potential predictability of drought and pluvial conditions over the central united states on...
TRANSCRIPT
Potential Predictability of Drought and Pluvial Conditions over the Central United States on
Interannual to Decadal Time Scales
Siegfried Schubert, Max Suarez, Philip Pegion,
Randal Koster and Julio Bacmeister
Global Modeling and Assimilation Office
Earth Sciences Directorate
29th Annual Climate Diagnostics and Prediction WorkshopMadison, Wisconsin18-22 October 2004
Problem and Approach
Does the predictability of Great Plains precipitation change on inter-annual and longer time scales? If so - why?
Examine the spread of an ensemble of century-long simulations forced with observed SSTs
AGCM: NSIPP-1 (NASA S-I Prediction Project)
Climatology and Skill (Bacmeister et al. 2000, Pegion et al. 2000, Schubert et al. 2002)Great Plains drought (Schubert et al. 2003; 2004)Global grid point dynamical core, 4rth Order (Suarez and Takacs 1995)Relaxed Arakawa-Schubert Convection (Moorthi and Suarez 1992)Shortwave/Longwave Radiation (Chou et al. 1994, 1999)Mosaic interactive land model (Koster and Suarez 1992, 1996)1st Order PBL Turbulence Closure (Louis et al. 1982)
C20C AGCM runs with Specified SST
HadISST and sea ice dataset (1902-1999)22 ensemble members - same SST, different ICs
(14 with fixed CO2, 8 with time varying CO2)
Model resolution: 3 degree latitude by 3.75 degree longitude (34 levels)
Idealized AGCM runs forced with composite SST patterns
Observations
Model ensemble mean
C20C runs
CO2 runs in blue
Quantities
- ensemble mean
2 - intra-ensemble variance
()2 - intra-ensemble coefficient of variation
Great Plains Precipitation (Normalized , Normalized 2 )
Great Plains Precipitation (Normalized , Normalized 2 )
Great Plains Precipitation (Normalized , Normalized )
Great Plains Precipitation (Normalized , Normalized )
JFM 0.11 -.33 0.02 -.15 .37
FMA 0.03 -.53 0.02 -.35 .71
MAM -.26 -.67 -.12 -.41 .75
AMJ -.55 -.76 -.23 -.38 .67
MJJ -.52 -.73 -.23 -.33 .53
JJA -.39 -.73 -.12 -.26 .45
JAS -.08 -.71 .04 -.26 .49
ASO 0.33 -.53 .19 -.29 .59
SON 0.54 -.46 .38 -.30 .70
OND 0.56 -.38 .32 -.30 .70
NDJ 0.41 -.28 .27 -.13 .61
DJF 0.19 -.23 .00 -.11 .23
(,) ((,) (,nino3) ((,nino3) (,nino3)
Summary of
Correlations
• Results show that periods of less rain have greater relative variability than periods of more rain– implies that droughts are less predictable than
pluvial conditions
• How do the SST influence precipitation variability in the Great Plains?– atmospheric variability– land/atmosphere coupling
Correlation Between Ensemble Mean () GP Precipitation and SST
Correlation between SST and GP Precipitation
Correlation between SST and GP Precipitation
Composites based on Great Plains Precipitation
200mb Z Composites Based On Largest/Smallest Values of
Coefficient of Variation of GP Precipitation
Largest Smallest
Difference in Composites of of 200mb Z
Dimensionless
Difference in Composites of of Evaporation
Model Runs with Idealized SST
• Focus on AMJ• Force model with 2 composite SST patterns
– Positive: GP precip > +1 STD– Negative: GP precip < +1 STD
• 100 ensemble members (March 1 - June30) for each composite
• Initial soil moisture conditions are from AMIP runs• Repeat both sets of runs with fixed soil moisture
(fixed beta)
SST Forcing Fields
°C
GP precip > +1 STD
GP precip < +1 STD
Differences in Idealized Runs-Precipitation
Fixed BetaInteractive soil
Differences in Idealized Runs-Evaporation
Fixed BetaInteractive soil
Soil Moisture
From C20C Runs
W W
E
E
W (soil moisture)
Idealized run -1stdIdealized run +1std
Interactive soil Fixed Beta
Idealized run +1std
Idealized run -1std
C20C runs
Conclusions and Implications
• In the Great Plains, simulated droughts are less predictable than pluvial conditions
• Differences in ensemble spread are associated with changes in the strength of the atmosphere/land coupling
• Should also be true in other “hot spots”• Future work - seasonality, model dependence,
other regions (e.g. SW US), SST uncertainty
JJA Land-Atmosphere Coupling Strength, Averaged Across AGCMs