Research On Key Issues Of Satellite Hyperspectral Infrared Data Assimilation

Satellite hyperspectral infrared sounder data has high spectral resolution and high precision which can detect atmospheric profile information and make a great contribution to changing the data assimilation analysis field.However,there are still problems such as the influence of infrared remote sensing data on cloud and the inability to assimilate high-noise band observations when we assimilate the hyperspectral infrared data.In terms of cloud detection,the traditional clear channel cloud-detection method relies on the background field problem.Therefore the “imager-assisted clear channel cloud detection algorithm” was realized.In addition,combined with the Logistic Regression method the "Cloud detection of IASI based on Logistic Regression" is proposed.Subsequently,Logistic Regression’s cloud detection method is combined with the traditional clear-sky channel cloud detection algorithm,and a "clear channel cloud detection algorithm based on Logistic Regression" is proposed.These two improved clear channel method cloud detection methods were introduced to the WRFDA assimilation system,and assimilation and prediction experiments were carried out for the typhoon "Haima".The experimental results show that: "Image-assisted clear channel cloud detection algorithm" and "clear channel cloud detection algorithm based on Logistic Regression" can introduce more clear observation data than traditional clear channel cloud detection method,thus effectively improving analyze the field and forecast effects.This paper uses the PCA compression and reconstruction method to calculate the statistics of the full-band PC coefficient and the sub-band PC coefficient.Then the noise of the infrared hyperspectral data is denoised using the obtained PC coefficient.The experimental results show that:(1)The PC coefficients of the full-band and sub-band have obvious noise reduction effects on the IASI radiation brightness temperature data.(2)In the long wave and medium wave bands,the sub-band PC coefficient has better denoise effect,and the full-band PC coefficient denoise effect in the short-wave band is better.In addition,based on the two essential characteristics of the "compression" and "denoiseing" of the PCA method,this study also introduced the reconstructed brightness temperature to the WRFDA assimilation system.The experimental results show that:(1)Only the reconstructed data of the 2M size principal component can achieve similar assimilation and prediction effects with the 249 M raw data.(2)Due to the excellent noise reduction effect of the PCA method,the data of five high noise channels was successfully introduced into the WRFDA,and the assimilation analysis field was improved.In summary,this paper focuses on the cloud detection and the data compression and denoise problem,which has certain effects on improving the assimilation effect of infrared hyperspectral data and improving the numerical prediction effect.Moreover this study can accumulated the valuable experience in researching Chinese FY series satellites.