JMA plans to begin operating Himawari-10, the Himawari-8/-9 follow-on satellite, in Japanese fiscal year 2029. In addition to improved observation functions of the imager, Himawari-10 will be equipped with a hyperspectral infrared sounder named GHMS (Geostationary HiMawari Sounder) as a new instrument. This is expected to be the third infrared sounder onboard a geostationary meteorological satellite, following GIIRS onboard the FY-4 series and IRS onboard the MTG-S.
One of the challenges in handling data from infrared sounders is the volume of data, which is especially problematic for satellite operators in terms of data distribution. One of the solutions to this problem is data compression by principal component analysis (PCA), which is also described in WMO's unified data policy. PCA is a statistical technique that uses the correlation between variables at hundreds of channels. By transforming correlated variables into a set of uncorrelated variables, random noise can be removed, and data can be represented with fewer variables.
In this presentation, we will present an investigation of information loss due to compression of hyperspectral infrared sounder data by PCA using existing hyperspectral measurements. Since data compression by PCA is lossy, it may also lose some of the atmospheric signal, although most of the removed data is random noise. We believe that a good understanding of the characteristics of data compression is necessary to ensure the quality of the delivered data.
One of the challenges in handling data from infrared sounders is the volume of data, which is especially problematic for satellite operators in terms of data distribution. One of the solutions to this problem is data compression by principal component analysis (PCA), which is also described in WMO's unified data policy. PCA is a statistical technique that uses the correlation between variables at hundreds of channels. By transforming correlated variables into a set of uncorrelated variables, random noise can be removed, and data can be represented with fewer variables.
In this presentation, we will present an investigation of information loss due to compression of hyperspectral infrared sounder data by PCA using existing hyperspectral measurements. Since data compression by PCA is lossy, it may also lose some of the atmospheric signal, although most of the removed data is random noise. We believe that a good understanding of the characteristics of data compression is necessary to ensure the quality of the delivered data.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner