1090 Analysis and Assimilation of CYGNSS Wind Data for Improved Tropical Convection Forecasts Xuanli Li1, Timothy J. Lang2, Brent Roberts2, John Mecikalski1 1University of Alabama in Huntsville, Huntsville, AL 2NASA Marshall Space Flight Center, Huntsville, AL 3. Model configuration and data assimilation Impact of CYGNSS data assimilation on 1. Introduction Intensity and track forecast • WRF ARW v3.8 and hybrid Ensemble 3D-Var DA system CYGNSS: The Cyclone Global Intensity MSLP Track Error • 9-km resolution, 06 UTC 24 August – 00 UTC 01 September 2017 Navigation Satellite System mission, • Observation: CYGNSS v2 beta Level 2 wind speed data with errors set as 2 m/s for windspeed 1005 400 OBS CTRL CTRL 350 launched in December 2016 1000 DA_hur_wslf < 20 m/s and 10% for windspeed > 20 m/s. DA_hur_wslf 300 ) 995 a P • DA: To assimilate the most available wind around Hurricane Harvey area. Cycles at h( P 990 )m 250 LS m k( ro 200 Instruments: 8 micro-satellite u 985 rrE T. Ashcraft 06 & 12 UTC 08/25, 12 UTC 08/27, 06 & 12 UTC 08/28, 06 & 12 UTC 08/29, min kc 150 iM 980 arT 100 observatories receive both direct and 06 & 12 UTC 08/30, 06 & 12 UTC 08/31. 975 50 reflected signals from GPS satellites 970 0 Experiments Data Assimilation CTRL No Observation: Retrieved ocean DA_wsfd CYGNSS FD wind speed Impact of CYGNSS data assimilation on surface wind with rapid revisit times in precipitation forecast DA_hur_wslf LF wind around Hurricane Harvey plus FD wind anywhere else regions of deep convection, in DA_thin_aver Thinned hur_wslf by taking average of all data within the model grid 00 UTC 2017-08-31 18 UTC 2017-08-31 particular TCs and MJO events. NEXRAD reflectivity DA_thin_max Thinned hur_wslf by using the maximum of all data within the model grid 4. Result 2. CYGNSS V2 beta data CYGNSS V2 beta L2 wind speed at 04 – 08 UTC 2017-08-25 CYGNSS V1 L2 winds featured significant errors due to a variety of reasons. CYGNSS V2 beta data (covering most of August and wsfd wslf hur_wslf thin_max thin_aver Strong winds around Harvey Stronger winds in thin_max than thin_aver CTRL September 2017) has been made available recently. The dataset includes two different GMFs, one suitable for high winds (LF) and one suitable for all other situations (FD). V2 beta winds show drastic improvement over V1, with RMSE close to the expected 2 m/s. Version 1 Version 2 beta FD Data assimilation 10-m wind increment from 1st DA cycle at 06 UTC 2017-08-25 DA_hur_wslf CTRL 10-m wind DA_thin_max DA_thin_aver DA_wsfd DA_hur_wslf Strong positive increment Larger positive increment NOAA Track-based analysis with IMERG When combined with precipitation data along the track, as well as simple filtering of the oversampled CYGNSS data, gust fronts and CYGNSS V2 beta L2 wind speed at 10 – 14 UTC 2017-08-27 Threat Scores (threshold = 2 mm/hr) other surface features near precipitation systems are readily apparent. against IMERG rain rate wsfd wslf hur_wslf thin_max thin_aver Time CTRL DA_hur_wslf Strong winds related to Harvey Stronger winds in thin_max than thin_aver 00 UTC 8/31/2018 0.16 0.18 06 UTC 8/31/2018 0.10 0.11 12 UTC 8/31/2018 0.10 0.12 18 UTC 8/31/2018 0.12 0.18 Initiate 10-m wind from different experiments after the 3rd DA cycle at 12 UTC 2017-08-27 5. Discussion and Further Works CTRL DA_wsfd DA_hur_wslf DA_thin_max DA_thin_aver • It seems promising to improve the forecast of Hurricane Harvey with assimilation of CYGNSS V2 beta L2 wind speed. • Different thinning procedure produce moderate Stronger surface wind field with assimilation of CYGNSS wind impact on data assimilation effect. CYGNSS vs. ECMWF+GDAS Since CYGNSS is mostly transparent to • A better CYGNSS data coverage (both temporal SLP and 10-m wind vector from different experiments at 12 UTC 2017-08-27 8/26/17- RMSE Bias precipitation, we hypothesize that and spatial) is needed to further improve 8/30/17 (m s-1) (m s-1) sharper gradients and increased offsets CTRL – 977 hPa DA_wsfd – 975 hPa DA_hur_wslf – 973 hPa DA_thin_max – 969 hPa DA_thin_aver – 971 hPa forecast. Assimilation of combined datasets, such between CYGNSS V2 beta winds and FD 2.7 +0.0 as IMERG rain, other scatterometer wind, rain ECMWF+GDAS winds in precipitation FD 2.0 -0.1 conventional data might be important for further norain could be due to model-unresolved cold LF 3.6 +0.7 forecast improvement. pools, gust fronts, and/or altered sea rain LF 2.8 +0.3 states associated with the convection. This research is supported by norain Differences in storm location are small NASA (NNH16ZDA001N-WEATHER) 22nd Conference on Satellite Meteorology and Oceanography, 98th AMS Annual Meeting, 7-11 January 2018, Austin, TX