Mobile atmospheric measurements and local-scale inverse estimation of the location and rates of brief CH4 and CO2 releases from point sources

Abstract

Abstract. We present a local-scale atmospheric inversion framework to estimate the\nlocation and rate of methane (CH4) and carbon dioxide (CO2)\nreleases from point sources. It relies on mobile near-ground atmospheric\nCH4 and CO2 mole fraction measurements across the corresponding\natmospheric plumes downwind of these sources, on high-frequency\nmeteorological measurements, and on a Gaussian plume dispersion model. The\nframework exploits the scatter of the positions of the individual plume\ncross sections, the integrals of the gas mole fractions above the background\nwithin these plume cross sections, and the variations of these integrals from\none cross section to the other to infer the position and rate of the\nreleases. It has been developed and applied to provide estimates of brief\ncontrolled CH4 and CO2 point source releases during a 1-week\ncampaign in October 2018 at the TOTAL experimental platform TADI in Lacq,\nFrance. These releases typically lasted 4 to 8\u2009min and covered a wide\nrange of rates (0.3 to 200\u2009g\u2009CH4/s and 0.2 to 150\u2009g\u2009CO2/s) to test\nthe capability of atmospheric monitoring systems to react fast to emergency\nsituations in industrial facilities. It also allowed testing of their\ncapability to provide precise emission estimates for the application of\nclimate change mitigation strategies. However, the low and highly varying\nwind conditions during the releases added difficulties to the challenge of\ncharacterizing the atmospheric transport over the very short duration of the\nreleases. We present our series of CH4 and CO2 mole fraction\nmeasurements using instruments on board a car that drove along roads\n∼50 to 150\u2009m downwind of the 40\u2009m\u2009×\u200960\u2009m area for\ncontrolled releases along with the estimates of the release locations and\nrates. The comparisons of these results to the actual position and rate of\nthe controlled releases indicate ∼10\u2009%–40\u2009% average\nerrors (depending on the inversion configuration or on the series of tests)\nin the estimates of the release rates and ∼30–40\u2009m errors in\nthe estimates of the release locations. These results are shown to be\npromising, especially since better results could be expected for longer\nreleases and under meteorological conditions more favorable to local-scale\ndispersion modeling. However, the analysis also highlights the need for\nmethodological improvements to increase the skill for estimating the source\nlocations.\n