Automating Coastal Cliff Erosion Measurements From Large-Area Lidar Datasets In California, Usa Geomorphology

Quantifying coastal cliff erosion is critical for improved predictions of coastal change and coastal management. However, few studies have been conducted at a scale (>100 km) and resolution (~1 m) sufficient to constrain regional change. Here, we quantified cliff erosion for 866 km of the California coastline using airborne LiDAR data collected in 2009–2011 and 2016. A semi-automated method was used to map cliff faces. Negative (volume loss) and positive (volume gain) change objects were created by grouping adjacent cells using vertical and areal change thresholds and surface optical signatures. We assessed the performance of five machine learning algorithms to separate erosion and deposition from other changes within the cliff face, notably vegetation loss and growth, and found that discriminant analysis performed best. After applying the classification method to the entire cliff change dataset, the results were visually inspected for quality control, producing a final dataset comprised of 45,699 erosion and 1728 deposition objects. The net volume loss from 2009–2011 to 2016 was 1.24 × 107 m3, equivalent to an erosion rate of 2.47 m3 yr−1 per meter of coastline, and an average cliff retreat rate of 0.06 m yr−1. Eroded volumes ranged from 6.43 m3 to 7.52 × 105 m3 and fit a power-law frequency distribution (β = 0.80; r2 = 0.99). Over this study period, 7% of eroded material remained on the cliff face. Cliff retreat rates varied spatially with the highest rates in Humboldt County (0.18 m yr−1) and the lowest in Orange County (0.003 m yr−1).