Estimating methane emission durations using continuous monitoring systems

We propose a method for estimating the duration of methane emissions on oil and gas sites, referred to as the Probabilistic Duration Model (PDM), that uses concentration data from continuous monitoring systems (CMS). The PDM probabilistically addresses a key limitation of CMS: non-detect times, or the times when wind blows emitted methane away from the CMS sensors (resulting in no detections). Output from the PDM can be used to bound the duration of emissions detected by snapshot measurement technologies, such as plane or satellites, that have limited ability to characterize emission duration due to the typically low temporal frequency (e.g. quarterly) at which they observe a given source. Linear regression indicates that the PDM has a bias of -4.9% (R^2 = 0.80) when evaluated on blinded controlled releases at the Methane Emissions Technology Evaluation Center (METEC), with 86.8% of estimates within a factor of 2x error from the true duration. We apply the PDM to a typical production site in the Appalachian Basin and use it to bound the duration of snapshot measurements. We find that failing to account for CMS non-detect times results in underestimated emission durations of up to a factor of 65x (6,400%) on this site.