Kravchenko O. Models, methods and informational technologies of vegetation and soil state monitoring

Thesis is devoted to the solution of scientific and applied problem of the development of informational technology of vegetation and soil state monitoring. Thesis is based on the application of methods of system analysis, creation of cascade of environmental models, inverse problems solving and data assimilation techniques. We developed a method for inverse problem solving for radiative transfer models of vegetation canopy that is based on the application of Mixture Density neural networks. Using the probability neural network model we developed a new approach to the estimation of vegetation water content from remote sensing data in optical electro-magnetic range. We improved Lucas-Kennedy's and Horn-Schank's methods for optical flow estimation for the analysis of sequence of images in the case of low frequency discretization. Modified versions of algorithms were used for the wind field assessment from the sequence of images acquired by meteorological geostationary satellites. We developed an informational technology of soil state estimation based on a cascade of hydrometeorological models that consists of Numerical Weather Prediction model WRF and land surface model NOAH. Theoretical results were implemented in the land surface satellite monitoring system. The developed services conform to the ISO and OGC international standards and were registered in the GEOSS Architecture Implementation Pilot.