for placing model results in context. To assess the variation between models and the process of understanding differences, estimates of GHG emissions from four farm-scale models (DailyWise, FarmAC, HolosNor and SFARMMOD) were calculated for eight dairy farming scenarios within a factorial design consisting of two climates (cool/dry and warm/wet) x two soil types (sandy and clayey) x two feeding systems (grass only and grass/maize). The milk yield per cow, follower cow ratio, manure management system, nitrogen (N) fertilisation and land area were standardised for all scenarios in order to associate the differences in the results with the model structure and function. Potential yield and application of available N in fertiliser and manure were specified separately for grass and maize. Significant differences between models were found in GHG emissions at the farm-scale and for most contributory sources, although there was no difference in the ranking of source magnitudes. The farm-scale GHG emissions, averaged over the four models, was 10.6 t carbon dioxide equivalents (CO(2)e)/ha per year, with a range of 1.9 t CO(2)e/ha per year. Even though key production characteristics were specified in the scenarios, there were still significant differences between models in the annual milk production per ha and the amounts of N fertiliser and concentrate feed imported. This was because the models differed in their description of biophysical responses and feedback mechanisms, and in the extent to which management functions were internalised. We conclude that comparing the results of different farm-scale models when applied to a range of scenarios would build confidence in their use in achieving ESR targets, justifying further investment in the development of a wider range of scenarios and software tools.

Abstract: Farm-scale models quantify the cycling of nitrogen (N) and carbon (C) so are powerful tools for assessing the impact of management-related decisions on greenhouse gas (GHG) emissions, especially on dairy cattle farms, where the internal cycling is particularly important. Farm models range in focus (economic, environmental) and the detail with which they represent C and N cycling. We compared four models from this range in terms of on-farm production and emissions of GHGs, using standardized scenarios. The models compared were SFarMod, DairyWise, FarmAC and HolosNor. The scenarios compared were based on two soil types (sandy clay versus heavy clay), two roughage systems (grass only versus grass and maize), and two climate types (Eindhoven versus Santander). Standard farm characteristics were; area (50 ha), milk yield (7000 kg/head/year), fertiliser (275 kg N and 150 kg N/ha/year for grass and maize, respectively). Potential yields for grass 10t dry matter (DM)/ha/year in both areas, maize 14 t DM/ha/ year in Eindhoven and 18t DM/ha/ year in Santander. The import of animal feed and the export/import manure and forages was minimized. Similar total farm direct GHG emissions for all models disguised a variation between models in the contribution of the different on-farm sources. There were large differences between models in the predictions of indirect GHG emission from nitrate leaching. Results could be explained by differences between models in the assumptions made and detail with which underlying processes were represented. We conclude that the choice of an appropriate farm model is highly dependent upon the role it should play and the context within which it will operate, so the current diversity of farm models will continue into the future. No Label