Performance of continuous emission monitoring solutions under single-blind controlled testing protocol.


  • Clay Bell Energy Institute, Colorado State University & BPX Energy ,
  • Chiemezie Ilonze Department of Mechanical Engineering, Colorado State University ,
  • Aidan Duggan Energy Institute, Colorado State University ,
  • Daniel Zimmerle Energy Institute, Colorado State University


Continuous emission monitoring (CM) solutions promise to accelerate fugitive methane emission mitigation in natural gas infrastructure by detecting large sources sooner than they would be detected in traditional leak surveys. Quantification by CM solutions has also been proposed as the foundation of measurement-based inventories. Herein we present the first published results of CM solutions tested using a consensus protocol. Single-blind testing was performed with controlled releases at a dedicated facility, using test conditions that were challenging but less complex than typical field conditions. Eleven solutions were tested, including point sensor networks and scanning/imaging solutions. Results indicate detection limits (emission rate with 90% probability of detection) of 3-30 kg CH4/h, with 6 of 11 solutions in the range of 3-6 kg CH4/h, with high uncertainty. Of the 11 solutions tested, 6 provided emission rates estimates and for emission rate of 0.1-1 kg/h, the solutions' mean relative errors ranged from -44% to +586% with single estimates between -90% and +300% (all uncertainties stated as empirical 95% confidence intervals) with most solutions' upper uncertainty exceeding +900%. Solutions' mean relative error for emission rates >1 kg/h is -40% to +93% with two solutions achieving mean relative errors below 20%. In this range, all solutions exhibited single-estimate relative error from below -80% to values above +240%. When test results were simulated using emissions data from a portfolio of facilities, two solutions estimated emissions to within ±20%, while 4 estimated emissions at least double the true emission rate. The large variability in performance between CM solutions, coupled with highly uncertain detection, detection limit, and quantification results, indicate that the performance of individual CM solutions should be well understood before relying on results for internal emissions mitigation programs or regulatory reporting. Additionally, given the rapid development of these solutions, additional implementation standards and regular testing will also be necessary.

Version notes

The most significant change from the previous manuscript is the correction of the probability of detection and localization results for solutions I and J. Other changes include improving the appearance to the legend of Figure 3 in the main paper and that of Figures 17, 18, and 19 in the SI reports. Authors' zip codes were added to their affiliations. The rest of the changes were fixing a citation error, a sign error in the SI, and a few grammatical errors. Finally, the reports of all the solutions are now in a zip folder to ease accessibility.


Supplementary material

Performance of continuous emission monitoring solutions under single-blind controlled testing protocol - Supplementary Information
The document contains relevant additional and supplementary information that was not captured nor discussed in the main paper.
Comprehensive reports on the performance of solutions
Report(PDF): Each report summarizes the demography of experiments performed to test a solution, the meteorological conditions during the experiments, and the performance of the solutions. To assess performance, metrics such as probability of detection, emission source localization, and emissions quantification established by the testing protocol were evaluated. Data column header definition(XLSX): Explains the information captured in each column of a solution's data table. Data(XLSX): Each data table contains both controlled releases and detection reports paired to form various classifications for a solution. Data table rows represents either true positive, false positive, or false negative detection classifications. A reported detection successfully paired with a controlled release was classified as a true positive detection. A reported detection which remained unpaired was classified as a false positive detection, while a controlled release which remained unpaired was classified as a false negative. Other excluded data make up the remaining rows of the data table.