Garbage in, garbage out
Michael Scott, SWIG, argues that, with limited or even no confidence in data from process and environmental modelling sensors, what's the point?Following the adoption of EC Directive 21, May 1994, the UK UWWTD (England & Wales) Regulations 1994 place new demands on both industry in general and, more specifically, those involved in water and waste treatment. There exists, then, some urgency in the drive to establish more confidence in, and comparability between, measurements. As Terry Long of the Environment Agency said at a recent SWIG Workshop: ³Self monitoring is very different from what went before, and now the Agency will want to assess the effectiveness of the process generating the discharge. There is a complicated set of rules to be addressed.²
One way to help establish confidence is to establish common practice and procedures via accepted standards. The Environment Agency has established a Monitoring Certification Scheme, MCERTS, to improve the quality of environmental monitoring data. MCERTS was initially introduced for Continuous Emission Monitoring Systems for chimney stacks, to provide regulators and industry with improved data quality on releases from industrial processes. WRc has a project underway, jointly funded by the Environment Agency and GAMBICA, to expand the coverage of MCERTS to include water monitoring instruments.
The basis of MCERTS is to provide the format for formal product certification of monitoring instruments. Product certification is based on laboratory and field testing, using International or European Standards where possible. In the area of continuous water monitors (CWMs) there are few standards which adequately define performance standards and conformance tests; with the exception of certain flowmeters. Therefore, there is an intention to develop the MCERTS performance standards and conformance tests to include CWMs.
Extensive work has been carried out on instrument standards in the past and work is ongoing at both European and intenational levels; notably including that of PISEG, CENELEC, the Agency, SCA, ETACS and ISO. Some individual water companies have produced instrument specifications, and others are known to have similar work in progress.
Performance standards and conformance tests need to be designed to demonstrate that the instrument is ³fit for purpose², and due care should be taken to ensure that the performance standards are based on the purpose of the measuring system not necessarily on what is achievable.
Lack of confidence
Some lack of confidence in the quality of sensor data arises from the different methods and cultures in laboratory determinations and measurements made on the process or in the environment. Operational standards must incorporate a thorough understanding of the analytical standards that are defined in regulation and the potential differences between the analytical methods used in the laboratory and those used by the CWM. This becomes particularly important when it is necessary to define conformance tests based on comparisons with a reference analytical method, as opposed to the use of a certified reference material (CRM).
The quality of data and the confidence in the data produced by on-line monitors is also due in part to the cost of ownership of the instrument. So whilst the cost and ease of operating, calibrating and validating instruments is not strictly the domain of a regulatory standard, these factors are important in obtaining good quality data over the long term and so may need to be addressed within the performance standards.
A wide variety of tools is used to generate process and environmental data and there is considerable business advantage in developing methods of combining the data from different sources, places, operators and times in a way which provides increased understanding and confidence.
At a recent SWIG Workshop, Dr Martin Lloyd, Farside Technology, emphasised that the title of his presentation, Combining Disparate Data, was important; the emphasis for his approach was one of ³combining², and not the currently fashionable ³fusion². A great deal of work has been undertaken by Oxford University on the SEVA validity codes, which describe the various aspects of a measurement and its robustness, but there was further work to be done extending these definitions. Using extended versions of SEVA it has been shown that any measurement data, regardless of source, uncertainty, time and spatial considerations, can be fitted into a single database. Visualisation tools can then be made available to cluster data for a particular site or investigation. Tools can be made available to make it easier to see whether the data set has deficiencies and where extra data might improve the information gained.
To gain confidence that a process or environmental process is properly
understood, monitored and controlled, we need to have benchmarks or
standards and we need to combine data from different sources so as to
increase our knowledge.