| In this project, we analyze the microwave emission measured by satellite channels from snow-covered ground. The estimation problem of snow characteristics is of great importance, because this estimation is necessary to gauge the supply and demand on water resources. This problem is even more important for agricultural industry and, in particular, for determining the rate of crop development. We need to use the satellite data, because the concerned involves the whole Ontario province where there are very few ground measurement stations, and hence most of the area are not covered by ground measurements. One of our objectives to complete the ground based information by satellite based information.; In this research, we apply various statistical and neural network approaches, including the cluster analysis, principal component analysis, factor analysis, pairwise plots of principal components and factor scores, neural network estimation by feed-forward models, to the above estimation problem.; As partial results of the research, we find the optimal classification of source variables and we reduce the dimension of the involved problem from 8 to 4. With pairwise plots, we estimate the ground measured variables by satellite information.; The research suggest that, for future analysis, we should apply smaller clusters, check the classification by our described neural network methods, and involve some geophysical constants and other remotely sensed data. |
