How Reliable are Atmospheric Temperature Trends from Climate Reanalyses?

Climate reanalysis development community generally understands that reanalysis datasets are not suitable for temperature trend investigations since large trend differences between reanalyses and satellite observations were found both in satellite bias correction practice in reanalysis development and direct comparisons between the two different types of data. However, the reasons for these differences are not well understood by the broad climate community that uses the datasets. Confidence or lack of it in temperature trends from the climate reanalysis requires a good understanding of reliability of the satellite observations as well as the methodologies used for development of the reanalysis datasets. Recently, NOAA/NESDIS inter-calibrated AMSU-A observations onboard NOAA polar orbiting satellites and found that pre-launch calibrations for most AMSU-A channels on NOAA-15 yield atmospheric temperature trends consistent with those from the post-launch inter-calibrated observations. This suggests that, although NOAA-15 incurred significant orbital drift during its nearly 14 years of operations, the pre-launch calibrated observations most likely yield reliable temperature trends for most atmospheric channels. Since NOAA-15 were used in most reanalysis datasets such as ERA-Interim, NASA MERRA, and NOAA CFSR, deviations in trends of these reanalyses from NOAA-15 observations may provide an insight on how reliable are reanalysis temperature trends when they are used for global change investigations. This study examines AMSU-A bias correction time series in ERA-Interim for NOAA-15 and other polar orbiting satellites for several important tropospheric and stratospheric channels. It was found that ERA-Interim agrees fairly well with the AMSU-A observations during 1998-2005. However, ERA-Interim departures from the AMSU-A observations after 2005 throughout the troposphere and lower stratosphere. This is consistent with previous finding by the ERA-Interim developers. We will analyze the reasons for the departures and provide recommendations for optimal use of the inter-calibrated AMSU-A data in future reanalysis development to minimize these departures for the climate reanalysis to have more reliable trends.