A new procedure for estimating observation error in AMSU data and its application to retrieval

Chien Ben Chou, Huei Ping Huang

Research output: Contribution to journalArticle

1 Scopus citations

Abstract

An accurate estimate of observation errors is crucial to the retrieval of atmospheric profiles from satellite data using a variational method. In practice, observation errors, both systematic and random, are often estimated from the difference between satellite observations and simulated satellite observations obtained from a radiative-transfer operator with a 12 h forecast as its input. Observation errors estimated by this approach may be contaminated by errors in the 12 h forecast. This work describes a practical way to eliminate the 12 h forecast error and improve the estimate of the observation error in the Advanced Microwave Sounding Unit (AMSU) data. Following the philosophy of the National Meteorological Center (NMC) method (that derives the statistics of forecast error from the differences between pairs of forecasts at disparate ranges valid at the same time), in this study the pairs of forecasts at different ranges in the NMC method are first converted to brightness temperatures in the AMSU channels by a radiative-transfer operator. The 12 h forecast errors are then determined from the representations of these forecasts in radiance space spanned by the AMSU channels. Since most AMSU channels have beam position-dependent systematic observation errors, the procedure further takes into account this dependence by performing the statistics separately for sub-groups of data in each AMSU channel with different beam positions. In a case-study, after eliminating the 12 h forecast error obtained by this procedure from the total estimated observation error, the remaining random error of the satellite observation is shown to be smaller than the background error (provided by 12 h forecasts of a numerical weather-prediction model) in most of the AMSU temperature sounding channels. Using the effor-corrected AMSU data in these channels, a retrieval experiment using a one-dimensional variational scheme shows an improvement of 0.2-0.4 K over the background error in the retrieved temperature profiles above 780 hPa.

Original languageEnglish (US)
Pages (from-to)79-101
Number of pages23
JournalQuarterly Journal of the Royal Meteorological Society
Volume130
Issue number596 PART A
DOIs
StatePublished - Jan 1 2004
Externally publishedYes

Keywords

  • Data assimilation
  • Remote sensing
  • Variational method

ASJC Scopus subject areas

  • Atmospheric Science

Fingerprint Dive into the research topics of 'A new procedure for estimating observation error in AMSU data and its application to retrieval'. Together they form a unique fingerprint.

  • Cite this