3-Dimensional surrogate cloud fields with measured structures Victor Venema Clemens Simmer...

3-Dimensional surrogate cloud fields 3-Dimensional surrogate cloud fields with measured structureswith measured structures

Victor Venema Victor Venema Clemens SimmerClemens Simmer

University of Bonn, GermanyUniversity of Bonn, Germany

Cloud measurements: Cloud measurements: Susanne Crewell, Ulrich LöhnertSusanne Crewell, Ulrich Löhnert

Radiative transfer: Radiative transfer: Sebastián Gimeno García, Anke Kniffka, Steffen MeyerSebastián Gimeno García, Anke Kniffka, Steffen Meyer

LES Modelling:LES Modelling:Andreas Chlond, Frederick Chosson, Siegfried Raasch, Andreas Chlond, Frederick Chosson, Siegfried Raasch,

Michael SchroeterMichael Schroeter

BBC campaignsBBC campaigns Keywords: boundary layer water clouds, Keywords: boundary layer water clouds,

structure & radiation, climate structure & radiation, climate

AirplanesAirplanes Tethered balloonsTethered balloons SatellitesSatellites Regional network with lidar, Regional network with lidar,

ir-radiometersir-radiometers Ground based remote sensingGround based remote sensing

BBC campaignsBBC campaigns BBC1: August, September 2001BBC1: August, September 2001 BBC2: May 2003BBC2: May 2003 Chaotic skies, typically Dutch cloudscapesChaotic skies, typically Dutch cloudscapes

BBC campaignsBBC campaigns BBC1: August, September 2001BBC1: August, September 2001 BBC2: May 2003BBC2: May 2003 Chaotic skies, typically Dutch cloudscapesChaotic skies, typically Dutch cloudscapes

– High quality measurements for testingHigh quality measurements for testing

Open databaseOpen database– ftp://bbc.knmi.nlftp://bbc.knmi.nl

Beautiful measurements, Beautiful measurements, but modellers need 3D fieldsbut modellers need 3D fields

Dimensionally challengedDimensionally challenged

Time [hr] UTH

eigh

t [k

m]

LWC template [kg/m3]

10.5 11

1.4

1.6

1.8

2

2.2

0

0.1

0.2

0.3

0.4

0.5

LWC Surrogate

0 2 4 60

2

4

6

0

2

4

6

0 2 4 61.5

2

Reff

Surrogate

0 2 4 60

2

4

6

0

2

4

6

0 2 4 61.5

2

Time [hr] UT

Hei

ght

[km

]

LWC template [kg/m3]

13.2 13.41

1.5

2

0

0.5

1

1.5

LWC Surrogate

0 2 4 6 80

2

4

6

8

0

2

4

6

8

0 2 4 6 81

1.52

Reff

Surrogate

0 2 4 6 80

2

4

6

8

0

2

4

6

8

0 2 4 6 81

1.52

Time [hr] UT

Hei

ght

[km

]

LWC template [kg/m3]

7 7.5

0.5

0.6

0.7

0.8

0.9

0

0.5

1

1.5

LWC Surrogate

0 5 100

5

10

0

5

10

0 5 10

Reff

Surrogate

0 5 100

5

10

0

5

10

0 5 10

Time [hr] UT

Hei

ght

[km

]

LWC template [kg/m3]

9 9.5

0.4

0.6

0.8

0

0.2

0.4

0.6

LWC Surrogate

0 5 100

5

10

0

5

10

0 5 100.40.60.8

Reff

Surrogate

0 5 100

5

10

0

5

10

0 5 100.40.60.8

MotivationMotivation Can not measure a full 3D cloud fieldCan not measure a full 3D cloud field Can measure many (statistical) cloud Can measure many (statistical) cloud

propertiesproperties

Generate cloud field based on Generate cloud field based on measurementsmeasurements

Emphasis on good structure for Emphasis on good structure for radiative transferradiative transfer

Structure & radiationStructure & radiation DistributionDistribution

– Amplitude (LWP, LWC, Amplitude (LWP, LWC, ) alone is already ) alone is already goodgood

– Success of Independent Pixel Success of Independent Pixel Approximation (IPA) at large scalesApproximation (IPA) at large scales

0 500 1000 1500 2000 2500 3000 35000

100

200

300

400

500

600

Number

LWP

(gr

/m2 )

MeasuredSurrogate

Structure & radiationStructure & radiation Power spectrumPower spectrum

– Spatial linear correlationsSpatial linear correlations

Full spectrumFull spectrum

Measured power spectrumMeasured power spectrum Scale breaksScale breaks WavesWaves Land sea mask Land sea mask ……

Satellite pictures: Eumetsat

Validation surrogate cloudsValidation surrogate clouds Is this statistical description good enough?Is this statistical description good enough?

3D LWC fields from LES modelling3D LWC fields from LES modelling Make surrogates from their statisticsMake surrogates from their statistics Calculate radiative propertiesCalculate radiative properties

– Radiances Radiances (remote sensing; Steffen Meyer)(remote sensing; Steffen Meyer)

– Irradiances Irradiances (radiative budget; (radiative budget; Sebastián Gimeno GarcíaSebastián Gimeno García))

– Actinic fluxesActinic fluxes (chemistry; Anke Kniffka)(chemistry; Anke Kniffka)

Compare themCompare them

LES Surrogates - LWCLES Surrogates - LWCLES originalsLES originals SurrogatesSurrogates

LES

: A

ndre

as C

hlon

dLE

S:

And

reas

Chl

ond

Validation resultsValidation results IrradiancesIrradiances

– RMSE mean: < 0.03 % of the radiative RMSE mean: < 0.03 % of the radiative budgetbudget

– Monte Carlo error, statistically not Monte Carlo error, statistically not significantsignificant

RadiancesRadiances– RMSE mean: < 0.3 % RMSE mean: < 0.3 % – Monte Carlo error, statistically not Monte Carlo error, statistically not

significantsignificant Actinic fluxActinic flux

– RMSE mean: < 1.4 %RMSE mean: < 1.4 %– No error estimate (SHDOM)No error estimate (SHDOM)

Validation resultsValidation results Fields with only one statistic Fields with only one statistic

(i.e. PDF (i.e. PDF oror Fourier) Fourier) – clearly worse clearly worse – statistically significantstatistically significant

The statistical description is probably The statistical description is probably good enoughgood enough– LES stratocumulus and within 1 %LES stratocumulus and within 1 %– Limitation is likely the amount of dataLimitation is likely the amount of data

Validation cumulusValidation cumulus Developed a more Developed a more

accurate Stochastic accurate Stochastic IAAFT algorithmIAAFT algorithm

Surrogates are Surrogates are copies of templatescopies of templates

In practise the bias In practise the bias is likely still there as is likely still there as you cannot measure you cannot measure the power spectrum the power spectrum that accuratelythat accurately

Validation cumulusValidation cumulus

LES originalLES original SurrogateSurrogate

Validation broken cloudsValidation broken cloudsLES originalsLES originals SurrogatesSurrogates

LES

: F

rede

rick

Cho

sson

LES

: F

rede

rick

Cho

sson

Time series

Iterative algorithm Iterative algorithm (Schreiber and Schmitz, 1996)(Schreiber and Schmitz, 1996)

DistributionFlow diagram Time series

Iterative algorithmIterative algorithm Original problemOriginal problem

– Generate a cloud similar to measurementsGenerate a cloud similar to measurements

New problemNew problem– Based on limited data estimate:Based on limited data estimate:

distributiondistribution power spectrumpower spectrum

fx (Hz)

f y (H

z)

Log 10

pow

er

(a)

0 0.1 0.2 0.3 0.4 0.50

0.1

0.2

0.3

0.4

0.5

5

10

15

20

Estimate spectrum – method 1Estimate spectrum – method 1 Add a dimensionAdd a dimension Assume isotropyAssume isotropy Rotate and scale Rotate and scale

power spectrumpower spectrumTime (s)

Tim

e (s

)

2000 4000 6000 8000

1000

2000

3000

4000

5000

6000

7000

8000 0

100

200

300

400

500

1000 2000 3000 4000 5000 6000 7000 80000

100

200

300

400

500

Time (s)

LWP

(gr

/m2 )

Pow

er

Frequency (Hz)

1D power spectrumLWP Measurement 2D power spectrum

Time (s)

LWP

Fre

quen

cy (

Hz)

Frequency (Hz)10

-310

-210

-1

10-5

100

105

k (Hz)

Pow

er

MeasuredSurrogate

Pow

er

Estimate spectrum – method 2Estimate spectrum – method 2 Scanning measurementScanning measurement Estimate 2D-autocorrelation functionEstimate 2D-autocorrelation function 2D anisotropic power spectrum2D anisotropic power spectrum

Estimate distribution - zenithEstimate distribution - zenith Option: PDF field is measured PDFOption: PDF field is measured PDF However!However!

– Inhomogeneous (non-stationary) fieldInhomogeneous (non-stationary) field– Underestimate the width distributionUnderestimate the width distribution

Time (s)

Tim

e (s

)

2000 4000 6000 8000

1000

2000

3000

4000

5000

6000

7000

8000 0

100

200

300

400

500

0 5 10 150

1

2

3

4

5

Reduction in variance (%)

Num

ber

Histogram cumulus

0 10 20 30 400

1

2

3

4Histogram stratocumulus

Reduction in variance (%)

Num

ber

Relative reduction in variance Relative reduction in variance Zenith measurement Zenith measurement Simulated on LES cloudsSimulated on LES clouds

Estimate distribution - zenithEstimate distribution - zenith

Estimate distribution - scanEstimate distribution - scan CorrectingCorrecting

– Frank EvansFrank Evans

More dataMore data SpreadSpreadScanningScanning

Estimate distribution - scanEstimate distribution - scan

-80 -60 -40 -20 0 20 400

1

2

3

4

5Scanning measurement - Sc

Reduction in variance (%)

Num

ber

-600 -400 -200 0 2000

2

4

6

8Scanning measurement - Cu

Reduction in variance (%)

Num

ber

Estimate distribution - heightEstimate distribution - height Distribution Distribution

as a function as a function of heightof height

Time [hr] UT

Hei

ght

[km

]

LWC template [kg/m3]

10.5 11

1.4

1.6

1.8

2

2.2

0

0.1

0.2

0.3

0.4

0.5

LWC Surrogate

0 2 4 60

2

4

6

0

2

4

6

0 2 4 61.5

2

Reff

Surrogate

0 2 4 60

2

4

6

0

2

4

6

0 2 4 61.5

2

Time [hr] UT

Hei

ght

[km

] LWC template [kg/m3]

13.2 13.41

1.5

2

0

0.5

1

1.5

LWC Surrogate

0 2 4 6 80

2

4

6

8

0

2

4

6

8

0 2 4 6 81

1.52

Reff

Surrogate

0 2 4 6 80

2

4

6

8

0

2

4

6

8

0 2 4 6 81

1.52

Time [hr] UT

Hei

ght

[km

]

LWC template [kg/m3]

7 7.5

0.5

0.6

0.7

0.8

0.9

0

0.5

1

1.5

LWC Surrogate

0 5 100

5

10

0

5

10

0 5 10

Reff

Surrogate

0 5 100

5

10

0

5

10

0 5 10

Time [hr] UT

Hei

ght

[km

]

LWC template [kg/m3]

9 9.5

0.4

0.6

0.8

0

0.2

0.4

0.6

LWC Surrogate

0 5 100

5

10

0

5

10

0 5 100.40.60.8

Reff

Surrogate

0 5 100

5

10

0

5

10

0 5 100.40.60.8

Added: Local valuesAdded: Local values Input from scanning Input from scanning

measurementmeasurement– Amplitude distribution Amplitude distribution – 2D power spectrum2D power spectrum– The measured The measured

values on the spiralvalues on the spiral

Added: Local constraint & maskAdded: Local constraint & mask

Surrogate with cloud mask

Added: Coarse meansAdded: Coarse means

Coarse fractionCoarse means

SurrogateOriginal

Added: Coarse means & maskAdded: Coarse means & mask

Coarse fractionCoarse means

SurrogateOriginal

ApplicationsApplications Closure studiesClosure studies

– Bring micro-physics and radiation togetherBring micro-physics and radiation together– Poster: 3D surrogates from in situ Poster: 3D surrogates from in situ

measurements (Sebastian Schmidt, measurements (Sebastian Schmidt, Ronald Scheirer, Francesca Di Giuseppe)Ronald Scheirer, Francesca Di Giuseppe)

Structure studiesStructure studies Fractal generatorFractal generator

o GeophysicsGeophysics

ApplicationsApplications Closure studiesClosure studies Structure studiesStructure studies

– How good is the fractal approximation?How good is the fractal approximation?– How accurate do you need to know the How accurate do you need to know the

PDF?PDF?

Fractal generatorFractal generator

o GeophysicsGeophysics

Structure studiesStructure studies 2D tdMAP clouds2D tdMAP clouds IAAFT surrogatesIAAFT surrogates

– Full structureFull structure– Only correlations at small scalesOnly correlations at small scales

Compare radiative propertiesCompare radiative properties

y

Optical depth -2.5 -2.5 0.0033333

50 100 150 200 250

50

100

150

200

250

0

5

10

15

20

25

30

y

Optical depth 0 -2.5 0.0033333

50 100 150 200 250

50

100

150

200

250

0

5

10

15

20

25

30

2D & 3D reflectance2D & 3D reflectance

RT

: S

ebas

tián

Gim

eno

Gar

cía

RT

: S

ebas

tián

Gim

eno

Gar

cía

ApplicationsApplications Closure studiesClosure studies Structure studiesStructure studies Fractal generatorFractal generator

– Sensitivity studiesSensitivity studies Retrievals and parameterisationsRetrievals and parameterisations

– Spectrum and PDF varied independentlySpectrum and PDF varied independently– Broad-band radiometer EarthCARE Broad-band radiometer EarthCARE

Jaime F. Gimeno (University of Valencia) & Jaime F. Gimeno (University of Valencia) & Howard BarkerHoward Barker

o GeophysicsGeophysics

ApplicationsApplications Closure studiesClosure studies Structure studiesStructure studies Fractal generatorFractal generator

o Soil temperature fields LES Soil temperature fields LES o Rain fields Rain fields

o Downscaling low resolution atmospheric models Downscaling low resolution atmospheric models to high resolution hydrological modelsto high resolution hydrological models

o Surrogate run-offSurrogate run-offo Wind stress fieldso Soil propertieso …

Comparison cloud generatorsComparison cloud generators3,2: 3D, 2D; S: Structure; P: PDF; 3,2: 3D, 2D; S: Structure; P: PDF; L: Local values; M: MaskL: Local values; M: Mask

IAAFT methodIAAFT method Cumulus fields (Evans; structure of a binary mask)Cumulus fields (Evans; structure of a binary mask)

CLABAUTAIR (Scheirer and Schmidt)CLABAUTAIR (Scheirer and Schmidt) Shift cloud (Schmidt; Los and Duynkerke)Shift cloud (Schmidt; Los and Duynkerke) 2D-2D Ice cloud (Liou et al.)2D-2D Ice cloud (Liou et al.)

tdMAP (tdMAP (A. Benassi, F. Szczap, et al.)A. Benassi, F. Szczap, et al.) Multi-fractal cloudsMulti-fractal clouds Bounded Cascade and other fractal cloudsBounded Cascade and other fractal clouds Fourier methodFourier method

– SITCOM (F. di Giuseppe; 2D structure)SITCOM (F. di Giuseppe; 2D structure)– Ice clouds (R. Hogan, S. Kew; 2.5D structure)Ice clouds (R. Hogan, S. Kew; 2.5D structure)

3SPLM3SPLM33SSP__P__

33SSPL_PL_2__L_2__L_2__L_2__L_

2S2SPP____2S2SPP____2S___2S___2S___2S___

ConclusionsConclusions Developed/extended an algorithmDeveloped/extended an algorithm

– Full 3D structureFull 3D structure– LWC height profileLWC height profile– Local measured constraints (fine or coarse)Local measured constraints (fine or coarse)– Cloud maskCloud mask

AdvantagesAdvantages– FlexibleFlexible– DimensionsDimensions– InstrumentsInstruments– Vary the statistics easily and independentlyVary the statistics easily and independently

More informationMore information HomepageHomepage

– Papers, Matlab-programs, examplesPapers, Matlab-programs, examples

http://www.meteo.uni-bonn.de/ http://www.meteo.uni-bonn.de/ venema/themes/surrogates/venema/themes/surrogates/

Google: surrogate cloudsGoogle: surrogate clouds

BBC campaign surrogatesBBC campaign surrogates– ftp://bbc.knmi.nl/bbc1/model/ftp://bbc.knmi.nl/bbc1/model/

Victor.Venema@uni-bonn.deVictor.Venema@uni-bonn.de

Constrained surrogatesConstrained surrogates Arbitrary constraintsArbitrary constraints Evolutionary search algorithmEvolutionary search algorithm

Better convergenceBetter convergence Try new statisticsTry new statistics Fractal geometry for cloud boundariesFractal geometry for cloud boundaries

Evolutionary search algorithmEvolutionary search algorithm

Constrained surrogatesConstrained surrogates height profilesheight profiles

– cloud basecloud base– cloud topcloud top– cloud covercloud cover– average LWCaverage LWC

HistogramsHistograms– LWPLWP– LWCLWC– number of layersnumber of layers

Power spectra & lengthPower spectra & length– LWPLWP– Highest cloud topHighest cloud top– Lowest cloud baseLowest cloud base

Coarse mean surrogatesCoarse mean surrogates InputInput

– Local coarse mean LWPLocal coarse mean LWP, , ττ

– Local coarse mean cloud fractionLocal coarse mean cloud fraction– Power spectrum, extrapolated to small Power spectrum, extrapolated to small

scalesscales

Combine satellite and ground based orCombine satellite and ground based orin situ measurementsin situ measurements– Measured small scale spectrum or PDFMeasured small scale spectrum or PDF

Downscaling modelsDownscaling models– A priory small scale spectrumA priory small scale spectrum

InstrumentsInstruments 3 microwave radiometers,3 microwave radiometers, 3 cloud radars, 3 cloud radars, 4 Micro Rain Radars (MRRs), 4 Micro Rain Radars (MRRs), 2 wind profiler-RASS systems, to measure 2 wind profiler-RASS systems, to measure

wind and temperature profiles, wind and temperature profiles, 4 lidar ceilometers, 4 lidar ceilometers, 2 lidars2 lidars numerous radiation, precipitation, turbulence numerous radiation, precipitation, turbulence

and meteorological instruments.and meteorological instruments.

LES Surrogates - RadiationLES Surrogates - Radiation

OriginalOptical depth

dist

ance

(km

)

0

1

2

IAAFT surrogateOptical depth

dist

ance

(km

)

0

1

2

Fourier surrogateOptical depth

dist

ance

(km

)

0

1

2

PDF surrogateOptical depth

distance (km)

dist

ance

(km

)

Optical depth

0 1 20

1

2

30 40 50 60 70

OriginalTransmittance

IAAFT surrogateTransmittance

Fourier surrogateTransmittance

PDF surrogateTransmittance

distance (km)

Transmittance

0 1 2

0.2 0.3

OriginalActinic flux

IAAFT surrogateActinic flux

Fourier surrogateActinic flux

PDF surrogateActinic flux

distance (km)

IAF

(W m-1 nm-1)

0 1 2

500 550 600

OriginalRadiance

IAAFT surrogateRadiance

Fourier surrogateRadiance

PDF surrogateRadiance

distance (km)

Radiance

(W m-2 sr-1 m-1)

0 1 2

0.08 0.1 0.12

Fourier &distribution

RadianceActinic fluxTransmittanceOptical depth

Fourier

Distribution

Original

LES Surrogates - RadiationLES Surrogates - Radiation