Land Surface Temperature Retrieval Using Nighttime Mid-Infrared Channels Data From Airborne Hyperspectral Scanner

Compared with thermal infrared (8-14 μm) spectrum, mid-infrared (MIR) spectrum is less sensitive to land surface emissivity (LSE) for estimating land surface temperature (LST). This work addressed the retrieval of LST from two adjacent MIR (3-5 μm) night-time airborne hyperspectral imager (AHS) simulated data with a split-window method, which can be expressed as a linear combination of the brightness temperature measured in two adjacent MIR channels with coefficients depending on LSE, view zenith angle (VZA), and water vapor content (WVC). Meanwhile, the LST retrieval accuracy for various channel combination was investigated and it was noted that the AHS channels 66 (3.5-4.25 μm) and 68 (4.25-5.0 μm) were the optimal channels for LST retrieval with a root-mean-square error (RMSE) less than 0.4 K for dry atmosphere and less than 0.5 K for wet atmosphere. Finally, the sensitivity analysis in terms of the instrumental noise, the uncertainties of LSE, and WVC were performed. It is worth noting that the combination of CH66 and CH68 performed well, and the LST retrieval errors were less than 0.5, 0.2, and 0.3 K caused by an noise equivalent delta temperature (NEΔT) of 0.33 K, WVC error of 20%, and LSE error of 0.01, respectively.