Berber R, Yuceer M, Karadurmus E
Department of Chemical Engineering, Ankara University, Tandogan, Turkey.
Water Sci Technol. 2009;59(3):515-21. doi: 10.2166/wst.2009.878.
Water quality models have relatively large number of parameters, which need to be estimated against observed data through a non-trivial task that is associated with substantial difficulties. This work involves a systematic model calibration and validation study for river water quality. The model considered was composed of dynamic mass balances for eleven pollution constituents, stemming from QUAL2E water quality model by considering a river segment as a series of continuous stirred-tank reactors (CSTRs). Parameter identifiability was analyzed from the perspective of sensitivity measure and collinearity index, which indicated that 8 parameters would fall within the identifiability range. The model parameters were then estimated by an integration based optimization algorithm coupled with sequential quadratic programming. Dynamic field data consisting of major pollutant concentrations were collected from sampling stations along Yesilirmak River around the city of Amasya in Turkey, and compared with model predictions. The calibrated model responses were in good agreement with the observed river water quality data, and this indicated that the suggested procedure provided an effective means for reliable estimation of model parameters and dynamic simulation for river streams.
水质模型有相对较多的参数，需要通过一项与诸多困难相关的重要任务，依据观测数据进行估算。这项工作涉及对河流水质进行系统的模型校准和验证研究。所考虑的模型由11种污染成分的动态质量平衡组成，源于QUAL2E水质模型，即将一段河流视为一系列连续搅拌釜式反应器（CSTR）。从灵敏度度量和共线性指数的角度分析了参数的可识别性，结果表明有8个参数在可识别范围内。然后通过基于积分的优化算法结合序列二次规划来估算模型参数。从土耳其阿马西亚市附近的耶希尔河上的采样站收集了包含主要污染物浓度的动态现场数据，并与模型预测结果进行了比较。校准后的模型响应与观测到的河流水质数据高度吻合，这表明所建议的程序为可靠估算模型参数和对河流进行动态模拟提供了一种有效方法。
