Grain size mapping in shallow rivers using spectral information: a lab spectroradiometry perspective

Every individual attribute of a riverine environment defines the overall spectral signature to be observed by an optical sensor. The spectral characteristic of riverbed is influenced not only by the type but also the roughness of substrates. Motivated by this assumption, potential of optical imagery for mapping grain size of shallow rivers (< 1 m deep) is examined in this research. The previous studies concerned with grain size mapping are all built upon the texture analysis of exposed bed material using very high resolution (i.e. cm resolution) imagery. However, the application of texturebased techniques is limited to very low altitude sensors (e.g. UAVs) to ensure the sufficient spatial resolution. Moreover, these techniques are applicable only in the presence of exposed substrates along the river channel. To address these drawbacks, this study examines the effectiveness of spectral information to make distinction among grain sizes for submerged substrates. Spectroscopic experiments are performed in controlled condition of a hydraulic lab. The spectra are collected over a water flume in a range of water depths and bottoms with several grain sizes. A spectral convolution is performed to match the spectra to WorldView-2 spectral bands. The material type of substrates is considered the same for all the experiments with only variable roughness/size of grains. The spectra observed over dry beds revealed that the brightness/reflectance increases with the grain size across all the spectral bands. Based on this finding, the above-water spectra over a river channel are simulated considering different grain sizes in the bottom. A water column correction method is then used to retrieve the bottom reflectances. Then the inferred bottom reflectances are clustered to segregate among grain sizes. The results indicate high potential of the spectral approach for clustering grain sizes (overall accuracy of 92%) which opens up some horizons for mapping this valuable attribute of rivers using remotely sensed data.