Methane emission quantification using UAV-based mass balance: Validation through controlled release experiments

Abstract. Uncrewed aerial vehicles (UAVs) offer flexible and cost-effective access to monitoring methane sources that can be otherwise difficult to sample, but confidence in UAV-derived fluxes requires robust validation and uncertainty characterisation. Here, we evaluate a variant UAV-based mass-balance method using single-blind controlled-release experiments at two UK sites: An aerodrome in Bedford and the Centre for Dairy Research (CEDAR), Reading. Controlled methane releases ranged from 0.021 to 2.16 kg h^-1 in Bedford and 8.3 to 40.6 kg h^-1 in CEDAR, including point and extended source configurations. Methane concentrations were measured in situ using an ABB GLA133-GPC analyser mounted on a hexacopter UAV equipped with a 2D sonic anemometer, and fluxes were derived from downwind horizontal transects.

Two validation approaches are considered: (i) using all individual flights to capture operational variability, and (ii) grouping flights by release rate to assess underlying method performance. Overall, fluxes calculated using the method show good correlation but substantial scatter at the individual flight level (CEDAR: slope = 1.01, r = 0.84, RMSE = 38.8 %; Bedford: slope = 0.71, r = 0.85, RMSE = 46.2 %) and substantially improved agreement when averaged by release rate (CEDAR: slope = 1.01, r = 0.94, RMSE = 19.9 %; Bedford: slope = 0.68, r = 0.99, RMSE = 38.4 %). This demonstrates that averaging reduces random variability while preserving systematic bias.

A method variant, which accounts for incomplete near-surface sampling by extrapolating the lowest available measurements and incorporating the associated uncertainty, substantially improves consistency, increasing the fraction of cases (46 % to 72 %) in which the known release rate falls within the one-standard-deviation uncertainty of the calculated release rate. We find that mean wind speed is the dominant environmental factor influencing the quality of agreement, with higher wind speeds associated with reduced bias (above 2.2 m s^-1, bias ≤ 50 %), likely due to reduced turbulence under such conditions. Wind direction variability alone shows little direct influence on flux precision; instead, a wide lateral extent of the measurement plane is noted to be critical for minimising plume sampling loss under changing wind conditions. Lower emission rates exhibit greater relative bias due to increased sensitivity to smaller signal-to-noise (due to both wind variability and lower plume concentration enhancements relative to natural background variability).

The improved performance after averaging is consistent with previous studies indicating that multiple flights per release rate (e.g. n ≥ 3) enhance representativeness by reducing the influence of environmental variability. These results demonstrate that UAV-based mass-balance methods can provide methane flux estimates accurate to within 20 %, for fluxes greater than 10 kg h^-1, under controlled conditions, but may suffer from systematic (but accountable) under-bias due to incomplete sampling. We also offer insights on methodological and environmental requirements to deliver robust measurements.