Efficient job list creation for long-term statistical modelling of combined sewer overflows

Abstract

The modelling of urban drainage systems is an important aspect of their design process and long-term statistical modelling using historical rain series is commonly used. The objective of this study is to determine whether logistic regression models that use rainfall event statistics can be a viable alternative to create job lists with fewer extraneous events. Two methods are used to develop a regression model; both use iterative stepwise algorithms to select the rain variables to include and both perform similarly. The resulting model is able to capture ∼90% of the relevant events with ∼50% fewer jobs compared to the reference job list. The results suggest that there is no right threshold to use, but instead this methodology facilitates balancing the number of jobs with the desired level of precision of the results. In all cases, it is possible to greatly decrease the number of jobs that need to be run. The methodology works relatively well on different nodes in the system, though node characteristics appear to impact the amount of CSOs captured. HIGHLIGHTS Job lists for LTS simulations can be targeted using logistic regression models.; The logistic regression models draws on statistical data from the rainfall input data.; The number of jobs can be drastically reduced while mostly maintaining the relevant jobs.; The proposed methodology can be used to run more efficient LTS simulations.;

Keywords

• cso
• logistic regression models
• lts simulations
• sewer system
• urban water management