Feature Wavelengths for Quantifying Methane Concentrations Using Shortwave Infrared Hyperspectral Imaging: A...

Methane (CH₄) is a significant greenhouse gas, and accurately quantifying its concentrations is essential for addressing climate change concerns. This study used controlled conditions to identify potential spectral regions or wavelengths within the short-wave infrared (SWIR) region that can be used for CH₄ quantification using hyperspectral imaging (HSI). This study also validated the efficiency of the wavelengths that are currently used in remote sensing. Glass containers with constant CH₄ flow rates were used for collecting HSI data (1010-2495 nm) at different CH₄ concentrations (N = 18; 0-2.5% CH₄). Partial least-squares regression (PLSR) was trained using the full 266-bands (1010-2495 nm). Regression coefficients and PLS weights were used to identify the potentially important regions and wavelengths. New PLSR models were developed using the important regions (multiband models). Individual wavelengths identified in the current study or previously used in remote sensing studies were also used to develop PLSR models individually or in two-band combinations. Potential overlaps between the identified spectral region and H₂O absorption bands were investigated. The results indicated that using spectral regions (multiband) or combinations of two bands provided more accuracy compared to when single bands were used. The following spectral regions can be used for the quantification of CH₄ in descending order: full 266-band (1010-2495 nm) > 1648 + 1670 nm > full 128-band (1010-1700 nm) > 2150-2243 nm > 1010-1185 nm. The results of this study were obtained under controlled conditions without interfering compounds. Testing these spectral regions in more complex environments will help confirm the best SWIR wavelengths.