Modeling Probability of Blockage at Culvert Trash Screens Using Bayesian Approach

Trash screens are commonly installed at culvert entrances to prevent the ingress of debris that might otherwise become lodged within the structure. However, these can be a hazard in themselves if not cleared at an appropriate interval. There are currently no tools available to support making decisions regarding screen inspection requirements based upon potential site-by-site blockage risks. The analysis reported here was performed to address this issue. The parameter considered as key in the decision-making process was the probability of screen blockage. To determine this for any screen under consideration, a stochastic predictive model was developed using inspection records, obtained from 140 screens in Belfast, Northern Ireland, to relate blockage probabilities to seven potential drivers extracted from channel, land-use, meteorological, temporal, and social deprivation factors, employing a logistic regression approach. To allow for randomness in the data set, a Bayesian framework was adopted through which the uncertainty associated with any prediction could be reported using appropriate credible intervals. The predictive accuracy of the model was also assessed using appropriate measures and, despite documented uncertainties, was shown to be well within acceptable limits.

Language

  • English

Media Info

Subject/Index Terms

Filing Info

  • Accession Number: 01484332
  • Record Type: Publication
  • Files: TRIS, ASCE
  • Created Date: Jun 12 2013 10:46AM