Identification of dynamic models for horizontal subsurface constructed wetlands

The current trend in horizontal subsurface constructed wetlands (HSSCW) modelling advocates structures of increasing complexity, which however have produced a limited improvement in the understanding of their internal functioning or in the reliable estimation of their parameters. Following a different approach, this paper proposes the combination of a set of simple structures and a robust identification method to approximate the dispersed flow and pollution reduction dynamics. The models are based on combinations of series/parallel CSTRs of unequal volumes in series with a plug-flow reactor. After introducing the model structures, their identifiability is assessed with a method based on approximate parameter confidence regions based on either the Fisher information matrix (FIM) or the Hessian matrix. Their relative agreement or disagreement yields information about model structural robustness and parameter identifiability. To demonstrate the method, the proposed models are calibrated with data sets from several constructed wetlands, with widely differing hydraulics and pollution removal characteristics. The identification method can assist in the selection of the best combination of hydraulics and kinetics to obtain robust and yet simple models for HSSCW. In particular, it is shown that the estimated volumes can be used for the wetland design and the pollution reduction kinetics are reassessed in terms of identifiability, providing a guideline to decide between first order and Monod kinetics.