Reducing the sampling periods required in protocols for establishing ammonia emissions from pig fattening buildings using measurements and modelling

Ammonia (NH3) emission factors for animal housing systems in the Netherlands are based on measurements using standardised measurement protocols. Both the original Green Label (GL) protocol and the newly developed multi-site sampling protocol are based on year-round sampling periods. The objective of this study was to develop a measurement approach in which sampling periods are substantially shortened (i.e. to six months) whilst minimising the loss of overall accuracy compared to year-round sampling. Two methods were investigated: method A that combined measurements with a statistical model obtained from an extensive dataset of GL measurements for housing systems of fattening pigs, method B that used only measurement data to calculate year-round emission levels. Both approaches were investigated for application on mechanically ventilated housing systems for fattening pigs. The accuracy of both methods was evaluated by applying them to another dataset from five different pig housing systems. A reduction of sampling period from twelve to six months led to small increases in the overall random measurement error of the mean emission. The increase was least for method A because of the modelling component in this approach. Shortening of the sampling period may result in the risk of systematic deviations from the reference values. These systematic deviations are lower (although statistically significant) for method A (−4.1 to 4.7%) compared to B (−6.8 to 6.4%). If a small (<5%) or moderate bias (<10%) is considered acceptable then both methods provide an attractive alternative to the full-year sampling approach with method A being preferred.

► Research into optimizing emission measurements strategies from pig fattening buildings.
► Reducing sampling periods from twelve to six months using measurements and modelling.
► Systematic deviations from the reference values were less than 5–10%.
► If bias (<5–10%) is considered acceptable, suggested approach provides an alternative for full-year sampling.