TEMPERATURE COMPARISONS AT 50 hPa

1 November 1997

METHOD:

Temperature comparisons were made between 18 years of NCEP/CPC
analyses, one year of the NCAR GCM, and one year (1995-1996)
of the DAO assimilation.  

This report examined 50 hPa temperatures in two latitude bins:
40N-50N and 60N-70N.

Temperature fields were binned by month on an equal area grid.
Mean and standard deviations were then calculated for each bin.

For the last step, results from NCAR and DAO transport model
levels were linearly interpolated to 50 hPa from the two 
surrounding levels.  This did not change single level results
greatly because vertical temperature gradients are weak in the
lower stratosphere.

RESULTS:

In the 40N-50N latitude region both NCAR and DAO agree fairly
well with NCEP throughout the year.  NCAR tracks the small
annual cycle better than DAO, however, the DAO has slightly
smaller biases and standard deviation.  In January and February,
the DAO shows slightly higher variability (larger standard
deviations) than NCEP, while NCAR shows substantially less
variability.

In the 60N-70N latitude region, NCAR shows a warm bias (compared
to NCEP) over most of the year (about 2.5 K).  The DAO summer
temperatures agree well with NCEP, while the winter temperatures
are about 3K colder than the NCEP climatology.  The DAO is much
more active (higher standard deviation) in January and February
than NCAR.  The DAO higher activity in January and February
fits NCEP better than NCAR.

DISCUSSION:

Both NCAR and DAO agree well with climatology.  The DAO has a 
more active January and February that agrees better with the
climatology, however, NCAR is not out of the climate range.
Both NCAR and the DAO are within about 2 degrees of the climatology
in the 40N-50N latitude region except in October when NCAR is
3.5 K too warm.  NCAR shows a warmer summer and colder winter,
in agreement with the small annual cycle seen in the climatology
and unlike the DAO where the warmest temperatures occur in February.

CONCLUSION:

Both NCAR and the DAO get a passing grade on the temperature.
It is difficult to chose between the two based on temperature.
The main difference appears to be a less active January and
February in the NCAR model output as reflected in the temperature
fields.