River Bathymetry Estimation

GeoFabrics supports the inclusion of river bathymetry data as part of the DEM generation process. The river bathymetry data can either be measured observations, or estimated using the RiverBathymetryGenerator processor module class. In both cases, bathymetry point data and a polygon defining the extents of the river are required. These are specified by the river_bathymetry and river_polygons options in the instruction file as described in the Instruction file contents. This page focuses on the estimation of river bathymetry data.

Estimating River Bathymetry

River bathymetry is estimated in several stages for a given channel.

1. The channel slope and width are estimates from LiDAR derived DEMs.
2. The channel reach level flow and friction estimates are derived from River Environment Classification (REC).
3. The channel depth is calculated from the river geometry and flow characteristics
4. The channel bed elevation is calculated based on the depth and water elevation.

A basic validation of the approach has been preformed as described at Bathymetry estimate validation

Pipeline for estimating and incorporated river bathymetry

Depth estimation equations

Two equations based on different hydrologic assumptions as used to calculate depth given river width, slope, flow and in one case bed friction.

Uniform flow theory

This is based on the approach published in Estimating River Channel Bathymetry in Large Scale Flood Inundation Models (or on sharepoint for those with access) by Neal et al.

This approach assumes:

• Uniform flow
• Bed friction and slope are equal
• Hydraulic radius = depth
• River flow dominates (i.e. no tidal)

Hydraulic geometry

This is based on the critique published in Comment on ‘‘Flow resistance equations withoutexplicit estimation of the resistance coefficient forcoarse-grained rivers’’ by Rau ́lLo ́pez, JavierBarraga ́n, and M. A ́ngels Colomer (or on sharepoint for those with access) by Rupp and Smart.

This approach assumes:

• The dominant processes are alluvial (i.e. not tidal or confined within flood defences)
• The published coefficients match the river (i.e. are applicable across all NZ rivers).

Flow and friction estimation

• Flow is estimated from the statistically based flood statistics web-map. A 1 in 2.33 year flood is used as that causing channel forming flow. This is estimated from the provided flood returns using a Gumbel GFD.
• Friction is estimated for use in the uniform flow theory approach using USGS tables that maps river type to n. We need to know the type of each reach. We do have a classification of bed sediment that could be used a proxy, it is a weighted average of proportional cover of bed sediment using categories: 1–mud; 2–sand; 3–fine gravel; 4–coarse gravel; 5–cobble; 6–boulder; 7–bedrock, predicted from a boosted regression tree model. In both cases, the values at each reach are mapped from REC version 1 (which is used by the flood statistics web-map, and has the reach type classification) to REC version 2.

River width and slope estimation

The river width and slope are estimated from 1 metre vegetation and ground LiDAR derived DEM along the river channel defined by REC 2.

The overall pipeline

This is done in three main steps:

1. Select channel to estimate from REC and create a ground and vegetation DEMs along the channel
2. Improve the REC defined channel centre-line alignment by detecting the actual channel location from sampled cross-section elevations along the channel.
3. Estimate the width and slope along from sampled cross-section elevations along the aligned channel.

The width estimation algorithm

In both the alignment (step 2) and width (step 3) steps the channel width is estimated from cross sections using a threshold detection algorithm.

Fixed threshold detection

In the initial width estimates as part of the alignment step are achieved with a fixed threshold detection algorithm. The sampled elevations along each cross section are traversed around the channel centre to detect adjacent samples that are either below the threshold over the estimated water surface elevation or NaN (as NaN can indicate water with no returns). The longest of these adjacent sections is then taken as the channel width and location for that cross section. This is done from the ground cross section elevations. The vegetation elevations are only used to ensure the NaN values are not just under a tree.

Variable threshold detection

In the final width estimation as part of the - more to come..