Evaluating and Extending the Ocean Wind Climate Data Record

Abstract

Satellite microwave sensors, both active scatterometers and passive radiometers, have been systematically measuring near-surface ocean winds for nearly 40 years, establishing an important legacy in studying and monitoring weather and climate variability. As an aid to such activities, the various wind datasets are being intercalibrated and merged into consistent climate data records (CDRs). The ocean wind CDRs (OW-CDRs) are evaluated by comparisons with ocean buoys and intercomparisons among the different satellite sensors and among the different data providers. Extending the OW-CDR into the future requires exploiting all available datasets, such as OSCAT-2 scheduled to launch in July 2016. Three planned methods of calibrating the OSCAT-2 σ_o measurements include 1) direct Ku-band σ_o intercalibration to QuikSCAT and RapidScat; 2) multisensor wind speed intercalibration; and 3) calibration to stable rainforest targets. Unfortunately, RapidScat failed in August 2016 and cannot be used to directly calibrate OSCAT-2. A particular future continuity concern is the absence of scheduled new or continuation radiometer missions capable of measuring wind speed. Specialized model assimilations provide 30-year long high temporal/spatial resolution wind vector grids that composite the satellite wind information from OW-CDRs of multiple satellites viewing the Earth at different local times.

Keywords: Radar cross section; remote sensing; satellite applications; sea surface; wind.

Fig. 1

Four decades of satellite wind measurements and scheduled future missions. The black lines show the series of microwave scatterometers that provide ocean vector winds (OVW). After HY-2B, CNSA plans to fly scatterometers not shown. The dotted line extending the QuikSCAT from 2009 onward denotes the nonspinning phase of operation. The blue lines show the SSM/I and SSMIS instruments flown on the series of DMSP satellite platforms numbered F8 to F20. These sensors only provide ocean wind speed, not direction. The pink lines show the microwave radiometers with the lower frequency channels needed for measuring sea surface temperatures in addition to wind speed, water vapor, clouds, and rain rates. The lower frequency channels also improve the wind speed accuracy. WindSat is the only microwave radiometer that also provides wind direction due to the inclusion of polarimetric channels. The green lines show the L-band radiometers SMOS, Aquarius, and SMAP, which are very insensitive to rain, and especially, well suited for measuring high winds in storms.

Fig. 2

Example of the local time of the ascending node for some of the sun-synchronous scatterometer and radiometer wind observations, from 1988 until present (solid lines). QuikSCAT and F08 (dash lines) differ in that their descending node is plotted. Sensors with rapidly precessing orbits (TMI, GMI, and RapidScat) are not shown in the figure.

Fig. 3

Global monthly time series of the rain-free wind speed differences between the ASCAT-A and the following sensors collocated to within four hours: QuikSCAT, TMI, WindSat, AMSRE, SSMI F17, AMSR2, GMI, and RapidScat. All of these satellite wind timeseries are RSS CDR products except for RapidScat (RSCAT), which is produced at JPL. NCEP GDAS model winds are also compared. The red star at the end of 2015 represents the ASCAT-RapidScat in the days after the hardware anomaly in August 2015. Note that the F17 SSM/I has a known wind speed drift which started in mid-2011. The origin of this drift is currently under investigation. Also, NCEP timeseries is not stable due to the frequent changes in the datasets it assimilates. As discussed in Section V-C, a calibration shift is apparent between ASCAT-A (version V1 displayed here) and the other datasets in September 2014. The data have now been reprocessed ([, version V2.1]) taking into account a calibration adjustment provided by KNMI [41].

Fig. 4

Existing and planned direct intercalibration of Ku-band σ_o measurements.

Fig. 5

Multiple paths for wind speed intercalibration. The bias and standard deviation are found by averaging over the pixels in the 1° latitude/longitude annual map of the wind speed difference.

Fig. 6

Wind speed differences of ASCAT-A minus GMI (top panel) and RapidScat minus WindSat (bottom panel). The ASCAT-A/GMI results are a 2-year average (2014–2015), and the time collocation is 2 h. The Rapid-Scat/WindSat results are averaged from October 2014 to August 2015 (i.e., up until the RapidScat gain anomaly), and the time collocation window is 1.5 h. Color scale is in units of m/s.

Fig. 7

Box-and-whisker figures illustrating the variability (5, 25, 75, and 95 percent quartile) of the Ku-band DPR backscatter over bare soil (left) and dense vegetation (right) at incidence angles up to 17° from nadir, using the Durden classification [62].

Fig. 8

(Top) Time series from 1998 to late 2015 for TRMM Precipitation Radar and GPM dual-frequency precipitation Radar surface backscatter cross section, and the individual periods of record for each of QuikSCAT, OSCAT-1, and RapidScat, for a location in the Amazon (2.41S 63.15W). Black is for lower zenith angles and red for higher zenith angles as indicated in each panel. (Bottom) Same as top panels, but for a location in the Congo (0.47N 21.57E).