بررسي و مدلسازي رطوبت خاك بدست آمده از سيگنال¬هاي بازتابي جي ان اس اس: مقايسه روش¬هاي تجربي و نيمه تجربي

Investigation an‎d modeling of soil moisture using GNSS reflected signals: comparison of empirical an‎d semi-empirical methods

مهندسي نقشه برداري

Recently, the innovative GNSS Reflectometry (GNSS-R) remote sensing technique has been increasingly used in various hydrological applications. This technique operates by utilizing signals from Global Navigation Satellite Systems (GNSS) that are reflected off the Earthʹs surface. The Cyclone Global Navigation Satellite System (CYGNSS) satellite mission is one of the operational GNSS-R missions. Despite its high potential for monitoring soil moisture an‎d vegetation status, rough surfaces significantly reduce the sensitivity of CYGNSS measurements to variations in these variables. Until now, no precise global dataset for soil surface roughness has been available to correct this negative effect. In this research, for the first time, a global dataset on the impact of surface roughness on CYGNSS observations was developed to improve the accuracy of soil moisture an‎d vegetation statue estimates derived from CYGNSS data. To achieve this, a semi-empirical method (L-MEB) was used to model the effects of surface roughness an‎d vegetation on CYGNSS observations. However, since separating the influence of surface roughness from vegetation in practice is challenging, these two factors were initially combined into a variable named VR. This new variable was then estimated using CYGNSS observations. Ultimately, the contribution of surface roughness to VR values was isolated using the Leaf Area Index (LAI) parameter an‎d assuming a linear relationship between this variable an‎d the vegetation optical depth. The eva‎luation showed that the CYGNSS mission’s surface roughness parameter was larger than the corresponding value provided by the SMAP mission. Additionally, in the second part of this dissertation, CYGNSS observations were corrected using the estimated effects of surface roughness an‎d applied to simultaneously estimate the vegetation optical depth an‎d soil moisture. For this purpose, the L-MEB model an‎d three supervised machine learning algorithms (ANN, RF, XGB) were used, an‎d their performances were compared. Our results indicate the potential of CYGNSS observations an‎d the estimated surface roughness parameter in monitoring spatial variations in SMAP soil moisture an‎d SMOS vegetation optical depth across the CONUS. Among the semi-empirical an‎d machine learning methods, the latter demonstrated superior performance, with the XGBoost (XGB) algorithm outperforming the others. Keywords: Soil Moisture(SM), GNSS Reflectometry, CYGNSS mission, Surface

9