Using Satellite Images for Inverting Surface Soil Moisture over Alishan Forest Area Taiwan

  • 肖 晶晶

Student thesis: Doctoral Thesis

Abstract

This research takes part in the Alishan area which aims to explore the use of Sentinel-1 SAR(Synthetic Aperture Radar) data and Sentinel-2 optical data to retrieve forest surface soil moisture The primary method used is the water-cloud Model which is a semi-empirical vegetation backscatter model based on the radiation transmission model proposed by Attema and Ulaby The second method is the gradient boosting decision tree algorithm proposed by Friedman in 2001 In the first step I developed 12 linear models for retrieving forest surface soil moisture through the water-cloud model Next through adding some other factors that may affect soil moisture such as elevation difference slope aspect solar radiation sunshine duration temperature obtained by interpolation surface roughness relief degree of land surface I used the gradient boosting decision tree algorithm to correct and improve soil moisture estimated by 12 linear models Finally we obtained a soil moisture map with high accuracy and a spatial resolution of 10 m in the study area The result shows that the most suitable parameters for retrieving soil moisture in the study area are the soil-adjusted vegetation index (SAVI) in the remote sensing index and the vv polarization backscatter coefficient obtained from the Sentinel-1 SAR data And through the gradient boosting decision tree algorithm the accuracy of all 12 linear models of the RMSE (Root Mean Square Error )value has been improved indicating the effectiveness of the research method Key words: Sentinel Data; Water-Cloud Model; Gradient Boosting Decision Tree; Forest Surface Soil Moisture
Date of Award2020
Original languageEnglish
SupervisorTing-Li Lin (Supervisor)

Cite this

'