Evaluation of sampling strategies for measurement of baseline methane and ammonia emission from pig housing

Accuracy and precision in agricultural emission factors depend on design of the experiments used for measurements. A Monte Carlo sampling simulation was applied to a large dataset of hourly methane and ammonia emission measurements from three farms including a total of 11 batches of finisher pigs to explore the impact of sampling intensity within and between farms and the design of the measurement campaign on precision and accuracy. Measurements showed correlation within farms, between-farm domination of variance, moderate diurnal cycles, and temporal autocorrelation. Sampling simulation results confirmed that a simple parametric approach is sufficient for approximate estimates of standard error of the mean emission rate (within 25%). Diurnal trends caused a small bias related to sampling time that was eliminated with 24 h measurement periods (or multiples of 24 h). With high autocorrelation, a single random 1 h measurement period performed nearly as well as much longer periods. Emission rate increased with pig batch age, but a systematic or stratified sampling plan provided no improvements in precision and sampling at predetermined weeks introduced bias. With a large between-farm variance, intensive sampling of methane or ammonia emission within farms is wasted effort if the objective of a campaign is to measure a mean emission rate for a country or region, and 3 measurement periods per farm is more than sufficient. Ammonia emission varied less between farms, but even so, results showed that more intensive sampling within farms will only marginally improve overall precision.