Main navigation

Let’s not get hung up on Precision – aim for Accuracy

In an ideal world everything we measure or predict would be the true value (accurate) and utterly repeatable (precise). Many of us are familiar with…

Let’s not get hung up on Precision – aim for Accuracy

In an ideal world everything we measure or predict would be the true value (accurate) and utterly repeatable (precise).

Many of us are familiar with the simple graphical representations of precision and accuracy…

Matrix of four examples: with low or high accuracy and low or high precision

…but in our day-to-day dealings with contaminated land we must accept that our actions can influence both precision and accuracy or should be governed by such principles.

As the graphic above shows, only when we have higher accuracy do we stand a chance of finding the true value – repeated measurements, even with low precision, will allow us to hone into the  true value.  Repeated measurements with low accuracy will steer us away from the true value even if the results have high precision.  Aim for accuracy (but don’t forget precision!).

Characterising soil contamination

When examining soil on all but the smallest of micro-scales, it is clear that it is heterogeneous in nature.  Once we add in the influences of humans and former industrial uses, soil contamination can vary spatially site-wide by a huge amount.

In order to characterise soil contamination on a site, we need precise chemical results at a number of locations (and depths) in order to attain an accurate representation of contamination.

The Environment Agency’s Monitoring Certification Scheme (MCERTS) has increased the precision of laboratory analysis across the board via increased reproducibility within and between laboratories – the results are also accurate representations of the sample they receive.  But how accurately we characterise contaminant concentrations over the WHOLE site is dependent on many aspects such as when/where/how we are sampling, the methods of sample storage & transport and subsequent assessment of the results.  For example, it is important that assessments can identify hotspots as well as widespread contamination in order to be deemed accurate.

So, we have precision taken care of with sound sampling protocols and an MCERTS certified laboratory, to ensure accuracy in our ground characterisation we need to ensure sufficient samples are obtained spatially otherwise we’ll be left with some precise snapshots of the site but no overall feel for the site – and thus be unable to accurately characterise contamination side-wide.

Deriving an Assessment Criteria

As an example, if we use specialist modelling software such as CLEA v1.071 to derive an assessment criteria for a given chemical through a clearly defined exposure scenario – the software might output a concentration in soil of say “452.142397007431 mg/kg”.  This is a very precise value, and is assumed to be an accurate estimation of the threshold at which risks to human health become unacceptable for the given assumptions/ scenario.  Presenting such a concentration as a Site Specific Assessment Criteria (SSAC) would be wholly inappropriate as the level of precision is meaningless for a number of reasons:

  • Can a laboratory report such precision?
  • Should a reported concentration of 452.142399 mg/kg (exceeding SSAC) really be treated any differently to 452.142394 mg/kg (below the SSAC)?
  • The level of uncertainty and conservatism inherent in the algorithms and input assumptions can impact results by orders of magnitude.

In most instances, two significant figures is considered appropriate precision – so in our example, the SSAC would be 450 mg/kg.  This is the level of precision adopted for C4SL (Defra) and other criteria published by advisory bodies such as LQM/ CIEH and EIC/ AGS/ CL:AIRE.  Where there is large uncertainty in modelling inputs, assumptions and/or algorithms, it may be more appropriate to present SSAC to just 1 significant figure (our example SSAC would be 500 mg/kg).

When creating assessment criteria, accuracy is driven partly by the underlying algorithms and partly by the inputs justified to fuel those formulae.  Once a criterion has been derived, the author must select an appropriate precision, cognisant of the quality of the data used to create it – this is usually to 2 significant figures.  Here, the bulk of the effort is in selecting appropriate modelling techniques and justifying inputs – these drive accuracy.

Measuring water levels

For precise water elevations we must have consistency in the datum we measure from; the equipment used must also be reliable and correctly calibrated.  Similarly, for the results to be accurate, the datum must be correctly surveyed and not change over time.  These, and many more commonly overlooked issues are explored in a forthcoming 3-part blog, stay tuned….

FREE EBOOK

2021

Phase 2 Site Investigations

Do you want an insight on Phase 2 Site Investigations? This eBook has been written specifically for those new to contaminated land and provides an overview for the uninitiated of what steps you need to follow.

Reader Interactions

Leave a Reply

Your email address will not be published. Required fields are marked *