Antonova, Sofia; Kääb, Andreas; Heim, Birgit; Langer, Moritz; Boike, Julia (2016): Principal Component Analysis of TerraSAR-X backscatter and coherence stacks one year (2012-2013) in the Lena River Delta, links to GeoTIFFs. PANGAEA, https://doi.org/10.1594/PANGAEA.872142, In supplement to: Antonova, S et al. (2016): Spatio-temporal variability of X-band radar backscatter and coherence over the Lena River Delta, Siberia. Remote Sensing of Environment, 182, 169-191, https://doi.org/10.1016/j.rse.2016.05.003
Always quote citation above when using data! You can download the citation in several formats below.
Principal Component Analysis (PCA) is a well-established technique in remote sensing for the visualization of multidimensional data. It reduces redundancy in multiband or multitemporal imagery, increases the signal-to-noise ratio and provides an opportunity to use multitemporal datasets for change detection. PCA transforms the axes of multidimensional data in such way that the new axes (the principal components) account for variances within the data, with the first PC accounting for the largest variance and the last PC accounting for the smallest variance.
In our study PCA of TerraSAR-X time stacks of backscatter intensity and interferometric coherence provided a good spatial overview of the essential information contained within the multiple time slices. The PC1 for both stacks showed the most common features of the contributing images and represented the means of the temporal stacks. The PC1 of the coherence stack accounted for 29% of the variance (or unique information) and mapped (i) water bodies (lakes and river), (ii) rocky outcrops, and (iii) the remaining land surfaces. The PC1 of the backscatter stack accounted for 35% of the variance and was contaminated by such effects as the presence or absence of lake ice and shadow/layover in the rocky outcrops region.
Anomalies in seasonal patterns were demonstrated by the higher PCs. The PC2 of the backscatter stack accounted for 22% of the variance and delineated water bodies. The PC3 of backscatter stack accounted for only 4% of the variance in the dataset and represented the spatial variance in river ice conditions during spring. The PC2 of coherence, which accounted for 9.5% of the variance in the coherence stack, represented the spatially variable snow conditions in spring (snowmelt to the south and stable snow cover to the north).
Median Latitude: 72.950000 * Median Longitude: 126.550000 * South-bound Latitude: 72.000000 * West-bound Longitude: 123.600000 * North-bound Latitude: 73.900000 * East-bound Longitude: 129.500000
Input dataset includes 35 repeat-pass StripMap TerraSAR-X images acquired over the central Lena River Delta, Siberia, from 3 August 2012 to 14 September 2013 every 11 days with a few gaps. TerraSAR-X is a Synthetic Aperture Radar operating in X-band (wavelength 3.1 cm, frequency 9.6 GHz). The scene size measured approximately 18 × 56 km. The orbit was in descending pass and the radar was right-looking. The acquisition incidence angle was approximately 31° and the polarization channel was HH for all used images. Local time of acquisitions was 08:34 (UTC: 22:34).
Based on this dataset, two stacks were created: (1) 35 backscatter intensity images and (2) 31 sequential 11-day interferometric coherence images. All images were geocoded to the WGS84 ellipsoid with a pixel size of 10 x 10 m in the Universal Transverse Mercator (UTM) projection Zone 52N. We applied PCA to the both stacks using the PCA tool in ArcGIS TM (ESRI). We provide here the first four PCs for each of the stacks as georeferenced rasters (.tif) as well as the text files with the covariance and correlation matrices, eigenvectors and eigenvalues.
10 data points