The identifiability analysis for setting up measuring campaigns in integrated water quality modelling

Abstract

Identifiability analysis enables the quantification of the number of model parameters that can be assessed by calibration with respect to a data set. Such a methodology is based on the appraisal of sensitivity coefficients of the model parameters by means of Monte Carlo runs. By employing the Fisher Information Matrix, the methodology enables one to gain insights with respect to the number of model parameters that can be reliably assessed. The paper presents a study where identifiability analysis is used as a tool for setting up measuring campaigns for integrated water quality modelling. Particularly, by means of the identifiability analysis, the information about the location and the number of the monitoring stations in the integrated system required for assessing a specific group of model parameters were gained. The analysis has been applied to a real, partially urbanised, catchment containing two sewer systems, two wastewater treatment plants and a river. Several scenarios of measuring campaigns have been considered; each scenario was characterised by different monitoring station locations for the gathering of quantity and quality data. The results enabled us to assess the maximum number of model parameters quantifiable for each scenario i.e. for each data set. The methodology resulted to be a powerful tool for designing measuring campaign for integrated water quality modelling. Indeed, the crucial cross sections throughout the integrated wastewater system were detected optimizing both human and economic efforts in the gathering of field data. Further, a connection between the data set and the number of model parameters effectively assessable has been established leading to much more reliable model results.