Synthetic Aperture Radar (SAR) data from the Copernicus Sentinel-1 mission were used in this practical. Sentinel-1 operates in the C-band and provides all-weather, day-and-night imaging capabilities. Due to its independence from cloud cover and illumination conditions, Sentinel-1 is particularly suitable for environmental monitoring, flood analysis, and feature extraction over mountainous regions such as Austria.
Sentinel-1 provides different processing levels:
For this practical, Level-1 GRD products were selected because the objective is feature extraction based on radar backscatter intensity rather than interferometric phase analysis.
Figure 1. Sentinel-1 GRD of the central region of Austria
Figure 2. Sentinel-1 acquisition criteria
Sentinel-1 Level-1 GRD products in Interferometric Wide Swath (IW) mode were selected for the central Austrian region. GRD data provide calibrated radar backscatter intensity suitable for feature extraction and land surface analysis without requiring interferometric phase processing. IW mode, the standard land acquisition mode, offers a 250 km swath width and approximately 10 m spatial resolution, ensuring sufficient spatial detail and coverage for regional-scale SAR analysis. Only scenes available for immediate download were selected to ensure data accessibility and processing continuity.
Figure 3: Clipping Sentinel-1 GRD
The Sentinel-1 GRD image was clipped to the defined study area in Austria using the Subset tool in SNAP. This reduced the data size and processing time by removing unnecessary regions outside the Area of Interest (AOI). After subsetting, the SAR bands (Amplitude VH) were visualized in grayscale and RGB composite format. The RGB combination enhanced the interpretation of different surface scattering characteristics, such as water, vegetation, and urban areas.
The scatter plot shows the relationship between Amplitude_VV (X-axis) and Amplitude_VH (Y-axis) for all pixels in the subset image. Each point represents one pixel, and its position indicates how strongly it reflects radar energy in both polarizations. The dense cluster in the lower-left region represents low backscatter areas (e.g., water or smooth surfaces), while points extending toward higher values indicate strong reflectors such as urban areas or rough terrain. The distribution pattern helps to understand scattering behavior and supports feature discrimination or threshold selection for classification.
Figure 4. Scatter plot of Sentinel‑1 GRD backscatter amplitudes in VV (x‑axis) and VH (y‑axis)
Principal Component Analysis (PCA) is a statistical transformation technique used to convert correlated SAR bands (such as amplitude and intensity of VV and VH) into a new set of uncorrelated variables called principal components. In SAR imagery, VV and VH bands are often correlated due to similar scattering responses. PCA reduces data redundancy, enhances contrast, and improves feature separability. The first principal component (PC1) contains the highest variance and mainly represents overall backscatter intensity, while PC2 highlights differences between polarization responses and is useful for feature discrimination. PC3 and PC4 contain progressively less variance and often represent subtle variations or noise. PCA, therefore, supports dimensionality reduction and improves classification performance.
Figure 5: Principal Component Analysis Output Result (PC1, PC2, PC3, PC4)
GLCM is a texture analysis method that measures how often pairs of pixel values occur together at a specific distance and direction. Rather than analyzing pixel intensity alone, it examines the spatial relationship between neighboring pixels.
This is particularly useful in SAR imagery because: