Hydrodynamic roughness of floodplain vegetation: Airborne parameterization and field validation

Abstract

Hydrodynamic modeling is a central tool for flood risk management and lies at the base for the determination of deposition of sediment and heavy metals. In recent years, considerable effort has been made on the development of 2D and 3D hydrodynamic models that accurately simulate overbank flow patterns and predict extreme flood water levels in rivers and floodplains (e.g., Baptist et al., (2007) and Stoesser et al. (2003). In addition to surface topography (Marks and Bates, 2000), hydrodynamic roughness of the floodplain surface is the key input parameter of these models. Currently, no accurate, spatially distributed and quantitative method exists to parameterize hydrodynamic roughness of the floodplains as input for models, leading to uncertainty in flood water levels as well as deposition patterns. Vegetation roughness is dependent on vegetation structural characteristics like vegetation height and density, rigidity of the stems and the presence of leaves (Kouwen and Li, 1980) . To provide hydrodynamic modelers with reliable input, the spatial and temporal distribution of surface characteristics is needed. This requires accurate and fast monitoring methods that can cover large floodplain areas. Various remote sensing data may provide information on vegetation type, structure and dynamics, using vegetation classification. While the spatial resolution and the level of detail of the classification vary with the type of remote sensing data, in all cases vegetation classes are converted to vegetation structure, which leads to undesirable loss of within-class variation. In contrast, Airborne Laser Scanning (ALS) enables direct extraction of vegetation structural characteristics such as vegetation height, biomass, basal area, and leaf area index (Cobby et al., 2001; Lim et al., 2003). However, ALS was never tested for floodplain vegetation under leaf-off conditions representative for winter floods, which has specific problems of inundated ground surface and small herbaceous vegetation elements which cannot be detected. Any mapping strategy requires accurate field reference data for validation of remote sensing information products. Vegetation density is a difficult parameter to measure in the field, due to the presence of side branches, complex stem shapes and leaves (Dudley et al., 1998; Zehm et al., 2003). In addition, none of the current field methods generates information on the threedimensional distribution of vegetation density. Especially for herbaceous vegetation, no accurate method exists to determine density in the field. Therefore a large uncertainty remains in the input to hydrodynamic models. On the other hand, the output of roughness models is mostly calibrated in flume facilities, where high flow velocities are used, combined with steep water surface slopes and low water depths. These circumstances are not representative for flow conditions on lowland floodplains. Current in situ measurements of vegetation roughness using fixed current meters and water level meters are inadequate to measure the relevant hydrodynamic parameters such as water depth, water surface slope and the 3D flow field. This lack of calibration data further increases the uncertainty in the hydrodynamic modeling.