A new algorithm for the downscaling of 3dimensional cloud fields

1
A new algorithm for the downscaling of
3-dimensional cloud fields

• Victor VenemaSebastián Gimeno García Clemens
  Simmer

2
Applications

• Downscaling 3D CRM/NWP model fields
• Downscaling of 2D satellite measurements
• Coarse mean LWC
• Coarse cloud fraction

3
Requirements downscaling method

• Nonlinear processes
• Sub (coarse) scale distribution
• IPA-bias if you average ? instead of
  (ir)radiances
• Non-local processes
• For example spatial correlations
• 3D bias ignore horizontal photon transport to
  low ?

4
Downscaling - Cumulus

• High resolution original gt
• Coarse means
• No clear subpixels
• 2 coarse fields
• Input downscaling
• Real application start with coarse fields
• Compare high-resolution fields
• Physical
• Radiative

Coarse means
Original
No. clear subpixels
5
Downscaling - Cumulus

• High resolution original gt
• Coarse means
• No clear subpixels
• 2 coarse fields
• Input downscaling
• Real application start with coarse fields
• Compare high-resolution fields
• Physical
• Radiative

Coarse means
No. clear subpixels
Surrogate
6
Downscaling - Cumulus

• High resolution original gt
• Coarse means
• No clear subpixels
• 2 coarse fields
• Input downscaling
• Real application start with coarse fields
• Compare high-resolution fields
• Physical
• Radiative

Coarse means
Original
No. clear subpixels
Surrogate
7
Cumulus validation data

• Diurnal cycle of Cu
• Land (ARM)
• 51 fields
• High resolution
• 64x64 pixels
• Horizontal resolution 100m
• Coarse resolution
• 16x16
• Horizontal resolution 400m
• Nc 300 cm-3

Brown, A.R., R.T. Cederwall, A. Chlond, P.G.
Duynkerke, J.C. Golaz, M. Khairoutdinov, D.C.
Lewellen, A.P. Lock, M.K. MacVean, C.H. Moeng,
R.A.J. Neggers, A.P. Siebesma and B. Stevens,
2002. Large-eddy simulation of the diurnal cycle
of shallow cumulus convection over land, Q. J. R.
Meteorol. Soc., 128(582), 1075-1093.
8
Stratocumulus validation data

• Dissolving broken Sc
• Ocean (ASTEX)
• 29 fields
• High resolution
• 200x200 pixels
• Horizontal resolution 50m
• Coarse resolution
• 20x20
• Horizontal resolution 500m
• Nc 200 cm-3

Chosson, F., J.-L. Brenguier and L. Schüller,
"Entrainment-mixing and radiative Transfer
Simulation in Boundary-Layer Clouds", J Atmos.
Res.
9
Algorithm

• Preparations
• Calculate power spectrum coarse LWC field
• Extrapolate spectrum to smaller scales
• Main iterative loop
• Adjust to the extrapolated spectrum
• Adjust to the coarse fields
• Remove jumps at edges of coarse field

10
Algorithm flow diagram
11
Extrapolation power spectrum

• Algorithm works with any power spectrum
• Cumulus clouds
• Assumption
• Intermediate to small scales are fractal
• follow power law (Varianceakb)
• Linear regression in log-log spectrum
• Fitting range
• small scales of coarse field (intermediate scales
  full field)
• Stratocumulus cloud
• Not fractal at intermediate scales
• Assumption
• Shape power spectrum same for all clouds
• Computed an average isotropic spectrum over all
  clouds
• Scaled by average variance at intermediate scales

12
Example 3D fields
Original Extrapolated
Surrogate Coarse field
Cumulus Stratocumulus
13
Example 3D fields
Original Extrapolated
Surrogate Coarse field
Cumulus Stratocumulus
14
Scatterplot irradiances Cu
Reflectance SZA 0 Reflectance SZA
60 Transmittance SZA 0 Transmittance SZA
60
Two Extrapolated Coarse field Interpolated
originals surrogate field
15
Scatterplot irradiances Sc
Reflectance SZA 0 Reflectance SZA
60 Transmittance SZA 0 Transmittance SZA
60
Two Extrapolated Coarse field Interpolated
originals surrogate field
16
RMS relative difference
17
RMS relative difference
18
Conclusions

• Downscaling algorithm works
• Large improvement for irradiancescompared to
  coarse cloud fields
• Extrapolation is a significant error source
• Low number of pixels in coarse fields
• Best extrapolation method is application
  dependent

19
Outlook

• Importance of the coarse cloud fraction field
• Include a distribution for the anomalies
• Wavelets, increment distributions?
• Applications
• Downscaling CRM/NWP model fields
• Anomalies, small-scale spectrum from LES or
  observations
• Downscaling of satellite measurements
• Coarse LWP fields
• High resolution in situ LWC, Reff measurements

20
Outlook

• Importance of the coarse cloud fraction field
• Include a distribution for the anomalies
• Wavelets, increment distributions?
• Applications
• Downscaling CRM/NWP model fields
• Anomalies, small-scale spectrum from LES or
  observations
• Downscaling of satellite measurements
• Coarse LWP fields
• High resolution in situ LWC, Reff measurements

Thank you for your attention!