Assessment of the CFOSAT scatterometer backscatter and wind quality Artist’s view of the CFOSAT satellite. Wenming Lin, Zhixiong Wang , Yijun He (NUIST) Xiaolong Dong, Risheng Yun, Xing-ou Xu, Di Zhu (NSSC) Marcos Portabella (ICM) Shuyan Lang, Jianqiang Liu (NSOAS)

3. Expected performances Backscatter Wind 4. Conclusions and outlook Outline 1. Introduction CFOSAT SCAT overview 2. Simulation L1B simulator L2 simulator 3. Expected performances Backscatter Wind 4. Conclusions and outlook 1. Introduction – overview of system defintion

1. Introduction Parameters Values Transmitted power  120 W Bandwidth 0.5 MHz Pulse duration 1.35 ms Pulse repetition rate (PRF) 75 Hz (VV) 75 Hz (HH) Antenna peak gain 32.0 dB (VV) 31.8 dB (HH) One-way 3 dB beam width Range: 14.5; Azimuth: 1.2 (VV) Range: 15.0; Azimuth: 1.1 (HH) Antenna spinning rate Default: 3.4 revolutions per minute (rpm) High spinning rate: 3.74 rpm Low spinning rate: 3.06 rpm Power consumption < 200 W Instrument mass < 70 kg Swath 1000 km The CFOSAT RFSCAT is a Ku-band (central frequency of 13.256 GHz) real aperture radar, with one vertically (V) polarized fan beam and one horizontally (H) polarized fan beam sweeping the Earth’s surface at medium incidence angles (~26 - 51). the VV and HH beams are indeed offset by 180 (see the lower panel), which is set to maximize the azimuth diversity of the observations acquired by the two different polarizations, with the objective of improving the retrieved wind quality Three-dimensional model view of the antenna architecture of the CFOSAT SCAT (upper panel); schematic illustration of the beams’ configuration (lower panel).

1. Introduction SCAT footprint and range-gated slices The onboard processing is implemented to sample the backscatter signal in regular ground range gates (slices) SCAT footprint and range-gated slices Schematic of the illumination geometry formed by two beams

2. Simulation The L1B simulator is a radar simulator developed on the basis of the defined instrument parameters, and is used to assess the precision of the RFSCAT backscatter measurements. This simulator consists of two relevant components, namely the orbit propagator and the signal (X-factor) simulator.

Output：L1B product in NetCDF format Input： Two-Line Element (TLE) set of satellite orbit CFOSAT SCAT instrumental parameters 3-hourly ECMWF forecasts as ‘true’ winds (spatially and temporally interpolated into SCAT slices’ position) Output：L1B product in NetCDF format Slices’ geometries Backscatters (0 ; GMF: NSCAT-4) Signal-to-Noise Ratio (SNR)、slice 0 precision (Kp)； Orbit parameters 0 flags Optional outputs： Power X factor Spatial response function (SRF) for research

Illustrations of SRFs for VV beam Forward looking 0 side- view 90 Side-view 270 Afterward looking 180  Offside view 225 Illustrations of SRFs for VV beam Descending pass Contour lines of 1 dB bin Black curves – footprint Color curves – slices (one in fine is shown)

2. Simulation “True” WVC winds Wind variability of 1.5 m/s

3. Expected performance Slice WVC 12.5 km WVC 25 km Two-dimensional histogram of the simulated 0 (VV beam) versus the relative azimuth angle g for  = 40 and wt= 4 m/s, and for slice 0 (a); WVC-mean 0 with grid resolution of 12.5 km (b); and WVC-mean 0 with grid resolution of 25 km (c)

3. Expected performance Slice WVC 12.5 km WVC 25 km Kp The slice Kp value is generally high (> 20%). Nonetheless, the WVC Kp is very much reduced after aggregating slices with similar incidence and azimuth angles into a specific WVC view. For the WVC-mean sigma0s with  < 45, the 12.5-km WVC-mean Kp is lower than 20% for wt > 5 m/s, and lower than 10% for wt > 10 m/s. This value is further reduced by at least 2% for the 25-km WVC. More interestingly, the Kp dependency on incidence angle is reduced after aggregating slices to WVC. Wind speed (m/s) Contour plots of the normalized measurement error (Kp, VV beam) as a function of wind speed (x-axis) and incidence angle (y-axis), and for a slice Kp (a); WVC-mean Kp with grid resolution of 12.5 km (b); and WVC-mean Kp with grid resolution of 25 km (c). Note that a contour spacing of 0.05 is used.

3. Expected performance “true wind field” 12.5 km

3. Expected performance Retrieved winds 12.5 km Retrieved winds 25 km Maximum solutions=4

3. Expected performance Retrieved winds 12.5 km Retrieved winds 25 km Since MSS provides solutions in the direction of the mean flow over the nadir regions, a more spatially consistent wind field is achieved after AR. Retrieved winds 12.5 km Retrieved winds 25 km Maximum solutions=144

3. Expected performance Inversion residual (MLE) @12.5 km This slide illustrates the MLE value of the retrieved wind field at 25 km and 12.5 km resolution, respectively. In the vicinity of the low pressure center, the area of large MLE values is larger at the lower grid resolution (Fig. 9a). This is due to the fact that when aggregating slices into an averaged 0 over dynamic (variable) areas, the larger the area, the less representative this 0 is of a mean “uniform” WVC-wind. Inversion residual (MLE) @12.5 km Inversion residual (MLE) @25 km Maximum solutions=144

3. Expected performance Resolution Geophysical noise (m/s) SD values of wind speed (m/s) dir () u v 25 km 0.0 0.34 13.6 1.15 1.21 0.5 0.35 1.14 1.22 1.0 0.37 12. 5km 0.43 15.1 1.23 1.33 15.2 1.24 1.34 0.45 15.3 1.25 Expected SCAT wind quality at different grid resolutions and geophysical noise conditions. The SD values of the difference between retrieved winds and input (“true”) ECMWF winds are estimated. Since the simulated 0 values are assumed to be well calibrated, the biases of wind speed, direction, u and v components are all close to zero (not shown). As expected, the wind quality of the 12.5-km WVCs is a little worse than that of the 25-km WVCs. Moreover, the statistical scores slightly degrade with increasing geophysical noise, for both 25-km and 12.5-km WVCs.    [m/s] SCAT scale Model scale ASCAT 0.34  0.001 0.80  0.003 ECMWF 1.25  0.007 1.02  0.005 Triple collocation results of collocated buoy-ASCAT-ECMWF data

3. Expected performance

4. Conclusions and outlooks SCAT 10-km slices Kp is generally high (above 20%); the WVC Kp however is generally fine (about 5%-20%), depending on the surface wind speed conditions. Geophysical noise (wind variability) impacts on the slice measurement, particularly for wind speeds below 5 m/s. However, geophysical noise does not substantially degrade the measurement precision of WVC 0s. Provided that the backscatter measurements are under rain-free condition and well calibrated, the expected wind quality is fully compliant with the mission requirements, i.e., a wind speed SD error < 2.0m/s and a wind direction SD error <20, as proposed by NSOAS. The L2 processing with multi-solution scheme and 2DVAR is too sensitive to the background wind quality, suggesting that further L2 processing developments are required at the postlaunch stage. The innovative observation geometry of SCAT is not fully exploited. Inversion, quality control, … Sea surface temperature effects on the Ku-band radar backscatter are not negligible. Improved GMF (large incidence range) with SST effect? This is because many slices are averaged to get a specific WVC 0 value, and the geophysical noise among those slices are assumed to be uncorrelated.

Thank You

2 Instrument characteristics Slice SNR VV, =25 VV, =30 VV, =35 Signal-to-noise ratio VV, =40 VV, =45 VV, =50

2 Instrument characteristics Slice SNR HH, =25 HH, =30 HH, =35 HH, =40 HH, =45 HH, =50