Potentials and limitations of Radarsat SAR in estimating Snow Water Equivalent (SWE) in the Great Lakes region of the United States
The basic hypothesis of this research is that the snow cover characteristics influences the underlying soil temperature which has a certain effect on the SAR backscattering since the dielectric properties of the soil are highly related to the soil temperature. Indeed, the backscattering ratio of the winter and summer images along with land cover characteristics are used in an Artificial Neural Network (ANN) algorithm to estimate snow water equivalent (SWE). To take the land cover effect into account, Normalized Difference Vegetation Index (NDVI) was added to the ANN as an input. The preliminary results have shown that the addition of vegetation-related information to the neural network model has a positive effect on the SWE retrieval accuracy.
The ground truth data are obtained from NOHRSC (National Operational Hydrologic Remote Sensing Center) SNODAS (Snow Data Assimilation System) products. SNODAS data are produced using snow gauge measurements along with Gamma radiation measurements run in a physical model. Comparing the estimated results to the SWE obtained from SNODAS, it is shown that in high latitudes there is a better potential to retrieve SWE than low latitudes.