The project investigates the application of neural networks for data analysis in remote sensing and Geographic Information Systems (GIS). It will concentrate on the extraction of classes with similar characteristics in an unsupervised mode. Multispectral and multitemporal remote sensing data, digital elevation models, topographic parameters derived from digital elevation models and combinations of data layers in a GIS, e.g. geomorphology and land use will be used. For classification based on training data critical parameters like network architecture, number of layers, completion criteria in the learning phase and error assessment will be investigated.
Neural networks will also be tested for data analysis and modelling in a GIS. Examples are the relationship between topography and climate parameters like temperature and precipitation, heavy rainfall events and resulting flooding risk, assessment of water resources in arid and semiarid areas, change detection. The procedures and techniques will be tested with data from Sweden and our international co-operation partners.
Artificial neural networks, Remote sensing, Geographic information systems (GIS), Topography