Embed Size (px)
Citation preview
Assimilation of (satellite) cloud-Assimilation of (satellite) cloud-
information at the convective scale in information at the convective scale in
an Ensemble Kalman Filteran Ensemble Kalman Filter
Annika Schomburg1, Christoph Schraff1, Africa Perianez1, Jason Otkin2, Roland Potthast1
1 DWD (German Meteorological Service)2 University of Wisconsin
WWOSC 16-21 August 2014, Montreal
Motivation: Growing renewable energy sector
2
• Weather dependencey of renewable energy: Increasing demands for accurate power predictions for a safe and cost-effective power system Project to improve forecast models
for the grid integration of weather dependent energy sources
Source: energymap.info
solarWindhydroelectricbiomassgasgeothermal
Energy production in Germany for week 30, 2014
Source: Fraunhofer ISE
Development of average renewable energy production in Germany since 2002:
For photovoltaic power a realistic simulation of cloud cover is crucial this talk: exploit cloud information sources for data assimilation
Motivation
• Difficult weather situations for photovoltaic power predictions:
• Cloud cover after cold front passes
• Convective situations
• Low stratus clouds / fog
• Snow cover on solar panels
Convective scale models needed to capture relief and atmospheric stability locally
3
Modelling system
4
COSMO-DE :
•Limited-area short-range numerical model weather prediction model•x 2.8 km / 50 vertical layers •Explicit deep convection
•New data assimilation system : Implementation of the Ensemble Kalman Filter: LETKF after Hunt et al. (2007)
Local Ensemble Transform Kalman Filter
LETKF
Analysis perturbations: linear combination of background
perturbations
First guess ensemble members are weighted according to their departure from the observations
OBS-FG
OBS-FG
OBS-FG
OBS-FG
R
Tbibk
i
bibb
k))((
)1(
1 )(
1
)( xxxxP
Background error covariance
Observation errors
Outline
• Assimilation of additional data to improve cloud cover:
• Satellite cloud products
• Cloudy infrared satellite radiances
• Photovoltaic power
6
Satellite cloud products
Satellite product: cloud top height
contains information on horizontal and vertical distributions of clouds
1 2 3 4 5 6 7 8 9 10 11 12 13
Satellite cloud product information
Geostationary satellite data: Meteosat-SEVIRI (Δx ~ 5km over central Europe, Δt=15 min)
Source: EUMETSAT
Height [km]
Cloud top height
MODEL EQUIVALENT
9
OBSERVATION: Satellite product: cloud top height
1 2 3 4 5 6 7 8 9 10 11 12 13
Method
• Extract information if a pixel is observed as cloudy:
Height [km]
Cloud top height
Cloud top height
Relative humidity at cloud top
height
Determine cloud top model equivalent: top of most humid layer k close to observation
100%
see Schomburg et al., QJRMS, 2014
Layer kRH(k)
height(k)
Observation
Model RH profile
Assimilated variables:
relative humiditycloud covercloud water
cloud iceobserved cloud top
3 lines in one colour indicate ensemble mean and mean +/- spread
• 1 analysis step, 17 Nov. 2011, 6 UTC (wintertime low stratus)
vertical profiles
Example: Single-observation experiments: missed low stratus cloud
10
First guess Analysis
Model equivalent:
11
Observation:Satellite product: cloud top height
1 2 3 4 5 6 7 8 9 10 11 12 13
Method
• Extract information if a pixel is observed as cloud-free:
Height [km]
Cloud top height
Cloud coverhigh clouds
Cloud covermid-level
clouds
Cloud coverlow clouds
0%
0%
0%
Z [km]
CLC
Maximum cloud cover in
high levels
Maximum cloud cover in
medium levels
Maximum cloud cover in
low levels
see Schomburg et al., QJRMS, 2014
Assimilated variables:
low clouds mid-level clouds high clouds‘false alarm’ cloud cover
(after 20 hrs cycling)
conventional+ cloud
conventionalobs only
12
Comparison “only conventional“ versus “conventional + cloud obs"
[octa]
conventional only conventional + cloud
Total cloud cover after 12 h free forecast
Observed cloud top height
Comparison of free forecast results
satellite obs
15. Nov 2011, 6:00 UTC
Cloudy infrared satellite radiances
Approach: Assimilate water vapour and window channels of Meteosat SEVIRI
• The goal is to obtain information on the cloud cover in the atmosphere
assimilate all-sky radiances from water-vapour channels and window channel(s)
•Challenges: Cloud dependent bias correction?
How to specify observation error
Thinning, localization in LETKF
.. etc..
15
Source: Schmetz et al, BAMS, 2002
SEVIRI channels
First step: Monitoring
16
Example for water vapour band WV7.3, sensitive to mid-level moisture (and clouds)
• Obs minus model statistics look promising with the exception of high-level clouds, semitransparent clouds should be excluded here
• Bias correction is needed: the model is 2-4 K too warm
Observation minus Background
RMSE
First assimilation results: 12 hours of cycling, assimilation of channel WV7.3
17
Bia
sR
MS
E
Without radiance assimilation With radiance assimilation
Photovoltaic power
Assimilation of photovoltaic power
Model variables:- surface irradiance- 2m temperature
- albedo
Model variables:- surface irradiance- 2m temperature
- albedo
Source: Yves-Marie Saint-Drenan, IWES
Forward operator:
Compute radiation on tilted
plane
Compute radiation on tilted
plane
Forward operator for PV module
Challenge: Data availability - Up to now only data for a few solar parks available for DWD
Synthetic PV power (clouds
main forcing factor)
Synthetic PV power (clouds
main forcing factor)
Example of simulated and observed photovoltaic power
Model forecast solar insolation at surfaceObserved photovoltaic powerSimulated photovoltaic power (based on model forecast radiation)
No
rmal
ized
po
wer
[W
/Wp
eak]
20
Conclusions
• Work on assimilating cloud information from various sources at the convective scale (∆x~3km) in a LETKF system ongoing:
• Satellite products, satellite radiances, PV power
• Challenges:
• Nonlinearity
• Presentation of clouds in the model
• Forward modelling
• PV data and meta-data availability and quality• Shading by trees, string failures, soiling…
• Improved cloud cover simulations is also expected to lead to a better onset of convection and temperature predictions
21
Thank you for your attentionThank you for your attention!
Determine the model equivalent cloud top
Avoid strong penalizing of members which are dry at CTHobs but have a cloud or even only high humidity close to CTHobs
search in a vertical range hmax around CTHobs fora ‘best fitting’ model level k, i.e. with minimum ‘distance’ d:
2
max
2 )(1
)(min obskobskk
CTHhh
RHRHd
relative humidity height ofmodel level
k
= 1
use y=CTHobs H(x)=hk
and y=RHobs=1 H(x)=RHk (relative humidity over
water/ice depending on temperature)as 2 separate variables assimilated by LETKF
use y=CTHobs H(x)=hk
and y=RHobs=1 H(x)=RHk (relative humidity over
water/ice depending on temperature)as 2 separate variables assimilated by LETKF
23
Z [km]
RH [%]
CTHobs
k1
k2
k3
k4
k5
Cloud top
model profile
•(make sure to choose the top of the detected cloud)
Example: 17 Nov 2011, 6:00 UTCObservations and model equivalents
RH model level kObservation Model
„Cloud top height“
• COSMO cloud cover where observations “cloudfree”
Example: 17 Nov 2011, 6:00 UTC
High clouds (oktas)Mid-level clouds (oktas)Low clouds (oktas)
Example: Missed cloud case:Effect on temperature profile
temperature profile [K] (mean +/- spread)
first guess analysis
observed cloud top
26
conventional only conventional + cloud
Total cloud cover of first guess fields after 20 hours of cycling
Satellite cloud top height
Results: Comparison of cycled experiments
satellite obs
12 Nov 2011 17:00 UTC
Low stratus cloud cover improved through assimilation of cloud products!Low stratus cloud cover improved through assimilation of cloud products!
Results III:Forecast impact
• 24 h free forecast starting after 21h cycling• 14 Nov. 2011, 18 UTC – 15 Nov. 18 UTC• Wintertime low stratus
conventional only conventional + cloud
Observed cloud top height
Results after 12 hours of free forecast
satellite obs
Total cloud cover after 12h forecast (15. Nov 2011, 6:00 UTC)
• The forecast of cloud characteristics can be improved through the assimilation of the cloud information
30
Results: Comparison of free forecast: time series of
errors
Conventional + cloud dataOnly conventional data
RMSE
Bias (Obs-Model)
Low cloudsMid-level cloudsHigh clouds
Mean squared error averaged over all cloud-free observations
RH (relative humidity) at observed cloud top averaged over all cloudy
observations
Assimilation of photovoltaic power
Model variables:- surface irradiance- 2m temperature
- albedo
Model variables:- surface irradiance- 2m temperature
- albedo
Source: Yves-Marie Saint-Drenan, IWES
Forward operator:
Compute radiation on tilted
plane
Compute radiation on tilted
plane
Forward operator for PV module
Sensitivities:
Direct solar irradiance [W/m²]
Diffuse part of solar irradiance
[W/m²]
Ambient air temperature [K]
