Deep Neural Networks with 3D Point Clouds for Empirical Friction Measurements in Hydrodynamic Flood Models

TL;DR

This work tackles the uncertainty in friction factor estimation for hydrodynamic flood models by predicting Manning's roughness $n$ from 3D point clouds. A laboratory-trained deep neural network, based on PointNet and augmented with subsampling and mixing of point clouds, maps surface topography to $n$ values derived from Manning’s equation $n=rac{R^{2/3} S^{1/2}}{V}$ with $R=rac{h w}{2h+w}$. The DNN is trained on flume data and applied to real 3DEP lidar data to drive 1D, 2D, and 1D/2D coupled models across regulatory and extreme storm conditions, including Hurricane Harvey, showing improved agreement with gauges, imagery, and claims relative to land-cover-based FFs. The approach reduces uncertainty, enables high-resolution, repeatable FF measurements from readily-available point clouds, and demonstrates notable improvements in flood predictions and model validation. The work also provides open-source tools and a clear pathway to standardize FF estimation across diverse flood regimes and geographies.

Abstract

Friction is one of the cruxes of hydrodynamic modeling; flood conditions are highly sensitive to the Friction Factors (FFs) used to calculate momentum losses. However, empirical FFs are challenging to measure because they require laboratory experiments. Flood models often rely on surrogate observations (such as land use) to estimate FFs, introducing uncertainty. This research presents a laboratory-trained Deep Neural Network (DNN), trained using flume experiments with data augmentation techniques, to measure Manning's n based on Point Cloud data. The DNN was deployed on real-world lidar Point Clouds to directly measure Manning's n under regulatory and extreme storm events, showing improved prediction capabilities in both 1D and 2D hydrodynamic models. For 1D models, the lidar values decreased differences with regulatory models for in-channel water depth when compared to land cover values. For 1D/2D coupled models, the lidar values produced better agreement with flood extents measured from airborne imagery, while better matching flood insurance claim data for Hurricane Harvey. In both 1D and 1D/2D coupled models, lidar resulted in better agreement with validation gauges. For these reasons, the lidar measurements of Manning's n were found to improve both regulatory models and forecasts for extreme storm events, while simultaneously providing a pathway to standardize the measurement of FFs. Changing FFs significantly affected fluvial and pluvial flood models, while surge flooding was generally unaffected. Downstream flow conditions were found to change the importance of FFs to fluvial models, advancing the literature of friction in flood models. This research introduces a reliable, repeatable, and readily-accessible avenue to measure high-resolution FFs based on 3D point clouds, improving flood prediction, and removing uncertainty from hydrodynamic modeling.

Figures (15)

• Figure 1: Isometric and side view of weir system used for uniform flow distribution. A dual weir was implemented to ensure near-zero velocity at hydraulic jump, and to remove flow turbulence from the final weir.
• Figure 2: Experimental setup to measure Manning's n. As shown, the smooth surface had low water depths, resulting in difficulties measuring water velocity with the Channel Velocity meter. Therefore, PTV was used. Figure modified from AGUPres
• Figure 3: Implementation of PointNet architecture. Conv1D stands for a 1D convolution with a Batch Normalization layer, and linear for linear layers with rectified linear unit nonlinearities. Figure adapted from AGUPres.
• Figure 4: Relevant watersheds for hydrodynamic model setups in Harris County.
• Figure 5: 1D Model Setup to study Riverine Flooding in Spring Creek.