Time series analysis of vegetation indices and radar coherence as precursors of landslide occurrence

Modern natural hazard monitoring systems, utilizing various platforms and sensors, support risk management and Early Warning Systems (EWSs). A crucial aspect of hazard prediction is detecting spatial and temporal changes in landslide areas and identifying their precursors. Despite the rapid development of modern measurement techniques, such as remote sensing, accurately monitoring landslide areas remains challenging. These challenges arise from the diversity of landslide types, the nature and density of vegetation cover, and the limitations associated with the spatial resolution of the acquired data, which may affect the detection of changes in the study areas. This study presents an analysis of optical images and radar interferograms for selected landslide areas to identify precursors and characterize landslide dynamics.

The analyses included a time series of changes in normalized vegetation indices and radar interferogram coherence. Vegetation indices such as the Normalized Difference Vegetation Index (NDVI), Green Normalized Difference Vegetation Index (GNDVI), Normalized Difference Water Index (NDWI), Normalized Difference Moisture Index (NDMI), Moisture Stress Index (MSI), and Normalized Moisture Stress Index (NMSI) were examined, along with radar image coherence. Integrating these data types enhances monitoring efficiency by combining information from different measurement techniques, providing complementary insights, and enabling a better understanding of landslide dynamics.

The conducted analysis of high-frequency measurement data revealed that normalized vegetation indices in many cases showed significant changes in landslide-prone areas before the landslides occurred. Decreases in coherence coefficient values over the same period also indicated significant changes in the analyzed areas, further confirming the occurrence of displacement in these areas. The observed correlation between the decrease in coherence and changes in vegetation index values suggests that landslide processes affected both the terrain structure and vegetation cover. Integrating optical and radar satellite data shows the potential for identifying landslide precursors and evaluating landslide activity. Such analyses can significantly support the development of landslide risk assessment tools and EWSs.