NONLINEAR DIFFUSION AND GEODESIC PATHS FOR AUTOMATIC CHANNEL NETWORK AND GEOMORPHIC FEATURE EXTRACTION FROM LIDAR

Knowledge of the channel network and channel/floodplain morphology in a river basin is of fundamental importance for flood prediction, watershed management, ecosystem analysis, and stream restoration. The availability of high resolution topographic data now permits the direct detection of channels, rather than the estimation of likely channel locations based on topographic features (slope, drainage area, or topographic curvature).

This talk will present a geometric framework for the automatic extraction of channel network and channel morphology attributes from high-resolution digital elevation data. The approach combines nonlinear diffusion for the pre-processing of the data and geodesic minimization principles for the extraction of channels. The nonlinear filtering operation removes small scale variability and enhances features that are critical to the channel extraction. Channels are then extracted as geodesics, or curves of minimal effort, where the effort is measured based on fundamental geomorphological characteristics such as flow accumulation and iso-height contours curvature. Results are shown from the application of the methodology to several high resolution data sets of different characteristics.