Uncertainty Analysis of Aircraft Based Mass-Balance Using a Synthetic Data Experiment

Aircraft data based mass balance has been extensively used in the literature as a standard GHG emission quantification method. In a recent 2016 paper, Conley et al. used an aircraft-based mass balance method to quantify the emissions from the Aliso Canyon underground storage facility leak outside of Los Angeles. The Aliso Canyon blowout incident and the aircraft data create a unique test bed to validate high-resolution atmospheric transport models. The primary objectives of this study are: 1) To improve and validate a NIST Large Eddy Simulation (LES) model simulation of the turbulent atmospheric transport of methane from the Aliso Canyon storage facility and, 2) To analyze the aircraft based mass-balance approach uncertainties within a synthetic data simulation experiment reproducing emission from the leak.

High-resolution LES simulations are performed on a 10km by 12km domain to simulate the natural gas leak from the Aliso Canyon storage facility. Comparison of LES results with aircraft measurement data indicates that the vertical and horizontal mixing of the plume are accurately simulated by the model. The aircraft flight paths (horizontal transects of the methane plume) are used to perform a one nearest neighbor search over the simulated surface to emulate aircraft measurements. Emulated data are lagged over time to create array of synthetic flight data with different flight starting times. These data sets are used to analyze impact of number of variables on recovering mass-flux rate such as, optimal sub-sampling of transects, time and height of flight paths and uncertainty as a function of the number and location of transects. The synthetic data experiment indicates that denser sampling of the core plume region reduces uncertainty in reproducing the leak.

We also present an analysis of the aircraft based mass-balance method using a spatial interpolation technique, kriging, ubiquitous in statistical literature. The kriging method is generally used to interpolate sparsely collected aircraft based data onto a regular grid. Three different kriging methods (Universal kriging, Thin Plate Spline and Wavelet-based kriging) are compared here in reconstructing the plume and emission from the leak. Uncertainties in these kriging methods are also studied. The kriging comparison indicates that the Thin Plate Spline is a fast and efficient method for reproducing the plume and leak with a standard error of prediction comparable to Universal kriging.