Principal component analysis and morphostructural characterization of a portion of the eastern continental shelf of Ceará, Brazil, using Landsat 5-TM images

Abstract

This study used Landsat 5 Thematic Mapper (5-TM) images in the morphostructural characterization of the shallow continental shelf of the eastern coast of Ceará (CE), in the Jaguaribe river mouth. The dam built near the river mouth to transpose the water towards Fortaleza, CE, ensures good water transparency since little of the river-carried sediment reaches the sea. The used image was captured on date and time to ensure low tide. Data using Secchi disc indicated good water transparency in the turbidity zone and seaside, coastal and marine areas. Bathymetry studies and underwater photos confirm the existence of many of the features described in this study. Digital image processing techniques have been applied in the study: colored compositions (RGB 124), 90° directional filter (band 1), and a colored composition employing the filtered band 1 and band 2 and 4. The filtered image of band 1 allowed recognizing different features when interpreted in detail scale. The principal component analysis (PCA) was used here, calculating eigenvalues and eigenvectors. When applied to bands 1 to 5 and 7, PCA generated good results in PC2, whose correlation with the blue band was 0.85 (explained cumulative variance of more than 96%). Applied to the visible bands, PCA produced very similar results in PC1, and the correlation between the blue band and PC1 was 0.86 (explained cumulative variance of 94%). These results show that band 1 is the main contributor in studies of submerged morphological features. Each process resulted in a new image, from which it was possible to produce the map of the area representing morphostructural characteristics. The use of digitally processed satellite images in the visible region greatly improves the characterization and mapping of submerged features in places with shallow and good transparency waters.