The study aims at designing a multi-sensor remote sensing method to support the identification and classification of snow cover using Sentinel-1 SAR and Sentinel-2 optical imagery over Hohe Tauern National Park, Austria, for the period 2018–2025. In particular, this research develops a PCA-based approach for the fusion of multi-sensor data and proposes a set of rules for hierarchical classification of non-snow, dry snow, and wet snow cover to assess their relationships with climatic variables for the four elevation classes.
The main objectives of the thesis are:
- To integrate Sentinel-1 SAR and Sentinel-2 optical data using Principal Component Analysis for improved snow cover detection across Hohe Tauern National Park.
- To classify dry and wet snow states by combining Sentinel-2-derived spectral indices with SAR backscatter thresholds.
- Explore relationships between snow states and climate variables (precipitation and temperature), and snowmelt and wet and dry snow.
To fulfil these goals, the work is divided into theoretical and practical components, which are reflected in the structure of the following chapters.
Theoretical sub-goals
- To review the physical basis of snow detection using the Normalized Difference Snow Index (NDSI) and the SAR backscatter response to snow properties.
- To examine the principles of PCA as applied to multi-source remote sensing data fusion.
- To provide an overview of long-term snow cover and climate trends in the Austrian Alps, with particular focus on the Hohe Tauern region.
Practical sub-goals
- To acquire and pre-process 1,650 Sentinel-1 and 1,117 Sentinel-2 images covering the accumulation season (15 December – 28 February) and depletion season (15 June – 15 September) over 2018–2025.
- To construct an eight-band composite (Sentinel-2 bands B2, B3, B4, B8, B11, NDVI, and SAR channels VV and VH) and apply PCA-based fusion, with NDSI excluded from the PCA input to prevent circularity in the subsequent classification step.
- To implement a hierarchical rule-based classifier using an NDSI threshold of ≥ 0.4 combined with SAR criteria (VV < −11 dB or temporal change ΔVV < −2.0 dB) for wet snow delineation.
- To validate the classification outputs against independent sensors (MODIS and Landsat 8/9) using the Area Under the ROC Curve (AUC) metric.
- To perform spatiotemporal analysis of snow cover extent across four elevation zones (Zone I: < 1,500 m; Zone II: 1,500–2,000 m; Zone III: 2,000–3,000 m; Zone IV: > 3,000 m).
- To correlate snow state dynamics with SPARTACUS gridded climate data for precipitation and temperature, and to assess inter-annual variability over the study period.