Mapping Hurricane Katrina's widespread destruction in New Orleans using multisensor data and the normalized difference change detection (NDCD) technique

Abstract

This paper introduces a new and general method for change detection based on the normalized difference change detection (NDCD) technique. A case study shows the use of the NDCD technique for flood mapping. Flood maps for the city of New Orleans (Louisiana, USA) resulting from the passage of Hurricane Katrina in 2005 were produced from the data processing of Satellite Pour l‘Observation de la Terre (SPOT)-4 High-Resolution Visible Infra-Red (HRVIR) and Landsat-5 Thematic Mapper (TM) images and the accuracies of the maps were verified using as ground truth the flood extension map of the city of New Orleans produced at the Dartmouth Flood Observatory (Dartmouth College, USA). The potentialities and performances of the NDCD technique in flood mapping were also compared to other standard change detection methods such as: (i) the near-infrared (NIR) normalized difference, (ii) unsupervised post-classification comparison, (iii) change vector analysis (CVA) and (iv) spectral–temporal minimum noise fraction (STMNF) transformation. When using the SPOT-4 HRVIR data, the NDCD technique led to better results than the other change detection methods considered here, while for the Landsat-5 TM data processing the closeness of the post-flood image to the Katrina landfall influenced the overall performances negatively. However, with respect to flood mapping in the urban area alone, which may be of major interest in most cases, the NDCD technique also performed better than all the other change detection methods considered here when using the Landsat-5 TM data.

Acknowledgements

This research study was funded partially by the Italian Ministry for University and Research, which was part of the OASIS Programme, financed by the European Commission – DG Research. We thank CNES and Spot Image for the SPOT-4 HRVIR image data. The Landsat-5 TM images and the NLCDIL were made available by the USGS through its website. We also thank Prof. G. R. Brakenridge and Dr. E. Anderson (Dartmouth College, USA) for providing the vector flood extension map of the city of New Orleans used for the accuracy assessment.