Using receptor-oriented modelling techniques to determine agricultural emissions of greenhouse gases

New Zealand’s greenhouse gas position is unusual. In CO₂-equivalent terms, about 45% of anthropogenic emissions are methane, 36% are carbon dioxide, and 19% are nitrous oxide. Ruminant livestock (sheep and cattle) account for about 90% of the methane, and agricultural activities account for essentially all nitrous oxide emissions. Estimates of the national emissions of these gases are currently based inventory calculations. The aim of the present research is to complement and improve regional emission estimates, using an approach based on the interpretation of direct atmospheric measurements.

This approach involves sampling boundary layer air using aircraft, tethered balloon and surface-based methods, and analysing samples by gas chromatography to determine methane concentrations. Reasonably uniform meteorological conditions are sought, and samples are taken upwind and downwind of the region of interest. The along-wind increase in methane – which is perhaps only 5 ppb in 1700 ppb – is thus due to surface agricultural sources located between the sampling sites.

The meteorology of the campaign day is simulated with the Regional Atmospheric Modeling System, and the results compared with local weather observations to ensure a good simulation. The Mesoscale Dispersion Modeling System is run in receptor-oriented mode, resulting in an influence function for each air sample. The influence function shows locations from which surface emissions can contribute to observed concentrations, and resembles a plume extending upwind from the sample point. Appropriate spatial patterns for methane emissions are specified and combined with the influence function to infer the emission rate itself.

Case studies are presented, including those from a coastal agricultural region in New Zealand under conditions of onshore winds. Methane concentrations are measured at six levels in the vertical, over the coast and inland. Each case provides six data points for which the ‘best fit’ regional methane emission rate is found.

Emission rates derived using this approach are consistent with those calculated using inventory methods. While this is encouraging, the level of uncertainty in both measurement and modelling makes it premature to view this as a validation of the emissions inventory calculations. These aspects are being improved upon, and application of the model to recent field campaigns, in which the air sampling is more extensive, should yield emission rates with increased confidence.

Receptor-oriented modelling is a powerful approach, and in future will be used to determine spatial and temporal changes in methane emission rates. Also, it will be applied to recent measurements of nitrous oxide, for which inventory-based emission rates have much greater uncertainty.