River bathymetry from conventional LiDAR using water surface returns

LiDAR-based DEMs are almost essential when it comes to high resolution floodplain modelling. Conventional LiDAR does not penetrate water bodies and additional bathymetric measurements are necessary to complete any parts of a DEM which were under water at the time of the LiDAR sortie. Bathymetric data collection is time-consuming, expensive and it requires careful processing to correctly interpolate between sounded points or cross-sections and to seamlessly integrate bathymetric data into a LiDAR-based DEM. Consequently, accuracy of underwater portions of a DEM is usually inferior to that of dry areas. When areas that are underwater at LiDAR collection time represent only a small portion of a flood domain, modelers often estimate a representative depth for these areas. By using LiDAR returns from the surface of flowing water, we investigate calculations to predict the level of the underlying river bed using a hydraulic model. To do this, there must be sufficient LiDAR returns from the water surface, the river flow must be known at the time of the LiDAR data collection and information on the roughness of the river bed is required. An assumed bed topography is used with the hydraulic model to predict the water surface elevation. The error between predicted water surface elevation and LiDAR-measured water surface elevation is then used iteratively to adjust the assumed bed position until the correct water surface elevation is calculated. The technique was applied to a 1024 by 512 meter reach of the gravel-bed Waiau River on the Canterbury Plains on the eastern side of the South Island of New Zealand. The river flow was 17.96 m(3)/s during the LiDAR data collection and about an 80 m width of the channel fairway was underwater. Comprehensive bathymetry measurements made with an echo sounder and RTK GPS were used to evaluate the model-predicted bed levels. The iterative hydraulic model technique predicted the bed level with high accuracy in locations where the Froude number was > 0.3. For these conditions 41% of the predicted bed levels lay within +/- 10 cm and 85% of the predicted levels lay within +/- 30 cm of the GPS-measured bed level. The Froude number influence may reflect the fact that the technique can not work if there is no water movement. The technique should not be applied to ponds, lakes or very slow flowing water.