| Summary: | 碩士 === 國立成功大學 === 資源工程學系 === 107 === To reduce the damage of landslide disasters caused by extremely heavy rainfall, rainfall threshold estimating is a way to predict landslides potential that may improve the early alert system. Transient Rainfall Infiltration and Grid-Based Regional Slope-Stability (TRIGRS) model (Baum et al. 2002) is widely used for rainfall-induced shallow landslide forecasting and also can be used for estimating rainfall threshold. But it was usually a hard work to obtain the required data cause the field survey is costly and time-consuming.
The dielectric constant is a measure of the electric properties of surface materials. It’s also highly dependent on the moisture content of the material and associated with radar reflectivity, which can be numerized as backscatter coefficient. This study attempts to retrieve soil water content with satellite image. By using synthetic aperture radar (SAR) from Sentinel-1 and hyperspectral from Sentinel-2 to obtain backscatter coefficient and vegetation description index, respectively. Then utilizing the water cloud model to correct the influence of vegetation scattering to estimate the surface soil water content. Other parameters needed in water cloud model mainly refered to Kam-Lon Chan (2018), which had buit a linear equation for estimating soil moisture in Taiwan, by regressing field data of soil water content and backscatter coefficient.
However, result of the estimation is not only showing a large numerical error but also lacking in the effective representation of the real situation. To figure out the reason, this study compares different conditions of model’s input to look for the factor that significantly influences the model process, and then, the terrain of the mountainous area is considered as the main reason. Limitations of the target area on the research of backscatter coefficient are be discussed and concluded.
|
|---|
