Called Change Detection and Thresholding (CDAT), the technique uses satellite images before and after flooding, and performs three calculations in order to produce maps of flood extent: band maths (the difference in the absolute values of the images); thresholding (classifying pixels of the different images based on threshold criteria); and segmentation (grouping larger areas of flooding).

Stephanie Long from Florida International University worked with colleagues from NASA’s Goddard Space Flight Centre to develop the technique and tested it on images of the Chobe floodplain in the Caprivi region of Namibia.

"NASA is looking to develop a flood extent mapping technique that can be automated to process images quickly, providing near real-time flooding information to relief agencies," Long told environmentalresearchweb. "I investigated several techniques but found that they were all too complicated and tied to specific types of data. CDAT, however, allows us to use data from a variety of imagery and can be used to identify flooding in vegetation. The technique also does not require any previous knowledge of the geology or hydrology of a region."

Long and the team chose to test the technique on the Chobe floodplain because this area experiences seasonal flooding and has seen a recent renewal of severe flooding after a long dry period in the 1990s. They used the CDAT with synthetic aperture radar (SAR) images from the European Space Agency’s ENVISAT/ASAR and the Canadian Space Agency’s Radarsat-2 and found that the technique worked well, even when looking at flooding in vegetation.

"As the area has not been flooded for several years, people started moving in and the 2009 flood caught them by surprise," said Long. "Our analysis has shown that since 2009 the flood extent has increased and our maps showed good correlation with Landsat flood extent classification."

However, Long is keen to stress the importance of good reference images. "Reference images provide knowledge of the dry or non-flooded scenes, and have great weight in the analysis of the flood extents," she said. "Our reference image from 2009 was taken during a flood and so the flood extent map actually illustrates the additional flooding." She also points out that SAR imagery can be expensive or coverage may be absent, making it difficult to obtain a new reference image.

Despite these limitations, Long believes that the technique is suitable for future automation for near real-time flood extent mapping. She plans to use CDAT to investigate flooding in Namibia further, in particular the hypothesis that there is an underlying multi-decadal cycle that drives changes in seasonal flows in the Zambezi River.

Long and colleagues reported their work in Environmental Research Letters (ERL).

Related stories

• Satellite data measure surface water in Amazon basin
•  Measuring water from space
•  Insight: drought and flooding in the upper Amazon
•  Satellites show effect of 2010 drought on Amazon forests
•  Scientists correct Amazon water level gauges from space
• Satellite data will be essential to future of groundwater, flood and drought management