Flood mapping through principal component analysis of multitemporal satellite imagery considering the alteration of water spectral properties due to turbidity conditions

We propose a methodology for flood mapping by remote sensing considering the alteration of the spectral response of water due to turbidity conditions using principal component analysis (PCA) of multitemporal satellite imagery limited to the visible and the near IR region of the electromagnetic spectrum. Crosta technique is applied to the resulting load matrix to select the PCs of interest, and to interpret the nature of the changes. Finally, it is proposed the use of linear unmixing, using the selected PCs as input, to classify the categories required for flood mapping. The proposed methodology is applied to mapping the flood areas after the tropical storm Manuel during September 2013 in Acapulco, Mexico. The proposed procedure is a more robust alternative to quotients or index-orientated approaches based on fixed spectral response of the water, e.g. the Normalized Difference Water Index. This approach can be useful to authorities, civil protection and other organizations dedicated to risk management during natural contingences to assess quickly the dimension of affected areas, without the expensive and complicated mobilization of recourses to the site, and to give a more efficient response