Airborne laser scanning for forested landslides: terrain model quality and visualization

Abstract

Mapping complex landslides under forested terrain requires an appropriate quality of digital terrain models (DTMs), which preserve small diagnostic features for landslide classification such as primary and secondary scarps, cracks, and displacement structures (flow-type and rigid-type). Optical satellite imagery, aerial photographs and synthetic aperture radar images are less effective to create reliable DTMs under tree coverage. Here, we utilized a very high density airborne laser scanning (ALS) data, with a point density of 140 points m−² for generating a high quality DTM for mapping landslides in forested terrain in the Barcelonnette region, the Southern French Alps. We quantitatively evaluated the preservation of morphological features and qualitatively assessed the visualization of ALS-derived DTMs. We presented a filter parameterization method suitable for landslide mapping and compared it with two default filters from the hierarchical robust interpolation (HRI) and one default filter from the progressive TIN densification (PTD) method. The results indicate that the vertical accuracy of the DTM derived from the landslide filter is about 0.04 m less accurate than that from the PTD filter. However, the landslide filter yields a better quality of the image for the recognition of small diagnostic features as depicted by expert image interpreters. Several DTM visualization techniques were compared for visual interpretation. The openness map visualized in a stereoscopic model reveals more morphologically relevant features for landslide mapping than the other filter products. We also analyzed the minimal point density in ALS data for landslide mapping and found that a point density of more than 6 points m−² is considered suitable for a detailed analysis of morphological features. This study illustrates the suitability of high density ALS data with an appropriate parameterization for the bare-earth extraction used for landslide identification and characterization in forested terrain.