Cost-effective erosion monitoring of coastal cliffs

• Published in: Coastal engineering (Amsterdam) 2018-08, Vol.138, p.152-164
• Main Authors: Westoby, Matthew J., Lim, Michael, Hogg, Michelle, Pound, Matthew J, Dunlop, Lesley, Woodward, John
• Format: Article
• Language: English
• Subjects: Coastal erosion; Coastal monitoring; Digital surface model (DSM); Remote sensing; Rock fall; Structure-from-motion (SfM); Terrestrial laser scanning
• ISSN: 0378-3839; 1872-7379
• DOI: 10.1016/j.coastaleng.2018.04.008

Structure-from-motion with multi-view stereo (SfM-MVS) methods hold the potential for monitoring and quantifying cliff erosion to levels of accuracy and precision which rival terrestrial laser scanning (TLS) and at a fraction of the cost. We benchmark repeat SfM-MVS against TLS for quantifying rock fall frequency, volume, and cliff face erosion rates for a ∼1 km section of coastal cliffs where cliff top infrastructure is threatened by erosion. First, we address a major unknown in these techniques, the number and configuration of control points. Surveys demonstrate that a sparse configuration along the cliff base and top, at spacing equivalent to the cliff height, provides suitable accuracy at acceptable logistic time and expense. Second, we show that SfM-MVS models match equivalent TLS data to within 0.04 m, and that the correlation between intersecting TLS- and SfM-derived rock fall volumes improves markedly above a detection threshold of 0.07 m3. Rock falls below this size threshold account for ∼77.7% of detected rock falls but only 1.9% of the calculated annual eroded volume. Annual erosion rates for the 1 km cliff face as calculated by repeat TLS and SfM differencing are 0.6 × 10−2 m a−1 and 0.7 × 10−2 m a−1, respectively. Kilometre-scale patterns of cliff erosion are dominated by localised zones of high-magnitude, episodic failure that are over an order of magnitude greater than background rates. The ability of non-specialist engineers, geologists, geomorphologists and managers to rapidly capture high quality, accurate erosion data in a cost-effective manner through repeat SfM-MVS has significant potential to inform coastal managers and decision makers. To further empower coastal authorities and communities, policy frameworks must be developed to incorporate and interpret these data. •SfM-MVS input photosets of coastal cliff faces can be acquired by non-specialists using a consumer-grade digital cameras.•Placing GCPs along the cliff top and base at a spacing approximate to cliff height produces acceptable model accuracy.•Correspondence between intersecting TLS- and SfM-detected rockfall volumes improves beyond a 0.07 m3 volumetric threshold.•Kilometre-scale, TLS and SfM derived erosion rates are comparable.•Erosion patterns are spatially variable and can locally exceed the background erosion rate by over an order of magnitude.