)
)
)
Milliff, R.F., M.H. Freilich, W.T. Liu, R. Atlas, and W.G. Large, 2002: "Global Ocean Surface Vector Wind Observations from Space" in Observing the Oceans in the 21st Century, C.J. Koblinsky and N.R. Smith (eds.), GODAE Project Office and Bureau of Meteorology, Melbourne., pgs. 102-119.
This chapter provides a summary of current capabilities and near-future plans for surface vector wind observation sampling frequency, record length, spatial resolution, and global coverage. These properties are put in proper physical oceanographic context by means of several examples described in the appendix.
The dependency of the slope on cross-swath cell number is demonstrated in the
following plots.
Shown are slope vs. cell number for the months of January, March, May, July,
September, and November 2000. The spectral slopes are computed from nonoverlapping,
along-track segments of zonal velocity spanning at least 30° of latitude.
The slopes were computed for 3 different regions:
North Pacific (40-50° N, 120-240° E), Subtropical Northern Pacific (20-30° N,
120-240° E), and the combined Western Tropical Pacific and Tropical Atlantic
(10° S-10° N, 120-180° E and 300°-360° E. The centers of tracks are
required to fall within these areas. All slopes are indicated by dots, the
averages are plotted with filled circles and lines, and standard deviations at each
cross-swath cell number are represented by red lines. The lower panels show bin
weights vs. cell number.
)
)
)
)
)
)
)
)
)
)
)
)
)
)
)
)
The PSD slope has been calculated for the regular QSCAT-1 data, the DIRTH product,
Remote Sensing Systems' KU-2001 product, and for collocated NCEP data (4xdaily,
T126 FNL Operational Analysis; courtesy of Mike Freilich and Barry Vanhoff at OSU).
DIRTH slope amplitudes are systematically (though perhaps not significantly) lower than
the R1 slope amplitudes for all cell numbers. They rise, however, much less in the nadir
region than for the R1 product. The cell-by-cell changes in slope match between R1 and DIRTH,
with a nearly constant offset of 0.2-0.3 outside the nadir region, and up to 0.5 - 0.6 around
nadir.
The KU-2001 slope amplitudes are as low as the DIRTH amplitudes in the "sweet-spots" (cell
numbers 10-25 and 52-67), but as large as the R1 amplitudes in the nadir region.
)
The first figure for each product shows the average slope of power spectral density (PSD) for the zonal wind component U averaged over all tracks in this region. The number of tracks averages 700-900 in the central swath region, but can be as low as 400 in the outer swath regions (cell numbers 3-9 and 68-74). Cell numbers 1, 2, 75, and 76 have no valid data. Standard deviations and the 5-point running mean are also shown. Bin weights are shown in the bottom panel of the first figure.
The second figure shows the average U PSD for wavelets 8°, 4°, 2°, and 1° at every third cell across the swath. The resulting slope at that cell number is represented by the line (linear fit to the 4 data points) and printed below the graphs. The gray background areas represent the across-swath average PSD values +/- the across-swath standard deviations. Note that the across-swath PSD standard deviations are smallest for the 8° and 4° wavelets, but larger for the 2° and even bigger for 1° wavelets. This is the case for the R1 product, but those standard deviations are almost constant (on the log plot) for the DIRTH product.
The third figure shows in four panels (with linear y-scale) the average U PSD vs. cell number. The solid line indicates the average PSD in the sweet spot regions (cell numbers 10-25 and 52-67). The dashed line represents the average of all cell numbers. Note that the PSD is almost constant with respect to cell number for the 8° and 4° wavelets, but increases in the nadir region for the 2° wavelet and shows the biggest relative nadir increase for the 1° wavelet. The same is true for the standard deviations (red bars).
The fourth figure is similar to the third figure, but shows average PSD for all wavelets in one graph on a logarithmic scale. Solid and dashed lines represent sweet spot and overall swath averages as in the third figure. The relative increases of PSD in the nadir region for 2° and 1° wavelets are more pronounced in this log plot.
Note that in the DIRTH product, the nadir region increase of PSD is eliminated for the
2° wavelet and significantly reduced for the 1° wavelet. In addition, PSD is
lower in DIRTH than in R1 for every cell number and for every wavelet. The resulting
sweet-spot average, however, is only 3% lower for 8° and 10% lower for 4°, but
21% lower for 2& and 32% for 1°. The bottom graph in the last figure for DIRTH
(log PSD vs. cell number for the 1° wavelet) indicates that the sweet-spot region,
i.e. cell numbers 10-25 and 52-67, corresponds with very constant values in PSD.
)
)
)
)
)
)
)
)
Both QSCAT and KU2000 exhibit greater variability (noise?) in the nadir region. This region is narrower in KU2000 than in QSCAT. The very central part
of the nadir region seems to be slightly less noisy in QSCAT than in KU2000. There is greater variability (true WVC-to-WVC wind variability or noise?)
in the outer swath regions in QSCAT than in KU2000.
)
)
)
)