A new approach for detection of surface water changes based on principal component analysis of multitemporal normalized difference water index

Abstract

Normally, to detect surface water changes, water features are extracted individually using multitemporal satellite data, and then, those data are analyzed and compared to detect their changes. In this study, a new approach, called normalized difference water index-principle components (NDWI-PCs), is introduced for detecting surface water changes. This approach is based on principal component analysis (PCA) of multitemporal, normalized difference water index (NDWIMF). To develop the new model, the performances of various satellite-derived indices, including the NDWIMF (the approach introduced by McFeeters), the NDWIGao (the approach introduced by Gao), the modified NDWI (MNDWI), the water ratio index (WRI), and the normalized difference vegetation index (NDVI), were initially examined for the extraction of surface water from Landsat data. The results indicated superiority of the NDWIMF for this purpose. Accordingly, the NDWIMF was calculated from the multitemporal Landsat images, and the obtained NDWIMF images were stacked into one composite file. The PCA technique was then applied to transform the composite image into a new PCA space. Finally, the resulting NDWI-PCs were classified and analyzed to detect surface water changes. Furthermore, the applicability of the proposed NDWI-PCs approach for detecting surface water changes was evaluated and proven in comparison with some common change-detection methods, including image differencing, image thresholding, principal component analysis, and postclassification comparison.