Keywords
decision support systems, integrated urban water system, priority pollutants, rule based regression model
Start Date
1-7-2010 12:00 AM
Abstract
A decision support system (DSS) for systematic control of priority pollutants (PP) sources, based on economic activities and production (release) processes in urban catchments was recently developed. One of the crucial functionalities of the DSS is evaluation of source control measures, which is based on mathematical models, used for simulating the fate of the PPs in urban catchments under different conditions. This work presents a methodology for efficiently building and integrating dynamic mathematical models into the DSS. A combination of two modelling approaches is proposed: empirical or data-driven and mechanistic or knowledge-driven. Data-driven methods, particularly those from the area of machine learning (ML), are proven to build simple and accurate models, but require a lot of measured data for their construction, which is a problem in the case of PP. Mechanistic models can overcome the data requirement problem by integrating expert domain knowledge in the model formulation. However, they tend to be too complex and computationally slow and thus, not appropriate for DSS. Within the proposed methodology a mechanistic dynamic integrated urban water system (IUWS) model for PPs is used independently of the DSS to simulate various scenarios in observed catchment. Simulated data are used by a ML algorithm for induction of rule-based regression model, which performs similarly as the mechanistic model and is integrated in the DSS. The procedure of model construction, integration, and use in the DSS is successfully illustrated based on semi-hypothetical data.
Integration of mathematical models in a decision support system for control of priority pollutants in urban catchments
A decision support system (DSS) for systematic control of priority pollutants (PP) sources, based on economic activities and production (release) processes in urban catchments was recently developed. One of the crucial functionalities of the DSS is evaluation of source control measures, which is based on mathematical models, used for simulating the fate of the PPs in urban catchments under different conditions. This work presents a methodology for efficiently building and integrating dynamic mathematical models into the DSS. A combination of two modelling approaches is proposed: empirical or data-driven and mechanistic or knowledge-driven. Data-driven methods, particularly those from the area of machine learning (ML), are proven to build simple and accurate models, but require a lot of measured data for their construction, which is a problem in the case of PP. Mechanistic models can overcome the data requirement problem by integrating expert domain knowledge in the model formulation. However, they tend to be too complex and computationally slow and thus, not appropriate for DSS. Within the proposed methodology a mechanistic dynamic integrated urban water system (IUWS) model for PPs is used independently of the DSS to simulate various scenarios in observed catchment. Simulated data are used by a ML algorithm for induction of rule-based regression model, which performs similarly as the mechanistic model and is integrated in the DSS. The procedure of model construction, integration, and use in the DSS is successfully illustrated based on semi-hypothetical data.
