Sensitivity and uncertainty analysis of sediment rating equation coefficients using the Monte-Carlo simulation (Case study: Zoshk-Abardeh watershed, Shandiz)

The sediment load estimation is essential for watershed management and soil conservation strategies. The sediment rating curve is the most common approach for estimating the sediment load when the observed sediment records are not available. With regard to the measurement errors and the limitation of available data, the sediment rating curve has a degree of uncertainty which should be accounted. Thus, in this work an stochastic Monte-Carlo simulation approach was employed to measure the degree of uncertainty and to analyze the sensitivity of sediment rating equation to changing the coefficients "a" and "b" at the Zoshk hydrometric station. Results of the Monte-Carlo simulation using 5000 samplings showed that in the hydrologic regimes of normal high flow, normal low flow and normal, the contribution of the coefficient "a" to variance is significantly higher than the contribution of the coefficient "b" to sediment load variance. However, this condition is conversely for the hydrologic regime of maximum flow. The sensitivity analysis confirmed a higher sensitivity of the sediment rating equation to the coefficient "a" than that to the coefficient "b" in the hydrologic regimes of normal high flow, normal low flow and normal. However, in maximum flow, the sediment rating equation showed a higher non-linear and more regulated sensitivity to the changes of the coefficient "b" than "a". In all flow regimes, the coefficient "a" has a wider 95% uncertainty band, in comparison with that for the coefficient "b", which establishes a higher level of uncertainty. The sediment rating equation showed a high degree of uncertainty in the hydrologic regime of minimum flow (lower than 0.3 m3.s-1). In this condition, all observed sediment records place outside the 95% and 97.5% of the uncertainty band. However, in medium and high flows, most of the observed sediment records are bracketed by the uncertainty band of 95% and 97.5%.