Comparison of the Remapping Algorithms for the Advanced Technology Microwave Sounder (ATMS)

Microwave radiometers have wide application in atmospheric remote sensing and provide essential inputs to numerical weather-prediction models. But the applications of these space-borne, multispectral measurements from multiple sensors are often plagued with the problem of nonuniform spatial resolution caused by the limited size of satellite instrument antenna and the frequency dependent microwave emission from the earth-atmosphere system. This mismatch in resolution becomes a critical issue when observations from multiple sources are combined to retrieve geophysical parameters. To address this issue, much efforts have been paid to develop remapping algorithms that can effectively unify the field of view (FOV) of measurements from various sources. This study compares the performance of two remapping algorithms that have been widely adopted in the operational ATMS data pre-processing. One is Backus-Gilbert inversion (BGI) method, implemented in ATMS Resampling Algorithm to produce ATMS brightness temperature at each Cross-track Infrared Sounder (CrIS) FOV. The other is the Filter algorithm, applied in ATOVS and AVHRR Pre-processing Package (AAPP) to remap ATMS data to AMSU-like FOV. The two algorithms are compared via both the simulated and actual ATMS data. ATMS Remapping Algorithms Reconstructed Point Spread Function (PSF) References qBGI Algorithm qFilter Algorithm Evaluation by Actual ATMS Datasets Evaluation by Simulated Datasets Summary BGI algorithm remaps the data in the spatial domain. It finds a set of optimal coefficients !"# for constructing a new observation $!%!&'(% with an expected FOV as a linear combination of adjacent original observations $!"# )&"'"*!+: $!%!&'(% ="./ * #./ * !"# $!"# )&"'"*!+ The coefficients are abtained by minimizing the following objective function: 0 = 01 234 5 + 789 4:; 5 <1 = ∫ ∑"./ * ∑#./ * !"# ?"# )&"'"*!+ − ?%!&'(% 8 AB (8 = ∆$&DE ∑"./ * ∑#./ * !"# 8 ?)&"'"*!+ / ?%!&'(%: original/target gain functions ∆$&DE: observation noise 9: scale factor set to be 0.001 5: noise tuning factor, determined by imposing the constraint ∑"./ * ∑#./ * !"#=1 Filter Algorithm is established based on the convolution theorem. Ta is regarded as the convolution of Tb with antenna gain function in spatial domain, which is equivalent to the multiplication of them in frequency domain. This algorithm manipulates the beam width in frequency domain: $!)&"'"*!+(G) I!)&"'"*!+(J) FFT Transform I!%!&'(%(J)=I!)&"'"*!+(J) K$I %!&'(% K$I)&"'"*!+ FFT Inverse Transform $!%!&'(%(G) K$I)&"'"*!+/ K$I%!&'(%are the Fourier transform of the original/target antenna pattern. Note that for image enhancement, a cutoff parameter is added to suppress the amplified noise. Ø BGI enhancement shows some improvement in the synthetic PSF. Compared to the original one, the synthesized PSF contains more high frequency components which is closer to the target one. Ø BGI degradation shows that the synthetic PSF perfectly matches the target PSF. Ø For Filter algorithm, the antenna pattern is not projected to earth surface and thus no reconstructed PSF is provided. Sp at ia l D om ai n Fr eq ue nc y D om ai n Original PSF Synthetic PSF Target PSF Original PSF Synthetic PSF Target PSF BGI Enhancement from Ch.1 with 3dB beam width of 5.2 to 3.3 BGI Degradation from Ch.3 with 3dB beam width of 2.2 to 3.3 To compare the performance of these two remapping algorithms and evaluate the noise characteristic of the remapped data, the model-simulated observations are generated by integrating the product of Tb and the ATMS PSF. The Tb field is simulated by Community Radiative Transfer Model (CRTM) from the geophysical field provided by Global Forecast System (GFS). The case of hurricane Dorian near South Florida at 1800 UTC August 31, 2019 is used in this study. Enhancement from Ch.1 with 3dB beam width of 5.2 to 3.3