Geosat Follow-On GDR Significant Wave Height

and Wind Speed Calibration


D J T Carter
Satellite Observing Systems


9 September 2005

Download the Word version, here

Download the pdf version, here

1. Introduction

Earlier calibration of GFO significant wave height (Hs) and wind speed at 10 m (U10) have been carried out using IGDR's.   In particular, David Cotton (pers. comm.) has put together a data set of co-located Hs and U10 values from US NDBC buoys and from IGDR's when the altimeter track passed close to a buoy, taking the nearest 1 Hz record to the buoy providing it was within 50 km and within 30 minutes of the buoy measurements.   Inspection indicates that the IGDR Hs values are the same as the GDR values.   But the U10 values are different, because of an adjustment of 0.63 dB, but the s0 values are the same, so I have derived the equivalent GDR U10 values by applying the algorithm in the GFO Users Handbook - derived from Witter & Chelton (1991) but incorporating the adjustment - to get U10 from the s0's.

These 'GDR' U10 values include some less than 1.83 m/s; although using the quality_1 flag in the records removes all s0 > 14 dB, and hence all U10 < 1.83 m/s.

 

2. Data

The data set contains values from January 2000 to August 2003.   The numbers for each year and the co-located NDBC buoys are given in Table 1.

 

NDBC buoy

all years

2000

2001

2002

2003

41001

50

7

21

20

2

41002

109

20

25

37

27

41010

61

12

20

14

15

42001

49

9

0

24

16

42002

104

21

16

40

27

42003

47

9

18

18

2

42019

106

15

31

35

25

42020

106

13

28

40

25

44004

48

1

19

21

7

44008

94

18

9

40

27

44011

13

0

1

7

5

44014

60

18

42

0

0

46001

121

19

37

38

27

46002

86

16

35

35

0

46005

117

23

28

40

26

46006

46

7

15

16

8

46035

25

15

6

0

4

51001

103

14

37

43

9

51002

131

22

42

41

26

51003

1

0

1

0

0

51004

109

19

40

28

22

 

Table 1. Buoy/GFO comparisons by year and by US NDBC buoy.

 

Buoys 42001, 42002, 42003 and 46035 measure wind speed at 10 m.   The others have anemometers at 5 m; the winds at 10 m were estimated from these using the air/sea temperature difference (Dobson, 1981: Appendix Table A-3); if either temperature was missing, the record was discarded.

 

Temporal variation

There seems to be some variation in the mean difference between buoy and altimeter Hs and U10 from year to year, as shown in Tables 2 and 3.

 

Year

Number

mean

s.d.

s.e.(mean)

2000

291

0.2217

0.2192

0.01285

2001

472

0.2307

0.2549

0.01173

2002

552

0.2603

0.2515

0.01070

2003

304

0.2626

0.2139

0.01227

 

Table 2. Statistics of buoy Hs - GFO Hs by year

 

Year

Number

mean

s.d.

s.e.(mean)

2000

278

-1.0405

1.6114

0.0966

2001

471

-0.1638

1.5307

0.0705

2002

537

-0.03296

1.4778

0.0637

2003

300

0.07437

1.3429

0.0775

 

Table 3. Statistics of buoy U10 - GFO U10 by year

 

The difference between buoy and altimeter Hs appears to increase slightly over the years, but this might be due to the changes in the buoys used. The drift in the U10 difference is more marked, and is significantly different in 2000 from the other years.   There were some problems with GFO during 2000; it was not finally accepted by the US Navy until November 2000. So I have discarded the 2000 data from further analysis.

 

3. Wave height calibration

Regressing the 1328 pairs of US NDBC buoy Hs and the GFO Hs from 2001 to August 2003 gives the result shown in Figure 1.

The orthogonal distance regression is:

       Hs(buoy) =1.088 Hs(gfo) + 0.093                                                    (1)

              (1.076 - 1.099)   (0.069 - 0.118)

where the numbers in brackets are 95% confidence intervals on the estimated slope and intercept.   The root mean square error, calculated from Hs(buoy) -(1.088 Hs(gfo) + 0.093), is 0.24 m, and the correlation coefficient is 0.98.

 

Figure 1. Comparison of significant wave heights from US NDBC buoys and the GFO altimeter. The solid line is the odr fit.

 

Queffeulou (2003, 2004) gives two calibrations for correcting GFO IGDR Hs based on comparisons against other altimeter values at ground-track crossings. Queffeulou (2003) gives:

       Hs(cor) =1.0812 Hs(gfo) + 0.0392                                                      (2)

in good agreement with (1). 

Queffeulou (2004) uses a larger data set of Topex Hs (B transmitter) at ground-track crossings from December 1999 to December 2003, which gives:

        Hs(cor) =1.0625 Hs(gfo) + 0.0754                                                      (3)

However, this is dependent on Queffeulou's calibration for Topex (B):

       Hs(cor) =1.0237 Hs(topex) - 0.0476                                                    (4)

Cotton (pers.comm, 2001) obtained, from a comparison against NDBC buoys:

       Hs(cor) =1.0376 Hs(topex) - 0.0674                                                    (5) 

If (5) is used instead of (4) then (2) becomes:

       Hs(cor) =1.0769 Hs(gfo) + 0.1921                                                     (6)

in much closer agreement with (1). But the difference between (1) and (3) is quite small: less than 0.1m for Hs<3m and less than 3% for Hs>3m.

 

4. Wind speed calibration

A comparison of the 1308 estimates of wind speed from GFO and the buoys, adjusted to 10 m, is shown in Figure 2 (left).   Removing the 19 'outliers' with differences of more than 3 s.d., gives the right hand figure. 

 

 

Figure 2. Comparison of significant wave heights from US NDBC buoys and the GFO altimeter. The solid line is the odr fit. 

The orthogonal distance regression (fitted assuming errors on the altimeter and buoy wind speeds are equal) from the right hand figure is:

       U10(buoy) = 0.953 U10(gfo) + 0.374                                                     (7)

                 (0.932 - 0.973)     (0.207 - 0.541)

with rrms of 1.22 m/s and correlation coefficient of 0.93.

The odr fit in the left hand figure of Figure 2 has slope and intercept of 0.954 and 0.292, within the 95% confidence intervals of (7), but the rrms is 1.43 m/s and the correlation coefficient is 0.90.   A similar analysis of the 827 data from 2002 and 2003 gives slope and intercept of 0.952 and 0.416, also within the 95% confidence intervals of (7).

 

5. Conclusions

GFO GDR significant wave heights appear to be about 9% low, when compared to US NDBC buoy values, while the GDR estimates of wind speed at 10 m are 4% high.  There is a need for further data to reduce the size of the confidence limits, and to investigate possible changes in calibration over the years. In the meantime, equations 1 and 7 will give reasonably accurate corrections, i.e.:

       Hs(cor) = 1.088 Hs(gfo) + 0.093

       U10(cor) = 0.953 U10(gfo) + 0.374

References

Dobson F. W. 1981
   Review of reference height for and averaging time of surface wind measurements at sea.
   Marine Met. & related Oceanographic Activities, Rpt No.3. WMO.

Queffeulou P. 2003
   Long-term quality status of wave height and wind speed measurements from satellite altimeters.
   Proc. ISOPE Conf., Hawaii, 25-30 May 2003 (CD-ROM) 

Queffeulou P. 2004
   Long-term validation of wave height measurements from altimeters.
   Marine Geodesy 27, 495-510

Witter D. L. & Chelton D. B. 1991
   A Geosat altimeter wind speed algorithm and a method for altimeter wind speed algorithm development.
   J. Geophys. Res. 96, 8853-8860.