School of Earth and Environment INSTITUTE FOR CLIMATE AND ATMOSPHERIC SCIENCE
Forecasting convection in West Africa
Douglas Parker, Cathryn Birch,
University of Leeds and Met Office
Dartington Hall, 28 January 2013
Contents
1. Some comments on forecasting issues for West Africa
2. Dynamics of convective triggering – a case study
3. Final remarks.
120 km
Drought?
2005 rainfall around Niamey (13.5N, 2E): courtesy Thierry Lebel.
Cambridge
Oxford*
* 1981-2010
Forecasting convection in West Africa
Given the synoptic state, and some existing pattern of convective storms,
... what will the pattern of convective rainfall be in the future?
Tools: • Current observations
• Numerical models
• Statistical relationships / climatology
• Conceptual models
It has been argued for some time, that the conceptual models for the tropics are still immature, relative to the understanding we have for midlatitudes.
School of Earth and Environment INSTITUTE FOR CLIMATE AND ATMOSPHERIC SCIENCE
African Monsoon Multidisciplinary Analyses Afrikanske Monsun: Multidisiplinære Analyser Afrikaanse Moesson Multidisciplinaire Analyse
Analisi Multidisciplinare per il Monsone Africano Afrikanischer Monsun: Multidisziplinäre Analysen
Analisis Multidiciplinar de los Monzones Africanos Analyses Multidisciplinaires de la Mousson Africaine
SOP1&2
SOP3
Around 800 scientists, engineers and operational specialists from 25 countries
7 aircraft AMMA observations, 2004 - 2009
Budget ~ 50 MEu
Over 500 papers and 7 special issues so far.
AMMA and weather forecasting
African Monsoon Multidisciplinary Analysis (AMMA): 2002-2020 programme ; intensive field programme 2005-2007.
Forecast development included in 2006 AMMA Operational Centre (AMMA-AOC) – Jean-Philippe Lafore.
Forecasters’ Handbook being written currently, for completion July 2013, published by Wiley-Blackwell. Editors Parker and Diop-Kane. Conceived in JET2000 – developed in AMMA.
Integration of recent AMMA research with operational forecasting methods We are writing a “Forecasters’ handbook for West Africa.
(JET2000) AOC, 2006 Trieste, 2009 Leeds, 2012 Dakar, 2013
WASF 1800 D+1
New “synthetic analysis” conventions: Lafore et al. West African Synthetic Analysis/Forecast (WASA/F)
WASA 1800 D
WASF 1800 D
MSG OLR 1800 28 July 2006
Some forecaster perspectives
Forecasters don’t trust NWP for convection (but they use it for the large-scale state ...)
• Is this fair? Some cases may be more predictable (e.g. 27-28 July 2006).
• Convergence and divergence are used – dangerous.
Lots of (minor?) differences between theory and practice, • e.g. Academics discuss 925 and 700 hPa ; forecasters like 850 hPa (=
steering level = closed vortices).
• Some concepts not well-defined for objective analysis (e.g. “monsoon trough”, African Easterly Wave trough, ITD, ...).
Availability of tools (e.g. Software) and training in their use, is variable.
Can NWP be skilful?
MSG cold cloud fraction and meridional winds (contours)
Söhne, Chaboureau and Guichard (2008): Meso-NH skill seems to be higher during periods of active synoptic African Easterly waves.
Meso-NH D+1 forecast
23 Aug 2006 23 Jul
Heidke skill score
MSG obs
Some dynamical perspectives 90% of rainfall at Niamey is from organised, propagating convection
(squall lines); ~ 50% in Benin. Need to separate problem of initiation from maintenance of convection (e.g. role of midlevel dryness).
The diurnal cycle is our friend. • Separates cause and effect • Separates isolated and organised convection
Remarkably, there is still no clear consensus on the physics of the relationship between convection and synoptic state in this region.
Synthetic charts need to be: • Repeatable (i.e. Objective rules) • Quantifiable, for evaluation (for instance in geographic position of
objects).
Link to synoptic state
Stein et al. 2011; See also Parker et al. 2005
CloudSat mean cloud-type distributions: Deep convection shaded. Contours = adiabats : Solid black dashed = African Easterly Jet
Dry Saharan air layer (SAL)
Monsoon layer
CIN and midlevel dry air suppress initiation but organise mature systems.
Link to synoptic state
Convective initiation is not closely tied to the African Easterly Waves, in comparison with midlatitude cyclones for example.
Initiation is strongly controlled by • Time of day
• Surface forcing • Mesoscale triggers (e.g. convergence
lines, gravity waves).
Can we improve on this if we define the synoptic environment better.?
Squall line initiations; Fink and Reiner 2003
10° X X
B223 case study: 31 July 2006 “The Mummy”
Parent storm Band of cloud Daughter storm
• Gravity wave emitted from the parent storm may have played a part in the development of the daughter storm
• Could have been a ‘bore’ which travelled along CBL-SAL interface
• Could have been a ‘wave-front’, covering the depth of the troposphere
Daughter storm
Gravity wave
Taylor et al. 2010. QJRMS
Mapes (1993)
Gravity wave-fronts
Marsham and Parker (2006)
n=2 n=3
• The waves extends over the entire troposphere
• n=1 mode travels the quickest • n=2 and 3 are slower • Important over the tropics – lots
of deep convection and influence of the Coriolis force is lower
• A surprisingly small number of model studies of observed cases (non-idealised), none over west Africa
θ (K)
n=1 n=2
MCS
Bores and solitary waves
Elevated inversion layer
Stable nocturnal BL
Density current
Observed in US storms and over Australia
ρ1
ρ2
ρ0
Gravity currents and waves in the lab.
Aaron O’Leary • Wave displaces air upwards – could trigger new deep convection • Waves could continue propagating long after density current has been eroded by the
surface fluxes
Exploring the dynamics with the Met Office Unified Model (MetUM)
• Met Office Unified Model global-12km-4km • 4km explicit convection • 12Z, 30th July 2006 with ECMWF analysis produced best results
See Birch et al. (2012), QJRMetS
15-‐18Z 30th July 2006
18-‐21Z 30th July 2006
21-‐00Z 30th July 2006
Storm initiation and development
TRMM MetUM - control
See Birch et al. (2012), QJRMetS
00-‐03Z 31st July 2006
03-‐06Z 31st July 2006
06-‐09Z 31st July 2006
Storm initiation and development
See Birch et al. (2012), QJRMetS
09-‐12Z 31st July 2006
12-‐15Z 31st July 2006
15-‐18Z 31st July 2006
Storm initiation and development
See Birch et al. (2012), QJRMetS
18-‐21Z 31st July 2006
21-‐00Z 31st July 2006
00-‐03Z 1st Aug 2006
Storm initiation and development
See Birch et al. (2012), QJRMetS
Rainfall rates
See Birch et al. (2012), QJRMetS
Organisation by a synoptic-scale trough (not shown).
• New convective cells initiate on dry soil but close to strong soil moisture gradients
Soil moisture and initiation
Observations Model
See Birch et al. (2012), QJRMetS
Gravity waves
• W1 travels towards the northwest at ~15.7 m s-1
• W2 travels towards the northwest at ~ 20.2 m s-1
• W1 sets up environment for W2 • Second example of initiation by waves
Wave speed calculation
• The waves appear to be wave-fronts rather than bores • Assuming that the tropopause acts as a rigid lid, the vertical
wavelength (λz) of the wave can be estimated using the equation:
where: n is the wave mode = 2 c is the phase speed = 17.5 m s-1
N is the Brünt-Väisälä frequency = 0.01 s-1
Gives a vertical wavelength of λz ≈ 11 km, which is approximately equal to the depth of the troposphere in this case.
See Birch et al. (2012), QJRMetS
Triggering of the B223 storm
• Larger-scale synoptic state determined the existence of the storm.
• Soil moisture pattern determined the location of the storm.
• Gravity wave determined the timing of initiation of the storm.
• The operational model was capable of capturing all this ... with considerable trial and error.
See Birch et al. (2012), QJRMetS
“Conclusions”
• Forecasting convection in this region is a balance between the synoptic control and the mesoscale triggering. Synoptic control seems to provide predictability.
• Can we do more to understand the synoptic control, through better conceptual models?
• The forecaster needs to deal separately with questions of initiation and organisation/maintenance.
• Models can capture these processes a posteriori, but for forecasting we have some way to go.
• We are preparing a “Forecasters’ Handbook” for this region: Chapter 3 (Lafore et al.) will be “Convective systems”.