A Generalized Linear Model Approach for Beach Characterization with Multi-temporal Lidar Data

Monitoring beaches and studying the processes that govern their change are critical to the future sustainability of these valued environments and the economies that depend on them. Airborne light detection and ranging (lidar) systems have revolutionized beach monitoring enabling high resolution sampling of nearshore topography over long segments of coastline quickly, accurately, and economically. Small-footprint, discrete-return systems enable beach and upland mapping with average spatial resolutions greater than 1 point per m, vertical accuracy (z) of 5– 10 cm, and horizontal accuracy (x,y) of 15-20 cm [1]. From the data, high resolution digital elevation models (DEMs) can be generated. By differencing DEMs or contours generated from repeat pass surveys, the change in volume or shoreline position for a beach can be measured respectively [2] [3]. Additionally, features can be extracted to measure changes in nearshore geomorphology [4] [5]. Data mining and pattern classification techniques offer great potential to move the analysis of lidar data beyond visual interpretation and simple (first order) measurements made from DEMs but to date have been relatively unexplored. This is particularly true for beach monitoring with lidar data because of the importance of subtle variation in topography and non-stationary processes along the beach, such as localized "hot spots" of anomalous erosion or accretion. When acquired with sufficient temporal coverage, the high spatial-resolution information in the lidar data can resolve non-stationary processes and reveal patterns in beach change otherwise unforeseen. Here, we present a systematic framework to mine morphologic features from time series lidar data acquired over a beach and characterize the joint effect of the features on the outcome of erosion. Our approach is stochastic in nature, and we use logistic regression to model the variation in morphology on probability of shoreline erosion along the beach (alongshore). Important features are methodically detected and the resultant models can be used for classification of high impact zones.