convection ecmwf sep10
TRANSCRIPT
8/3/2019 Convection ECMWF Sep10
http://slidepdf.com/reader/full/convection-ecmwf-sep10 1/33
Modelling and predictability of
weather systems in West Africa :lessons learnt from AMMA
Jan Polcher, LMD/IPSL/CNRSWith contributions from : C. Taylor, S. Bastin, N.
Kalthoff, A. Agusti-Panareda, P. Ruti, ...
African Monsoon Multidisciplinary AnalysesAnalyses Multidisciplinaires de la Mousson Africaine Afrikanischer Monsun: Multidisziplinäre Analysen
Analisi Multidisciplinare per il Monsone Africano Afrikanske Monsun : Multidisciplinære Analyser
Analisis Multidiciplinar de los Monzones Africanos Afrikaanse Moesson Multidisciplinaire Analyse
8/3/2019 Convection ECMWF Sep10
http://slidepdf.com/reader/full/convection-ecmwf-sep10 2/33
Outline
Introduction to the climate of West Africa (WA)
The impact of enhanced atmosphericobservations during AMMA on predictability
Is there a potential for predictability in surface
heterogeneities ?
Small scale moisture gradients.
Cool highs and its impact on African EasterlyWaves.
The impact of inland wetlands on convection.
Can models respond to these surfaceheterogeneities ?
8/3/2019 Convection ECMWF Sep10
http://slidepdf.com/reader/full/convection-ecmwf-sep10 3/33
8/3/2019 Convection ECMWF Sep10
http://slidepdf.com/reader/full/convection-ecmwf-sep10 4/33
Rainfall is mostly brought byconvection
This type of rainfall can leave strong contrastsat the surface between wet and dry surfaces.
[Tao and Moncrieff, 2009]
Using TRMM afew feature of WArainfall can beidentified :
The region has
the largestprecipitatingsystems.
Over 70% of the rainfall in the Sahel originates fromsystems wider than 100km.
In Niamey most of the rainfall falls in 15 to 20 days withmore than 10mm/d.
8/3/2019 Convection ECMWF Sep10
http://slidepdf.com/reader/full/convection-ecmwf-sep10 5/33
2005 average rainfall over the Nigermesoscale site.
2005 Rainy season for 3 stations (LF17)
of the AMMA-Niger Mesoscale site
0
100
200
300
400
500
600
700
800
1/3 1/4 2/5 2/6 3/7 3/8 3/9 4/10 4/11
C u m u l ( m m )
Niamey IRD Kafina Kollo Moy 51/70 Moy 71/90
Decadal déficit
51/70 vs.71/90
The strong spatial variability of rainfall
[C. Bailleul, L. Descroix, pers. comm.]
8/3/2019 Convection ECMWF Sep10
http://slidepdf.com/reader/full/convection-ecmwf-sep10 6/33
Surface characteristics in WestAfrica
May
August
July
Early in the the season, spatial structures ofsoil moisture could offer predictability.(This is information which can be remote sensed.)
Later in the season, soil moisture memory
is likely to play a more important role.
May and June the surface energybalance is driven by bare soilevaporation. Runoff and ponding isimportant.
July vegetation starts to smooth outevaporation. Role of infiltration increases.
August and September the vegetationdrive the surface processes. The roots
extract deeper water.
8/3/2019 Convection ECMWF Sep10
http://slidepdf.com/reader/full/convection-ecmwf-sep10 7/33
Observing the evolution of thesurface
In June the surface keeps amemory of the last rain eventfor 2-3 days => Spatialcontrasts of surface fluxesbetween wet and dry patchesare maintained.
But this evolves through theseason as the soil moistens andthe vegetation grows.
Can RCMs or GCMs, and theirassociated land surfaceschemes, represent theseprocesses correctly ?
[Schwendike et al., 2010]
8/3/2019 Convection ECMWF Sep10
http://slidepdf.com/reader/full/convection-ecmwf-sep10 8/33
Enhanced Observing networks for 2005-2007Special observing periods in 2006
8/3/2019 Convection ECMWF Sep10
http://slidepdf.com/reader/full/convection-ecmwf-sep10 9/33
Improved atmospheric observationsand their impacts on forecasts
8/3/2019 Convection ECMWF Sep10
http://slidepdf.com/reader/full/convection-ecmwf-sep10 10/33
The radio-sounding network upgradeaccomplished by AMMA
During 2006 the AMMA research project has demonstrated that
West African organizations (ASECNA, GMET, NIMET) can providethe same density of radio-sounding as seen in other regions.
It needs to beensured that this
effort issustained ...but this can not bedone by researchprojects !
8/3/2019 Convection ECMWF Sep10
http://slidepdf.com/reader/full/convection-ecmwf-sep10 11/33
Bias corrections of humidity
The humidity bias correction was one of the mostimportant contributions to improving the analysis.
[A. Agusti-Panareda et al. 2009, M. Nuret al.2008]
The GPS stations allowed to verify the precipitable waterobserved by radio sounding.
The large diversity of sounds made the correction of biasesa difficult task.
8/3/2019 Convection ECMWF Sep10
http://slidepdf.com/reader/full/convection-ecmwf-sep10 12/33
8/3/2019 Convection ECMWF Sep10
http://slidepdf.com/reader/full/convection-ecmwf-sep10 13/33
Impact of AMMA data on precipitationforecasts
Météo-France evaluated the impact of the AMMA data on the forecasts over themonth of August (4 day forecasts startingat 0hUTC).
The model continues to have asoutherly placed convergence zone.
Intense precipitation features arebetter represented.
[C.Faccani et al. 2010]
The equitablethreat scores are
improved – butscore remain weak for intenserainfall events.
8/3/2019 Convection ECMWF Sep10
http://slidepdf.com/reader/full/convection-ecmwf-sep10 14/33
Remote impact of improved West Africananalysis
RMSE Z500t0+48h
RMSE Z500
t0+72h
The improvements in the forecastare advected into the mid-latitudes.This leads to significantimprovement of the Z500 scoresafter 48 and 72 hours.
Some questions remain :
Does this benefit exist throughthe entire season ?Is this particular to the 2006
conditions ?
Averaged over the period 1-Aug to 14 Sep
8/3/2019 Convection ECMWF Sep10
http://slidepdf.com/reader/full/convection-ecmwf-sep10 15/33
Is there a potential for predictability insurface heterogeneities ?
8/3/2019 Convection ECMWF Sep10
http://slidepdf.com/reader/full/convection-ecmwf-sep10 16/33
The rôle of small scale soil moistureanomalies
Potential temperature
Convection onthe 31th of July2006 left a traceon surface
temperature.
Atmospheric moisture
Boundarylayerconditionson theaircrafttrack.
≈ 250 km
8/3/2019 Convection ECMWF Sep10
http://slidepdf.com/reader/full/convection-ecmwf-sep10 17/33
Impacts on the atmospheric dynamics
The moist patches modify thestructure of the PBL and inducedivergent circulations.
[Taylor et al., 2007]
15N 16N
Potential
temperature
Mixing
ratio Westerlies
Southerlies
Wind
Temperaturegradient
8/3/2019 Convection ECMWF Sep10
http://slidepdf.com/reader/full/convection-ecmwf-sep10 18/33
The spatial scales inducing circulations
The analysis of a wider setof flights has shown thatthe mesoscale circulations(phase and intensity)depend on the background wind
The spectral coherency analysis allows to identify the spatialscales at which the surface influences the atmosphere.
10Km patches and larger generate significant atmosphericstructures. Θ
eis in phase with surface
structures while wind is
out of phase.
Coherence(T s,V) Coherence (T
s,Θ
e )
8/3/2019 Convection ECMWF Sep10
http://slidepdf.com/reader/full/convection-ecmwf-sep10 19/33
Simulations of the 11th of June 2006
Mali
Burkina Faso
Ghana
Ivory Coast
Benin
Model setup:COSMO model (former LM) from DWDResolution: 0.025 ° (2.8km)No convection
parameterisation scheme50 layers in the vertical(up to 28 km)Multi-layer SVAT model
TERRA_M (7 soil layers)
Initialisation and boundaryconditions:
Operational analyses fromECMWF
Initiated on 11 June 2006 at 00UTC
[Gantner & Kalthoff, 2010, Adler et al., 2011]
8/3/2019 Convection ECMWF Sep10
http://slidepdf.com/reader/full/convection-ecmwf-sep10 20/33
Soil moisture conditions for modelinitialisation
MOI AMSR-E
CTRL BAND /65% of MOI HOM
COSMO initial volumetric soil moisture in % in the uppermost layer on 11 June 2006 at
0000 UTC for the cases MOI , CTRL, and BAND/HOM case.
8/3/2019 Convection ECMWF Sep10
http://slidepdf.com/reader/full/convection-ecmwf-sep10 21/33
Model results: initiation of convection
Orography in m, vertical velocity (omega) in Pa s-1 averaged between 1200 and 1500 UTC (dotted isolines, interval 0.1, start value 0.2), horizontal wind vector in m s-1 at 1700 UTC at 950 hPa, and precipitation in mm h-1 (solid isolines) at 1700 UTC on 11 June 2006.
The monsoon flow is out of the South West.Some precipitating cells developed in the lee of slopes
(upward motion)But only the surface conditions explain the triggering
needed.
CTRL HOM MOI
8/3/2019 Convection ECMWF Sep10
http://slidepdf.com/reader/full/convection-ecmwf-sep10 22/33
Model results: initiation of convection
Volumetric soil moisture in % at 1500 UTC in the uppermost layer and precipitation inmm h-1 on 11 June 2006 at 1700 UTC.
HOM contains the heterogeneities of vegetation
HOM and MOI are wetter and thus have shallowerCBLs and higher CIN.
Cells developed over dry patches where thermallyinduced converges destroyed the CIN.
HOM and MOI developed thermal contrast through
vegetation and soil texture heterogeneities.
CTRL HOM MOI
8/3/2019 Convection ECMWF Sep10
http://slidepdf.com/reader/full/convection-ecmwf-sep10 23/33
The impact of a dry spot on a system
24-h accumulated precipitation in mm starting from 0600 UTC on 11 June 2006 Hovmöller diagram of precipitation in mm h-1 averaged between 10.5 and 13.5 °N
CTRL
PBL is very high over the drier area =>downward mixing of momentum
Thermally induced circulation and opposingbackground winds => convergence zone
Superposition of both effect => enhancedconvergence over the western part of thedry area & triggering of convection
8/3/2019 Convection ECMWF Sep10
http://slidepdf.com/reader/full/convection-ecmwf-sep10 24/33
Initiation of convection during 2006
Using the ISIS system of Météo-France and Meteosatimages the initial location of storms could be traced back .
The locations are then placed onthe maps of surface temperature
anomalies to compute gradients.
AMMA deliverableD1.3.3.i, Taylor et al.
(ISIS picks-up most systems around 16Z but first cold clouds arevisible 2 hours earlier.)
8/3/2019 Convection ECMWF Sep10
http://slidepdf.com/reader/full/convection-ecmwf-sep10 25/33
Initiation and temperature gradients
Gradients on randomly chosen points are computed todefine a reference distribution.
75 initiations 123 initiations
73 initiations 45 initiations
Maximum initiation when :
Ts
gradient is
opposite to the
direction ofbackground wind.
Wind opposes thesoil moistureinduced circulation.
The length scaleof grad. is 40km.
In this region soil moisture gradients enhance initiationof convection by 13% compared to 12% by orography.The role of soil moisture changes during the season.
June June July
August September
8/3/2019 Convection ECMWF Sep10
http://slidepdf.com/reader/full/convection-ecmwf-sep10 26/33
Surface processes and synopticvariability
Compositing anomaliesof hot surfacetemperatures allows tolink them toatmospheric features.
A deep and dry PBL formsover the hot spots.A heat low is generates.
To the East southerlies (@925hPa) intensify at 06Z.A cool and moist high
follows
Meteosat Obs. ECMWF Analysis-10 Longitude 10
T i m
e
[Taylor et al., 2005]
8/3/2019 Convection ECMWF Sep10
http://slidepdf.com/reader/full/convection-ecmwf-sep10 27/33
Wet spots at the larger scale
Using remote sensed soil wetness anomalies(microwave polarization) and meteorological analysis
products over 9 wet seasons, composites of theatmospheric response to wet patches can beconstructed.
The impact of wetpatches is consistent with high cloud coverevents.The wet events
modify PBLtemperatures for 3-4days.The cool high
seems to be a robust
feature. Taylor (2007)
Soil wetness
8/3/2019 Convection ECMWF Sep10
http://slidepdf.com/reader/full/convection-ecmwf-sep10 28/33
Impact of the cool high on circulation
The cool high induces a low level vortex whichincreases the southerlies in front of the wet patch and
might help propagate the system.
These dynamicalstructures are similarto those associatedwith the intra-seasonalfluctuations of rainfall[Janicot and Sultan, 2001]
The estimatedintensity of the cool
high correspondswell to the analysedfields.
Heat flux anomaly [W/m 2 ]
A theoretical model : surface cooling → geostrophic vortex
R i l l t li
8/3/2019 Convection ECMWF Sep10
http://slidepdf.com/reader/full/convection-ecmwf-sep10 29/33
Regional scale wetness anomalies
In order to determine therole of the delta ingenerating rain 24 years of meteosat data (for Augustand September) were used.
Using the morning warningrates the data was split inyears with large and smallwetlands (wet/dry years).
9W 0W
16N
12N
[Taylor, 2009]
Koulikoro
Niamey
Initiation of convection in the Niger
8/3/2019 Convection ECMWF Sep10
http://slidepdf.com/reader/full/convection-ecmwf-sep10 30/33
Initiation of convection in the Nigerdelta.
During wet years more coldclouds are observed in thevicinity of the delta.
The clouds propagate Westat 17ms-1.
The increase in high cloudsoccur in the afternoon.
The increase in initiations ismaximum to the West of thedelta.
This is consistent with a
wetland breeze opposing thebackground wind (S.W.) andinitiating convection.
Cold cloud cover change (wet - dry)
% of dry
Numberofinitiat
ions(3-6.5W)
W E C
(wet - dry)
The inner delta initiates 54%
more storms during wet years.
8/3/2019 Convection ECMWF Sep10
http://slidepdf.com/reader/full/convection-ecmwf-sep10 31/33
Are models able to exploit thispredictability ?
Most model compared inAMMA-MIP were globalmodels, except for :
PROMESENSEMBLES : RCMsforced by ERA-Interim
There is an indication that deficiencies in rainfall anddynamics are related.
All models misrepresent the diurnal cycle of
convection. [Ruti et al., 2010]
Exploiting this predictability is a
8/3/2019 Convection ECMWF Sep10
http://slidepdf.com/reader/full/convection-ecmwf-sep10 32/33
Exploiting this predictability is amodeling challenge
Soil moisture gradients
High resolution models are needed and the possibilityof sub-grid parametrizations need to be explored.Time constants of bare soil evaporation have to be
correct.Phenology of the vegetation needs to be predicted.
Cool highsThe phasing of rainfall and the diurnal cycle is criticalfor generating the vortex.
Regional scale land surface anomaliesMore complete representations of the hydrological
cycle are needed.Rivers, lakes, flood plains and irrigated areas need tobe able to interact with the atmosphere.
The most immediate challenge for models remains thediurnal cycle of precipitation.
(3-4 hours of strong solar radiation before dusk will erode thesurface gradients.)
8/3/2019 Convection ECMWF Sep10
http://slidepdf.com/reader/full/convection-ecmwf-sep10 33/33
Conclusions
A denser radiosounding network increases
predictability … but its impact is limited by thesystematic biases in models.
As in most regions where CAPE and CIN are high, thetriggering mechanisms are key to predict the initiationof convection.
A better representation and initialization of landsurface are likely to increase predictability in WestAfrica (spatial structures as well as time scales).
This would require model improvement in thefollowing areas :
A more detailed representation of surfaceprocesses and their diversity : phenology, fullhydrological cycle.
A higher resolution.Diurnal cycle of convection.