What’s in a water sample? How does it change over time? How can we evaluate the change? How does the data produced represent the sample at the source? All of these are questions to which there is no fixed answer to cover all different water samples and types; but they are all equally important and need to be considered when sampling, preserving and considering which analysis should be scheduled.

Ultimately, the sample analysed should be representative of the water body at the time of sampling. However, this presents challenges analytically that cannot be simply overcome by using a more cutting edge piece of technology or a more sophisticated extraction process.

It is widely known that analyte concentrations change with time, essentially the sample becomes degraded. This may be through bacterial activity, chemical interactions, exposure to air or light, or even through the loss or chemical degradation of volatile chemicals. It is therefore critical to establish the length of time an analyte remains stable without any statistically significant changes in concentration, i.e. to establish the holding time of the sample for any particular analyte. It is essential to analyse the sample within the stability timeframe and, where this is not possible, to highlight any samples and target analytes that may have deviated from this timeframe.

Deviating samples – what’s in a name?

In 2013, following guidance issued by the Laboratory Committee of the European Co-operations for Accreditation (EA), the United Kingdom Accreditation Service (UKAS) issued clear guidelines in the form of a technical policy statement (TPS63) defining and instructing laboratories how to deal with the subject of deviating samples.

The definition given is as follows: “Deviating samples can be defined as those which are not (correctly) cared for, for example they may have exceeded their maximum holding time, lack the date and time of sampling and/or other relevant information, have not been retained at appropriate temperature, are presented in inappropriate containers/packaging, have inappropriate headspace, be denatured through heat, light or humidity, have rotted or suffered microbiologically, have become cross contaminated, been damaged in transit, have been supplied in insufficient quantity (or with incorrect dimensions) and so on.”

It is up to a laboratory to ascertain at sample receipt, if the sample submitted is likely to be deviating and at this point to inform the customer. It is important to understand that if a sample is considered to be deviating, any test methods carried out are still accredited activities – it is the sample that is deviating and not the analysis.

Obviously, considering the list of factors which can lead a sample to deviate, the responsibility for ensuring the validity of the sample submitted for analysis does not just lie with the laboratory and the customers must understand the steps that they have to take. The laboratory has little influence over the timing of sample submission, the sampling date, the temperature of the environment the samples are held in prior to receipt and indeed if any samples are submitted have headspace. The laboratory can help by advising the appropriate containers for testing and then informing the customer how they should be used and the relevant holding times.

“the laboratory has little influence over the timing of sample submission, the sampling date, the temperature of the environment the samples are held in prior to receipt and indeed if any samples are submitted have headspace”

It is important to remember that the holding time relates to the interval between sampling and beginning the analysis. This could be acidification, digestion, or solvent extraction, which provide a more stable environment for the analyte of interest.

Established holding times and references

There are numerous published sources of stability times and some good widely available documents include the USEPA methods within SW846 and ISO 5667-3:2012 “Water quality – Sampling – Part 3: Preservation and handling of water samples”. Alternatively further information can be found in the Methods for the Examination of Waters and Associated Materials (Blue Books); Standard Methods for the Examination of Water and Wastewater. These documents indicate recommended sampling containers, preservation steps and also the holding time prior to analysis or preservation. These are best practice guidelines and in some cases have extensive validation data informing the holding time that is stated; however, in other cases there is no published validation data. Indeed in some cases no holding time may even be established. With so many variables, UKAS have not specified a protocol for establishing analyte stability times.

As each laboratory will have their own set of preferred containers, preservatives and sample transportation protocols, it’s critical that each laboratory evaluate its own holding times wherever practicable. For example, a laboratory performing mercury analysis using cold vapour atomic fluorescence spectroscopy may select potassium dichromate as an “in the field” preservative, but this chemical is being phased out for use due to its harmful and toxic nature; an alternative of gold and nitric acid may be used. The impact of the preservative on the analytical processes, as well as the stabilisation effect, needs to be carefully assessed by the laboratory before putting in to practice.

“laboratories are undertaking their own in house stability trials in order to establish valid holding times for the types of samples that are submitted”

Where no documentation can be found, however, or where the documented storage conditions are not consistent with a laboratory’s facilities, then it becomes necessary for the laboratory to carry out their own stability trials. Laboratories may also wish to undertake stability trials in order to demonstrate that their storage conditions offer extended stability beyond the documented timescales. This will consequently provide their customers with a more manageable timeframe in which to deliver the samples to the laboratory, without compromising sample stability.

In-house stability trials

Laboratories are now widely undertaking their own in house stability trials in order to establish valid holding times for the types of samples that are submitted using each laboratory’s unique container types and tested by their unique accredited methods. Samples analysed in the trials must be stored in the same manner as commercial samples, i.e. normal storage conditions. This should include temperature, exposure to light, sample containers, headspace etc.

The following practical parameters require consideration:

  • Storage conditions
  • Sample matrices
  • Sample characteristics
  • Testing frequency and test period – including establishing a baseline (T0)
  • Analyte concentration (range)
  • Number of replicates
  • Interpretation of test data

Each of these is considered in detail in the following sections.

Storage conditions

A stability trial is designed to establish the length of time that samples can be held under normal storage conditions, without evidence of significant analyte degradation or change in analyte concentration.

Within this definition, the term ‘normal storage conditions’ means that the test samples are stored under identical conditions to routine samples, including bottle type, sample volume, storage temperature, preservation (where applicable) and head-space, such that as far as practicable, the test samples are managed under storage conditions encountered by routine samples within the laboratory.


The matrices to be tested should cover all typical matrices, and must include accredited matrices. Care should be taken in preparing the matrices – i.e. it may be necessary to pool the resource before taking sub-samples to store for different lengths of time, it may also be necessary to spike a large volume of the matrix before dividing the sample (see sample spiking).

“care should be taken in preparing the matrices – i.e. it may be necessary to pool the resource before taking sub-samples to store for different lengths of time”

Test intervals

The test interval set should be daily or more frequent if the literature holding time given is less than 24 hours. In this case, the testing should be carried out more frequently for the first 24 hours. Care should be taken to ensure lack of measurement at weekends is taken into account.

Establishing a baseline

The first measurement should be taken as soon as possible after sampling, this is T=0 and is the reference point for measuring any changes. In some instances it may be possible to test samples initially within an hour of sampling (e.g. tap water matrix), whereas ground water (borehole) samples are likely to be sourced from client samples and there may be a consequential delay, however, this should be minimised as much as possible.

Sample spiking

For any degradation to be monitored, it may be necessary to spike the analyte of interest at a known (low) concentration in the target matrix. For example, the full range of VOCs are unlikely to be present in samples used for stability trials, so it would be necessary to spike each matrix. The spiking should be at 10 % of the calibration range. T=0 is the concentration immediately after spiking.

Number of replicates

When deciding if an analyte concentration has changed, it is advisable that a mean concentration is determined through multiple measurements. An average concentration can be established once any outliers within the data set are removed.

As a general rule of thumb, at each time period a minimum of six replicates is recommended. Where possible, 11 or more replicates should be considered, as this provides in excess of 10 degrees of freedom for statistical interpretation of the data.

It should be noted that once the stability of an analyte has been established, this stability time is applicable, irrespective of the analytical method used, provided the prevailing storage conditions are identical. For example, if experimentation shows that zinc within a one litre glass bottle at 5°C is stable for 21 days, this is true whether the zinc is then analysed by ICP-MS, ICP-OES or AAS. Individual stability assessments against each analytical method are therefore not necessary.


The data that is generated by the protocol needs interpreting in order to identify whether a significant change in analyte concentration has occurred. In its simplest form, this may just be a maximum percentage change in analyte concentration, e.g. the analyte is outside stability once the concentration difference between T=0 and T=X is greater than 5%. For an analyte such as pH, an absolute target may be set (eg 0.1 pH units).

Other statistical tests exist that allow the comparison of mean concentrations values, such as a t-test. These are generally straightforward to apply and give meaningful interpretation from a broad data set.

It is for each laboratory to decide how they wish to interpret their data. Different critical t-values will be obtained if different numbers of replicates are used, due to the change in degrees of freedom. Likewise, while some laboratories use critical t-values based at the 95% confidence levels, other laboratories may wish to use alternatives such as those at the 90% or 99% confidence levels.

All of these factors will affect the outcome of the test. Similarly, if a simple maximum percentage change is chosen as the indicator, then the value chosen will affect the outcome of the interpretation. Overall, the laboratory must choose their own ‘critical’ values in order to assess whether a sample has remained within stability or not. Ultimately, for accredited methods, these stability trials are assessed by the accreditation body for robustness and fitness for purpose.

“for accredited methods, stability trials are assessed by the accreditation body for robustness and fitness for purpose”

What have we learnt?

When first setting about designing an experiment to establish a sample stability, it’s wise to consider what criteria the data is to be assessed against as this often informs the study design. Once this has been established, factors such as number of replicates, frequency of testing and matrix types may become more obvious, as these are often dictated by the mathematical processing at the end.

Most commercial laboratories have now chosen to evaluate some of their holding times in house rather than refer to established reference documents. This is a commendable exercise and shows greater understanding of the requirements of the laboratory to produce data which is truly representative of the water body at the time of sampling.