OCEAN SURFACE WIND FIELD RETRIEVAL FROM GNSS-R DELAY-DOPPLER MAPS USING A SEQUENTIAL ESTIMATOR AND COLLECTION OF AN AIRBORNE GNSS-R DATASET FOR FUTURE MODEL IMPROVEMENT

<p dir="ltr">Global Navigation Satellite System Reflectometry (GNSS-R) has emerged as a powerful microwave technique for observing ocean-surface wind fields with wide spatial coverage and high temporal resolution. Spaceborne GNSS-R from the Cyclone Global Navigation Satellite System (CYGNSS) constellation can provide better spatial and temporal sampling of ocean winds in the tropics than conventional polar orbiting scatterometers. Although Delay Doppler Maps (DDMs) generated from GNSS-R capture information about regions of the ocean surface spanning approximately a 90~km by 90~km area on the ocean's surface, these observations are not generally invertible except at the specular point and along the ambiguity-free line. CYGNSS wind speed products are DDM-based single-point retrievals of wind measurements on the ocean surface that utilize a limited portion of the individual DDM around the specular point over which the wind speed can be assumed to be uniform. The size of this region sets the spatial resolution.</p><p><br></p><p dir="ltr">This thesis develops an approach for the sequential processing of a sequence of DDMs, allowing information from overlapping ocean surfaces to be combined to produce more consistent and accurate estimates of surface conditions. A forward scattering model that relates the wind field to the DDM is used in an extended Kalman filter (EKF) to process the DDM sequence and generate wind speed retrievals over a uniformly gridded ocean surface covering a 90 km by 90 km swath with a 10 km by 10 km resolution. The sequential algorithm was applied to multiple storms spanning the 2019-2021 Atlantic hurricane seasons and the wind speed retrievals were compared against four independent data sources: wind measurements from the Soil Moisture Active Passive (SMAP) satellite, the Stepped Microwave Radiometer (SFMR) airborne instrument, Dropsondes, and winds from the Hurricane Weather Research and Forecasting (HWRF) model. For comparison, accuracy of the complete wind field retrieval was compared with the storm-centric v3.0 CYGNSS Level 3 product. For accuracy at specular point, the wind speed product was compared with the v3.0 CYGNSS Level 2 Young-Sea Limited Fetch (YSLF) winds. Wind speed retrieval accuracy of the sequential processor was shown to improve upon the CYGNSS products across the entire swath, especially at higher wind regimes of > 30 m/s.</p><p dir="ltr"><br></p><p dir="ltr">An effective resolution of the sequential estimator product was evaluated by fitting a half-Gaussian function to the change in the equivalent wind output during coastal crossings and was found to be approximately between 10 and 20 km, showing an improvement over the reported 25 km resolution of the standard CYGNSS L2 products (and even though CYGNSS L3 products sample wind speeds at upto $0.1^\circ$/$0.2^\circ$ latitude longitude grids, they are aggregated and averaged from the L2 winds) and suggesting that the resolution was limited by the defined spatial sampling and could be improved through defining a denser grid.</p><p><br></p><p dir="ltr">The Katzberg empirical model, defining the probability density function of surface slopes as a function of wind speed, which makes up a critical component of the forward scattering model, was found to produce more accurate retrievals than the empirical model function independently developed for the CYGNSS mission and did not show a significant saturation with wind speed. We hypothesize that this improvement is the result of the PDF model capturing the variation of the scattered power with the changing local bistatic angle.</p><p dir="ltr"><br></p><p dir="ltr">In addition to this, the thesis also documents an airborne GNSS-R experiment conducted with the help of the National Oceanic and Atmospheric Administration (NOAA) to collect reflected GNSS signals from the ocean surface. This dataset provides an independent platform for generating DDMs and for future improvement of scattering-model parameterizations that relate ocean-surface slope statistics to wind speed. Together, the CYGNSS sequential retrieval and the NOAA experiment establish a foundation for a unified GNSS-R framework that bridges satellite and airborne observations for next-generation ocean wind remote sensing.</p>