)
Ocean surface wind direction and speed observations from 81 drifters in several regions of the
world ocean are analyzed. The data periods for all drfiters are presented below.
Days are numbered starting with January 1, 1998 = day 1; "A8" means August 1988, and "J0"
means January 2000, etc.The drifter's wind direction and speed observations will be compared with
QSCAT data, which start on July 19, 1999.
Several drifters were eliminated due to bad position data; and bad wind direction or speed data were identified by comparison with collocated QSCAT observations. As a result, 66 drifter contain acceptable data. Those are shown below.
)
)
)
)
)
)
)
)
)
)
)
)
)
)
)
)
)
)
)
)
)
The search for the best-fit coefficients is done in several steps, illustrated here for drifter 6686 (see fig. 5.1).
All collocated data (Ndat=79) are plotted. The small numbers within the plot symbols represent the time differences of
the collocated data, when divided by 10min. The average difference and RMS with respect to collocated QSCAT are listed
above the figure: for drifter 6686, avDiff=-20.1°, and RMS=34.8°.
Certain data are excluded from the subset of data used for finding the correction coefficients:
For drifter 6686, there are 79 collocated data. Of those data, Ndiff=3, Nsp=16, and Ndist=8, resulting in Nsig=56. Four data points are excluded due to meeting several exclusion criteria simultaneously. Three data points are eliminated due to large st.dev. values, leading to a subset of Nfit=53, which are used to compute the correction coefficients: C1 = -12.10°, C2 = 4.14°, C3 = 3.74°, C4 = -21.32°, and C5 = -7.48°. These coefficients are listed below the figure, as well as their standard deviation in the best-fit routine to determine their values (in parentheses).
Also plotted in the figure are the average difference of all collocated data (dashed line), the offset coefficent C1 (dashed line), the correction function (solid curve: correction coefficients are listed as "(60min)" below the figure), and an alternative correction function (dotted curve: "30min") for a subset of data consisting of those collocated data when the time differences are only 30min or less. For drifter 6686, Nfit(60min)=53, and Nfit(30min)=30. For all drifters, these two correction functions are not very different from each other.
The second figure shows the corrected data (plotted as QSCAt-drifter differences), and listed are the resulting average difference and RMS of all collocated data, based on the 60min coefficients. For drifter 6686, avDiff=-4.6° and RMS=20.8°. Five data points (red symbols) fall outside the range of +/-2*RMS (red lines).
)
)
There are drifter data, however, for which the sine/cos correction method fails. This happens when the wind direction data is clustered in a relatively narrow range of wind direction. As an example, the data for drifter 18905 are plotted below (fig. 5.2). Here, with the sine/cos correction, the corrected data have an RMS of 77.4° with respect to QSCAT data. In this case, a simple offset correction is applied. The resulting offset-corrected data have an RMS of 17.0° with respect to QSCAT. There are three drifters that were corrected with an offset correction, instead of with the sine/cos correction.
)
)
)
In total, 61 drifters with wind direction data could be corrected. The summary statistics of comparing the corrected data with collocated QSCAT is shown below, as histograms of the resulting average differences and rms with respect to QSCAT data.
For wind direction, there are a total of 61 drifters, with 11,233 collocated data pairs. The average of all differences is -2°, and the average of all rms is 33°.
)
)