A Multi-Domain Compression Radiative Transfer Model for the Fengyun-4 Geosynchronous Interferometric Infrared Sounder (GIIRS)
Mingyue Su,Chao Liu,Di Di,Tianhao Le,Yujia Sun,Jun Li,Feng Lu,Peng Zhang,Byung-Ju Sohn
DOI: https://doi.org/10.1007/s00376-023-2293-5
IF: 5.8
2023-07-09
Advances in Atmospheric Sciences
Abstract:Forward radiative transfer (RT) models are essential for atmospheric applications such as remote sensing and weather and climate models, where computational efficiency becomes equally as important as accuracy for high-resolution hyperspectral measurements that need rigorous RT simulations for thousands of channels. This study introduces a fast and accurate RT model for the hyperspectral infrared (HIR) sounder based on principal component analysis (PCA) or machine learning (i.e., neural network, NN). The Geosynchronous Interferometric Infrared Sounder (GIIRS), the first HIR sounder onboard the geostationary Fengyun-4 satellites, is considered to be a candidate example for model development and validation. Our method uses either PCA or NN (PCA/NN) twice for the atmospheric transmittance and radiance, respectively, to reduce the number of independent but similar simulations to accelerate RT simulations; thereby, it is referred to as a multi-domain compression model. The first PCA/NN gives monochromatic gas transmittance in both spectral and atmospheric pressure domains for each gas independently. The second PCA/NN is performed in the traditional spectral radiance domain. Meanwhile, a new method is introduced to choose representative variables for the PCA/NN scheme developments. The model is three orders of magnitude faster than the standard line-by-line-based simulations with averaged brightness temperature difference (BTD) less than 0.1 K, and the compressions based on PCA or NN methods result in comparable efficiency and accuracy. Our fast model not only avoids an excessively complicated transmittance scheme by using PCA/NN but is also highly flexible for hyperspectral instruments with similar spectral ranges simply by updating the corresponding spectral response functions.
meteorology & atmospheric sciences