1. Characterization of Ocean Wind Vector Retrievals Using ERS-2 Scatterometer High-Resolution Long-term Dataset and Buoy Measurements
Supervisor:
Prof. Frank S. Marzano
Candidate:
Federica Polverari
Assistant Supervisors:
Eng. Raffaele Crapolicchio
Dr. Marco Talone
Master’s Degree in Electronic Engineering
curriculum in Complex Systems and Remote-Sensing for the Environment
Image of ESA

2. Outline The ERS-2 Scatterometer Data Data Selection Methodology
L1 Validation
L2 Validation
Main Results L1 L2
Conclusions

3. ERS-2 Scatterometer
The European Remote-sensing Satellite (ERS)-2 was involved in Earth Observation activities and it was launched in 1995 as follow-on of the ERS-1 mission
ERS-2 embarks an Active Microwaves Instrument (AMI) in which it is incorporated a radar instrument called Scatterometer working at 5.3 GHz
The scatterometer is designed to measure the Normalized Radar Cross Section (NRCS) of the sea surface
Image from ESA Website
Image from Stoffelen, 2008
The NRCS is obtained by the scattering mechanism (Bragg Scattering) of wind-generated capillary waves which are in equilibrium with the near-surface wind over the ocean
R: distance target-radar
Wr: received power
Wt :transmitted power
F: scatterometer foootprint
Lampda: radar wavelenght
D: Directivity gain
Thanks to the Geophysical Model Function and Inversion Algorithms, from NRCS measurements it is possible to estimate the WIND SPEED and WIND DIRECTION at 10 m above the mean sea level

4. Image from PIRATA Website
Data
The European Remote sensing Satellite (ERS)-2 Wind Speed (WS) and Direction (WD)
This database covers the period during
The products available for the users are the Advanced Scatterometer Processing System (ASPS ) Level 2.0 products
The investigation has been carried out using the High Resolution (25 km) products
The In-Situ Wind Speed and Direction provided by
the Prediction and Research Moored Array in the Atlantic (PIRATA)
Site
Date first deployed
0 0
Feb 1998
0 10W
Sept 1997
2N 10W
Nov 1999
2S 10W
5S 10W
Jan 1999
6S 10W
Mar 2000
10S 10W
0 23W
 Dec 2001
0 35W
Jan 1998
15N 38W
12N 38W
Feb 1999
8N 38W
4N 38W
Image from PIRATA Website

5. Data Selection One buoy measurement several satellite measurements
The spatial and temporal collocation is the way to find satellite and buoy data pairs which can be compared.
Spatial Collocation:
Nodes whose distance from a buoy position is less than 1 degree (~100km) are selected
Temporal Collocation:
From the spatial collecated data, nodes whose time difference from buoy acquisition time is less than 10 min are selected
One buoy
measurement
several satellite measurements

6. Data Selection One buoy measurement several satellite measurements
Repetition of buoy WS and WD for each scatterometer measurement
Average of scatterometer WS and WD
Weighted average on square distance

7. Methodology – L1 σ° σ°* A0, A1, A2 A0*, A1*, A2* remotely sensed
Validation at Level1: comparison between predicted and estimated coefficients of the C-band Geophisical Model Function (GMF) for equivalent neutral wind CMOD5.N
remotely sensed
in situ
WS,WD
Repetition
WS*,WD*
CMOD5.N σ°
σ°*
Fitting
Fitting
A0, A1, A2
A0*, A1*, A2*

8. Mean Difference between predicted and estimated A0
Main Results – L1
4 m/s (blue), 8 m/s (red),12 m/s (black)
The estimated (dots) and predicted (full line)
A0 exhibit a similar behaviour for each beam
Mean Difference between predicted and estimated A0 WS
3m/s
8m/s
12 m/s
Fore
0.6216
1.4104
Mid
0.2212
1.1618
Aft
0.5547
1.9064

9. Mean Difference between predicted and estimated
Main Results – L1
Mean Difference between predicted and estimated A1 WS
3m/s
8m/s
12 m/s
Fore
1.4522
Mid
0.0060
0.0647
Aft
0.0295

10. Mean Difference between predicted and estimated
Main Results – L1
Mean Difference between predicted and estimated A2 WS
3m/s
8m/s
12 m/s
Fore
0.0179
0.0026
0.2658
Mid
0.1983
0.3526
1.7009
Aft
0.1036
0.1315
1.1892

11. Main Results-L1:Fore Beam
Additional analysis:
Spatial collocation 50 km
Benefit: reduction of the Representativity Error
Number of samples is strongly reduced
Threshold:
80 measurements
Spatial collocation: 100 km
Spatial collocation: 50 km

12. Main Results-L1:Mid Beam
Spatial collocation: 100 km
Spatial collocation: 50 km

13. Main Results-L1: Aft Beam
Spatial collocation: 100 km
Spatial collocation: 50 km

14. Methodology – L2
Comparison and statistical analysis between buoys and scatterometer in term of wind speed and wind direction
The case of buoy 10S 10W (out of 13): with the greatest number of matches
In order to have collocated measurements couples, we have considered the following criteria :
Mean of satellite WS and WD related to each bouy measurements
Weighted average on square distance

15. Wind Speed Misfit (considering the mean)
Main Results – L2
Comparison and statistical analysis between buoys and scatterometer in term of wind speed and wind direction
The case of buoy 10S 10W (out of 13): with the greatest number of matches
Wind Speed Misfit (considering the mean)
Normalized Histogram
σ = 1.5

16. Main Results – L2 Seasonal Trend of Wind Speed 0.2058 0.2133 1.3150
STATISTICS OF WS MISFIT
Mean Misfit (m/s)
Mean Misfit (weighted average) (m/s)
Misfit Std
(weighted average)
0.2058
0.2133
1.3150
1.3600
Seasonal Trend of Wind Direction
STATISTICS OF WD MISFIT
Mean Misfit (deg)
Mean Misfit (weighted average) (deg)
Misfit Std
(weighted average)
3.6564
3.1420

17. Main Results-L2 Wind Speed Misfit indeces: 0 - 0.2 m/s 0.2 - 0.5 m/s
Combining the collocated measurements from all the available buoys
Wind Speed Misfit indeces:
m/s
m/s
m/s
> 1m/s
Wind Direction Misfit indeces:
0° - 3°
3° - 10°
10° - 30°
> 30°

18. Main Results-L2 To characterize the three wind speed ranges:
Combining the collocated measurements from all the available buoys
To characterize the three wind speed ranges:
Buoys and ERS-2 wind speed difference against buoy wind speed
WS < 4 m/s:
negative differences thus the satellite tends to overestimate the lower wind.
WS > 9 m/s:
positive differences thus the satellite tends to underestimate higher wind
4 m/s < WS < 9 m/s
greatest concentration of measurements whose differences are around 0 m/s

19. Main Results-L2
Characterization of the misfit in term of buoy wind speed
Characterization of the misfit in term of ERS-2 wind speed
Buoy Wind Speed
Range (m/s)
Number
of data
Mean
Difference
(m/s)
Standard Deviation
0 - 2
134
1.7028
2 - 4
428
1.6602
4 - 6
973
1.5019
6 - 8
1778
0.3142
1.3062
8 - 10
1269
0.7557
1.2543
242
1.5226
1.8969

20. Conclusions From the validation at Level 1
From this first analysis of the ERS-2 and PIRATA buoys datasets during the period between , with the chosen criteria we can conclude that:
From the validation at Level 1
The validity of the ERS-2 GMF has been checked:
the buoy wind measurements are able to retrieve the characteristics of ERS-2 backscatter coefficient measurements with main discrepancies for higher winds;
Not significant changes can be observed with the reduction of the spatial collocation.
From the validation at Level 2
The ERS-2 measurements are consistent with the buoys wind measurements :
the Wind Speed difference is less than 1 m/s on average;
the Wind Direction difference is less than 30 degree on average;
ERS-2 overestimates lower winds and underestimates higher winds.
Reference:
A., Bentamy, D. Croize-Fillon, C., Perigaud, “Characterization of ASCAT measurements based on buoy and QuikSCAT wind vector observations”, Ocean Sci., 4, , 2008.

21. Thank you for your attention!
Acknoledgements
Thanks to:
Prof. Frank S. Marzano
Raffaele Crapolicchio
Marco Talone
All colleagues of Serco SpA
Thank you for your attention!

22. Focusing On
CMOD5.N: similar to a second order Fourier expression, the power 1.6 was included because it avoids the need of higher harmonics.
2) Neutral Wind: is the wind at 10-metre heigh for given surface stress that would be observed if there was neutral atmospheric stratification [Geernaert and Katsaros 1986].

23. Statistics for each buoy
Main Results – L2
Statistics for each buoy
Buoy Site
WS Mean Misfit (m/s)
WD Mean Misfit (Deg)
WS Misfit Std
WD Misfit Std
0 0
0.3267
1.6617
0N 10W
0.2618
1.4139
0N 23W
0.2896
3.3124
1.7453
0N 35W
0.1407
1.6601
2N 10W
0.1279
1.5944
2S 10W
0.2580
1.3981
4N 38W
0.0337
1.8407
5S 10W
0.1442
1.1880
6S 10W
0.2851
3.7537
1.4149
8N 38W
0.0725
1.8389
12N 38W
0.1166
1.6466
ON AVERAGE
0.1870
1.5821