Extending surface temperature and emissivity retrieval to the mid-infrared (3-5 μm) using the Multispectral Thermal Imager (MTI)

Surface-temperature (T) and emissivity (ε) estimation from remotely sensed mid-infrared (MIR: 3-5 μm) data requires modifications to existing long-wave infrared (LWIR: 8-12 μm) T/ε separation algorithms because of the significantly different characteristics of Planck's function between the MIR and LWIR wavelength regions and the strong effects of reflected solar irradiance in the MIR. A modified version of the normalized emissivity method (NEM), utilizing independently scaled maximum emissivities in each channel, was applied to thermal data acquired by the Multispectral Thermal Imager (MTI) with two MIR channels, J (3.81 μm) and K (4.97 μm), and three LWIR channels, L (8.26 μm), M (8.65 μm), and N (10.51 μm). Atmosphere-free simulations of T and ε retrieval over a wide variety of terrestrial surfaces yielded T values within ± 0.75 K of 'true T' in the range of 270-330 K and ε values within ± 0.011 of true ε in channels L, M and N, and ± 0.019 and ± 0.023 in channels K and J, respectively. The algorithm was tested successfully using MTI data over the Mauna Loa caldera in Hawaii. Unconstrained effects of shading and unresolved shadows in channel J day time data, and the strong atmospheric effects in channel K limit the application of the algorithm, in its present form, to night-time data.