Jacob, Daniel JBruno, Jack Henry2026-10-1920262026-10-192026Bruno, Jack Henry. 2026. Quantifying Methane Emissions Across Spatial Scales Using Satellite and Aircraft Remote Sensing. Doctoral Dissertation, Harvard University Graduate School of Arts and Sciences.32863971https://dash.harvard.edu/handle/1/42760632Current methods for detecting atmospheric plumes and inferring point-source rates from high-resolution satellite imagery are labor-intensive and not scalable with regard to the growing satellite dataset available for methane point sources. We present a two-step algorithm called U-Plume for automated detection and quantification of point sources from satellite imagery. The first step delivers plume detection and delineation (masking) with a U-Net machine learning architecture for image segmentation. The second step quantifies the point-source rate from the masked plume using wind speed information and either a convolutional neural network (CNN) or a physics-based integrated mass enhancement (IME) method. The algorithm can process 62 images (each measuring 128$\times$128 pixels) per second on a single 2.6~GHz Intel Core i7-9750H CPU. We train the algorithm using large-eddy simulations of methane plumes superimposed on noisy and variable methane background scenes from the GHGSat-C1 satellite instrument. We introduce the concept of point-source observability, $O_{\mathrm{ps}} = Q/(UW\Delta B)$, as a single dimensionless number to predict plume detectability and source rate quantification error from an instrument as a function of source rate $Q$, wind speed $U$, instrument pixel size $W$, and instrument-dependent background noise $\Delta B$. We show that $O_{\mathrm{ps}}$ can powerfully diagnose the ability of an imaging instrument to observe point sources of a certain magnitude under given conditions. U-Plume successfully detects and masks plumes from sources as small as 100~kg\,h$^{-1}$ in GHGSat-C1 images over surfaces with low background noise and successfully handles larger point sources over surfaces with substantial background noise. We find that the IME method for source quantification is unbiased over the full range of source rates, while the CNN method is biased towards the mean of its training range. The total error in source rate quantification is dominated by wind speed at low wind speeds and by the masking algorithm at high wind speeds. A wind speed of 2--4~m\,s$^{-1}$ is optimal for detection and quantification of point sources from satellite data (chapter 1). The MethaneAIR aircraft remote sensing instrument observes methane dry air column mixing ratios (X$_{\mathrm{CH_4}}$) over $\sim$100$\times$100~km$^2$ scenes with sub-km resolution, from which methane emissions can be inferred by inverse analysis with an atmospheric transport model. It emulates the MethaneSAT satellite instrument that operated March 2024--June 2025 to quantify emissions from oil/gas production regions. We show here how the single-day MethaneAIR observations can be integrated with the global continuous but relatively coarse and sparse observations from the TROPOMI satellite instrument into a common Integrated Methane Inversion (IMI) platform for optimizing methane emissions. The IMI, originally designed for TROPOMI, is used here with 12$\times$12~km$^2$ spatial resolution and lognormal error probability density functions (PDFs) for prior estimates. Application to two scenes in oil/gas production basins of the western US shows general consistency between independent MethaneAIR (single day) and TROPOMI (monthly) inversions, including for emission hotspots, with some differences that may reflect temporal variability of emissions and/or unrecognized error correlations in the observations. The IMI is able to optimize emissions even when starting from a very poor prior estimate, although the prior estimate contributes information toward the optimized emissions. Using TROPOMI inversion results as prior estimate improves the MethaneAIR inversions by correcting emissions upwind of the MethaneAIR observation scenes and by adding information to the original prior estimate. (chapter 2). The MethaneSAT satellite instrument, operating from March 2024 to June 2025, observed methane column concentrations over selected oil and gas basins with an unprecedented combination of sub-km spatial resolution and high precision. Here we conduct analytical Bayesian inversions of MethaneSAT observations with the Integrated Methane Inversion (IMI) to infer 13-month mean methane emissions at 25-km resolution and compare them to inversions of TROPOMI daily satellite observations over the same domain and period. We focus on the Permian basin, which MethaneSAT observed on 28 days, more than any other basin, and which has a high TROPOMI retrieval success rate (33\% for the basin). The instruments have different vertical averaging kernel sensitivities, but we find that they are consistent in their observed column variability when using the GEOS-Chem forward model of the IMI as a common intercomparison platform. We find that IMI inversions of TROPOMI and MethaneSAT data agree in their basin and sub-basin totals but have large differences at fine resolution that could reflect variable point sources. MethaneSAT provides higher information than TROPOMI on its observation days but is limited by its return time and its small viewing scenes. The best inversion results are obtained when using the TROPOMI inversion results as prior estimates for the MethaneSAT inversion. Comparison to daily emission estimates from the MethaneSAT Level-4 product shows good agreement in the basin-scale magnitude of emissions, except for a subset of scenes where weak and/or variable winds complicate the treatment of large point sources. (chapter 3).application/pdfenGEOS-ChemMachine LearningMethaneMethaneSATSatelliteEnvironmental scienceRemote sensingThesis or Dissertation2026-10-190009-0002-5621-7981