Aquarius Wind Speed Retrievals and Implica6ons for SMAP Ocean Vector Winds Alex Fore, Simon Yueh, Wenqing Tang, Julian Chaubell, Gregory Neumann, Akiko Hayashi, and Adam Freedman 214 California Ins6tute of Technology, Government Sponsorship acknowledged
Model Func6on Genera6on Collocate with both SSMI/S and WindSat wind speed, rain rate, and NCEP wind vectors. Require < 1 hour offset; average all data that fall within 2 km of Aquarius loca6on. Exclude non- zero rain- rates. Use SSMI/S / WindSat wind speed with NCEP wind direc6on. Have included correc6ons for significant wave height (SWH). Rain correc6ons have also been included in the model func6on.
9 Aquarius / SSMI/S Match [%] 4 Latitude 4 9 18 13 9 4 4 9 13 18 Longitude 1 2 3 4 6 7 8 9 1 Most regions have > 6 % match- ups with SSMI/S
Ocean Comparison Aquarius HH / PALSAR HH Plot of PALSAR HH GMF (black square) and our Aquarius HH GMF ( red o). - Compute wind speed PDF weighted mean ra6o of Aquarius GMF divided by PALSAR GMF. Beam 1 2 3 Mean Ra6o [db].7 -.9 A [db] Aq. A / Palsar A [db] 1 Beam 1; Aquarius vs PALSAR A; Mean Ratio [db]:.7 2 Aquarius PALSAR 2 2. 7. 1 12. 17. 2 22. 2 1. 1.. 1 1. 2. 7. 1 12. 17. 2 22. 2 1 Beam 2; Aquarius vs PALSAR A; Mean Ratio [db]:. Beam 3; Aquarius vs PALSAR A; Mean Ratio [db]:.9 A [db] Aq. A / Palsar A [db] 2 2 Aquarius PALSAR 3 2. 7. 1 12. 17. 2 22. 2 1. 1.. 1 1. 2. 7. 1 12. 17. 2 22. 2 A [db] Aq. A / Palsar A [db] 2 2 3 Aquarius PALSAR 3 2. 7. 1 12. 17. 2 22. 2 1. 1.. 1 1. 2. 7. 1 12. 17. 2 22. 2
Aquarius Ac6ve Model Func6on HH (top),vv (boiom); Beam 1 (lej), 2 (middle), and 3 (right) Note flat / non- monotonic behavior near ±9 degrees rela6ve azimuth 2 Beam 1 HH Model Sigma [db] 7 2 Beam 2 HH Model Sigma [db] 12 2 Beam 3 HH Model Sigma [db] 16 17. 8 9 17. 13 14 17. 17 18 1 19 12. 1 7. 11 12 13 14 12. 1 7. 16 17 18 19 2 12. 1 7. 2 21 22 23 24 16 21 2 2. 17 18 2. 22 23 2. 26 27 18 12 9 6 3 3 6 9 12 18 Relative Azimuth Angle [deg] 19 18 12 9 6 3 3 6 9 12 18 Relative Azimuth Angle [deg] 24 18 12 9 6 3 3 6 9 12 18 Relative Azimuth Angle [deg] 28 2 Beam 1 VV Model Sigma [db] 6 2 Beam 2 VV Model Sigma [db] 9 2 Beam 3 VV Model Sigma [db] 11 17. 7 8 17. 1 11 17. 12 13 9 12 14 12. 1 7. 1 11 12 13 14 12. 1 7. 13 14 16 17 12. 1 7. 16 17 18 19 18 2 2. 16 17 2. 19 2 2. 21 22 18 12 9 6 3 3 6 9 12 18 Relative Azimuth Angle [deg] 18 18 12 9 6 3 3 6 9 12 18 Relative Azimuth Angle [deg] 21 18 12 9 6 3 3 6 9 12 18 Relative Azimuth Angle [deg] 23
Aquarius Passive Model Func6on H emissivity (top), V emissivity (boiom); Beam 1 (lej), 2 (middle), and 3 (right) 2 Beam 1 eh.3 2 Beam 2 eh.3 2 Beam 3 eh.3 17..27 17..27 17..27.24.21.24.21.24.21 12. 1 7..18..12 12. 1 7..18..12 12. 1 7..18..12.9.6.9.6.9.6 2..3 2..3 2..3 18 12 9 6 3 3 6 9 12 18 Relative Azimuth Angle [deg] 18 12 9 6 3 3 6 9 12 18 Relative Azimuth Angle [deg] 18 12 9 6 3 3 6 9 12 18 Relative Azimuth Angle [deg] 2 Beam 1 ev.2 2 Beam 2 ev.2 2 Beam 3 ev.2 17..22 17..22 17..22.2.17.2.17.2.17 12. 1 7...12.1 12. 1 7...12.1 12. 1 7...12.1.7..7..7. 2..2 2..2 2..2 18 12 9 6 3 3 6 9 12 18 Relative Azimuth Angle [deg] 18 12 9 6 3 3 6 9 12 18 Relative Azimuth Angle [deg] 18 12 9 6 3 3 6 9 12 18 Relative Azimuth Angle [deg]
Aquarius Scaierometer- Only Wind Speed Algorithm Use σ HH and VV polariza6on. With only one azimuthal look, require addi6onal informa6on to constrain problem. Assume the wind direc6on from NCEP Retrieve best wind speed given the σ HH, σ VV, and NCEP wind direc6on. Flatness in model func6on causes significant problems at cross- wind (±9) rela6ve azimuth angles. Scaierometer- Only Cost Func6on: " J =! $ # $! gmf obs (,HH!!,HH ) obs kp HH!,HH 2 % ' &' obs (!!,VV ) "!! gmf,vv $ obs # $ kp VV!,VV 2 % ' &'
3 27 24 21 Histogram of SCAT vs SSMI/S Speed 11 1 9 8 SCAT Speed [m/s] 18 12 7 6 4 Log(Counts) 9 3 6 2 3 1 3 6 9 12 18 21 24 27 3 SSMI/S Speed [m/s]
4 Bias STD SCAT Speed compared to SSMI/S SCAT Speed Bias / STD [m/s] 3 2 1 Overall Bias:.71 Overall STD: 1.7 1 2. 7. 1 12. 17. 2 22. 2 SSMI/S Speed [m/s]
Triple- Colloca6on Results (SCAT) The scaierometer- only wind speed product has performance within.1 m/s RMS of QuikSCAT Bias Slope RMS Error SSMI 1.6767 ECMWF.27.99.8461 SCAT 2.1.1 -.2193 1.386.9337 Bias Slope RMS Error SSMI 1.6373 QuikSCAT.489.9477.93 SCAT 2.1.1.418 1.36.9448
Aquarius Combined Ac6ve / Passive Retrieval Use σ HH, σ VV, T BH, and T BV observa6ons. Requires ancillary sea- surface temperature values. Requires accurate es6mate and removal of galaxy. We include significant wave height and rain in the model func6on when ancillary data are available. Simultaneous retrieval of Ocean vector wind and sea surface salinity. F ap Combined Ac6ve/Passive Cost Func6on: 2 2 2 2 2 2 ( T T ) ( T T ) ( σ σ ) ( σ σ ) ( w w ) sin (( φ φ ) / 2) BV BVm BH BHm VV VVm HH HHm NCEP NCEP ( SSS, w, φ) = + + + + + ΔT ΔT k σ k σ Δw δ 2 2 2 2 2 2 2 2 pc VV pc HH Q = T T BV BH I = T + T BV BH
3 27 24 21 Histogram of CAP vs SSMI/S Speed 11 1 9 8 CAP Speed [m/s] 18 12 7 6 4 Log(Counts) 9 3 6 2 3 1 3 6 9 12 18 21 24 27 3 SSMI/S Speed [m/s]
4 Bias STD CAP Speed compared to SSMI/S CAP Speed Bias / STD [m/s] 3 2 1 Overall Bias:.6 Overall STD: 1.41 1 2. 7. 1 12. 17. 2 22. 2 SSMI/S Speed [m/s]
Triple- Colloca6on Results (CAP) The CAP wind speed product has performance that is about.1 m/s beier than QuikSCAT, and is nearly the same as SSMI/S Wind Speed Bias Slope RMS Error SSMI 1.728 ECMWF.227.966.87 CAP 2.1.1 -.262 1.61.7291 Wind Speed Bias Slope RMS Error SSMI 1.6398 QuikSCAT.4868.948.92 CAP 2.1.1.1 1.339.8364
Triple- Colloca6on Results (CAP) The vector triple- colloca6on results suggest that the CAP wind direc6on is not as good as that from QuikSCAT. U Component Bias Slope RMS Error ECMWF 1.7882 QuikSCAT -.146 1.9 1.273 CAP 2.1.1 -.17 1.238 1.3166 V Component Bias Slope RMS Error ECMWF 1.784 QuikSCAT -.38 1.222 1.1817 CAP 2.1.1 -.934 1.34 1.697
Poten6al of Combined Ac6ve/Passive L- band Hurricane / High winds One plarorm for Ocean vector winds, salinity and soil moisture. We are very interested in SMAP data over the ocean to see what L- band is capable of! Hurricanes L- band scaierometery
Hurricane Ka6a Aquarius CAP Speed (red) H*Wind (cyan) NCEP Wind (black)
Aquarius CAP Data Distributed via PO.DAAC Aquarius CAP product is processed at JPL Sea surface salinity Ocean wind speed Ocean wind direc6on CAP V2.8.1 L2 and L3 products available at PO.DAAC hip://podaac.jpl.nasa.gov/seasurfacesalinity Follow FTP Data Access link
Summary Model func6ons for Aquarius generated using SSMI/S, SWH, and NCEP wind direc6on Wind performance: The scaierometer- only wind speed performance is slightly worse than that for QuikSCAT The Combined ac6ve/passive wind speed performance is beier than that of QuikSCAT, however the direc6on is worse. Salinity Performance: CAP has a monthly RMS difference with ARGO that is less than the V2.1.1 data product. Including the effects of rain in the salinity model func6on seem to give addi6onal improvements. We seem to have achieved the.2 psu salinity performance that was the goal for Aquarius monthly data products. There is an overall zonal bias that sll needs to be corrected.
Water Cycle in Indian Subcon6nent Illustrated by Aquarius Data
Sea Surface Salinity Comparison with ARGO monthly maps (Sept. 211 Dec. 213) V2.1.1 RMS Difference Bias (Aquarius- ARGO)
SSS Monthly RMS difference w.r.t. ARGO between 4S- 4N, V2.1.1
SSS Monthly root- mean- square difference between 4S- 4N, excluding EPFP and ARO, V2.1.1
ADPS 2 N, 9 W CAP 8 N, 38 W CAP_RC
Aquarius SSS, Wind, and Soil Moisture Products Aquarius soil moisture from Jackson and Bindlish; SSMI/S rain from RSS