Exploitation of homogeneous isotropic turbulence models for optimization of turbulence remote sensing
Delft University of Technology
A.C.P. Oude Nijhuis, C.M.H. Unal, O.A. Krasnov, H.W.J. Russchenberg and A. Yarovoy
Outline
1 Introduction
The UFO project Research question
2 Overview of turbulence models Stochastic turbulence models Parametric turbulence model 3 Simulation of radar observables
Zephyros
4 Simulated radar measurements Scanning radar
Staring radar
Polarimetric radar spectra Canting angles
Next section
1 Introduction
The UFO project Research question
2 Overview of turbulence models
Stochastic turbulence models Parametric turbulence model
3 Simulation of radar observables
Zephyros
4 Simulated radar measurements
Scanning radar Staring radar
Polarimetric radar spectra Canting angles
Inertia effect
Ultra Fast wind SensOrs project
The UFO project
A solution to mitigate weather haz-ards and increase airport capacity. Prospect: WVs and weather hazards can be monitored under all weather conditions by using UFO scanning radars and lidars??!!
Research question
Research question
How does turbulence look like from the radar?
1 How can turbulence models be used for simulation of radar
measurements?
2 How does turbulence look like in polarimetric radar
measurements?
The problem with radar turbulence retrievals
Radar Doppler spectral width / Radar Doppler mean velocity variance
σ2 v = σ2d+ σ02+ σ2α+ ζI2σ2T Assumption of homogeneous isotropic turbulence moderate turbulence ζ2 Iσ2T∝ σ2d, σ20, σα2 light turbulence ζ2 Iσ2T σ2d, σ02, σ2α strong turbulence ζ2 Iσ2T σ2d, σ02, σα2 large footprint ζI = 1 small footprint ζI = ? DSD DSD
turbulence intensity turbulence intensity turbulence intensity
turbulence too small to measure
Next section
1 Introduction
The UFO project Research question
2 Overview of turbulence models Stochastic turbulence models Parametric turbulence model 3 Simulation of radar observables
Zephyros
4 Simulated radar measurements
Scanning radar Staring radar
Polarimetric radar spectra Canting angles
Inertia effect
Overview of turbulence models
Stochastic turbulence models
• Random initialization.
• With given parameters, the outcome is always different.
• Different applications and complexity. Parametric turbulence model
• With the given parameters, the outcome is always the same.
Stochastic turbulence models
Stochastic turbulence models
Example (Careta and Sagues (1993))
• Start from auxilary random scalar, which is transformed to a divergence free 2D field. • Used for study
of diffusion processes.
Stochastic turbulence models
Example (Mann (1998)) • 3D wind field, spectral tensor is used. • Used in wind engineering.Stochastic turbulence models
Example (Pinsky and Khain (2006))
• 2D windfield, obeys the continuity equation • Used for simulation of diffusion processes
Stochastic turbulence models
Example (Oude Nijhuis et al. (2014))
• 1D windfield, large dynamic range (mm-km scale) • Used for turbulence intensity retrieval studies.
Parametric turbulence models
Parametric turbulence models
Defining parameter: turbulence broadeningσT, which depends on
turbulence intensity,, and the radar resolution volume parameters.
Solutions to integrate it in the radar measurements:
1 Smearing/smoothening/broadening of spectra. [typical approach,
not considered]
Parametric turbulence models
Ensemble of isotropic vectors
For each particle, an ensemble of isotropic vectors is used:
~vp,∗=~vterminal+~vair +~vturbulence,∗, (1)
where for each 3D unit direction ˆk∗,
vturbulence,∗ =σT
√
3ˆk∗. (2)
The average cross section is obtained by averaging over all directions: σ = 1
n X
σ(~vp,∗). (3)
• The particle symmetry axis (/minor axis) is parallel to vp,∗.
Parametric turbulence models
Ensemble of isotropic vectors
For each particle, an ensemble of isotropic vectors is used:
~vp,∗=~vterminal+~vair +~vturbulence,∗, (4)
0.1 mm 1.0 mm 5.0 mm
light turbulence, σ = 0.1 ms−1.
heavy turbulence, σ = 2.0 ms−1.
Overview of turbulence models
Reference Application Features
CA93 Diffusion processes 2D
MA98 Wind turbine engineering 3D
PI06 Droplet tracks 2D
CTM14 EDR retrievals 1D, scale
symmetric PA15 (ensemble of
isotropic vectors)
Radar observables No field sim-ulation.
Next section
1 Introduction
The UFO project Research question
2 Overview of turbulence models
Stochastic turbulence models Parametric turbulence model
3 Simulation of radar observables Zephyros
4 Simulated radar measurements
Scanning radar Staring radar
Polarimetric radar spectra Canting angles
Zephyros
Zephyros
• Zephyros is a software package under development for wind/turbulence simulation and retrieval.
• Development version publicly available on GIThub.
• Written mostly in C, bit of Fortran. Interfaces for Python/Matlab.
• Has fast mode for retrievals (small ensemble of scatterers) and slow mode (large ensemble, better accuracy) for simulations.
Next section
1 Introduction
The UFO project Research question
2 Overview of turbulence models
Stochastic turbulence models Parametric turbulence model
3 Simulation of radar observables
Zephyros
4 Simulated radar measurements Scanning radar
Staring radar
Polarimetric radar spectra Canting angles
Scanning radar
Details of the simulation
• Radar scans in the xy-plane.
• No noise is added.
• frequency 3.298 GHz radar (S-Band)
• EDR = 0.1 m2s−3. (very strong turbulence!)
• LWC = 1 gm−3, rain with gamma distribution parameters, µ = 5, D0 = 2mm.
• resolution volume parameters, FWHM = 2.1 ◦, range resolution 30 m.
Scanning radar
Doppler mean velocities
CA93 MA98 PI06
CTM14 0.1 mm1.0 mm 5.0 mm light turbulence, σ = 0.1 ms−1. heavy turbulence, σ = 2.0 ms−1. PA15
Scanning radar
Doppler spectral width
CA93 MA98 PI06
CTM14 0.1 mm1.0 mm 5.0 mm light turbulence, σ = 0.1 ms−1. heavy turbulence, σ = 2.0 ms−1. PA15
Staring radar
Details of the simulation
• Radar staresalong the x line.
• No noise is added.
• frequency 3.298 GHz radar (S-Band)
• EDR = 0.1 m2s−3. (very strong turbulence!)
• LWC = 1 gm−3, rain with gamma distribution parameters, µ = 5, D0 = 2mm.
• resolution volume parameters, FWHM = 2.1 ◦, range resolution 30 m.
Staring radar
Doppler spectra for hh
CA93 MA98 PI06
CTM14 0.1 mm1.0 mm 5.0 mm light turbulence, σ = 0.1 ms−1. heavy turbulence, σ = 2.0 ms−1. PA15
Polarimetric radar spectra
Details of the simulation
• Radar stares at 45 ◦ in the xz-plane.
• No noise is added.
• frequency 3.298 GHz radar (S-Band)
• EDR is varied in the simulations.
• LWC = 1 gm−3, rain with gamma distribution parameters, µ = 5, D0 = 2mm.
• Range is 1 km, resolution volume parameters: FWHM = 2.1 ◦, range resolution 30 m.
Polarimetric radar spectra
Polarimetric variables for Mann (1998)
[m2s−3] 10−5 10−2 10−1
σv ,hh [ms−1] 0.750 0.802 9.67
dBZdr 0.601 0.631 0.668
dBLdr -67.3 -46.0 -37.7
• Zdr, Ldr and spectral width contain turbulence intensity information.
< ZDR >= dB(< ηhh> / < ηvv >) (5)
Polarimetric radar spectra
Polarimetric variables for the parametric model [m2s−3] 10−5 10−2 10−1
σv ,hh [ms−1] 0.750 0.847 1.14
dBZdr 0.597 0.589 0.555
dBLdr -63.5 -43.2 -36.4
• Parametric model reproduces the same dependencies as the stochastic model!
< ZDR >= dB(< ηhh> / < ηvv >) (7)
Polarimetric radar spectra
hh Doppler spectrum
Polarimetric radar spectra
specific Zdr
Polarimetric radar spectra
specific Ldr
Canting angles
canting angle spread
MA98 PA15 Experiment
Inertia effect
Solution
The inertial velocity term~v0
p is assumed to be small in comparison to
the total particle velocity:
~vp=~vterminal +~vair +~vp0 (9)
The solution is found by solving the equations of motion for a backward small trajectorie:
dvp,z
dt = Fg − Fb− Fd ,z =ηI ,zv
2
Inertia effect
Solution
µ = 5, = 10−2 µ = 5, D
0= 5 mm
Next section
1 Introduction
The UFO project Research question
2 Overview of turbulence models
Stochastic turbulence models Parametric turbulence model
3 Simulation of radar observables
Zephyros
4 Simulated radar measurements
Scanning radar Staring radar
Polarimetric radar spectra Canting angles
Inertia effect
Conclusions
• Zdr and Ldr contain information on the turbulence intensity, which can be used in turbulence retrievals in precipitation. It depends however on the DSD.
• The inertia effect can alter the Doppler spectral width.
• A parametric model is found that can reproduce the features of radar measurements for stochastic turbulence.
• The parametric model implicitly reproduces the canting angle distribution.
References
K.V. Beard, V.N. Bringi, and M. Thurai. A new understanding of raindrop shape. Atmospheric research, 2010. A. Careta and F. Sagues. Stochastic generation of homogeneous isotropic turbulence with well-defined spectra. physical
review E, 1993.
R. J. Doviak and D. S.. Zrnic. Doppler radar and weather observations second edition. 2006. J. Mann. Wind field simulation. Prob. Eng. Mech., 1998.
A.C.P. Oude Nijhuis, C.M.H. Unal, O.A. Krasnov, H.W.J. Russchenberg, and A. Yarovoy. Simulation of atmospheric turbulence: Fractal turbulence. Poster presentation at the 21st Symposium on Boundary Layers and Turbulence, 2014.
M. Pinsky and A. Khain. A model of a homogeneous isotropic turbulent flow and its application for the simulation of cloud drop tracks. Geophysical & Astrophysical Fluid Dynamics, 2006.
The end
Backup slides: Radar filter
Radar filter
1 User input: drop size distribution, atmospheric temperature
profile, parameters that characterize the radar.
2 Dividing resolution volume in subvolumes. The resolution
volume is divided into equally weighted subvolumes.
3 Coordinate transformations BEAM → AZEL → ENU.
4 Interpolation of user input For each subvolume, the user input
that is required is interpolated.
5 Radar cross sections are calculated for each subvolume and for
each particle.
Backup slides: Radar filter
Radar filter
The weather radar equation can be written as (details→ Doviak and Zrnic (2006)): < P∗(r0)>= Cr r2 0 η∗, (11)
where< P∗ > is the expected received power, Cr a radar constant, r0
the range at the resolution center andη∗ the reflectivity. The
reflectivity,η∗ is defined as:
η∗=
Z
Ω
σ∗(De)N(De)dDe, (12)
Backup slides: Radar filter
Radar filter radar reflec-tion < η∗>= X i η∗,i n reflectivity factor < Z∗>= 10· log10( λ4 < η ∗ > π5|K w|2 ) Doppler ve-locity < v∗,D >= X i η∗,i n vD,i . < η∗ > Doppler spectral width < σ2 v ,∗,D >= X i η∗,i n (vD,i− < v∗,D >) 2.< η ∗ > differential reflectivity < ZDR >= 10· log10(< ηhh > / < ηvv >) depolarisation ratio < LDR >= 10· log10(< ηhv > / < ηvv >)Backup slides: Radar filter
Radar filter copolar cor-relation co-efficient < ρc >= X i Shh,iS † vv ,i s X i |Shh,i|2 X i |Svv ,i|2 cross polar correlation coefficient < ρh cx >= X i Shh,iS † hv ,i s X i |Shh,i|2 X i |Shv ,i|2 < ρv cx >= X i Svv ,iSvh,i† s X |Svv ,i|2 X |Svh,i|2Backup slides: Radar filter
Radar filter specific differential phase < KDP >= 103λ Z ∞ 0 Re(Shhf − Sf vv)N(De)dDeBackup slides: Radar filter
Radar filter, spectra Doppler spectrum < η∗,1,2>= X i ,v1<vD,i<v2 η∗,i n /(v2− v1) specific dif-ferential re-flectivity < ZDR,1,2>= 10· log10(< ηhh,1,2> / < ηvv ,1,2>) specific de-polarization ratio < LDR,1,2>= 10· log10(< ηhv ,1,2 > / < ηvv ,1,2>) specific cor-relation co-efficients < ρ1,2 >= “· · · ·00 for v1< vD,i < v2