Application of AVHRR data to a one-dimensional variational retrieval scheme for cloudy TOVS data

Chien Ben Chou, Huei Ping Huang

Research output: Contribution to journalArticlepeer-review


The use of the Advanced Very High Resolution Radiometer (AVHRR) data in a one-dimensional variational scheme is examined to retrieve cloud parameters and atmospheric profiles. The variational scheme used TIROS Operational Vertical Sounder radiance data for retrieval. The AVHRR data were used in the partly cloudy and cloudy cases to provide initial guesses for cloud parameters in the iterative scheme, to detect the presence of cirrus clouds, and to determine the sea surface temperature used in retrieval. Sensitivity tests showed that the error in the initial guesses of cloud parameters has substantial impact on the accuracy of the retrieved fields; this sensitivity increases with increased cloudiness. Cloud parameters deduced from AVHRR data are nearly optimal, in terms of maximizing the efficiency of convergence, as the initial guesses for the retrieval scheme. In the absence of cirrus cloud, a retrieval procedure incorporating AVHRR initial guesses produced temperature and humidity profiles for partly cloudy cases that are about as accurate as those for clear cases. In both cases the maximum improvement made in the retrieval procedure over background error was about 0.2 K in the temperature profile, and 0.05 (in logarithm of mixing ratio) in the humidity profile. For partly cloudy cases, best retrieval results were obtained for a low cloud top, or a middle cloud top but with small cloud fraction. Cirrus cloud remains a problem, as its presence generally degrades the quality of retrieval.

Original languageEnglish (US)
Pages (from-to)3867-3878
Number of pages12
JournalMonthly Weather Review
Issue number11
StatePublished - Nov 2000
Externally publishedYes

ASJC Scopus subject areas

  • Atmospheric Science


Dive into the research topics of 'Application of AVHRR data to a one-dimensional variational retrieval scheme for cloudy TOVS data'. Together they form a unique fingerprint.

Cite this