Investigating methods of stream planform identification

Stream planforms are used to map scientific measurements, estimate volumetric discharge, and model stream flow. Changes in these planforms can be used to quantify erosion and water level fluctuations. This research investigated five cost-effective methods of identifying stream planforms: (1) consumer-grade digital camera GPS (2) multi-view stereo 3D scene reconstruction (using Microsoft Photosynth (TM)) (3) a cross-sectional measurement approach (4) a triangulation-based measurement approach and (5) the 'square method' - a novel photogrammetric procedure which involved floating a large wooden square in the stream, photographing the square and banks from numerous angles and then using the square to correct for perspective and extract the outline (using custom post-processing software). Data for each of the five methods was collected at Blackhawk Creek in Davenport, Iowa. Additionally we placed 30 control points near the banks of the stream and measured 88 lengths between these control points. We measured or calculated the locations of these control points with each of our five methods and calculated the average percent error associated with each method using the predicted control point locations. The effectiveness of each method was evaluated in terms of accuracy, affordability, environmental intrusiveness, and ease of use. The camera equipped with GPS proved to be a very ineffective method due to an extremely high level of error, 289%. The 3D point cloud extracted from Photosynth was missing markers for many of the control points, so the error calculation (which yielded 11.7%) could only be based on five of the 88 lengths and is thus highly uncertain. The two non-camera methods (cross-sectional and triangulation measurements) resulted in low percent error (2.04% and 1.31% respectively) relative to the control point lengths, but these methods were very time consuming, exhausting, and only provided low resolution outlines. High resolution data collection would require even smaller increments, drastically increasing field work time. These two non-photo methods also disturb the sediments in the stream, which may be undesirable. Finally, the 'square method' provided a high resolution outline with only a 1.51% error and was relatively easy to complete in the field. It did, however, require a great-deal of manual post-processing time. Despite this caveat, we recommend the square method as an inexpensive procedure for obtaining fine resolution planform data with minimal disturbance to the stream environment. Stream planforms and calculated control point locations with respective average percent errors