surface flux

horizontaladvection

7

8

a

8

7

5

3

6

6

2

7

2

8

3

1

Mechanisms of Summer Subtropical Southern Indian Ocean Sea Surface Temperature Variability

The Importance of Meridional Humidity Advection

A.M. Chiodi1 and D.E.Harrison2 NOAA, Pacific Marine Environmental Laboratory / Joint Institute for the Study ofthe Atmosphere and Ocean, Univ. of Washington, Seattle, WA, USA

(1) Andy.Chiodi@noaa.gov P:206 526-6758(2) D.E.Harrison@noaa.gov P:206 526-6225

10o S

20o S

30o S

40o S

20o E 60o E 100o E 140o E 20o E 60o E 100o E 140o E

oC

SST AnomalyDecember 1997 December 1998

-2 -1 0 1 2

Background

1

Results

40o S

20o S

30o S

40o S

20o S

30o S

oC

10o S

10o S

Observations (NOAA OISST)

Mixed Layer Model (Price et al. 1986)1997 1998

6

4

2

1

0.5

0

2

40o S

20o S

30o S

50o E 100o E 50o E 100o E40o S

20o S

30o S

50o E 100o E 50o E 100o E

a b c

e fd

Components of Monthly LatentHeat Flux Anomalies

Total Humidity Wind Speed

anomalyclimatology( )

q

∆

( )

November 1997

November 1998SurfaceWarming

SurfaceCooling

-0.6 -0.4 -0.2 0 0.2 0.4 0.6

3

Monthly Temperature Tendency(November)

The warm SST anomalies are largely caused byabrupt (2-4oC/mon), coherent warming that occurson scales of roughly 2500 km (see Fig. ). Theshape and magnitude of these warming regions arereasonably well reproduced by a mixed layer modeldriven with surface fluxes of heat and momentum.

100o E50o E 50o E 100o E

Only latent heat anomalies have the magnitudeand shape necessary to drive these SST anomalies. Aperturbation analysis of the standard bulkparameterization of latent heat flux has shown thatthe total normalized latent heat flux anomaly(anomaly/climatological mean) is approximately equalto the sum of two parts; the normalized wind speedand q anomalies ( q = near surface humidity -surface saturated humidity). A pattern similar to theone seen in the SST and latent heat flux anomalies isusually most evident in the q anomaly (see Fig. ).

10o S

20o S

30o S

40o S20o E 60o E 100o E 140o E20o E 60o E 100o E 140o E

10261002 1010 1018(hPa)SLP

November 1997 November 1998

10 m/s

Monthly SLP and Wind (10m)

Data from NCEP Reanalysis

SST/Surface Heat Flux RegionsSLP Regions

Index Averaging Regions

Latent HeatMeridional Wind

SLPSSTAdvective HumidityTendency

1992 1994 1996 1998 2000 2002 2004

2

1

0

-1

-2

3

-3

SST (FM)

2

1

0

-1

-2

Latent Heat (NDJFM)Meridional Wind (NDJFM)

SLP (NDJFM)Advective Tend. (NDJFM)

November

Seasonal

SST/SLP Indices

120o E0o60o E

40o S

20o S

20o N

0o

120o E0o60o E

0 0.3 0.4 0.5 0.6 0.7 1.0

db c e

f ig h j

40o S

20o S

50o E 100o E

-80 -60 -40 -20 0 20 40 60 80 (W/m2)

4

5a

Figure 1: December 1997 monthly mean SST anomaly (left).December 1998 monthly mean SST anomaly (right). Referenceperiod 1990-2004. Data from NOAA OISST.

Figure 2: The observed temperature change duringNovember 1997 (upper left) and November 1998(upper right). The November temperature changeestimated by a mixed layer model using NCEPsurface fluxes (1997 bottom left; 1998 bottomright)

Figure 3: November 1997 mean normalized latent heat fluxanomaly (a), normalized q anomaly (b) and normalizedwind speed anomaly (c). Anomalies are normalized by theirclimatological monthly means. (d), (e) and (f) are thesame as (a),(b) and (c), respectively, except for 1998.

∆

∆

Total ≅ Humidity Component + Wind Speed Component

∆

Visual inspection of SLP and near surface windfields suggest that conditions that are favorable towarming are created within the western/central flankof the subtropical atmospheric anticyclone (Fig. ).4

Figure 4: November 1997 mean SLP (color field) and10m winds (left). November 1998 mean SLP and 10mwinds (right).

A simple index for anticyclone position is significantlycorrelated with SST and latent heat flux variability. Thisindex also correlates well with meridional wind and advectivesurface humidity variability, suggesting that atmosphericadvection drives these anomalies (see Fig. )

Figure 5a: Area averaging regionsfor SLP (yellow) and SST,latentheat flux, meridional wind andhumidity advection (orange).Indices are formed by differencingtwo regions of the same type.

5b

Figure 5b: Southwestern minuscentral basin differences of 30-daymean SST, SLP, advective humiditytendency and meridional windanomalies (upper panel). Themeans are centered on 15 Novemberfor all variables except SST, whichis centered on 30 November.Timeseries are normalized to haveunit variance. Lower panel is thesame, except that SST is averagedfrom Feb. through Mar. and theother variables are averaged fromNov. through Mar.

20o N

20o S

40o S

0o

0o E 50o E 100o E8 m/s

hPa - 1000

W/m2

0o E 50o E 100o E

0o E 50o E 100o E

m/s

20o S

40o S

0o

20o N

kg/kg/s oC

0o 50o E 100o E0o 50o E 100o E0o 50o E 100o E

a) b)

c) d) e) f)

4 8 12 16 20 24

4 2 0 -2 -4 -21 -14 -7 0 7 14 21x10-9

-1.5 -1 -0.5 0 0.5 1 1.560 0 -3030 -60

Composite Winds (10m) and SLPData from NCEP Reanalysis

Composite DifferencesLatent Heat Flux Meridional Wind Humidity Tendency SST Anomaly( -V qa).

h

∆

1992 1994 1996 1998 2000 2002 2004

Composite averages based on the extrema of this SLPindex clearly show a bi-modal pattern of anticyclone variability(Fig. , upper panel). Composite differences show thatthe anomaly pattern of interest is seen in latent heat flux,meridional atmospheric advection and SST variability (Fig. , lower panel).

Figure 6: (a) Mean SLP and 10m wind composite during high western/loweastern SLP intervals. Intervals are 15 days long with 25 intervals percomposite. Period 1992-2004. (b) Same as (a), except for low western/higheastern SLP intervals. (c) Composite difference of latent heat flux (highwestern minus high eastern SLP). (d) Composite difference of meridionalwind speed. (e) Composite difference of advective humidity trend. (f)Composite difference of SST anomaly (lagged 1 week).

Figure 7: Probability that anomalies of q are more importantthan of wind speed to latent heat flux variability. Left panelresults from analysis of the NCEP reanalysis, right panel is fromthe idealized boundary layer model described here.

Figure 8: The boundary layer model latent heat flux anomaly predicted from theNovember 1997 10m wind anomaly (a). Other top row panels are the same as(a) except that; (b) zonal wind anomalies are neglected, (c) wind speed anomaliesare neglected, (d) meridional wind anomalies are neglected, (e) humidity advectionis neglected. The lower panels are the same as those above them, except forNovember 1998.

The way in which such latent heat flux anomalies are created was examinedwith a simple model for low level humidity. The humidity tendency in thisatmospheric boundary layer model is determined by advection, air-sea flux anda generic removal of moisture (e.g. mixing with overlying air mass). Monte Carlosimulations show that the importance of q variability in this model is consistentwith the NCEP reanalysis, but estimated conservatively (see Fig. ).

Experiments with this model show that the model-estimated latent heat fluxanomalies (found from imposed wind speed anomalies and calculated humidityanomalies) are caused primarily by humidity anomalies that are driven byanomalous meridional humidity advection (see Fig. ).

Conclusion

The southern subtropical Indian Ocean SST variability that is known tocorrelate with African rainfall is mainly caused by the following mechanism: (i) meridional wind anomalies advect water vapor and cause low level humidityanomalies which, (ii) cause latent heat flux anomalies that drive the SSTvariability of interest.

Zonal wind anomalies sometimes have important effects on SST variability(as previously reported), but results strongly suggest that the mechanism forthis SST variability primarily depends upon meridional wind-dependentmechanism described here. These meridional wind anomalies appear to bedependent upon the (bi-modal) position of the subtropical anticyclone. Thecause of this anticyclone variability is unknown.

Model Estimated Monthly Latent Heat Flux Anomalies

The Relative Importance of q and Wind Speed Variability

6

Atmospheric Boundary Layer ModelSchematic : In this analytic model,changes in humidity are driven bywind anomalies, which causevariations in surface evaporation andhorizontal advection.

turbulentexchange

∆

∆

∆

We reexamine the mechanism responsible fora type of subtropical Indian Ocean sea surfacetemperature variability (see Fig. for an example)that is known to correlate with rainfall over certainregions of Africa that depend on rainfall for theireconomic well being.

Recent studies have determined that zonalwind speed anomalies are important to the formationof this type of SST variability. Reexamination ofthe mechanism, using ocean mixed layer modeling(see Fig. ), analyses of operational air-sea fluxes(see Fig. ), and consideration of simpleatmospheric boundary layer physics (see Figs.and ) has shown that meridional wind anomaliesare crucial to the formation of the SST variabilityconsidered here. A novel mechanism that is simplydependent upon meridional wind anomalies ispresented.

∆