UK Model Cloud and Convection
Presentation to MOSAC 2016, Paper 21.10
Adrian Lock
Kwinten van Weverberg, Ian Boutle, Cyril Morcrette,
Mohamed Zerroukat, Nigel Wood, Chris Smith, Anne McCabe,
Kirsty Hanley, Michael Whitall, Alison Stirling,
Humphrey Lean, Simon Vosper
UK km-scale NWP
• The first steps towards scale-aware parametrization have been
implemented operationally
• Stratocumulus and diurnal convection represent two important and
still challenging regimes for km-scale NWP
• Stratocumulus: largely unresolved turbulence, albeit with resolved
mesoscale variability
• Convection: currently entirely resolved and reliant on aliasing smallscale convection to resolved scale
• Both are sensitive to the representation of subgrid turbulence:
• Blended (grey-zone) PBL and 3D Smagorinsky schemes (Feb 2015)
• Stochastic PBL perturbations (Mar 2015, Nov 2016)
• Free-tropospheric mixing (Nov 2016 but UK only, not suitable for tropics)
Low cloud
Surface SW fluxes over the UK
• Comparing with observations suggests UK models are too “cloudy”
• MOGREPS-UK even more so than UKV, likely due in part to the use of
the same RHcrit profile despite coarser resolution (2.2 km cf 1.5 km grid)
July 2014 mean diurnal
error in surface
downwelling SW flux
UKV
MOGREPS-UK control
MOGREPS-UK mean
Thanks to Cyril Morcrette
Diagnosing RHcrit
van Weverberg et al (2016)
• Subgrid variance and co-variances of T
and q can be diagnosed from turbulence
scheme → diagnosis of RHcrit
• Used with PC2 in GA7
• Developed new techniques for diagnosing
RHcrit in the same way but from lidar
observations
Mean PBL depth
• Initial tests in UKV (with Smith scheme)
are encouraging
• Adaptive to flow and vertical
resolution
• Missing larger scales of variability?
Diurnal and PBL depth averages of RHcrit for
clear skies during the MC3E campaign
(Oklahoma, spring 2011)
Representation of (layer) clouds
in UK models
• Continue evaluation of turbulence based RHcrit parametrization
• PC2 also an option but variable RHcrit likely to be crucial
• Will include evaluation of turbulence scheme itself
• Continue evaluation of the impact of enhanced vertical resolution
• SCM and case study modelling indicate impact on PBL entrainment
through improved representation of inversion structure and radiative
fluxes across cloud-top
• Further comparisons with lidar planned with various sites round
the world (Azores, Darwin, SGP, Germany)
• Involving external collaborations with PNNL, Univ of Hohenheim
Convection
Case study test of PS38 physics
14UTC 27th Aug 2015, T+11
PS38 initiates showers more readily
(significant improvement)
PS36
Visible satellite image
Radar
PS38
Radar
Case study test of PS38 physics
14UTC 27th Aug 2015, T+10
PS36
PS36+PS38 physics
Visible satellite image
PS38
PS36
Radar
• Focused cells of
often too intense rain
• Not enough area of
lighter rain
Active areas of research to
improve km-scale convection
1. Semi-Lagrangian advection
2. Subgrid turbulent mixing
3. Grey-zone convection parametrization
Active areas of research to
improve km-scale convection
1. Semi-Lagrangian advection
• Regional model Priestley-style conservation
• Already in SINGV, planned for PS39 UK models
UKV case study testing of
regional model conservation
Instantaneous rainfall distributions
(47 case studies)
• Adding conservation significantly
reduces the heavy tail of the
precipitation distribution
• Also reduces domain mean
precipitation by around 10%
• Still only an approximate
correction locally
• We would like to try and
address the underlying issue
with semi-Lagrangian
advection
© Crown copyright Met Office
Idealisation of well
resolved/continuous plume
Initial conditions: convergent flow+blob of moist air
z
x
© Crown copyright Met Office
Thanks to Chris Smith
Idealisation of well
resolved/continuous plume
Blob diluted by surrounding dry air and is
transported vertically
z
x
© Crown copyright Met Office
Poorly resolved semi-Lagrangian plume
Pointwise dept. point
exactly correct but does
not see convergence.
(U++ U- )/2 = 0
z
x
© Crown copyright Met Office
U-
U+
Poorly resolved semi-Lagrangian plume
Pointwise dept. point
exactly correct but does
not see convergence.
(U++ U- )/2 = 0
No dilution, low-level
blob acts as infinite
source
[“Global” conservation
ameliorates the problem
but not a solution]
z
x
© Crown copyright Met Office
U-
U+
“Cheap” cell-average advection
Use multiple (eg, 2), or
stochastically varying,
dept. points per cell
Should give improved
dilution and better local
conservation
z
x
© Crown copyright Met Office
Active areas of research to
improve km-scale convection
1. Semi-Lagrangian advection
• Regional model Priestley-style conservation
• Alternative departure point strategies
2. Subgrid turbulent mixing
• Add Leonard term (Moeng et al, 2010) for subgrid fluxes
arising from correlations between resolved and
unresolved scales:
k  2 w 
2 w  

F   x
 y
12 
x x
y y 
with k a semi-empirical constant
• Will be potentially significant around poorly resolved updraughts
(where resolved w and q, for example, are likely to have correlated
gradients)
Impact of Leonard term in UKV
Case study of widespread showers: 15th June 2016
Kirsty Hanley and Michael Whitall
• Inclusion of Leonard term suppresses heavy rainfall
• Also suppresses maximum vertical velocities
• Has little impact on lack of light rain
Average rainfall over southern UK
for rain rate < 4mm hr-1
Radar
PS38
PS38+Leonard (k=1)
PS38+Leonard (k=2)
Light rain
Average rainfall over southern UK
for rain rate > 4mm hr-1
Heavy rain
Radar
PS38
PS38+Leonard (k=1)
PS38+Leonard (k=2)
Active areas of research to
improve km-scale convection
1. Semi-Lagrangian advection
• Regional model Priestley-style conservation
• Alternative departure point strategies
2. Subgrid turbulent mixing
• Add Leonard term (Moeng et al 2010)
• Alternative subgrid models (eg., higher order closures or
dynamic)
3. Grey-zone convection parametrization
• Build on developments in global convection parametrization
• Add scale-dependency to closure (stochastic),
entrainment/detrainment, trigger function
• See Michael Whitall’s poster (#13)
Summary
• Significant progress has been made to improve the
representation of physical processes in km-scale UM
• A very active area of research
• Lots of innovative ideas actively being developed
• Both in model development and use of observations
• Involving significant collaborations
• RMED (UM Partnership) Convection Working Group
• ParaCon
Thank you
Questions?