ECMWF will be performing a 40-year ReAnalysis (ERA-40) for the period 1958-present using a higher resolution model (T159) and a more advanced data assimilation system (3DVAR) than was used in the previous 15-year reanalysis (ERA-15). One purpose of ERA-40 is to provide a complimentary data set to the 50-year NCEP/NCAR reanalysis for use in climate change studies. While it is clear that climate-scale noise due to changes in the model and data assimilation systems have been eliminated, another source of noise, due to changes in the observing system, remain in both reanalyses. This paper addresses the impact of observing system changes, in particular from satellites, on the low frequency components of the ERA-15 and NCEP reanalyses and plans for compensation in ERA-40.
Two changes are analysed: 1) the transition from no satellite to satellite data that occurred around 1979 for the NCEP reanalysis; and 2) drifts in the NOAA 9 and 10 TOVS radiances in 1986 and the transition from one- to two-satellite coverage in 1989 for ERA-15.
In November, 1986 a subtle change in the NOAA 9 TOVS cloud-cleared radiance data occurred. This change, combined with the loss of NOAA 9 and the start of NOAA 10 in January, 1987 and the removal of surface pressure observations in western Brazil at the same time, appear to have caused a profound and persistent impact on the ERA-15 temperature/moisture of lower tropical troposphere and the hydrologic cycle. A similar, but higher amplitude temperature bias was found in the NCEP upper troposphere/lower stratosphere during the transition to satellite data in 1979. Thus, application of reanalysis to climate change studies will be more complicated and of less accuracy than previously expected. Further, detection and compensation for observing system variation will be a major focal point of the ERA-40 pre-production experimentation and monitoring tool development