Publication Date:
2023-12-11
Description:
GNSS Reflectometry (GNSS-R), referring to exploiting the GNSS signal of opportunity reflected off the Earth surface, has emerged as a novel remote sensing technique for monitoring geophysical parameters. The Cyclone GNSS (CYGNSS), launched on December 15th, 2016, is a constellation of eight microsatellites with cost-effected receivers, fully dedicated to the GNSS-R applications, and can track reflected signals from multiple GNSS satellites. Compared with traditional optical and radar remote sensing, GNSS-R can provide massive datasets with global coverage and improved temporal resolution, which offers unique potential for characterizing the complex Earth system.With the increase of GNSS-R observation data volume, deep learning techniques show their strong capability in retrieving ocean surface wind speed by extracting features from the Delay-Doppler Maps (DDMs). Furthermore, it is shown that deep learning models significantly improve the quality of existing GNSS-R wind speed products. The model achieves an overall RMSE of 1.31 m/s compared with the ERA5 reanalysis data and leads to an improvement of 28% in comparison to the operational retrieval algorithm based on the empirical geophysical model functions (GMFs).However, some known geophysical parameters, such as precipitation, are theorized to be impacting the reflected signals, altering the pattern of the DDMs, and consequently biasing the retrievals. The correction of such bias is not trivial because of its nonlinear dependency on various environmental and technical parameters. Therefore, we explore how deep learning-based fusion on additional precipitation data can correct the bias and further investigate the potential of deep learning models to retrieve precipitation.
Language:
English
Type:
info:eu-repo/semantics/conferenceObject
Permalink