No abstract

No abstract

A FCN was successfully trained to delineate methane plumes from RGB and matched filter time series data. IMEs derived from the predicted plume labels were slightly smaller than IMEs calculated using manual labels. Persistent false positives in the matched filter data were rarely identified as plumes by the FCN. The success of this research presents an exciting case for utilizing deep learning to accurately delineate methane plumes as time series data for point source emitters becomes more readily available.