A practical method for angular normalization on land surface temperature using space between thermal radiance and fraction of vegetation cover

Land surface temperature (LST) retrieved from thermal infrared (TIR) remote sensing data displays strong directional anisotropy. This characteristic makes the LST at different angles incomparable and limits its application. Multi-angle data are usually required to model or retrieve the angular anisotropy, but most of the TIR images are observed from one single angle. This study investigates the relationship between fractional vegetation coverage (FVC) and thermal radiance, and proposes a practical method for angular normalization of LST using an FVC-radiance space LST angular normalization (called the FRAN method) to perform LST angular normalization for two-component (leaf and soil) vegetated canopy using only single-angle, single-temporal TIR data. In this method, the single-angle data are used to construct the space and build a linear relationship between FVC and radiance. Nadir radiance and LST are obtained by using FVC data at nadir. The validity and accuracy of the FRAN method are tested by using simulated dataset. Results show that 60%-70% of the angular effect of LST can be corrected, and the maximum LST correction reaches 2.0 K. The LST biased error distribution changes to unbiased distribution after angular normalization. The method sensitivity to the vertex of FVC-radiance space and the parameter in FVC calculation is analyzed. The FRAN method is applied to Moderate Resolution Imaging Spectroradiometer (MODIS) data, and obtains the maximum correction of LST as 1.2 K. Viewing zenith angle and land cover affect the corrected LST. The FRAN method has the potential for practical angular normalization of LST with single-angle data.