What Do Cave Orcs and Riverbeds Have in Common?
In this post we look at how a hand-held laser scanner is being used in the water lab to capture high-resolution elevation data of a scale model of the Waiho River.
Lincoln has a long history of using physical models of to study gravel bed river systems, and the West Coast’s Waiho River in particular. At the moment, we’ve got a postgrad, Rose Beagley, doing just that (and working with Tim Davies from Canterbury University).
For many years now, the Waiho River has been rapidly aggrading, and large floods threaten to overtop the existing stopbanks, putting a holiday park, amongst other things (like Franz Josef township, as if they didn’t have enough to worry about...), at high risk. (And Tim would happily tell you that the stopbanks have created the problem.) Rose is interested in looking at the long term effects of potential new stopbank and highway alignments and a 1:3,000 scale physical model is just the ticket. With a physical model, we can experiment with a range of different configurations without spending much money or time and no one ends up in danger. The image below shows the area of interest – Rose’s model covers the area from the true right side of the Waiho (the Franz side) and out to the furthest-out set of yellow/white broken lines:
The model itself is pretty simple. To get the braiding effects we see on gravel bed rivers, all one really needs to do is let sediment laden water run out over a sloped surface and, with enough time, braiding just naturally occurs. And the basic idea is that what happens on the scale of the Rakaia River (or the Brahmaputra for that matter) is the same as what happens on the scale of a small flow of water running over Sumner Beach, or a model in the water lab (sounds very fractal, you might say. I wouldn’t disagree). Here’s what Rose’s model looks like:
Water and sediment are mixed together at the top of the model (the far end of the image) and then flow downslope within the confines of the polystyrene walls (which simulate the natural landform boundaries of the floodplain).
Using riverbed elevation data is an effective way of comparing the results of different scenarios. In the old days, data were collected by placing a graduated beam across the model and measuring the distance from the beam down to the riverbed surface at regular intervals, say every 10 cm or so: time consuming, prone to error and very low resolution.
Several years back, we purchased a hand-held laser scanner from an outfit in town, Applied Reseach Associates NZ (ARANZ). The early Weta Workshop had used their gear to scan in scale models of things like cave orcs and then animate those models to great effect in the Lord of the Rings movies.
We thought we could do the same thing with our scale models. The system is similar to LiDAR in that it uses a laser to capture the elevations. The laser is mounted in the middle of a “wand” with two video cameras mounted on the wings.
The cameras capture the laser beam striking the surface and can take those images and determine how far away from a fixed point that location is. With the aid of radio transmitters linking all the components together, the collected points fit into a 3-D coordinate system that’s accurate to +/- 1 mm: perfect for our purposes.
To make this all work, Rose first runs the model for a period of time so that it gets to some equilibrium. Here’s a short movie of the model in action (21 seconds):
When enough time has elapsed, she turns off the water and sand (because the laser beam will be absorbed by the water rather than reflected and then starts her scanning. In the short video (42 sec) below you can see the points captured on on screen as she goes – what we get out of this step is a “cloud of points”, each with an X, Y and Z coordinate.
Here’s a close up of the actual data collected on screen. You can easily a lot of the detail of individual channels:
From the capture software she can then export the data in a format that’s easily imported into ArcGIS (a DXF file to be exact). Most of her large scans have up to 900,000 points, so these are not insignificant datasets. Once converted over to a shapefile, we’re ready to create the digital elevation models (DEMs). What we want to end up with is a raster elevation grid based on the point cloud so to go from vector points to raster grid cells we have to use the interpolation tools in ArcGIS. There are a range of possible tools and for this project Rose has been using the Natural Neighbors (sic) tool. At this scale, most methods will return very similar results so we haven’t stressed out too much on the exact tool used. So here’s an example output from a scan. On the left is an interpolated DEM. On the right is a hillshade of that grid:
The hillshade layer simulates the fall of shadows over the surface to help us better see the detail fo the terrain.
So, job done for now. After capturing the data at this point in time, Rose switches the model back on and lets it run for a few more hours and then repeats scans all over again. In this way, she builds up snapshots in time of how the riverbed is evolving. To better appreciate the differences between time steps, a simple raster calculation allows us to see which areas of the bed have eroded, which have stayed the same, and which have aggraded. And all it takes is simply subtracting one grid from another. With a bit of colour coding we can then easily see which areas of the bed have been changed and in which direction. The image below shows the sequence of steps from 1) interpolated DEM from scanned points, 2) the hillshade layer, 3) polygons used to clip out the bed area of interest, 4) the clipped grid and 5) the resulting subtraction grid (green is areas of erosion, blue are areas of no change, and orange shows areas of aggradation) :
As a next step, new stopbank configurations can be inserted, the model run and their effects evaluated and compared.
The laser scanner has been a very effective way of capturing high-resolution elevation data for these scale models. GIS provides the environment for working with the raw data to get some meaningful measures of how the system is evolving. The scanner does have its limitations though, as Rose would probably be happy to tell you about… The scanner is very sensitive to metal so the whole frame of the model is made of wood (thanks Warwick!) and any metal kept a safe distance away (a challenge in the water lab…). There are also limits as to how large an area can be scanned. With a bit of judicious equipment placement (and more than a modicum of patience), she can just scan the whole area of the model but I suspect we’re right at the limits. Nonetheless, we’re getting some good results out of all this and it’s a great example of remote sensing in action, though may not be on the scale we’re used to thinking about it.