Geostatistical Processes for Flood Reporting Established

Different functions have been developed that perform different types of raster statistical analysis. In particular these functions generate histograms of pixels of a input image file that intersect either vector geometries or raster masks. The processes have been established as Web processing Services (WPS).

As a first application these processes have been combined to processing chains that allow for the flood reporting on a provincial level. The resulting products provide tabular and map data giving information on how much area in respective landcover classes have been potentially flooded hazardously.

Watermask time series from TerraSAR-X and ASAR radar sensors were completed for the rainy season 2008 and 2009 for the WISDOM test sites and the Mekong Delta using automatic watermask processing (WamaPro). The data products were integrated in the WISDOM information system and accomplish the existing time series.

The above mentioned data products allowed for the generation of inundation timeseries at different scales and for different periods. Since the periods covered appeared to be "normal" flood events (no disaster floods) this information could be used for calculating reference datasets showing the distribution of permanent flooded and seasonal flooded (e.g. due to irrigation and flooding of rice fields) areas.

The developed geostatistical processes in a first step compare new watermasks or flood predictions against the inundation references and with that identify those regions which are potentially affected by a hazardous flood. Those can then be combined e.g. with landcover information for the same region and to calculate the area of those hazardous flooded landcover areas. Such quantitative information can also be calculated at provincial or district levels and with that user customized reporting can be established.

This next figures show products of individual processes in the deveoped processing chain for flood reporting as derived by combining different remote sensing and modelling data.

Permanently or seasonal inundated areas have been derived by utilizing watermask time series derived from the ENVISAT ASAR satellite sensor. This dataset has been compared to the result of hydrologic modelling of the flood in 2004 provided by the German Research Centre for Geosciences (GFZ) resulting in a discrimination of hazardous flooded areas.

Those have been geo-statistically analyzed against a Land cover classification derived from Landsat ETM+ satellite image from 2002.

The product shows not only a map of flooded areas but also the quantitative information of flooded area by Land cover class.