new developments in soil moisture analysis at dwd filemoisture through stomata resistance....
TRANSCRIPT
New developments in Soil Moisture Analysis at DWD
M. Lange, R. Hess, W. WergenDeutscher Wetterdienst,
Offenbach, Germany
LM user meeting, Langen 2007
LM user meeting, Langen 2007
Outline
• Motivation for SMA in GME• Parameterisation of the variational SMA
scheme.• Experiences from operations
– Impact on screen level temperature– Accumulation of soil ice by the SMA
LM user meeting, Langen 2007
Almo exceeds LME Precipitation due to soil moisture initialisation
Running monthly mean precipitation and evaporation Daily soil moisture top and bottom layers
Top layer LME
Top layer ALMO
Bottom layer LME
Bottom layer ALMO
Positive feedback of dry/wet soil on precipitation forecast
Dry soil
Less cloudsand rain, strong
radiation input
EnhancedDry out
Desiccationof plants,
Weakevaporation
High vegetationcover, Strongevaporation
Refillingof soil
Wet soil
More cloudsand
precipitation,
LM user meeting, Langen 2007
Long term dry out in GME leads to positive Bias in T2m during summer
Development of a soil moisture analysisfor GME
• Soil moisture measurements are not available on a dense network, Radar observations even not on theglobal scale.
• Explore forecast errors in screen level parameter to correct initial soil moisture.
• A variational soil moisture assimilation scheme runsdaily for the LME model. This requires at least twoadditional model forecast runs. This iscomputationally too expensive for GME.
LM user meeting, Langen 2007
LM user meeting, Langen 2007
2d var (z,t) soil moisture analysis applied in LME
)()()()( 221
221 obs
mmTobs
mmbT
b TTOTTwwBwwJ −−+−−= −−
44 344 21errorfcmT
bmobs
mT
mTmTT
mTbana wTTOBOww2
221
211
21
2 ))(()( −Γ+ΓΓ+= −−−−
Cost function penalizes deviations from observations and initial soil moisture content
Analysed soil moisture depends on T2m forecast error and sensitivity ∂T2m/∂w
0=∇J
)00:0()00:15,00:12(2
wT m
∂∂
Parameterisation of ∂T2m(12:00,15:00)/∂w(0:00) with ∂Lhfl/∂w from penman eqn. at obs time
ShflLhfl Δ−≅Δ
a
msp r
TTcShfl )( 2Δ−Δ=Δ ρ
GShflLhflRn −−−=0Surface energy balance equation
Soil moisture variation:
ma
p Trc
Lhfl 21
Δ⎟⎠⎞
⎜⎝⎛ −
−≅Δααρ
LM user meeting, Langen 2007
ΔLhfl (W/m2)
ΔT2m
(10*
K)
)00:12(~)00:12( 2mTLhfl ΔΔ
LAI
slaf f
rrr
+=
wathumtemradssss FFFFrrrr ***)( 1max,
1min,
1max,
1 −−−− −+=
fa
apotsnowipl rr
rEfffLhfl+
−−= )1)(1(
LM user meeting, Langen 2007
tlppwpfcap
pwprootwat fww
wwF
)( −
−=
Evaporation from vegetation depends on soilmoisture through stomata resistance
Sensitivity of 2m temperature on soil moisture
root
rootk
pwproot
s
s
s
LAIfak zdz
wwr
rr
frrLhfl
wLhfl ,
max,
)1(1)( −
−+
=∂∂
LM user meeting, Langen 2007
root
rootk
tlppwpfcapk
wat
zdz
fwwwF ,
)(1
−=
∂∂ Root density constant with
soil depth!
root
rootk
pwproot
s
s
s
LAIfap
a
k
m
zdz
wwr
rr
frrLhfl
cr
wT ,
max,
2 )1(1)(1 −
−+
⎟⎠⎞
⎜⎝⎛−
=∂∂
αα
ρ
LM user meeting, Langen 2007
Validation against model forecast
• Variaton of initial soil moisture at selectedgridpoints in the whole range betweenplant wilting point and field capacity.
• Compare Sensitivity of T2m on soilmoisture with parameterisation.
LM user meeting, Langen 2007
Horizontal decoupling hypothesis is valid
LM user meeting, Langen 2007
All gridpointsLhfl(11:00-12:00) > 200 W/m2
Parameterisation can be applied during radiation conditions in the whole range between plant wilting point and field capacity
No further need for additional model runs !dT2m(12:00)/dwb(0:00) (model)
dT2m/dw
b(12:00) (theorie)
dT2m/dw
b(theorie)
dT2m/dwb (model)
Date: 20050525
LM user meeting, Langen 2007
Status and further plan
• Development phase almost finished.• First validation experiments with the LM
model are planned to start soon.• Implementation in GME should be ready
end of summer 2007.
LM user meeting, Langen 2007
Experiences from operations
• Impact of SMA in LME on bias in 2m temperature.
• SMA affects soil ice content.
LM user meeting, Langen 2007
Routine No SMA
Bias T2m on LM domain, avg 12:00, 15:00 Rmse T2m on LM domain, avg 12:00, 15:00
SMA reduces effectively Bias and Rmse of T2m in LME
LM user meeting, Langen 2007
LM user meeting, Langen 2007
Accumulation of soil ice by adding water in SMA
Water and ice conten layer 5 (27-81cm), LME, gridpoint (323,354)
H2O
(mm
)
Soil water
Soil ice
LM user meeting, Langen 2007
Modification of SMA
1.) Frozen soil at initial time of the next forecast: Tso(vv=24hr)<T0Apply only negative SMA increments:Wso(ana)=Min( Wso(mod)+ δwa, Wso(mod) )
2.) Soil unfreezed at begin of next forecast but soil frozenat time of analysis: Tso(vv=24hr)>T0, Tso(t=0)<T0
Positive soil moisture increments are possible if maximum liquid water content is notexceeded:
Wso(ana)=Min(Wso(mod)+δwa, Wlmax)
LM user meeting, Langen 2007
Modification of SMA
If soil is frozen at begin of the next forecast:Apply only negative SMA increments:Wso(ana)=Min( Wso(mod)+ δwa, Wso(mod) )
If soil is unfreezed at begin of next forecast but soiltemperature below freezing point at time of analysis:
Positive soil moisture increments are possible if maximum liquid water content is notexceeded and soil ice occurs.
Wso(ana)=Min(Wso(mod)+δwa, Wlmax)
LM user meeting, Langen 2007
Soil ice content layer 5 (27-81cm)
Routine No SMA
New SMA
Fast reduction of soil ice when T2m is sensitive on soil moisture
LM user meeting, Langen 2007
Conclusions• Soil moisture analysis is appropriate to prevent from permanent
drifts and to reduce forecast errors in screen level parameter. • A parameterisation of the SMA scheme is developed for GME to
save computational costs for additional model runs required in thevariational method.
• The parameterisation can be applied successfully under conditionswhen transpiration from vegetation dominates.
• Side effects should be considered when model changes areintroduced.
• Accumulation of soil ice through the SMA is fixed by suppressingpositive increments in frozen soil.
Acknowledgement• Thanks to Emanuele Zala from meteo
swiss who provided the Almo data for2006.
LM user meeting, Langen 2007
LM user meeting, Langen 2007
Almo exceeds LME Precipitation due to soil moisture initialisation
Running monthly mean precipitation and evaporation