Assessing the assessment
Martin Allitt, of Sussex-based consultancy Richard Allitt Asssociates, looks at modern techniques for checking quality of flow survey data
Historically, surveys carried out in relation to wastewater network modelling have focused on catchment-wide studies, and many of the standard specifications for surveys were written when surveys were carried out across whole catchments; this was especially the case with manhole surveys.
Nowadays, most parts of the UK have been surveyed to a fairly high standard in most urban areas, though with odd pockets of incomplete or missing data.
The requirements for surveys nowadays are predominantly for the infill variety. The specifications for quality control checks are in many cases not appropriate or practical for such infill surveys, and as a consequence quality control checks are generally no longer undertaken in the field. This places a greater burden on the modeller to assess whether the quality of the data is adequate - a potentially difficult task from behind a desk or computer. This article focuses on modern techniques for checking the quality of data from flow surveys, and is based on a recently presented paper entitled Modern Techniques for Checking Data Quality, a copy of which can be found at www.raaltd.co.uk.
Getting it right
Flow surveys are expensive and time-consuming, and it is important that the data collected is of the highest quality; sites giving poor-quality data need to be recognised at the earliest opportunity so monitors can be re-located to alternative sites. The current requirement for flow surveys is to follow the Model Contract Document for Short-term Sewer Flow Surveys1, which states that in-sewer checks should be carried out as part of the data retrieval visits and also at the end of the survey, to check that the monitors are recording correctly.
It is becoming increasingly important that the modeller takes a greater role in understanding and checking the quality of the data during the currency of the survey. This involves carrying out more external quality checks on the data. The quality of the flow survey can be broken down in to three sections: the quality of the installation site; the quality of the data at the site; and the location and spacing of raingauges.
To stand a chance of collecting good-quality flow data, it is important that flow monitors are installed at good sites. A pre-installation survey is therefore essential, to confirm that the sites chosen in the office will allow good-quality data to be collected.
Before the pre-installation survey, time should be spent in the office to understand the catchment and to identify the key areas of the network where flow monitors need to be installed. It is important that the sites chosen in the office are ones that look as if they are suitable. For example they should meet the following criteria:
n no change in pipe size from incoming to outgoing,
It is important to have a preferred site for each monitor, and ideally at least two alternative sites which are also suitable on paper, before going into the field and carrying out the pre-installation survey. Thus, if for some reason the preferred site turns out to be unusable or unsatisfactory, alternative sites will already have been identified.
Once the flow monitors have been installed it is advisable to monitor the data from each monitor on a weekly basis to check that good-quality data is being recorded.
Scattergraphs are a good way of checking that the data being recorded is of good quality and that there is nothing downstream or upstream in the network which is having an adverse effect on the flow regime at the monitoring site. It is worth plotting scattergraphs each week if the flow survey contractor can send you the flow and depth data on a weekly basis. Scattergraphs can easily be plotted using standard spreadsheet programs but it is worthwhile plotting the data with a log-log scale. You can see from the example in Figure 2 that it is worth plotting the measured data along with the theoretical flow-depth curve (calculated from the Colebrook-White formula). Ideally the data recorded by the installed flow monitor should follow the theoretical line (as in Figure 2).
If the downstream pipeline and the upstream pipeline are the same size but have slightly different gradients, the capacities will be slightly different. It is then worthwhile plotting the theoretical line for both pipes; ideally the data should lie between both theoretical lines.
Scattergraphs are useful as they can show lots of information. The scattergraph in Figure 3 shows the data recorded at one site for two verification events. It indicates that as the depth increases initially there is a linear increase in the flow in the pipeline and that the recorded data fits between the theoretical lines for the upstream pipeline and the downstream pipeline. As the depth increases further the scattergraph shows that the flow no longer increases linearly and that there is virtually no increase in flow. This particular situation was created by a blockage in the pipeline downstream causing about 80% of the pipe cross-sectional area to be lost. After this was identified from the scattergraph a CCTV survey was made, followed by jetting. This was done whilst the flow survey was underway and, because of the early intervention, subsequent rainfall events (another two were recorded) in this sewer were unrestricted.
Another scattergraph example can be seen in Figure 4. This flow monitor was located just upstream of a CSO and shows four stages of flow. At point A there is a linear increase in depth as the flow increases, this matches well to the theoretical Colebrook-White relationship. In this part of the scattergraph, up to a depth of about 150mm, there are free-flow conditions. The next part of the scattergraph shows the downstream network start to surcharge and cause backing up to the monitor site; at point B it can be seen that flow has hardly changed, while depth has increased substantially. Once the surcharge level reaches the level of the weir at the CSO, at a depth of about 340mm, there is a rapid increase in flow for very little increase in depth - this can be seen in part C of the scattergraph. Finally, once the overflow has operated, it is clear that the capacity of the overflow pipe (rather than the weir) starts to restrict the flow, and at point D the data follows the line for the overflow pipe rather than the main pipe.
Figure 4 shows clearly that just because the data does not follow exactly the line of the theoretical depth-flow curve, the data is not automatically of poor quality. The important thing is to understand what the scattergraphs are showing and to act accordingly. In the case of Figure 3 the appropriate action was a CCTV and jetting exercise. In the case of Figure 4 the appropriate action was to ensure that the CSO was modelled correctly and also that the overflow pipe as well as the weir was modelled.
Scattergraphs are very useful tools in understanding the flow in the sewers, and with experience, much information can be gained about the flow regimes in the sewerage network up- and downstream of the monitor location.
Reading the rain
Rainfall data is the hardest on which to carry out quality checks. The Model Contract Document for Short Term Sewer Flow Surveys2 states the contractor should ensure each raingauge is operating correctly at each visit and that the data logger records tips in both directions. These checks only ascertain that the raingauge itself is recording correctly.
The WaPUG code of practice3 sets out acceptable criteria in terms of rainfall depths, rainfall intensities and variation between adjacent raingauges.
It is getting harder these days to find suitable locations at which raingauges can be installed; as a result it is becoming more difficult to achieve an adequate coverage across a catchment. Therefore it is important that the quality of data recorded is of a high standard.
There are only a limited number of checks which a modeller can undertake to check the quality and consistency of rainfall data. One very useful method is to plot graphs of cumulative rainfall both weekly and ongoing for the whole survey period. This should be done throughout the flow survey period so that an early indication can be gained if one raingauge is recording less than the others. Figure 5 shows an example of three raingauges, but raingauge 2 was installed one week earlier than the other two, and also raingauge 1 was inoperative for a period of nearly one week due to the filter becoming blocked. Figure 6 shows the same information but with the first week for raingauge 2 removed - it can now be seen how close the measurement is at raingauge 2 and raingauge 3. This technique can become cumbersome with very large surveys, and in those situations it is probably best to plot a series of graphs with only adjacent raingauges shown on each graph. There is scope for better methods of quality checking of rainfall data, and maybe there is a role for weather radar in this.
1. Model Contract Document for Short Term Sewer Flow Surveys. WRc, December 1993.
3. The WaPUG Code of Practice for the Hydraulic Modelling of Sewer Systems Version 3.001. 2002.
© Faversham House Group Ltd 2005. edie news articles may be copied or forwarded for individual use only. No other reproduction or distribution is permitted without prior written consent.