Determination of methane emissions from livestock at the farm scale using observed meteorological and concentration profiles and inverse modelling techniques

Agricultural greenhouse gas emissions account for 50% of New Zealand’s carbon dioxide-equivalent emissions, a prominence unmatched by the emissions profile of any other Annex I party to the United Nations Framework Convention on Climate Change. Methane emitted by livestock through enteric fermentation accounts for one third of our carbon dioxide-equivalent emissions. The methane inventory in NZ is based on per-animal emission rates, scaled up to provide regional and national totals. Top-down emissions estimation techniques are used here as a check on the validity of inventory totals, with the aim of reducing emission flux uncertainties through a combination of methods.

In principle, methane concentrations increase as air is blown across a region grazed by livestock. If vertical profiles of methane are measured both upwind and downwind, the contrast between them reflects the strength of the intervening sources. At the farm scale, near-surface contrasts in methane are easily detectable. This enables individual farm emissions to be assessed, and potentially provides a tool to verify emission mitigation measures provided that sufficient precision is attainable.

Concentration profiles upwind and downwind of the source area are determined by collecting air time-integrated air samples in bags on 3m masts. Profiles up to an altitude of around 50m are measured by collecting samples through tubing raised by helikite or kytoon at the site of the downwind mast. Additional near-ground sampling along a downwind transect normal to the wind direction confirms the trajectory and breadth of the methane-rich plume. The air samples are analysed subsequently for methane concentration in the laboratory by gas chromatography. The meteorology is characterized by near-continuous soundings transmitted by a radiosonde suspended beneath the helikite or kytoon, and also by anemometer readings from the 3m masts.

Dispersion modelling is carried out in receptor-oriented mode using a Lagrangian particle dispersion (LPD) model. From the downwind sampling locations (the receptors), the released model particles are transported against the wind, backwards in time. The ensemble of intersections with the ground surface defines a set of influence functions. Each influence function describes the near-surface origins of the sampled air. Methane emission fluxes are derived using Bayesian inversion techniques, finding the best fit between modelled and observed concentrations. The analysis can incorporate a priori emissions estimated from the upscaling of per-animal emission rates, and also generates confidence intervals for the calculated methane fluxes.

The challenge for the modelling techniques is to provide a full profile of meteorological parameters in the surface layer, based on measurements at a limited number of levels above the ground. Using surface-layer similarity theory, profiles of mean wind and temperature in the layer may be derived from the observed wind and temperature at one or two levels, plus solar radiation. Meteorological measurements at more than two levels have not yet been made. They will be considered for future field campaigns, and the profile calculations will need to be suitably adapted. Boundary-layer parameters important to dispersion calculations, such as friction velocity, convective velocity scale, Obukhov length, stability function and mixing depth, are also calculated. The LPD model requires that the variance of each velocity component be specified in order to simulate the turbulent part of the dispersion. Empirically-derived turbulence profiles through the mixed layer are used, and validated against high-frequency measurements where possible.

The dispersion modelling system has been applied to a case study from June 2003 in the Wairarapa region of New Zealand, a flat area of land near sea level bounded by hills between 300m and 700m in height. Without incorporating a priori emission estimates, an emission rate of 71kg per day has been calculated from a paddock of 1.76ha being grazed by a herd of 160 cows. This equates to 440g per day per animal, but is around 70% higher than inventory estimates of the annual average emission. A fuller analysis will provide confidence limits based on measurement and modelling uncertainties.

An important aspect of this work is the validation of the dispersion model and analysis techniques. This will include field campaigns where, in place of livestock, an artificial source of methane will be installed. Methane will be released at a uniform, metered rate from a tank through a lattice of tubes in the paddock to simulate a distribution of cows. An identical measurement campaign and modelling exercise will be carried out to validate the inferred methane flux against the known source strength.