1. Comparisons of Two Preliminary Windsat Vector Wind Data Sets with
QuikSCAT and Buoys
M.H. Freilich
B.A. Vanhoff
8 February 2005
COAS/OSU

2. OVERALL OUTLINE
NDBC Buoy comparisons (best-quality
Windsat data)
Rain Flag – accuracy of flagged data
Windsat Rain Flags
Comparisons with NDBC, SSM/I
QuikSCAT-NESDIS_0 global comparisons
Directional comparisons

3. Windsat Data Initial Editing
For Betten_1 data sets, DELETED all data having:
Geographic Flags set (ice[6],land[7],lakes[9],salinity[10])
No Wind Retrieval Flag set ([0])
10 GHz RFI Flag set ([12])
Sun Glint Flag set ([13])
For NESDIS_0 data sets, DELETED all data having:
Land Contamination Flag set ([29])
Any EDR Quality Flag [4-27] set
For all data sets, IGNORED Windsat Speed Range Flag
(analyzed all speeds)

4. QuikSCAT Data & Initial Editing
Speed, Direction measurements on ~25 km centers from standard
L2B processing
Vector wind estimates have 25 km resolution, 30 km land mask
Used “selected” ambiguity
DELETED all data having:
Autonomous (MUDH Rain Flag set (eliminates ~6% of global data)
Far Swath (outer 100 km on each side, v-pol only retrievals with
little azimuthal diversity); WE DID KEEP NADIR SWATH!
This results in “research quality” QuikSCAT data

5. Windsat: Selected NDBC Validation Locations (22)

6. Windsat, QSCAT & Buoy: Collocated winds, non-raining
(Dashed line: NDBC collocated with NESDIS_0)
All NDBC winds adjusted to 10 m equivalent neutral stability
Betten_1 speed histograms have unrealistic concentration ~9-11 m/s
NESDIS_0 speed histogram narrower than QSCAT or NDBC (or Betten_1)
Betten_1 directional histos have unrealistic small peaks near 0, 120 deg
All Windsat directional histos have sharp drop-offs below deg

7. Windsat & QSCAT (dir. edit) vs. NDBC -- Rain-Free
50 km distance threshold, Non-Raining
Speed bias statistics nearly identical (0.75 m/s along-wind random component error
for QuikSCAT and Betten_1, 1.0 m/s along-wind rce for NESDIS_0)
All Windsat products overpredict wind speed for buoy speeds > 15 m/s
Windsat dir. std. dev. larger than QSCAT for wind speeds < 10 m/s (all 3 WS data sets)
NESDIS_0 dir. std. dev. less than QSCAT for wind speeds > ~15 m/s
Betten_1 dir. std. dev. smallest of all data sets for speeds m/s

8. Direction Comparisons w/ NDBC (Non-Raining)
Thin lines indicate random component error model simulation (1.6, 2.0, 2.4, 2.8 m/s)
Betten_1 (blue) = 2.6 m/s NESDIS_0 = 2.8 m/s QSCAT (red) = 2.0 m/s
Windsat directional std. devs. consistent with random comp. error model for spds < 20 m/s
Windsat may have better ambig removal skill at high wind speed (QSCAT departs from
model because of residual (< 90 deg) ambiguity removal errors)

9. Betten_1 & QSCAT: Vector Wind Accuracies vs. NDBC
QSCAT statistics and random component errors smaller than Windsat

10. Windsat, SSM/I F13: Rain-Flagged
Total Flagged
14.4%
17.4%
9.5%
7.9%
Windsat collocated w/ NDBC
SSM/I F13 similarly collocated
with NDBC
F13 has 6 AM crossing time
(similar to Windsat)
F13 considered “rain flagged”
if rain rate > 0 mm/hr
SDR-based rain flags overflag significantly, systematically with increasing wind speed
(both NESDIS_0 and Betten_1 data sets)
Betten_1 EDR-only flag rate is similar to SSM/I

11. Windsat/Buoy: Raining vs. Non-raining accuracies
50 km distance threshold
2.8 m/s rce
Rain-flagged data have much larger speed bias for buoy speeds < 11 m/s
(ie., Windsat overpredicts speed in rain)
Generally larger directional noise at all speeds for the rain-flagged data

12. NESDIS_0/QuikSCAT: Non-raining coverage
1 hour, 50 km distance threshold
QuikSCAT collocated w/ Windsat

13. NESDIS_0/QSCAT Comparisons
50 km, 1 hour collocation thresholds
Random component error sims in blue/red solid lines
0.75/1.0 m/s alongwind rce, 2.0/2.8 m/s cross-wind rce (QS/WS)
High wind speed saturation for Windsat?
Increase in directional s.d. probably due to QS errors

14. NESDIS_0 & QuikSCAT Directional Distributions
Selected direction (rel to N.)
Collocated global data set; no QS or WS rain (by flags)

15. NESDIS_0/QuikSCAT: Directional Histograms

16. Conclusions
Non-raining Windsat retrievals are somewhat less accurate than QSCAT
Betten_1 more accurate overall than NESDIS_0
Non-raining Windsat directional accuracy better than QSCAT at high
speeds (> 15 m/s)
Windsat SDR-based rain flags are too conservative
EDR-based flag is better
Rain causes speed bias and directional accuracy degradation
Global comparisons suggest Windsat underpredicts high wind speeds
(> 20 m/s)
NESDIS_0 direction retrieval issues (discontinuities in histogram)