ENVI Plays Key Role in Flood Detection
Story written and submitted by Mária Putsay and Judit Kerenyi of the Hungarian Meteorological Service
Customer Challenge
During the recent years large and extreme floods occurred frequently in many parts of the world. In the past, the flood forecasts and defense-related information provided by Romania to Hungary (downstream) have been entirely based on the ground-observed data, which were mostly collected by non-automatic hydrometeorological stations. An international team needed to produce a high-resolution land use map of the study area and develop a method for floodwater detection.
Solution Achieved
During the recent years large and extreme floods occurred frequently in many parts of the world. One of the Central-Eastern European regions which regulary suffers from flood damages is the transboundary area of the White Körös and Black Körös rivers flowing from Romania into Hungary.
In the past, the flood forecasts and defense-related information provided by Romania to Hungary (downstream) have been entirely based on the ground-observed data, which were mostly collected by non-automatic hydrometeorological stations.
In view of 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 NATO Science for Peace (SfP) Programme.
As a result of this project, remote sensing technologies, a GIS (Geographic Information System) database merging the surface and remote sensing data, and numerical hydrological prediction calculations supported by traditional means have greatly improved the management of flood hazards.
The task of the Hungarian Meteorological Service was to produce a high-resolution land use map of the study area and develop a method for floodwater detection. We used Aster data to produce land use map and MODIS data for flood detection. MODIS and ASTER are both sensors on the board of Terra (EOS AM) and Aqua (EOS PM) satellites. For image processing we used the ITT (now Exelis Visual Information Solutions) ENVI software, which has proved to be an excellent tool for these purposes.
Processing of ASTER images with ENVI
Land use maps are used as background information together with topographical, physiographical and hydrological parameters in the GIS database. We have derived the land use map by classifying 15 m spatial resolution ASTER images. Both ASTER and MODIS data could be opened in ENVI as external files. In that case we immediately have calibrated data for each channels. This makes the processing faster and easier .
For the land cover classification the unsupervised classification module of the ENVI software was used. Several Aster images have been classified separately, and transformed afterwards to a common map projection using the mosaicing module of ENVI.
As examples we show an ASTER composite image and the final land use map. (see corresponding images)
Processing of MODIS images with ENVI
MODIS is a passive imaging spectroradiometer, acquiring data in 36 spectral bands. For our purpose we used Level 1B data, visible bands with 250m spatial resolution and near infrared channel data with 500m spatial resolution.
Two methods were developed for water detection. Both methods are threshold techniques. One is automatic, quick, but less accurate, it may contain misdetected cloud-shadowed pixels. The other algorithm is a semi-automatic method, more accurate but interactive correction is needed if the investigated area is partly cloudy (some cloud shadows can be detected as water).
An IDL macro was included into the ENVI software (creating a new menu line) to perform the bow-tie correction for MODIS images.
The visualization of different composite images and channels with the possibility of linking the displays provided us a very comfortable tool at the test phase.
For water detection we used (beside many other channel data) NDVI (Normalized Difference Vegetation Index). ENVI offers a special module to compute NDVI very comfortably.
The Band math module makes it easy to combine different images and/or channel data. Water mask was created using our special formulas then it was converted to vector file. The mask can be easily verified by overlaying the water vectors on the original image which can also be done very effectively with the ENVI software. By overlaying the water vectors of several successive days we can see how the inundated area varied.
Water vectors on land use map
The water vectors concerning successive days were visualized in a common map and on land use map as well in order to depict the inundated areas and their changes.
ENVI is a perfect tool for interactive image processing for many different purposes. It has several assets: it is reliable, well documented, and in our case it has proved to be a very flexible package that enabled us to monitor the floodwater.