N0NSNt2 CVNt SN02 CVN

where S is the absolute component of the error, dominating close to the DL, and CV the relative component dominating at higher values. However, because S is not known for most of the PRCs the numeric value of DL was used instead. The values DL and CV were taken from Table 19.2 and apply for N0 as well as Nt. In practice only those RS values for which the result from Eq. (19.7) was lower than 0.2 (20%) were used.

Both error profiles are plotted as a function of Nt for a hypothetical PRC with a spike level of "1" in Fig. 19.4. The right-hand graph indicates the importance of sufficiently low CV. Where the measuring error exceeded 12% no values would meet the criteria. The left-hand graph shows how better sensitivity extends the range where the precision meets the assessment criterion of 0.2. When the detection limit is

Fig. 19.4. Error profiles for (N0/Nt -1) versus Nt. Left graph shows the dependence on the DL and the right graph the CV of the measurement. The horizontal line at 0.2 (20%) indicates the assessment criteria for use or not use of the corresponding RS.

10 times lower the range extends by almost a full order of magnitude. An effect similar to that observed with reduced DL can be achieved by increasing the amount of PRC spiked. The maximum amount of PRCs that can be added is governed by the linear range of the analytical method, but environmental factors should also be considered as the PRCs will deplete to the environment. During the monitoring programme, the number of PRCs used was increased, and higher concentrations were spiked. Thus as the programme continued, more PRCs fell more frequently in the application range.

19.4.4 Artefacts in sampling rates

In addition to inaccurate analytical performance, sample properties and sampling processes may also affect the reliability of the estimates of sampling rate. As for internal standards used in analytical chemistry, it is important that PRCs do not occur in environmental samples. In the left-hand graph in Fig. 19.5 the RS calculated for D10-fluoranthene is plotted versus the one for D10-pyrene. The bubble diameter represents the amount of pyrene that was collected by the sampler. The majority of the RS values show an excellent agreement with a slope not distinguishable from 1. However, the RS values from samplers where the pyrene levels, and all other PAHs, are rather high show a much lower RS for D10-pyrene. A likely explanation is that the D10-pyrene

0 0

Post a comment