A Multidimensional Analysis Approach Toward Sea Cliff Erosion Forecasting

Erosion poses a significant threat to infrastructure and ecosystems on coastlines worldwide. Public infrastructure such as US 101—a critical conduit linking coastal communities and renowned destinations—can be costly to maintain due to erosion hazards. Erosion is episodic and varies both spatially and temporarily; hence, forecasting erosion patterns to identify vulnerable infrastructure is immensely challenging. This study presents an innovative Geographic Information Systems (GIS) algorithm to forecast sea cliff erosion progression utilizing imagery datasets (hereafter referred to as ‘rasters’). This approach is demonstrated for an approximately 300 m segment of sea cliffs near Spencer Creek Bridge in Beverly Beach State Park, Oregon, USA. First, Digital Elevation Model (DEM) rasters are created from multiple epochs of terrestrial lidar point clouds using two approaches: Triangular Irregular Networks (TINs) and Empirical Bayesian Kriging (EBK). These DEMs were integrated into a multidimensional raster to generate trend rasters. Based on these trend rasters, forecast DEMs were created based on several different combinations of training and forecast epochs. The forecast DEMs were evaluated against the original lidar data, to calculate residuals to determine optimal model parameters. It was revealed that four combinations warrant particular attention: EBK with harmonic and linear regression of trend rasters, and TIN with harmonic and linear regression of trend rasters. These methods demonstrate consistent decreases in residuals as the number of epochs used for interpolation increases. Under these circumstances, it is expected that the forecasting DEMs will exhibit residuals lower than 10 cm. This outcome is contingent on the condition that the time between the epochs used for prediction and the forecasted epochs does not increase.