Paper No. 2
Presentation Time: 9:15 AM


KUCHTA, Matthew, Department of Physics, University of Wisconsin - Stout, 410 10th Avenue East, 126F JHSW, Menomonie, WI 54751 and ZIMMERMAN, Todd, Department of Physics, University of Wisconsin - Stout, 410 10th Avenue East, 126A JHSW, Menomonie, WI 54751,

Mobile bed stream models are effective tools for research and education, but obtaining quantitative 3D data is challenging. The Microsoft® Kinect™ sensor is a low-cost (ca. $100 USD), 3D infrared scanner that collects both RGB color and distance to sensor data (640 by 480 pixel field of view) at 30 Hz. Previous investigators have demonstrated effective geoscience field applications of the Kinect scanner. We have adapted these methods for measuring experiments using a stream table (Em2 built by Little River Research and Design). The point cloud data generated by the Kinect was imported into a Python script where 60 seconds worth of scan data (ca. 1,800 measurements) are combined into an averaged scan to reduce measurement errors. Preliminary results demonstrate the accuracy of the Kinect scanner is sufficient to track millimeter-scale changes of the streambed. Comparing scans from different time stops in the experiment highlights areas of aggradation and degredation as well as sediment contribution from stream banks and hillslopes. Using the Kinect, we are able to show how changes in sediment supply, discharge, and base level affect the evolution of a model stream system and how the model compares to the lower reach of the Red Cedar River in Dunn County, Wisconsin (a real-world fluvial system).

The current version of the Kinect sensor is limited by a marked decrease in accuracy as scanning distance increases beyond one meter. Combined with the sensor’s field of view, this limits individual scans to a one-meter wide section of the model. We also compared the Kinect methods to other techniques, including profile measurement with a laser level and photometric analyses. Stream table long profiles are similar to those obtained from real-world mapping, but are limited to two dimensions and don’t account for lateral changes in bed elevation. Photometric analyses such as time-lapse photography can effectively show landscape evolution over the entire time frame of the model run, but optical distortions hinder quantitative measurements, particularly of volume. Although our methods were developed specifically for the 2-meter long Em2 stream table with color-coded ground melamine plastic media, these methods are easily adapted to other mobile bed geomodels.