OR/17/009 Summary
| Tye, A M, Kirkwood, C, Dearden, R, Rawlins, B G, Lark, R M, Lawley, R L, Entwistle, D, and Mee, K. 2017. Environmental factors influencing pipe failures. British Geological Survey Internal Report, OR/17/009. |
This report details work carried out under NERC grants NE/M008339/1 and NE/NO13026/1 which were collaborations between the British Geological Survey and Yorkshire Water, with an additional knowledge transfer component involving Scottish Water and Dŵr Cymru Welsh Water. The work examines whether models developed using environmental, topographical and geohazard information could complement existing management tools, and increase the understanding as to how pipe networks of different materials interact with their broader environment. This can be seen as a first step in identifying ways in which greater resilience could be built into pipe networks.
Broad groups of pipe materials were examined, these being the cast iron and plastic pipe networks for clean water and the concrete and clay networks for waste water. Modelling was undertaken using the spatial model package ‘SPATSTAT’ in the ‘R’ statistical platform. Initially Null Models were established to predict the ‘expected number of bursts’ per unit area (100 x 100 m cells) for each pipe type (construction material) and water type (clean or waste) of interest based on the density (length) of pipe present in each cell. Single covariates, identified as being important in pipe failure obtained through an Expert Elicitation process with YW, were tested against the null model. Those significant covariates (P<0.05) were then included in a sequential model where covariates were added and kept. In the second part of the modelling exercise, a selection of additional environmental and geohazard information were added to the expert elicitation model and a similar modelling exercise undertaken.
Overall, the modelling exercise demonstrated that for the YW region, covariates such as roads, water source, and number of dwellings (as a likely proxy for pressure changes and use) were of greater influence on the pipe network than many geohazard factors. This may be due to the YW region being fairly benign to typical factors that are recognised as damaging pipe networks such as shrink swell clays and compressible ground. For both of these covariates the highest class of hazard wasn’t found in the YW region. However, soluble (gypsum bearing rocks) ground were identified as a problem for the concrete waste water network and the potential presence of sulphide was important for the cast iron network. In addition, outputs showed that pipe networks associated with the coal measures and some areas of lacustrine clay appeared to have greater than expected pipe failures, which the covariates used in the model could not explain. Results were presented as heat maps, by combining the outputs from each 100 x 100 m cell cell using the model coefficients from the final sequential model for each pipe network. This allows a spatial assessment of the overall environmental, geological and topographical ‘hostility’ towards the pipe network. In addition, individual heat maps for each significant (P<0.05) covariate were created, placed on a single scale so that the intensity of each environmental, topographical and geohazard covariate could be compared across the YW region. End of project meetings were held between BGS and the water companies to disseminate results and discuss the benefits and possibilities of the modelling approach used.