In the case of accidental methane leakage on a gas production industrial site, it is essential that the risks associated with an explosion of escaped clouds are assessed. By combining spectral and spatial information, hyperspectral technology is an attractive solution for the detection of such a cloud and for the quantification of its concentration. Total has started in 2014 a research program in partnership with Onera, called NAOMI (New Advanced Observation Methods Integration) to develop new tools for remote characterization of accidental methane plumes, especially over areas inaccessible to the personnel. From one of the results of this partnership, Onera is developing an algorithm, IMGSPEC, especially designed for this purpose, using hyperspectral acquisitions in the LWIR domain. The principle of IMGSPEC consists of estimating the spectral transmission of the gas cloud using an image of the background. An acquisition image of the same scene without gas is not necessarily available however. The strong point of the algorithm is its ability to recover the signal of the background. The integrated concentration is subsequently estimated pixel by pixel constituting a ppm.m concentration map. Finally, the flow rate of the leak is calculated considering the mass of the cloud, combining concentration estimation and methane density, and the wind speed which is measured with a meteo-station for instance. This algorithm was tested in June during a specific test campaign on the Lacq platform, a Total R and D industrial site. Methane leaks have been performed regulating the following flow rates: 1g/s, 10 g/s and 100g/s. Flow rate was estimated by IMGSPEC in near real-time following hyperspectral datacube acquisitions. Acquisition and processing times were both 4s, constituting a global flow rate estimation time below 10s
The determination of the aerosol type in a plume from remotely sensed data without any <i>a priori</i> knowledge is a challenging task. If several methods have already been developed to characterize the aerosols from multi or hyperspectral data, they are not suited for industrial particles, which have specific physical and optical properties, changing quickly and in a complex way with the distance from the source emission. From radiative transfer equations, we have developed an algorithm, based on a Look-Up Table approach, enabling the determination of the type of this kind of particles from hyperspectral data. It consists in the selection of pixels pairs, located at the transitions between two kinds of grounds (or between an illuminated and a shadow area), then in the comparison between normalized estimated Aerosol Optical Thicknesses (AOTs) and pre-calculated AOTs. The application of this algorithm to simulated data leads to encouraging results: the selection of only six pixels pairs allows the algorithm to differentiate aerosols emitted by a metallurgical plant from biomass burning particles, urban aerosols and particles from an oil depot explosion, regardless the size and the aerosol concentration. The algorithm performances are better for a relatively high AOT but the single scattering approximation does not enable the characterization of thick plumes (AOT above 2.0). However, the choice of transitions (type of grounds) does not seem to significantly affect the results.