Joint IPSL/NCAS-Climate meetingParis, 14-16 May 2007
The impact of basic state errors on monsoon intraseasonal variability and the
intraseasonal-interannual relationship
A. G. Turnera,b, P. K. Mohantyc, J. M. Slingoa,b, P. M. Innessa,b
a National Centre for Atmospheric Science
b Walker Institute, University of Reading, UKc Department of Marine Sciences, Meteorology and Physical Oceanography
Laboratory, Berhampur University, India
Outline
Model framework
Spatio-temporal behaviour of intraseasonal variations
Active-break composition using an OLR index
Evolution of selected fields
Combined EOF analysis in model integrations
Intraseasonal-interannual relationships and their link to the basic state
Summary
Model framework
Control integration (HadCM3) run for 100 years with daily diagnostics.
A further integration uses limited-area heat flux adjustments to counteract the effect of systematic model error (HadCM3FA). Also run for 100 years (only 85 years daily data available).
Comparisons made with ERA-40 reanalysis (Uppala et al. 2005)
Heat flux adjustments
Traditionally used in older models (e.g. HadCM2) to prevent climate drift; HadCM3 does not have this problem.
Used here to counteract biases in the mean state.
Devised by Inness et al. (2003) to investigate the role of systematic low-level zonal wind and SST errors on the MJO.
Coupled model run for 20 years, Indian and Pacific SSTs within 10S-10N relaxed back towards climatology.
Anomalous heat fluxes generate a mean annual cycle which is applied to a new 100 year integration (HadCM3FA, described in Turner et al. 2005).
Heat flux adjustments
Large fluxes (up to 186Wm-2 at 120W) into the cold tongue.
Much smaller (~30W.m-2) over Maritime Continent and Indian Ocean.
Annual Mean
Amplitude of annual cycle
Small annual cycle apart from upwelling region off African coast, and central Pacific.
Improvements to the mean state
HadCM3FA mean summer (JJAS) surface temperature
differences with HadCM3
HadCM3 differences with ERA-40
Spatio-temporal behaviour in intraseasonal bands
Daily anomalies to seasonal cycle split into 10-20 day and 30-60 day bands using Lanczos filter.
Bands chosen representative of observed modes: models show reasonable power spectra at these frequencies.
10-20 day band commonly associated with westward propagation (recently suggested this is movement of depressions along the monsoon trough; Krishnamurthy & Shukla 2007).
30-60 day band displays northward and eastward propagation (the NPISO, boreal summer ISO, related to MJO?).
Spatial variation in intraseasonal bands
Percentage of total intraseasonal variance explained in each band in reasonable spatial agreement with reanalysis (850hPa zonal winds).
Similar pattern for OLR.
10-20day30-60day
ERA-40 HadCM3 HadCM3FA
Temporal variation in intraseasonal bands
Lag regressions of u850 averaged 5-15°N (top) and 70-90°E (bottom) against reference timeseries (85-90°E, 5-10°N) after Goswami & Xavier 2005.
Some evidence of westward and northward propagation.
10-20day30-60day
HadCM3FAHadCM3ERA-40
Active-break composites using an OLR index
OLR index of Vecchi & Harrison (2002)1 used to define active and break events.
Those events which persist for five or more days σ above or below the mean are selected as active and break events respectively.
Composites at different lags generated with respect to event onset.
1difference between normalized 7-day boxcar smoothed OLR anomalies (10-30°N, 65-85°E) and (10°S-5°N, 75-95°E) minus 50 day centred mean.
Evolution of active and break events: 10-20 day precip
HadCM3
Active Break
HadCM3FA
Active Break
-10
-5
-3
-1
0
Evolution of active-break events:30-60 day precip
HadCM3 HadCM3FA
Active ActiveBreak Break
-20
-15
-10
-5
0
Active-break composites: mixed layer depth anomalies
HadCM3 HadCM3FA
Act
ive
Bre
ak
Active-break composites: wind stress curl anomalies
HadCM3 HadCM3FA
Act
ive
Bre
ak
Evolution of active-break events:30-60 day wind stress curl anomalies
HadCM3FA
Active ActiveBreak Break
-20
-15
-10
-5
0
HadCM3
Evolution of active-break events:30-60 day mixed layer depth anomalies
Active ActiveBreak Break
-20
-15
-10
-5
0
HadCM3 HadCM3FA
Combined EOF analysis
EOF analysis performed on combined wind and precipitation fields.
Data from HadCM3 and HadCM3FA combined into one timeseries to find common modes of variation and assess differences in interannual-intraseasonal relationships.
Anomaly timeseries calculated using two methods (after Krishnamurthy & Shukla 2000):
1: daily rainfall anomaly including seasonal anomaly R’(m,n) for day n of year m, calculated by removing daily rainfall climatology.
Combined EOF analysis #1
First two combined EOFs show evidence of seasonal mean monsoon behaviour
Combined EOF analysis #1: stratify PCS by JJAS mean AIR
PC1 much more strongly perturbed by JJAS mean AIR in HadCM3FA.
Suggests basic state errors influence projection of IS onto IA variability.
red=+1σ, blue=-1σ.
HadCM3
HadCM3FA
Combined EOF analysis #1: stratify PCS by JJAS mean DMI
Both PC1 and PC2 more strongly perturbed by JJAS mean DMI in HadCM3FA.
Suggests basic state errors influence projection of IS onto IA variability.
red=+1σ, blue=-1σ.
HadCM3
HadCM3FA
Combined EOF analysis #1: stratify PCS by JJAS mean Niño-3
PC1 more strongly perturbed by JJAS mean ENSO forcing.
Suggests basic state errors influence large scale forcing on monsoon ISV.
red=+1σ, blue=-1σ.
HadCM3
HadCM3FA
Combined EOF analysis
Anomaly timeseries calculated using two methods (after Krishnamurthy & Shukla 2000):
2: daily rainfall anomaly with removed seasonal residual R’’(m,n) for day n of year m, calculated by removing daily rainfall climatology and seasonal mean anomaly from rainfall timeseries.
Combined EOF analysis #2
First two combined EOFs show evidence of seasonal mean monsoon behaviour
Pattern correlations with other method:
0.97, 0.96, 0.95, 0.83
Combined EOF analysis #2: stratify PCS by JJAS mean AIR
Less obvious difference in PC1 perturbation by JJAS mean AIR.
Suggests most of IS-IA link is due to seasonal residual in the anomaly.
red=+1σ, blue=-1σ.
HadCM3
HadCM3FA
Combined EOF analysis #2: stratify PCS by JJAS mean DMI
PC2 more strongly perturbed by JJAS mean DMI in HadCM3FA, similar to method #1.
PC1 unperturbed by stratification, like method #1.
red=+1σ, blue=-1σ.
HadCM3
HadCM3FA
Combined EOF analysis #2: stratify PCS by JJAS mean Niño-3
PC2 more strongly perturbed.
Is seasonal residual completely removed?
red=+1σ, blue=-1σ.
HadCM3
HadCM3FA
Combined EOF analysis
Initial stratification of PCs suggests stronger links between interannual and intraseasonal variations in HadCM3FA
Removal of residual of seasonal mean anomaly by simple method tends to suggest little projection of intraseasonal behaviour onto interannual variation.
Projection of IS onto IAV occurs most in HadCM3FA, where interannual variability of the monsoon-ENSO system is much stronger (Turner et al. 2005).
More analysis required (careful choice of filtering) to remove any remaining residual from the daily precipitation anomalies.
Summary
Basic state has limited impact on spatio-temporal evolution of monsoon intraseasonal variability using flux adjustment method.
Coherent active/break structures in wind stress curl, MLD, and other ocean fields.
Suggestion of greater projection of intraseasonal monsoon variability onto interannual variations in HadCM3FA, however this may be due to residual from seasonal mean.