Skip to main content
ARS Home » Southeast Area » Florence, South Carolina » Coastal Plain Soil, Water and Plant Conservation Research » Research » Publications at this Location » Publication #323207

Title: Monte Carlo uncertainty analyses of a bLS inverse-dispersion technique for measuring gas emissions from livestock operations

item Ro, Kyoung
item Martin, Jerry
item FLESCH, THOMAS - University Of Alberta
item VIGURIA, MAIALEN - Neiker-Tecnalia

Submitted to: Meeting Abstract
Publication Type: Abstract Only
Publication Acceptance Date: 10/26/2015
Publication Date: N/A
Citation: N/A

Interpretive Summary:

Technical Abstract: The backward Lagrangian stochastic (bLS) inverse-dispersion technique has been used to measure fugitive gas emissions from livestock operations. The accuracy of the bLS technique, as indicated by the percentages of gas recovery in various tracer-release experiments, has generally been within ± 10% of actual emissions with appropriate sensor placement and data filtering. However, standard deviation of the emission calculations can be as high as 32%. This uncertainty is due to the combined uncertainties in the field measurements, model assumptions, and so on. The objective of this study is to investigate the uncertainty of the bLS technique by performing multiple Monte Carlo simulations using the bLS model. While typical analytical error-propagation techniques utilize first-order approximation, the complexity of the bLS dispersion model requires a more robust analysis. A Monte Carlo method is a more appropriate method for the task. It numerically produces output distributions propagated by input uncertainties. It is simple, but requires high computing resources. Three 15-min datasets from a previous waste lagoon study were used to assess the uncertainty of the emission rates calculated by a bLS model (WindTrax). Randomly generated input values for concentration and wind statistics were selected assuming uniform probability distributions based on instrument’s specifications. Up to 100,000 bLS model iterations were performed to generate accuracy distributions. The resulting distributions showed smaller dispersion of bLS accuracy than typically found in field validation studies, with standard deviations less than 12%. This suggests that factors other than sensor uncertainties, such as model idealization of terrains, contributed to the overall uncertainties of the bLS technique in field settings. More experiments are currently underway isolating different sources uncertainties (in terms of sensors, degree of uncertainties of each sensors, and so on). These results will be presented at the meeting.