Automatic and real-time monitoring of sediment concentrations in eroded runoff is an effective way to accurately assess soil erosion process on slopes. It is assumed that low sediment concentrations could be inferred from turbidity and spectral characteristics, which are two simple and economical observation methods, and the soil properties would affect this kind of measurements. Four kinds of soil are used to obtain water samples with low sediment concentrations (0.1-10 mg/L), namely, black soil (BS), albic soil (AS), cinnamon soil (CS), and brown soil (BRS). The turbidity and spectral characteristics of the samples are measured to evaluate the relationships by calibration and validation between sediment concentrations and turbidity (method 1), sediment concentrations and reflectance with several transformations (method 2), and sediment concentrations and both indicators (method 3). The influences of soil properties on these relationships are also discussed. The linear relationship between sediment concentrations and turbidity is significant in each sample (method 1). Method 2 has a lower accuracy than method 1, in which the obtained characteristic bands and fitting models are different among samples, with a poor result for BS samples and acceptable results for the other samples. Overall, method 3 has the highest accuracy. The order of simulation accuracy from high to low is generally BS > AS > BRS > CS. The influences of soil properties are obvious and various. The effects of soil median diameter (d50) and specific surface area (SSA) on turbidity coefficients and characteristic spectral bands are not significant, but the soil organic matter (SOC) contents are. The results indicate that measurement of sediment concentration within low ranges by both turbidity and spectroscopic techniques has good accuracy in method 3, and different samples should be calibrated in application due to the effects of various soil properties.
Keywords: Characteristic bands; Soil properties; Spectral reflectance; Spectral transformation.