NOAA Center for Satellite Applications and Research banner
 

Demonstrating Suomi NPP CrIS Full Spectral Resolution SDR Products using STAR Offline Processing System

Principal Investigator: Yong Han

An example of CrIS FSR (red line) and normal spectral resolution (black line, offset by 50 K) spectra

Click to enlarge - an example of CrIS FSR (red line) and normal spectral resolution (black line, offset by 50 K) spectra. The spectral resolutions of the FSR mid-wave and shortwave bands are increased by factors 2 and 4, respectively, in comparison to those of the NSR SDRs. Channels sensitive to the carbonate gases in the two bands are labeled with CH4 (1210 - 1400 cm-1), CO (2155-2190 cm-1) and CO2 (2300-2370 cm-1).

NOAA/STAR has developed a ground processing system to generate the full spectral resolution (FSR) Sensor Data Records (SDRs) for the Cross- Track Infrared Sounder (CrIS) instrument onboard the Suomi National Polar-Orbiting Partnership (S-NPP) satellite. This achievement laid the ground work for the data production, and demonstrated the S-NPP FSR SDR products. Started on December 04, 2014, when the CrIS was commanded to the FSR mode operation from the normal spectral resolution (NSR) mode operation, STAR routinely produces the FSR SDRs, available to the public via the STAR FTP site with a data latency of less than 12 hours. The FSR SDR includes a total of 2211 spectral channels, a substantial increase from the 1305 channels of the normal spectral resolution (NSR) SDRs, which continue to be generated by the NOAA operational processing system after the CrIS FSR transition. Figure 5 shows an example of the FSR and NSR spectra. The spectral resolutions of the FSR mid-wave and shortwave bands are increased by factors 2 and 4, respectively, in comparison to those of the NSR SDRs. The FSR radiance data are critical for retrieving carbonate products such as carbon monoxide (CO), carbon dioxide (CO2) and methane (CH4). In addition, the increase of the information content from the FSR radiance data on atmospheric temperature and water vapor improves the numerical weather prediction (NWP). STAR is utilizing the CrIS FSR data in the NOAA-Unique CrIS/ATMS (Advanced Technology Microwave Sounder) Processing System to generate carbonate products.

The development of the FSR SDR processing system was challenging in both algorithm and software, because the current SDR algorithms designed for the NSR SDRs do not work correctly for the FSR SDR processing and the current SDR software was not scalable to accommodate the increased dimensions of the FSR data. The new and improved algorithms include self-apodization correction, spectrum resampling, radiance noise estimation and dynamic update of the Correction Matrix Operator (CMO). A modification of the self-apodization correction algorithm was made in order to remove large radiance ringing artifacts in the shortwave band due to an inappropriate expansion factor, which results in the breakdown of energy conservation when applied to the FSR SDR processing. The resampling matrix calculation is reformulated in the un-decimated spectral domain rather than decimated domain to avoid the negative impact of the guard bands in which the signals are contaminated with noise. The FSR noise calculation includes the spectral calibration, which is ignored in the NSR processing, to take into account the noise contributions from spectral calibration in the mid- and short-wave bands due to increased spectral resolutions in these two bands. The new CMO handling algorithm is developed to improve spectral accuracy by more frequent updates of the CMO matrix since the spectra with increased spectral resolution is more sensitive to spectral calibration uncertainty.

Substantial work was done to upgrade the SDR software and create a computational environment within which the FSR processing is automatic and efficient. The FSR SDR software includes a new framework that resolves the key issue of the backward compatibility and one source code for both FSR and NSR SDR processing. The STAR FSR SDR processing system is required to process the data in near real-time on an ordinary Linux computer. To meet this requirement, an efficient computational environment was developed to manage input and output data and data processing in a parallel computing fashion.