Influence of selecting secondary settling tank sub-models on the calibration of WWTP models – A global sensitivity analysis using BSM2

• WWTP model outputs are very sensitive to settling parameters.
• Sensitivity measures differ using 1st- or 2nd-order secondary settling sub-models.
• Different calibration procedure proposed using either of settling sub-models.
• 2nd-order model: calibration focuses on measurable settling parameters.
• 1st-order model: high uncertainty by adjusting non-measurable settling parameter.

This study investigates the sensitivity of wastewater treatment plant (WWTP) model performance to the selection of one-dimensional secondary settling tanks (1-D SST) models with first-order and second-order mathematical structures. We performed a global sensitivity analysis (GSA) on the benchmark simulation model No. 2 with the input uncertainty associated to the biokinetic parameters in the activated sludge model No. 1 (ASM1), a fractionation parameter in the primary clarifier, and the settling parameters in the SST model. Based on the parameter sensitivity rankings obtained in this study, the settling parameters were found to be as influential as the biokinetic parameters on the uncertainty of WWTP model predictions, particularly for biogas production and treated water quality. However, the sensitivity measures were found to be dependent on the 1-D SST models selected. Accordingly, we suggest a different optimum parameter selection for the calibration of WWTP models when either of the 1-D SST models is used. Using first-order models, the calibration should give equal importance to the adjustment of the hindered settling and slow settling parameter values. The adjusted hindered settling parameters have, however, no physical meaning, and might additionally obtain unrealistic values. In contrast, using second-order SST models, the focus of calibration should be on providing measured values for the hindered settling parameters. This approach is in close agreement with the recommendations made in the good modelling practice protocol on activated sludge modelling. Besides, based on the GSA results, adjusting the non-measurable slow settling parameter produces significantly less variance in the model outputs using the second-order model than using the first-order model.