Optimization of Instrumental Spectral Configurations for the Split-Window Method in the Context of the TRISHNA Mission

We propose an original approach to optimize the Thermal infraRed Imaging Satellite for High-resolution Natural resource Assessment (TRISHNA) instrument spectral configuration for the split-window (SW) method. First, we consider as input of end-to-end simulations an emissivity data set that accounts for cavity effect within vegetation canopy. Second, we propose a bidimensional approach where both locations of TRISHNA SW channels, namely
$\lambda _{c}^{\text {TIR3}}$
and
$\lambda _{c}^{\text {TIR4}}$
, can slide within predefined spectral intervals. We report a large sensitivity to channel positions, with variations of root mean square error (RMSE) on retrieved land surface temperature (LST) up to 3 K. Our bidimensional approach shows that this sensitivity is consistent with the underlying assumptions of the SW method. Indeed, two regions are observed in the
$(\lambda _{c}^{\text {TIR3}},\,\lambda _{c}^{\text {TIR4}})$
space: 1) an unfavorable region corresponding to
$\lambda _{c}^{\text {TIR3}}\leq 10.0~\mu \text{m}$
, where large RMSE values are ascribed to large differences between emissivities in both SW channels, and 2) a favorable region corresponding to
$\lambda _{c}^{\text {TIR3}}\geq 10.3~\mu \text{m}$
, where differences between emissivities in both SW channels are small and RMSE values are driven by the differences between atmospheric transmittance in both SW channels. Overall, it is necessary to better account for the difference in surface emissivities between the two SW channels, whereas disregarding the cavity effect within vegetation canopy is not critical. Eventually, our bidimensional approach permits to define an optimal position for
$\lambda _{c}^{\text {TIR3}}$
at
$10.6~\mu \text{m}$
, which induces larger robustness to uncertainties on channel positions. By applying our study on two structurally different SW formulations and addressing impacts of uncertainties on land surface emissivity (LSE) and atmospheric water vapor content (AWVC), we show that these results can be generalized to other SW formulations.