Comparison between linear and nonlinear trends in NOAA-15 AMSU-A brightness temperatures during 1998–2010

Abstract

Brightness temperature observations from Microwave Sounding Unit and Advanced Microwave Sounding Unit-A (AMSU-A) on board National Oceanic and Atmospheric Administration (NOAA) satellites have been widely utilized for estimating the global climate trend in the troposphere and stratosphere. A common approach for deriving the trend is linear regression, which implicitly assumes the trend being a straight line over the whole length of a time series and is often highly sensitive to the data record length. This study explores a new adaptive and temporally local data analysis method—Ensemble Empirical Mode Decomposition (EEMD)—for estimating the global trends. In EEMD, a non-stationary time series is decomposed adaptively and locally into a sequence of amplitude-frequency modulated oscillatory components and a time-varying trend. The AMSU-A data from the NOAA-15 satellite over the time period from October 26, 1998 to August 7, 2010 are employed for this study. Using data over Amazon rainforest areas, it is shown that channel 3 is least sensitive to the orbital drift among four AMSU-A surface sensitive channels. The decadal trends of AMSU-A channel 3 and other eight channels in the troposphere and stratosphere are deduced and compared using both methods. It is shown that the decadal climate trends of most AMSU-A channels are nonlinear except for channels 3–4 in Northern Hemisphere only and channels 12–13. Although the decadal trend variation of the global average brightness temperature is no more than 0.2 K, the regional decadal trend variation could be more (less) than 3 K (−3 K) in high latitudes and over high terrains.