Evaluating the Accuracy of Gridded Satellite Wind Vector Maps

We evaluated several QuikSCAT based vector maps of near surface winds, stress and pseudostress with respect to winds at the research mooring from CLIMODE and from the KEO mooring near the Kuroshio Extension in the North Pacific. The analyses were done with the versions of the products listed below that were available in May 2010.

The evaluations focused on the accuracy of the wind product in reproducing the anemometer wind anomalies. An additional goal was to determine the temporal and spatial resolution of the product in comparison with the product's grid. We also show a comparison of the temporal means.

Product Name	Variable	Temporal Grid	Spatial Grid	Color (plots)	Source
Kelly & Dickinson	Wind	daily	1/2^o	cyan	suzanne@apl.washington.edu
Ifremer	Wind	daily	1/2^o	blue	ftp.ifremer.fr
NCEP2 Reanalysis	Wind	daily	gaussian, ~1.9^o	gray	ftp.cdc.noaa.gov
ECMWF	Wind	daily	gaussian, ~1.1^o	orange	http://dss.ucar.edu/datasets/ds111.1/
Tang & Liu	Wind	12-hourly	1/2^o	purple	http://airsea-www.jpl.nasa.gov/seaflux/
CCMP	Wind	6-hourly	1/4^o	magenta	http://podaac.jpl.nasa.gov/
Milliff	Wind	6-hourly	1/2^o	green	http://dss.ucar.edu/datasets/ds744.4/
NCEP2 Reanalysis	Wind	6-hourly	gaussian, ~1.9^o	black	ftp.cdc.noaa.gov
Kelly & Dickinson	Stress	daily	1/2^o	cyan	suzanne@apl.washington.edu
Ifremer	Stress	daily	1/2^o	blue	ftp.ifremer.fr
COAPS	Pseudostress	6-hourly	1^o	olive	http://coaps.fsu.edu/scatterometry/

Comparison of the mean vectors (wind/stress/pseudostress) from CLIMODE (elft) and KEO (right) anemometers. The product vectors below are shown in four groups from top to bottom: daily wind, 6-hourly wind, stress and pseudostress. The Tang & Liu product is in the daily wind group. The colors are the same as indicated in the above Table.


Mean vectors at CLIMODE Mooring	Mean vectors at KEO Mooring

Anomaly Comparisons Using a Taylor Diagram

We use a Taylor diagram to compare the products with the "truth," here, the anenometer mooring winds. For the correlations we use the magnitude of the complex correlation between vectors. The complex correlations also give a mean angle difference for the product relative to the anemometer anomaly (not shown here). The distance between the color dot for each product and the red dot on the x-axis is the normalized error, a simple measure of the accuracy of the wind product anomalies. For the CLIMODE comparisons the best products are EDMWF, K&D, Ifremer and CCMP. Results are similar for wind products at KEO, except that the Milliff product is similar to the other four products. The T&L and both NCEP products have magnitudes that are too large and lower correlations.

Wind products have a fixed temporal grid,which is often (incorrectly) assumed to correspond to the temporal resolution of the fields. To find the actual product temporal resolution, we computed the normalized errors relative to smoothed anemometer winds. The amount of smoothing required to attain a minimum in normalized error is a measure of the product's actual temporal resolution. By this test ECMWF products attain the lowest errors when compared with anemometer winds smoothed over one day for CLIMODE; CCMP has a temporal resolution of 6 hours, consistent with its grid, somewhat lower for KEO. The K&D winds have a temporal resolution between 1 and 2 days for either site. The T&L winds have a temporal resolution of 2-3 days, compared with its 12-hr grid. The accuracies are consistent with the results above: CCMP, Ifremer, K&D and ECMWF have the lowest errors overall.