Improvements In Wind Speed Retrieval for the CYGNSS Mission

The NASA spaceborne Cyclone Global Navigation Satellite System (CYGNSS) mission is a constellation of 8 microsatellites focused on tropical cyclone (TC) inner core process studies. CYGNSS will be launched in October 2016, and will use GPS-Reflectometry (GPS-R) to measure ocean surface wind speed in all precipitating conditions, including those experienced in the TC eyewall, and with sufficient frequency to resolve genesis and rapid intensification. Here we present some recent developments in the wind speed retrieval algorithm and we illustrate an improved version of the current baseline Level 2 (L2) wind speed retrieval algorithm designed for CYGNSS. The baseline approach makes use of two different observables computed from delay-Doppler Maps (DDMs) of radar cross section, called Delay-Doppler Map Average (DDMA), and Leading Edge Slope (LES). The first is the averaged radar cross section over a delay-Doppler window around the DDM peak, while the second is the leading edge of the Integrated Delay Waveform (IDW), obtained by integrating the DDM along the Doppler dimension. The observables are calculated over a limited range of time delays and Doppler frequencies to comply with baseline spatial resolution requirements for the retrieved winds, which in the case of CYGNSS is 25 km. The relationship between the observable value and the surface winds is described by an empirical Geophysical Model Function (GMF), and the wind estimates from each observable are combined to obtain a Minimum Variance (MV) estimator, which produces improved wind retrievals compared to DDMA and LES alone. The main novelties introduced in the wind speed retrieval approach are: 1) The inclusion of the Principal Component Analysis (PCA) as a further observable into the MV estimator, which contributes to decrease the total RMS error between true and retrieved winds, 2) A “smart” time averaging of the observables based on the so-called stare-processing approach, which aims at a better mitigation of the speckle noise affecting the observables while still complying with spatial resolution requirements, and, 3) A simple mathematical modification of the observables, which stems from considerations related to the scattering mechanisms, and which linearizes the relationship between ocean surface roughness and the observables; An analysis of the performance, quality and error in the winds retrieved using this improved algorithm is presented, using GPS-R synthetic data simulated through the End-to-End Simulator (E2ES) developed for CYGNSS, and using real data from the TechDemoSat-1 (TDS-1) GPS-R experiment. The analysis highlights the improved relationship between observables and wind speed, a decreased RMS error, and a reduction in the sensitivity to Significant Wave Height (SWH) which has been observed recently in GPS-R data from TDS-1.