72 巻 (2016) 4 号 p. I_1039-I_1044
This study proposes an improvement to the method for evaluating suspended sediment (SS) transport from different regions of a catchment to the downstream area using X-ray fluorescence (XRF) analysis. The target area, the Oromushi River basin in Hokkaido, has SS yields among the highest in Japan due to the predominance of soils originated from volcanic rock and high contents of silt. The geochemical composition of soil and deposited sediment samples collected from 12 land groups, each with an area of 2.45 km2, was analyzed by XRF analysis. Kruskal-Wallis H test results pointed out that among the detected geochemical components by XRF only Na2O, SiO2, Al2O3, CaO, Fe2O3, MgO, P2O5, K2O, MnO and TiO2 were statistically suitable to conduct SS estimations. A modified Neural Network Analysis (NNA) was applied to estimate the SS yields from the different land groups. The results suggested that land groups G1, G3, G11 and G12 representing 33.45% of the total area produce about 64.70% of the total SS yield.