Methane is invisible in the visible spectrum and nearly invisible in most satellite bands. It becomes detectable in the shortwave infrared, where the molecule absorbs sunlight that has passed down through the atmosphere, reflected off the ground, and passed back up again. Instruments do not see the gas directly. They see the shadow it casts in the reflected solar spectrum.
The absorption windows
Methane has usable absorption structure in two regions of the shortwave infrared, near 1.65 microns and near 2.3 microns. The 2.3 micron region is the one most imaging spectrometers exploit, because the absorption is stronger there and the feature has a distinctive comb of narrow lines rather than a single smooth dip. That structure is what makes methane separable from surface materials, which vary smoothly across the same wavelengths.
Detecting it requires contiguous narrow bands. Broad multispectral bands average the comb away.
Matched filtering
The standard retrieval approach is a matched filter. The method builds a statistical model of what the scene normally looks like across the relevant bands, typically the mean spectrum and covariance of the surrounding pixels, then tests every pixel for the presence of a known target signature: the methane absorption pattern scaled by concentration. Pixels whose deviation from the background aligns with that signature score high. Pixels that deviate in unrelated ways score low.
This works well because the background is estimated from the scene itself, so it adapts to whatever surface is present. It also explains the characteristic failure modes. Surfaces that mimic the methane pattern, some hydrocarbon rich materials, certain painted roofs and calcite bearing rocks, can produce false enhancements. Dark surfaces, water and deep shadow return too little light to test at all. Clouds break the assumption entirely.
Enhancement is not flux
A matched filter output is a column enhancement, usually expressed in parts per million multiplied by metres. It says how much extra methane sits in the vertical path above that pixel compared to the background.
Converting an enhancement map into an emission rate in kilograms per hour requires additional information: the plume geometry, and above all the wind speed at the plume. Wind is normally taken from a meteorological reanalysis product at coarse resolution, and it dominates the uncertainty budget. A well constrained enhancement map paired with a poorly constrained wind field produces a flux estimate with error bars far wider than the detection itself. Reporting the two separately is more honest than folding them together.