Determination of spatial-temporal distribution of bedload transport areas in physical model

Bedload transport is leading to erosion and deposition processes that shape the rivers morphology. Detecting areas of active bedload transport in rivers is essential for understanding morphodynamic processes within river systems. Previous research has employed various methods to determine sediment transport both in field and in laboratory settings. Field measurements are often limited due to difficult and non-predictable boundary conditions. Laboratory experiments provide opportunities to study sediment transport in a controlled environment with reproducible boundary conditions. However, it remains challenging to non-intrusively detect bedload transport areas across different temporal and spatial scales. In this context, we explore the efficacy of an image processing method to detect bedload transport areas within physical models through the water surface.

The measurements were carried out in a flume with mobile bed, approximately 30 m long and 3.6 m wide, with a longitudinal slope of 0.003. The mobile bed and the feed material consist of the same grain size distribution and represent a well-graded gravel bed river (D_m = 1.50 mm and D₉₀ = 3.06 mm). The simulated hydrographs varied between a HQ₂ flood (Q = 36.4 l/s) and a HQ₅ flood (Q = 49.4 l/s). The discharge was kept constant during the measurement period (approximately 10 minutes). Three cameras were mounted approximately 3 m above the flume covering and area of approximately 18 m x 3.6 m.

The three cameras continuously recorded pictures of the physical model at different temporal resolutions (0.033 Hz – 1 Hz) with a spatial resolution of 1 px/mm². By subtracting the intensity values of consecutive images, spatial values indicating sediment transport intensity could be obtained. The comparison between bedload transport areas identified from image processing and those discerned through visual observation reveals a strong alignment, suggesting the potential of image processing to reflect in-stream bedload transport areas accurately.

Through a combination of image processing methods, visual discrimination, and measurements in the physical model, two threshold values can be depicted. Values in subtracted images exceeding the lower threshold value indicate the initial signs of sediment transport, while values surpassing the larger threshold, signify full sediment transport.

The chosen time interval of image recording requires careful consideration, because it significantly influences the resulting threshold values. A prolonged time (30 seconds) interval with the analysis of many images facilitates the determination of average sediment transport over time, while shorter intervals (1 second), provide a snapshot insight into the distribution of bedload transport areas.

The results of this study reveal the potential of using image processing techniques in laboratory experiments to identify bedload transport areas. With further calibration, these methods hold promise for measuring bedload transport quantity and other more intricate parameters at different temporal and spatial scales.