monitoring algorithms
& their performances
23/04/2015
NLR
Amsterdam
2
/
2
/
•
Work has been done in the framework of the EU FP7 program,
the UFO project.
•
Presentation contributors: Albert Oude Nijhuis, Agnes
Dolfi-Bouteyre, Cedric Rahatoka, Ludovic Thobois, Richard Wilson
Introduction and motivation
Contemporary EDR/wind retrieval methods
EDR from radar and lidar profilers
EDR/wind from scanning lidar
EDR/wind from scanning radar
Wind vector retrieval development
Conclusions
4
/
4
/
We would like to improve wind vector/turbulence intensity
retrievals and improve wake vortex monitoring.
Introduction and motivation
Improve
turbulence
retrievals
Improve
wake vortex
monitoring
Improve
wind
retrievals
Evaluation of
contemporary
methods
New methods?
Improve
lidar/radar
technology
Evaluation of
contemporary
methods
Instruments at the Toulouse trials for remote sensing of
turbulence intensity / wind vectors.
Windcube V2
UPMC Radar
Scanning
windcubes
Thales X-Band
weather radar
Leosphere
windcube
profiler
X-band
Doppler radar
profiler from
LATMOS
(UPMC)
Leophere
Windcube UFO
- Glide slope
or
- Air volume
Scanning at
three
elevations.
- Air volume
6
/
6
/
Wind vector retrieval
We measure Doppler velocities, which are radial velocities along the
line of sight.
Radar/Lidar Doppler velocities within an analysis volume
are converted into a wind vector.
Wind field is assumed to be linear.
Wind field is retrieved by
a least squares fit or;
by single value decomposition.
Figure adapted from `Doppler radar and
weather observations’ (Doviak and Zrnic, 2006)
𝒗 𝑥, 𝑦, 𝑧 = 𝒗 𝑥
0
, 𝑦
0
, 𝑧
0
+
𝜕𝒗
𝜕𝑥
(𝑥 − 𝑥
0
)
+
𝜕𝒗
𝜕𝑦
…
Turbulence intensity retrieval
Turbulence is quantified by the Eddy dissipation rate (EDR)
8
/
8
/
Turbulence intensity retrieval
EDR
can be derived from velocity measurements from radar, lidar or
sonic anemometers.
EDR
retrievals
A sequence
of velocities
Doppler
Spectral
width
Power spectrum
Structure function
Variance
More processing steps,
e.g. Doviak (2006) / White
(1999).
Profilers at the Toulouse trials
CURIE: a X-band Doppler radar from LATMOS (UPMC)
Leosphere WindCube 7v2: a Doppler lidar profiler
Doppler spectra are obtained each 6 s (radar) or 0.8 s (lidar)
Measurements consist in backscattered power,
10
/
10
/
WindCube 7v2 lidar profiler
Different methods show consistency.
Vertical and off-vertical velocity
variances
=> isotropic fluctuating field
Good agreament between different
EDR estimates
(here total variance and spectral
level).
Curie radar profiler
Tilme-height plot showing the diurnal cycle of turbulence activity during the
Toulouse ATB trial as observed by CURIE radar: Cn2 (top), vertical wind
12
/
12
/
Comparison of radar and lidar profilers
Time series of EDR estimates from radar (blue) and lidar (red) measurements
during day-time.
Diurnal cycle of turbulence activity observed simultaneously by radar (left)
and lidar (right).
Remarks on EDR from profilers
EDR from radar and lidar profilers show consistency.
Some discrepancies still need to be clarified (likely related to low signal
to noise ratio).
The ability of these instruments to measure turbulence intensity in the
atmospheric boundary layer during non rainy periods is clearly
demonstrated.
14
/
14
/
Lidar Scanning strategy
3D WIND
Windcube
UFO
EDR/wind from scanning lidar
Volume Wind Measurements with UFO Lidar
Provide 2D Wind measurements every 500m from the ground to 500m of
10
8
6
4
2
0
2
4
6
8
10
0
50
100
150
200
250
300
350
400
450
500
Distance from Lidar (km)
H
ei
g
h
t
(A
G
L
m
)
Power Law
Extrapolation
Illustration: Artificial Terrain
Spatial Interpolation/
Power Law
Interpolation
Cross-sectional View
Legend: Zones of Interpolation
**Buildings near the immediate vicinity
of lidar, obstructed the lower two
elevation angle scans
Lines indicate various
elevation angles
2°
2,5°
3°
4°
5°
6°
8°
20°
30°
Volume Wind Measurements with UFO Lidar
Methodology used for retrieving 2D wind vectors according to available
16
/
16
/
05/05/2014 – 11h02
5x PPI of Radial data
8x CAPPI of 2D Winds
Volume Wind
Algorithm
2°
3°
5°
8°
20°
22m
87m
217m
347m
477m
EDR/wind from scanning lidar
Volume Wind Measurements with UFO Lidar
Glide Path Wind Measurements with UFO Lidar
Windcube
UFO
18
/
18
/
7x part PPI of Radial data
2D Winds along glide path
Glide Path Wind
Algorithm
EDR/wind from scanning lidar
Glide Path Wind Measurements with UFO Lidar
EDR from azimuthal structure function
-10000
-5000
0
5000
-6000
-4000
-2000
0
2000
4000
6000
15-Apr-2014 09:19:39-6
-4
-2
0
2
4
Vr (m/S)Azimuthal structure function
07:12 08:10 09:09 10:08 11:07 12:06 13:05 14:04 15:03 16:02
0
0.1
0.2
0.3
0.4
Time
ED
R
1
/3
(
m
2
/3
.s
-1
)
averaged over 10 mn, no error terms 24-Apr-2014... : EDR1/3 z=0050 m to0100 m z=0100 m to0150 m z=0150 m to0200 m z=0200 m to0250 m z=0250 m to0300 m z=0300 m to0350 m z=0350 m to0400 m
EDR1/3 (m2/3.s-1)averaged over 10 mn, no error terms 24-Apr-2014... : EDR1/3
a
lt
it
u
d
e
(m
)
time
08:12
09:09
10:07
11:04
12:01
12:59
13:56
14:53
50
100
150
200
250
300
350
0
0.05
0.1
0.15
0.2
0.25
0.3
EDR1/3 (m2/3.s-1) 24-Apr-2014. EDR1/3 (m2/3.s-1) averaged over 10 mn, no error termsAzimuth from 47° to 293°
Elevation = 6°
0
100
200
300
400
500
600
700
800
0
0.05
0.1
0.15
0.2
0.25
15-Apr-2014 09:14:47 averaged 10 mn
m
m
²/
s
²
EDR 1/3 = 0.0476 m 2/3. s-1
L0 = 505 m
20
/
20
/
EDR from azimuthal structure function
EDR^1/3 (m2/3.s-1) averaged over 10 mn, no error terms 06-May-2014
a
lt
it
u
d
e
(m
)
time
07:00
08:01
09:02
10:02
11:03
12:04
13:04
14:05
15:06
16:06
50
100
150
200
250
300
350
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
EDR^(1/3) (m2/3.s-1)06:00
0
06:58
07:57
08:56
09:55
10:54
11:53
12:52
13:51
14:50
15:49
16:48
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
Time
ED
R
1
/3
(
m
2
/3
.s
-1
)
06-May-2014... : EDR1/3 averaged over 10 mn, no error terms
z=0050 m to0100 m
z=0100 m to0150 m
z=0150 m to0200 m
z=0200 m to0250 m
z=0250 m to0300 m
z=0300 m to0350 m
z=0350 m to0400 m
0.1
0.2
0.3
0.4
0.5
ED
R
1
/3
(
m
2
/3
.s
-1
)
16-Apr-2014... : EDR1/3 averaged over 10 mn, no error terms
z=0050 m to0100 m
z=0100 m to0150 m
z=0150 m to0200 m
z=0200 m to0250 m
z=0250 m to0300 m
z=0300 m to0350 m
z=0350 m to0400 m
16-Apr-2014. EDR1/3 (m2/3.s-1)averaged over 10 mn, no error terms
a lti tu d e (m) 50 100 150 200 250 300 350 0.05 0.1 0.15 0.2 0.25 0.3 EDR1/3 (m2/3.s-1)
Remarks on EDR/wind from scanning lidar
Wind vectors can be retrieved in the glide path.
EDR from azimuthal structure function gives a consistent picture
22
/
22
/
Obtaining EDR from radar.
Radar Doppler velocity variance / spectral width is a combination of
factors.
Validation of EDR by big data
144 days of EDR values compared with 80 parameters at meteorological
supersite in Cabauw.
Small scale effects (DSD, drop inertia) on retrieved EDR mitigated by
using a large footprint.
Comparison of EDR from vertically profiling radar (TARA) shows good
agreement with sonic anemometer in convective mixed boundary layer.
Comparison of EDR fails in nocturnal boundary layer, when
stratification is important.
January 19
th
, 2012
24
/
24
/
X-Band Radar : Analysis - 24/04/2014
2D Wind Field (top) and EDR spectral width Retrieval (bottom)
EDR value (in dB)
Elevation : 3.5°
Elevation : 5°
Elevation : 6.5°
Wind main direction
Heading ~0°
Wind main radial velocity ~8m/s
Unrealistic high EDR values
Elevation : 5°
Elevation : 6.5°
2 wind main directions
Wind radial velocity span : 0 to 20m/s
Wind main direction : -50° wrt north
Elevation : 3.5°
X-Band Radar : Analysis - 06/05/2014
2D Wind Field (top) and EDR spectral width Retrieval (bottom)
EDR value (in dB)
26
/
26
/
EDR/wind from scanning radar
X-Band Radar : Analysis - 06/05/2014
For a large footprint, the effects of drop size distribution, inertial effects
are of less importance. Here a sequence of mean Doppler velocities is
used.
EDR is retrieved from velocity variance, 2nd and 3rd order structure
functions.
Collaborative Radar/Lidar Wind Monitoring in Glideslope
RADAR
LIDAR
RADAR
LIDAR
06/05/2014
No rain,
cloudy conditions
24/04/2014
Raining, strong winds
28
/
28
/
EDR from radar and lidar profilers
Remarks on EDR from scanning radar and lidar
With EDR from spectral width mostly realistic EDR values are found but
some unrealistic high values at low elevation.
EDR from spectral width remains a challenge as it is a combination of
many factors. In the future polarimetric radar, or additional information
from models / ground sensors, can help to enhance the retrievals.
EDR from combination of radar cells (large footprint) gives realistic
values.
Under some conditions, the radar and lidar are able to observe air
space simultaneously.
Wind vectors are retrieved from Thales X-band radar
Wind vectors are retrieved from Thales X-band radar
Least squares fit with linear wind model (See Doviak et al.)
Optimal estimation with forward model is used.
30
/
30
/
Contemporary wind vector retrieval
Radar/Lidar Doppler speeds within an analysis volume
are converted into a wind vector.
Wind field is assumed to be linear.
Wind field is retrieved by
a least squares fit or;
by principal component analysis.
Figure adapted from `Doppler radar and
weather observations’ (Doviak and Zrnic, 2006)
𝒗 𝑥, 𝑦, 𝑧 = 𝒗 𝑥
0
, 𝑦
0
, 𝑧
0
+
𝜕𝒗
𝜕𝑥
(𝑥 − 𝑥
0
)
+
𝜕𝑦
𝜕𝒗
…
4D-Var wind vector retrieval
4D-Var (optimal estimation) is applied by minimization of a cost
function.
The cost function consist of three parts
- Reflection (measurable)
- Radial velocities (measurable)
- Parameters
Error covariance matrices S are given.
Correlation length is defined for the parameter space
32
/
32
/
Comparison of wind vector retrieval techniques
Wind vector retrieval development
Linear wind model
4D-Var wind retrieval
Unrealistic jumps
Resolution volume analyzed:
-> Spurious result at large distance
Unclear how small scale wind
variations are resolved.
Time variation not included
Controllable smoothness
Complete scan [or multiple] solved
at once:
-> No spurious results.
Ambiguity of measurements can be
shown:
• Horizontal direction variation
• Horizontal speed variation
Time variation included,
Wind vectors are retrieved from Thales X-band radar
Spurious wind vectors can be detected by looking at average
circulation and convergence.
The radar is not at the center of the universe …
… therefore the solution should not depend on its position!
← Circulation
Divergence →
34
/
34
/
Summary
Summary
EDR from lidar/radar profilers show a consistent picture, with some
possible discrepancies due to low SNR. The lidar/radar profilers can be
applied in non-rainy conditions.
Wind speeds from scanning lidar show good agreement with
TUB
research aircraft and reference sensor (lidar profiler).
EDR from scanning lidar, with azimuthal structure function, gives a
consistent picture.
EDR from scanning radar spectral width sometimes gives unrealistic
high values. EDR from scanning radar with a large footprint shows
consistent picture for different algorithms.
Under some conditions, the lidar and radar can record simultaneously.
Wind retrieval based on optimal estimation can give more realistic
retrieved wind vectors, if we look at circulation and divergence of the
resulting vectors.
References / further reading
Babb, D.M., Verlinde, J., Rust, B.W., 2000. The removal of turbulent broadening in radar Doppler spectra using linear inversion with double-sided constraints. Journal of atmospheric and oceanic technology .
Barbaresco, F., Juge, P., Klein, M., Ricci, Y., Schneider, J., Moneuse, J., . Optimising runway throughput through wake vortex detection, prediction and decision support tools.
Bouniol, D., Illingworth, A.J., Hogan, R.J., 2003. Deriving turbulent kinetic energy dissipation rate within clouds using ground based 94 GHz radar. Conference on radar meteorology .
Bringi, V., Chandrasekar, V., 2004. Polarimetric Doppler weather radar.
Careta, A., Sagues, F., 1993. Stochastic generation of homogeneous isotropic turbulence with well-defined spectra. physical review E .
Chan, P.W., 2011. Generation of an eddy dissipation rate map at the Hong Kong international airport based on Doppler lidar data. Journal of atmospheric and oceanic technology .
Cohn, S.A., 1995. Radar measurements of turbulent eddy dissipation rate in the troposphere : A comparison of techniques. Journal of atmospheric and oceanic technology .
Doviak, R.J., Zrnic, D.S., 2006. Doppler radar and weather observations second edition.
Emanuel, M., Sherry, J., Catapano, S., Cornman, L., Robinson, P., 2013. In situ performance standard for eddy dissipation rate.
Frech, M., 2007. Estimating the turbulent energy dissipation rate in an airport environment. Boundary layer meteorology .
Heijnen, S.H., Ligthart, L.P., Russchenberg, H.W.J., 2000. First measurements with TARA; an S-Band transportable atmospheric radar. Physics and Chemistry of the Earth .
Krishnamurthy, R., Choukulkar, A., Calhoun, R., Fine, J., Oliver, A., Barr, K., 2013. Coherent doppler lidar for wind farm characterization. Wind Energy .
Mann, J., 1998. Wind field simulation. Prob. Eng. Mech. .
Meischner, P., Baumann, R., Holler, H., Jank, T., 2001. Eddy dissipation rates in thunderstorms estimated by doppler radar in relation to aircraft in situ measurements. Journal of atmospheric and oceanic technology .
O’Connor, E.J., Illingworth, A.J., Brooks, I.M.,Westbrook, D., Hogan, R.J., Davies, F., Brooks, B.J., 2010. A method for estimating the turbulent kinetic energy dissipation rate from a vertically pointing Doppler lidar, and independent evaluation from balloon-borne in situ measurements. Journal of atmospheric and oceanic technology .
Oude Nijhuis, A., Unal, C., Krasnov, O., Russchenberg, H., Yarovoy, A., 2013. Dynamics of turbulence in precipitation: Unraveling the eddies. IPC2013 .
Oude Nijhuis, A., Unal, C., Krasnov, O., Russchenberg, H., Yarovoy, A., 2014a. Optimization of turbulence measurements for radar, lidar and sonic anemometers. ERAD2014 .
Oude Nijhuis, A., Unal, C., Krasnov, O., Russchenberg, H., Yarovoy, A., 2014b. Outlook for a new wind field retrieval technique: The 4d-var wind retrieval. Radar2014 .
Pinsky, M., Khain, A., 2006. A model of a homogeneous isotropic turbulent flow and its application for the simulation of cloud drop tracks. Geophysical & Astrophysical Fluid Dynamics .
Pope, S., 2000. Turbulent flows.
Rodgers, C.D., 2000. Inverse methods for atmospheric sounding - Theory and practive, vol. 2 of Atmospheric, Oceanic and Planetary Physics, World Scientific, Singapore.
Siebert, H., Lehmann, K., Wendisch, M., 2005. Observations of small-scale turbulence and energy dissipation rates in the cloudy boundary layer. Journal of atmospheric sciences .
Unal, C., Dufournet, Y., Otto, T., Russchenberg, H., 2012. The new real-time measurement capabilities of the profiling TARA radar. Seventh European conference on radar in meteorology and hydrology (ERAD) .
White, A.B., Lataitis, R.J., Lawrence, R.S., 1999. Space and time filtering of remotely sensed velocity turbulence. Journal of atmospheric and oceanic technology .
Yanovsky, F., 1996. Simulation study of 10 ghz radar back scattering from clouds and solution of the inverse problem of atmospheric turbulence measurements. Computation in Electromagnetics, IET .
Yanovsky, F., 2002. Phenomenological models of Doppler-polarimetric microwave remote sensing of clouds and precipitation. Geoscience and remote sensing symposium .
Yanovsky, F., Russchenberg, H., Unal, C., 2003. Doppler-polarimetric radar observations of turbulence in rain. Scientific Report: IRCTR-S-006-03 .
36
/
Development of an algorithm to provide vertical wind
profiles every 500m up to 500m and up to 10km from
the LIDAR every 5 minutes
Comparisons with certified WindcubeV2 LIDAR show
good agreement better than 1m/s
LEO 2
