T - LMD

Atmosphere-surface coupling: heat and momentum flux
Anton Beljaars
1.
Thermal coupling in stable situations
2.
Sensitivity to momentum flux
Thanks to: Gianpaolo Balsamo Emanuel Dutra and Irina Sandu
DEPHY workshop Toulouse: 19-20 Jan 2015
Slide 1
Mean absolute error of minimum 2T in ECMWF short
range forecasts for January 2011
Zonal mean average
DEPHY workshop Toulouse: 19-20 Jan 2015
Slide 2
How is the winter and night time cooling at the surface controlled?
Atmospheric temperature
Tair
SH
Tsk
Radiation intercepting/emitting level:
e.g. vegetation canopy, litter layer on top of bare
soil, snow layer, or combination of these in a
heterogeneous configuration
G
Deep soil
Qnet
Tsoil
1. Which fraction of radiative cooling
is taken from the atmosphere
through sensible heat flux and
which fraction from the land
surface?
2. Over what depth is the cooling
distributed in the atmosphere
(boundary layer depth)
DEPHY workshop Toulouse: 19-20 Jan 2015
Slide 3
The strength of the coupling is hidden in a number of parametrizations
Radiation is affected by:
• Clouds
• Aerosols
• Water vapor
Coupling between lowest model level and surface (skin layer) is affected by:
• Wind speed
• Roughness lengths
H = ρ c p CH | U | (θl − θ sk )
• Stability function
k2
• Heterogeneity
CH =
ln( z / zom ) ln( z / zoh )
Boundary layer diffusion above the lowest model level is affected by:
• Wind shear
• Stability
• Meso-scale variability
• Asymptotic mixing length
w 'θ ' = − K H
dθ
,
dz
l −1 = (κ z ) −1 + λ −1
DEPHY workshop Toulouse: 19-20 Jan 2015
Slide 4
FH ( Rib )
 dU

KH = l2 
+ S m  f H ( Ri )
 dz

Coupling coefficients are hidden in a number of parametrizations
Lowest model level (10 m)
Coupling between skin level and deep soil is affected by all
the details of the land surface scheme:
Tskin
• Soil thermal properties
• Presence of snow and snow properties
• Representation of land cover (skin or canopy to ground
conductivity in ECMWF model)
• Soil water freezing and thawing
• Heterogeneity
DEPHY workshop Toulouse: 19-20 Jan 2015
Slide 5
Increased diffusion of heat in stable situations
Stability (Richardson number) dependence of heat and
momentum diffusion coefficients
Heat
T-profiles after cooling a neutral boundary layer
profiles for 9 hours with 25/50 W/m2
25 W/m2
Revised
Old
Old
Revised
Momentum
50 W/m2
Revised
Old
Old
Revised
DEPHY workshop Toulouse: 19-20 Jan 2015
Slide 6
Soil water freezing
Soil heat transfer equation during freezing
∂θ I
∂T ∂
∂T
( ρC) s
= λT
+ L f ρw
∂t ∂z ∂z
∂t
θ I Soil frozen water
θ I = θ I (T ) = f (T )θ

∂f  ∂T ∂
∂T
(
ρ
C)
−
L
ρ
θ
=
λ


s
f w
T
∂T
∂t
∂z
∂z


Apparent heat capacity
DEPHY workshop Toulouse: 19-20 Jan 2015
Slide 7
Difference in 2m temperature for January 1996
From long “relaxation” integrations starting 1 Oct 1995
1994 model version
Revised BL - Control
DEPHY workshop Toulouse: 19-20 Jan 2015
Revised BL & soil freezing - Control
Slide 8
Difference in 2m temperature for January 1996
From long “relaxation” integrations starting 1 Oct 1995
Effect of revised LTG in 1994 model version
DEPHY workshop Toulouse: 19-20 Jan 2015
Effect of revised LTG in 2011 model version
Slide 9
Difference in 2m temperature for January 1996
From long “relaxation” integrations starting 1 Oct 1995
old snow scheme – new snow scheme
The new snow scheme
(Dutra et al. 2010)
has lower conductivity
and therefore the
winter temperature
drops more over snow.
Insulating snow also
increases the model
sensitivity to boundary
layer diffusion.
DEPHY workshop Toulouse: 19-20 Jan 2015
Slide 11
Energy budget over 6 hours before the “minimum temperature” (Feb 2009)
1 + H/Qn = G0 /Qn
-H/Qn
G0/Qn
DEPHY workshop Toulouse: 19-20 Jan 2015
Slide 12
Energy budget over 6 hours before “minimum temperature” (Feb 2009, land only)
Qn + LE + H = G0
U < 3 m/s
U > 3 m/s
H
LE
G0
Lne
t
DEPHY workshop Toulouse: 19-20 Jan 2015
Slide 13
Heat fluxes over snow (Eric Brun)
DEPHY workshop Toulouse: 19-20 Jan 2015
Slide 14
The surface energy budget over snow in the polar night over Antarctica is
very simple
Tair
SH
For Antartica, most of the
gradient is in the surface layer
Qnet
Tsk
G=0
= -LWD+σT4
= CH |U| (Tair-Tsk)
LWd, Tair and U are ok; Tsk is biased
Conclusion: CH is biased
ECMWF uses zom=1.3 10-3 m and
zoh=zom/10
Etienne Vignon and Christophe Gethon find
from observations at Dome C:
zom= 10-4 m and zoh=zom/2 to zoh=zom/5
DEPHY workshop Toulouse: 19-20 Jan 2015
Preliminary experimentation by
Emanuel Dutra and Irina Sandu
has shown:
1. Reduction of the roughness
helps
2. G is not zero in ECMWF model,
snow properties need updating
+ multi-layer snow
Slide 15
Summary
•
•
•
Strong sensitivities have been demonstrated: temperature can be controlled by
changing coupling to the atmosphere and by changing coupling to the deep soil
Reasonable results for temperature are obtained by optimization, although errors are
still substantial with large-scale geographical patterns
Given the large uncertainty in a many coupling parameters, it is likely that
compensating errors exist
Way forward:
• Verify fluxes with flux towers (e.g. over Antarctica, Eric Brun)
• Consider atmosphere and land as a coupled problem and analyze relations
between variables to demonstrate realism of the full system (papers by Alan Betts)
• Use tracers as an additional constraint on the problem of atmospheric diffusion
DEPHY workshop Toulouse: 19-20 Jan 2015
Slide 16
Model and observations at Cabauw (3-hourly)
Data kindly provided by
ECN/KNMI
DEPHY workshop Toulouse: 19-20 Jan 2015
Slide 17
CO2: model and
observations at Cabauw
(daily amplitudes of
diurnal cycle)
Wind speed: model and
observations at Cabauw
(daily averages)
DEPHY workshop Toulouse: 19-20 Jan 2015
Slide 19
Momentum fluxes
DEPHY workshop Toulouse: 19-20 Jan 2015
Slide 20
Turbulence scheme
Micro-meteorological
range (turbulence)
Macro-meteorological
range (weather)
4-days
1-hour
5-min
1-min
5-sec
Horizontal wind speed spectrum at Brookhaven at z=100m (van der Hoven, 1957)
Should be non-controversial:
• Clear scale separation
• Solid scaling relation between
fluxes and profiles
Businger et al. (1971)
DEPHY workshop Toulouse: 19-20 Jan 2015
Slide 21
But in practise
•
The atmosphere has a lot of meso-scale variability,
which is missing in the ECMWF model.
•
Houchi et al. (2010) analyse a large volume of
radio sonde data and conclude that the ECMWF
model underestimates shear by a factor 2.5
This might be the background of “long tail
formulations”, which makes the uncertainty large
•
And how to handle the
heterogeneous surface boundary
condition ?
Houchi et al. (1971)
Concepts are well established:
• Use “effective roughness” ( Mason, 1988; Grant and
Mason, 1990; Wood and Mason, 1991)
• Turbulent Orographic Form Drag TOFD (Wood et al.,
2001; Beljaars et al. 2004)
• The uncertainty in the coefficients that characterise the
landscape and the sub-grid orography is large
Llanthony valley, S. Wales
DEPHY workshop Toulouse: 19-20 Jan 2015
Slide 22
Terrain heterogeneity is everywhere
KNMI surface station during MESO-GERS 84
10 km North-West of Vic Fezensac (Weill et al., 1988)
Aerial picture taken from CNRM research aircraft Hurel-Dubois
Picture: Gerard van der Vliet
DEPHY workshop Toulouse: 19-20 Jan 2015
Slide 23
Changes implemented in 40R1 building on Sandu et al., 2013
Turbulence closure for stable conditions:
Up to 38R2
-
long tails near surface, short tails
above PBL
-
λ =150m
-
non-resolved shear term, with a
maximum at 850hPa
K M ,H
∂U 2
=
l f M , H ( Ri ),
∂Z
1 1
1
=
+
l kz λ
From 40R1
-
long tails everywhere
-
λ = 10% PBL height in stable
boundary layers
-
λ = 30 m in free shear layers
+
Increase in drag over orography
Consequence: net reduction in diffusion in stable boundary layers, not much
change in free-shear layers, except at 850 hPa
DEPHY workshop Toulouse: 19-20 Jan 2015
Slide 24
Impact of diffusion / sub-grid orography changes
•
•
•
•
Reduction of wind direction bias over Europe by 3° in winter, 1 ° in summer (out of 10 °)
Improvement in low level jets
Improvement of the large-scale performance of the model in winter N Hemisphere
Deterioration of tropical wind scores (against own analysis, not against observations!)
DEPHY workshop Toulouse: 19-20 Jan 2015
Slide 25
Predictive skill since cycle 40r1
DIFF - CTL
DIFF+GWD - CTL
January 2012, Z1000hPa 38R2
38R2 + DIFF+GWD
-
Change to orographic drag affects planetary
waves, anticyclones (e.g. too weak over EA)
The improvement in averaged NH scores
mostly located over E. Asia, NA
I. Sandu; ECMWF Newsletter, 138
DEPHY workshop Toulouse: 19-20 Jan 2015
Slide 26
Boundary layer
Sub-grid orography
WGNE/DRAG project
(Ayrton Zadra)
SO+BL
Comparison of surface
torque over land from
different models for January
2012
DEPHY workshop Toulouse: 19-20 Jan 2015
Slide 27
East-west surface stress averaged over 6-hr interval of daily forecasts in January 2012
EC
6-12 UTC
MO
18-24 UTC
EC
MO
DEPHY workshop Toulouse: 19-20 Jan 2015
Slide 28
Effect of orography resolution on 2T (T1279-T639)
Cooling over 9 hrs with 50 W/m2
Hypothesis; with high resolution orography:
• Resolved gravity waves are stronger
• Local shear is increased
• Turbulent mixing is stronger
• Stable layers are deeper
• Near surface temperature is higher
DEPHY workshop Toulouse: 19-20 Jan 2015
Strong mixing
Weak mixing
Slide 29
Concluding remarks
-
-
Near surface temperature errors increase with latitude and are sensitive to
atmospheric and subsurface thermal coupling
A realistic forecast of the diurnal cycle may still have the wrong balance between
atmospheric and subsurface coupling (does it have impact on Arctic amplification
in climate models?)
Only a detailed analysis of observations can resolve this issue
Meso-scale variability is likely to influence turbulent diffusion (long versus short
tails), but this is hardly documented and requires further research
Forecast errors show strong sensitive to the representation of land surface drag
Surface drag can not be verified and different models show large differences
particularly in the distribution over turbulent and sub-grid orography schemes.
This has consequences for the geographical distribution of drag and for the
dependence of drag on stability
High resolution models for orography or canopy flow (for heterogeneous terrain
characterization) can help
DEPHY workshop Toulouse: 19-20 Jan 2015
Slide 30
Thanks for your attention
DEPHY workshop Toulouse: 19-20 Jan 2015
Slide 31
Regression on daily summer
data from the ECMWF model
[non-tropical basins: 10700
days]
Betts (2006): JGR, 111, D07105
Diurnal temp range
DTRsc = DTR/ΔTR
Strength of NBL
ΔTNsc= ΔTN/ΔTR
Scaled heat flux
Hsc= HN/(-LWnetN)
DEPHY workshop Toulouse: 19-20 Jan 2015
Slide 32
Dependence of scaled energy budget on wind
speed
For NBL:
Hsc+ Gsc≈ 1
Partitioning changes with wind
speed, but basins show different
slope
DEPHY workshop Toulouse: 19-20 Jan 2015
Slide 33
Towards higher resolution
-
Convection becomes partially resolved
–
–
–
–
-
Orography will be increasingly resolved
–
–
–
-
Non-equilibrium closure (some progress but do we understand it ?)
Communication with neighbouring points (use of momentum equation with source
term from mass residual of mass flux entrainment/detrainment is non-trivial)
3D turbulence (LES to support CRM sub-grid models)
Shallow convection still needed (stratocumulus to cumulus transition still not well
parametrized)
Flow separation over steep orography (numerically non-trivial)
More resolved gravity waves (are they handled correctly by the numerics?)
Is hand-over from parametrized to resolved momentum fluxes correctly handled ?
(momentum budget studies that cover a range of resolutions)
Meso-scale variability, is it important?
–
–
Are long tail stability functions necessary because of meso-scale variability (Can high
resolution simulations support a parametrization of meso-scale variability?)
How to characterize land surfaces for drag (can LES studies of canopy flow “measure”
drag over heterogeneous terrain ? )
DEPHY workshop Toulouse: 19-20 Jan 2015
Slide 34
Conclusions
•
To develop models that simulate realistic projections it is required to
represent the current climate with the correct mechanisms
•
To verify it is necessary to compare with observations at the process level
•
Clouds and their phase are crucial for the radiation
•
The ratio of sub-surface / atmospheric energy fluxes requires careful
evaluation
•
NWP environment has advantages for model development because the
comparison with observations is simpler than in climate mode
•
Priority areas for research and further model development are:
●
●
●
●
Mixed phase clouds (models do not necessarily have a physically realistic
representation)
Cloud / radiation interaction
Boundary layer / Land surface + snow interaction (including heterogeneous
terrain)
The ECMWF snow model needs more layers to represent different time
scales
DEPHY workshop Toulouse: 19-20 Jan 2015
Slide 35