Michael G. Bosilovich and A li d M d SilArlindo M. da SilvaNASA GSFC GMAO
Reanalyses assimilate a broad suite of Reanalyses assimilate a broad suite of observations, yielding a single consistent productp
The presence, or lack, of observations often determines the quality of reanalysis data
Observation reports from the analysis are usually in observation‐space format, or diverse bi ii d t f tbinary or ascii data format
Research studies do not usually indicate the presence of observations and their impact on presence of observations and their impact on reanalysis
NASA’s most recent reanalysis –Modern‐Era l f h d lRetrospective analysis for Research and Applications
(MERRA) – Rienecker et al. (2011, J. Clim) 1979 – present; more than 300 diagnostic variables1979 present; more than 300 diagnostic variables
Gridded Innovations and Observations (GIO) Derived from GSI “diag” (feedback) files, and includes the
b ti b k d f t (O F) d l i observation, background forecast error (OmF) and analysis error (OmA)
Conventional obs (sonde, stn pressure, etc) and also Radiance data
Gridded to MERRA’s native pressure coordinate –½°×⅔°, 42 pressure levels for every 6 hour analysis time4 p y y
How deep can we dig into the analysis and observations with this kind of post processed observations?
Warm biases in commercial aircraft –Cardinali et al. Warm biases in commercial aircraft Cardinali et al. (2003) and Ballish and Kumar (2008) Use forecasts or increments to evaluate obs
f Some limitations – few seasons, small regions Recommend aircraft T bias corrections
GIO ComparisonGIO Comparison Develop collocations for a 31 year period over the continental USI iti ll l ki t b C i l l ll ti Initially looking at 200mb – Cruise level, more collocations
Collocations are based on the 6 hourly analysis and MERRA analysis grid
Can we use these collocations to derive the characteristics for bias corrections in reanalyses?
Some grid points are screened due to low number of collocations
Apr 1991 – MDCRS Start
A‐F ~ (O‐F) + β( ) β ‐ Effective Gain: A measure of how much the analysis draws to an observing systemβ C t t l Bi β – Contextual Bias: a measure of bias of an observing system observing system relative to other contributing 91‐09 (200mb) β
Density
(340K Collocations)
contributing observing systems
91 09 (200mb) βAirCraftT 0.50 ‐0.04
RAOB T 0.41 0.21
As Aircraft observations 0 20.30.4
Contextual Bias (K)
As Aircraft observations become prevalent their β is reduced and Gain ‐0 3
‐0.2‐0.10.00.10.2
βincreases
RAOBs have somewhat ‐0.6‐0.5‐0.4‐0.3
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
continuous gain, but become biased
RAOB Acraft
0 450.500.550.60
Effective Gain
Aircraft observations control the analysis at
b th US 0 250.300.350.400.45
200mb over the US after 1991
0.200.25
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
RAOB Acraft
Effective Gain
0 5
0.6
0.7
RAOB
0.3
0.4
0.5 RAOB
ACraft
N15 ch7
N16 ch7
0.1
0.2N16 ch7
N18 ch7
N14 ch3
N14 HIRS40
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
N14 HIRS4
Near‐200mb peaking channels from MSU, AMSU and HIRS
Radiosonde obs more consistent in time than aircraft Some vertical variations not yet accounted for
Process at higher resolution?Process at higher resolution? For example, better locate data near tropopause
Integrate radiance and conventional Integrate radiance and conventional observation comparisons
Assuming Radiosonde output is accurate Assuming Radiosonde output is accurate, aircraft bias correction needs to be developed but will need to consider developed, but will need to consider seasonality and altitude.
Many uses in identifying observational Many uses in identifying observational influence on the reanalysis
Thanks for any [email protected]@nasa.gov