LM(E) user meeting, Langen 2006 Soil moisture analysis at DWD M. Lange, R. Hess, W. Wergen DWD Experiences from operations and ELDAS
LM(E) user meeting, Langen 2006
Soil moisture analysis at DWD
M. Lange, R. Hess, W. Wergen
DWD
Experiences from operations and ELDAS
Outline• Motivation for soil moisture analysis• The method• Experiences with LM(E)• Results from Eldas• Conclusions
LM(E) user meeting, Langen 2006
LM(E) user meeting, Langen 2006
(E-TESSEL) – TESSEL: JJA
Increased
Latent heat flux(Wm-2)
increased
Low cloudcover
Small impact onHigh cloud coverMedium cloudcover
Due to:
Lower canopy resistance
LM(E) user meeting, Langen 2006
x
2d var (z,t) soil moisture analysis
LM(E) user meeting, Langen 2006
)()()()( 221
221 obs
mmTobs
mmbT
b TTOTTBJ −−+−−=−− ηηηη
))(()(
0
221
211
21
2 bmobs
mT
mTmTT
mTbana TTOBO
J
ηηη −Γ+ΓΓ+=
=∇−−−−
QMAMB
BOJATt
mTT
mT
+=
+ΓΓ=∇=+
−−−
1
112
12
2 )(
Cost function penalizes deviations from observations and initial soil moisture content
Calculate minimum of cost function analytically
Analysis error and update of Background error covariance matrix:
)00:0(/)00:15,00:12(),()( 22222 dwbdTTT mmTbmTbmm =Γ−Γ+= ηηη
Assumption: Linearity between variation of T2m and wb
• LME: SMA against No SMASix week Test Period June/July 2004
• LME vs. LMParallelsuite April-November 2005
• Climatology of soil moisture increments
LM user meeting, Langen 2006
Experiences from operations
SMA (T2m) reduces effectively Bias and Rmse of T2m in LME
LM user meeting, Langen 2006
No SMA vs. SMA-T2m, Bias, Rmse T2m(12:00, 15:00)
SMA-T2m: LME beats LM
LM user meeting, Langen 2006
LM user meeting, Langen 2006
Hydrological annual budget of SMA far from being closed
LM user meeting, Langen 2006
Large positive bias for T2m in 2004, 2005
in spite of soil moisture analysis
2001 2006
Eldas
• Combine european expertise in soil moisture assimilation to buildand implement a soil moisture assimilation system at a number ofeuropean weather centers.
• Build a demonstration database covering at least one seasonalcycle.
• Validate the soil moisture fields against independent observations.
• Assess the added value of soil moisture assimilation to predict thehydrological cycle on the continental scale with focus on predictionof droughts and the risk of floodings.
Development of a European Land Data Assimilation System to predict droughts and floods
LM user meeting, Langen 2006
Eldas model experiments covering the periodApril 15 – December 31 2000
3 hourlyPrecipitation
data
Externalforcing
EcoclimapSMA(T2m, Rh2m)depending on
experim.
TERRA(2 layers)
LM rotated(0.2° x 0.2°)
EuropeGME / LM
Physiograficdata (LAI, plcov,
zgeo…)
Soil moistureassimilation
system
Land surfacescheme
GridEldasDomain
Model base
LM user meeting, Langen 2006
Experimental setup
Model experiments
++Precipìtationforcing
++Rh2m++++T2m
No SMA(Soil moisture fromprevious forecast, Initialdata from SMA-T2m)
SMA-T2m+Rh2m+Precip.
SMA-T2m+Precip.
SMA-T2m+Rh2m
SMA–T2mUsed
Observations
LM user meeting, Langen 2006
LM user meeting, Langen 2006
Significant impact on T2m with analysed precipitation, small effect with Rh2m in SMA
LM user meeting, Langen 2006
Positive Impact on Rh2m even more pronounced.Again higher impact from analysed precipitation data.
LM user meeting, Langen 2006
Improvement of 24 hr precipitation forecast with precipitation forcing
LM user meeting, Langen 2006
Eldas domain average daily background errorvariance top layer (mm^2)
Eldas domain average daily background errorvariance bottom layer (mm^2)
Confidence in soil moisture Analysis best during vegetation period for thebottom layer in top layer highest before and behind the vegetation period.
Bg
erro
r var
ianc
e to
p la
yer
(mm
2 )
Bg
erro
r var
ianc
e bo
ttom
laye
r (m
m2 )
Gridpoint validation of soil moisture at LindenbergUsing analysed precipitation in the LM forecast
LM user meeting, Langen 2006
LM user meeting, Langen 2006
Soil moisture is climate monitor
Summer 2003
Conclusions• SMA improves the forecast of screen level parameter and
precipitation. Using analysed precipitation has strongest positiveimpact. A high qualityprecipitation product at analysis time thereforewould be very much appreciated.
• SMA compensates for deficits in the flux parameterisation. Thisprevents from recovering errors in the land surface scheme duringoperations and leads to drifts in the soil moisture content.
• The interdisciplinarity between modellers, validators, hydrologistsand agronoms should be enforced to come to more realistic soilmoisture values applicable for flood and drought prediction.
• Future satellite informations on soil moisture content and heatingrates available from microwave and thermal infrared radiation datashould be used in the analysis. Resources are necessary to realisethis!
LM user meeting, Langen 2006
Thanks to …
• Dr. Gerd Vogel who provided validationdata from Lindenberg.
• Thomas Hanisch who provided assistancewith model experiments.
• Ulrich Damrath for verification results.
LM user meeting, Langen 2006