Measuring methane emission rates of a dairy cow herd (II): results from a backward-Lagrangian stochastic model

We apply a backward-Lagrangian stochastic (BLS) model to determine methane (CH4) emission rates from a herd of dairy cows freely grazing pasture within a fenced paddock. We assess how model characteristics and measurement errors of the input variables contribute to the error of the emission rate. This error is of order 20%. We find a scattered but systematic trend for the predicted CH4 source strengths to increase with the ratio of concentration detector height to source–detector distance, z/r, by around 20% across a 20-fold range of data for z/r. We then compare the CH4 source strengths from BLS to those from two other micrometeorological techniques. Compared to the flux-gradient technique (FG), there is little bias (slope of linear regression 0.98) but large scatter (squared correlation coefficient, R2, of 0.53). We believe this high degree of scatter is largely due to the deficiencies of FG. By contrast, comparison of BLS to the integrated horizontal flux technique (IHF) yields a larger bias (regression slope 1.09) but much better correlation (R2 = 0.80). When we express the z/r dependence of the BLS source strength by the excess relative to IHF source strength, we find that BLS source strengths vary from about 10% higher than IHF, for z/r ≈ 0.005, up to about 35% higher, for z/r ≈ 0.12. We compare the z/r dependence of the BLS results to the dependence of the turbulent Schmidt number on z/r as we determined earlier from analysis of field measurements. We find the results from the BLS model are less sensitive to the implicit choice of the Schmidt number than expected. We conclude that BLS is useful to corroborate IHF (since it requires no additional measurements), or as a viable alternative when technical requirements for IHF are too demanding and a slightly larger error in CH4 emission rate is acceptable.