LAND USE MAP FROM ASTER IMAGES AND WATER MASK ON MODIS IMAGES

摘要

In recent years, extreme floods occurred frequently in many parts of the world. One region, which suffers from flood damages on a regular basis, is the transboundary area of the White Koros and Black Koros rivers flowing from Romania into Hungary. For further improvement of flood management in this area, an international team was formed, with representatives of Hungary, Romania and USA and a project started on "Monitoring of Extreme Flood Events in Romania and Hungary Using Earth Observation (EO) Data" in the framework of the NATO Science for Peace (SfP) Programme. Remote sensing technologies, GIS (Geographic Information System) database merging the surface and remote sensing data, numerical hydrological predictions and the traditional means can greatly improve the management of flood hazards. The paper provides a brief overview of the satellite data processing tools needed for the project, developed at the Hungarian Meteorological Service (HMS). Our task was to produce a high-resolution land use map for the study area and develop a method for floodwater detection. The land use map will be used as background information with topographical, physiographical and hydrological parameters in the GIS database. We have derived it by applying 15-m spatial resolution ASTER images. The method for the land cover / land use mapping included two steps: (a) Geo-referencing and (b) Data classification. During post-processing, the classes were merged into 8 groups: 4 agricultural areas, and forest, water, cloud, and shadow. Urban areas could not be separated in a discrete class so they had to be classified manually. Two methods were developed for water detection using multispectral MODIS images. Both methods are threshold techniques. One is automatic, quick, but less accurate, because it may contain misdetected cloud shadow pixels. The other algorithm is a quasi-automatic method, more accurate but interactive correction is needed if the investigated area is partly covered by clouds. Water masks were produced for the period of the spring 2000 flood. The water vectors concerning successive days were visualised in a common map (and on the land use map as well) to see how the inundated area varied.