Day 4

Detailed paper information

Back to list

Paper title Optical flow compared to phase correlation of UAS images used to analyse displacements of a high-alpine, fast landslide
  1. Doris Hermle TU Munich Speaker
  2. Michele Gaeta NHAZCA S.r.l.
  3. Michael Krautblatter
  4. Paolo Mazzanti “Sapienza” University of Rome
  5. Markus Keuschnig GEORESEARCH Forschungsgesellschaft mbH
Form of presentation Poster
  • C4. HAPs/UAVs
    • C4.01 Innovative UAV applications
Abstract text Remote sensing analyses of high–alpine landslides are required for future alpine safety. In critical stages of alpine landslides, both high spatial and temporal resolution optical satellite and UAS (unmanned aerial system) data can be employed, using image registration, to derive ground motion. The availability of today’s high temporal optical satellite (e.g. PlanetScope, Sentinel-2) data suggests that short-term changes can possibly be detected; however, the limitations of this data regarding qualitative, spatiotemporal, and reliable early warnings of gravitational mass movements have not yet been analysed and extensively tested.
This study investigates the effective detection and monitoring potential of PlanetScope Ortho Tiles (3.125 m, daily revisit rate) satellite imagery between 2017 and 2021. These results are compared to high accuracy UAS orthoimages (0.16 m, 7 acquisitions from 2018-2021). We applied the image registration of phase correlation (PC), a robust area–based algorithm implemented in COSI-Corr, and an intensity–based dense inverse search optical flow (DIS) algorithm performed by IRIS. We investigated mass wasting processes in a steep, glacially–eroded, high–alpine cirque, Sattelkar (2’130-2’730 m asl), Austria. It is surrounded by a headwall of granitic gneiss with a cirque infill, characterised by massive volumes of glacial and periglacial debris, rockfall deposits, and remnants of a dissolving rock glacier. Since 2003, the dynamics of these processes have increased, and between 2012-2015 rates up to 30 m/a were observed.
Both algorithms PC and DIS partially estimate false-positive ground motion, due to poor satellite image quality and imprecise image– and band co–registration. This calculated displacement from satellite data can be estimated if compared to results by UAS imagery. These results are qualitatively supported by manually traceable boulders (< 10 m) from UAS orthophotos.
Displacement calculations from UAS imagery provide knowledge about the extent and internal zones of the landslide body for both algorithms. For the very high UAS spatial resolution data, however, PC is limited to 12 m of ground motion because of decorrelation and ambiguous displacement vectors, and these are a result of excessive ground motion and surface changes. In contrast, DIS returns more coherent displacement rates with no upper displacement limit but some underestimated values. Compared to displacement rates derived from PlanetScope, there are zones of different ground motion similar to the UAS results, while at the same time there is no decorrelation. Nevertheless, for some image pairs, the signal–to–noise ratio is poor, and hot-spots can only be detected based on existing UAS results and the option of the high temporal data.
Knowledge of data potential and applicability is required to detect gravitational mass movements reliably and precisely. UAS data provides trustworthy, relative ground motion rates for moderate velocities, thus enabling us to draw conclusions regarding internal landslide processes. In contrast, satellite data returns results which cannot always be clearly delimited due to the lack of quality spatial resolution, precision, and accuracy. Nevertheless, applying optical flow to landslide displacement analysis improves the results’ validity and shows its great potential for future use. Because the robust PC returns noise when correlation is lost, while DIS does not, true displacement values of DIS are actually underestimated.