Habitat and Land Cover Change Detection in Alpine Protected Areas: A Comparison of AI Architectures

TL;DR

This paper tackles habitat change detection in alpine protected areas by systematically comparing post-classification and direct change detection paradigms using state-of-the-art geospatial foundation models (Prithvi-EO-2.0, Clay) and CNN baselines (U-Net, ChangeViT) on a long-term HabitAlp dataset from Gesäuse National Park. It assesses performance under realistic data constraints, including multimodal inputs (RGB, CIR, LiDAR, terrain), long temporal coverage, and imbalanced, fine-grained habitat classes. Key findings show that Clay v1.0 provides consistent gains in multi-class changes and temporal robustness, while direct CD yields higher IoU for binary changes; LiDAR substantially improves semantic segmentation but yields limited gains for multi-class CD. The results underscore the potential and limitations of GFMs for operational habitat monitoring in complex alpine environments and point to future work integrating post-processing and physical constraints to enhance practical applicability.

Abstract

Rapid climate change and other disturbances in alpine ecosystems demand frequent habitat monitoring, yet manual mapping remains prohibitively expensive for the required temporal resolution. We employ deep learning for change detection using long-term alpine habitat data from Gesaeuse National Park, Austria, addressing a major gap in applying geospatial foundation models (GFMs) to complex natural environments with fuzzy class boundaries and highly imbalanced classes. We compare two paradigms: post-classification change detection (CD) versus direct CD. For post-classification CD, we evaluate GFMs Prithvi-EO-2.0 and Clay v1.0 against U-Net CNNs; for direct CD, we test the transformer ChangeViT against U-Net baselines. Using high-resolution multimodal data (RGB, NIR, LiDAR, terrain attributes) covering 4,480 documented changes over 15.3 km2, results show Clay v1.0 achieves 51% overall accuracy versus U-Net's 41% for multi-class habitat change, while both reach 67% for binary change detection. Direct CD yields superior IoU (0.53 vs 0.35) for binary but only 28% accuracy for multi-class detection. Cross-temporal evaluation reveals GFM robustness, with Clay maintaining 33% accuracy on 2020 data versus U-Net's 23%. Integrating LiDAR improves semantic segmentation from 30% to 50% accuracy. Although overall accuracies are lower than in more homogeneous landscapes, they reflect realistic performance for complex alpine habitats. Future work will integrate object-based post-processing and physical constraints to enhance applicability.

Figures (6)

• Figure 1: Map of the study area
• Figure 2: Flow chart of the experimental framework
• Figure 3: Binary change detection performance comparison across test scenarios
• Figure 4: Multi-class change detection performance comparison across test scenarios
• Figure 5: Visual comparison of change mapping between best performers (In-domain test dataset 2003-2013)