Abstract
Suspended sediment load in relatively smaller rivers draining agricultural areas was monitored and the suspended sediment rating curve was established using two regression analysis approaches—applying data stratification to ameliorate the prediction model equations. The sediment load data were observed for forty-five months in three rivers in an agricultural area in southern Ehime Prefecture, Japan. The data were analyzed using the power function and detransformed logarithmic function regression methods, while testing and elucidating the appropriateness and efficiency of these regression methods. Results showed that data stratification significantly improved the discharge-sediment load correlation and reduced curve-fitting errors, thereby, improving the efficiency of the derived model equation. Moreover, data stratification was found necessary in the analysis to account for nil sediment concentration observed during low flow periods. Between the two regression analysis methods, power function regression appears to have better predictive capability and, thus, more appropriate to smaller rivers. Specifically, as compared to the detransformed logarithmic function regression, power function yields models with significantly higher correlation and efficiency coefficients, as well as predicted sediment load closer to the observed sediment load.
