Université de Lyon, INSA de Lyon, LGCIE, 34 avenue des Arts, F-69621 Villeurbanne cedex, France.
Water Sci Technol. 2011;64(9):1926-34. doi: 10.2166/wst.2011.187.
An empirical model for TSS event mean concentrations in storm weather discharges has been derived from the analysis of data sets collected in two experimental catchments (Chassieu, separate system and Ecully, combined system) in Lyon, France. Preliminary tests have shown that the values of TSS EMCs were linked to the variable X =TP ×ADWP (TP rainfall depth, ADWP antecedent dry weather period) with two distinct behaviours under and above a threshold value of X named λ: EMCs are increasing if X < λ and are decreasing if X > λ. An empirical equation is proposed for both behaviours. A specific calibration method is used to calibrate λ while the 4 other parameters of the model are calibrated by means of the Levenberg-Marquardt algorithm. The calibration results obtained with 8 events in both sites indicate that the model calibration is satisfactory: Nash Sutcliffe coefficients are all above 0.7. Monte Carlo simulations indicate a low variability of the model parameters for both sites. The model verification with 5 events in Chassieu shows maximum levels of uncertainty of approximately 20%, equivalent to levels of uncertainty observed in the calibration phase.
从法国里昂两个实验流域（沙西厄，独立系统和埃库利，综合系统）收集的数据集中，得出了 TSS 事件平均浓度在暴雨天气排放中的经验模型。初步测试表明，TSS EMC 值与变量 X = TP × ADWP（TP 降雨量，ADWP 前期干燥天气期）有关，X 值低于和高于称为 λ 的阈值时表现出两种不同的行为：如果 X < λ，则 EMC 增加；如果 X > λ，则 EMC 减少。针对这两种行为，提出了一个经验方程。使用特定的校准方法来校准 λ，同时使用 Levenberg-Marquardt 算法来校准模型的其他 4 个参数。在两个地点的 8 次事件中进行的校准结果表明，模型校准令人满意：纳什-苏特克利夫系数均高于 0.7。蒙特卡罗模拟表明，两个地点的模型参数具有较低的可变性。在沙西厄的 5 次事件中进行的模型验证表明，最大不确定性水平约为 20%，与校准阶段观察到的不确定性水平相当。
