Composite proportional sample

Figure 3.3 Results of duplicate composite proportional samples, showing variations in some...

The composite proportional sample (COMP) procedure is a method that actually aims to collect samples of each draw of water used for dietetic purposes at the monitored tap. This is, therefore, the only procedure that takes into account all variations within one week. Thus, the composite proportional sample should be considered to be the only really representative sample, i.e. representative for the average weekly intake of the consumers sharing the tested tap during the test week. 

The composite proportional sample is taken with a consumer operated device fitted to the drinking water tap, which splits off a small constant proportion ( 5 %) of every volume of water drawn for dietetic purposes during one week. Figure 3.5 gives a schematic view of a proportional sampling device. 

A consumer-operated device, attached to the tap, collects the composite proportional sample. The composite proportional sampling device is connected to the kitchen tap through a coupling nut or a hose clip. In practice, this might cause problems, as some taps are shaped in a w that makes it very difficult to attach the sampling device (e.g. mixer and spr -type taps). The gap between device and tap should be as short as possible. The device should be coimected horizontally. 

By definition, the proportional sampling procedure scores 100 % for the criteria of representativeness and reproducibility. However, this applies only when consumers operate the proportional device correctly. To test whether the consiuner has used the device correctly, the volume of the composite proportional sample can be checked. In some test areas the volmne of the composite proportional sample was noted and used to calculate the average daily water consumption per person. Figure 3.7 shows the water consumption distribution for these test areas, and Figure 3.8 shows the... 

Figure 3.7 Water consumption distribution based on composite proportional sampling...

Figure 3.7 shows that the modal average daily consumption as calculated from the volumes of composite proportional samples was about 2 litres,. This value corresponds with data from other sources. For some households an average daily... 

Both the modal daily consumption per occupant and the dependency on the household size are in accordance with data from other studies. This indieates that, in general, the composite proportional sampling deviee was operated correetly in the field test. There is some doubt over the apparently high eonsumption in one-person properties. These high values might stem from real high water use (e.g. tea, visitors, elderly people string at home all d ), but also from improper use of the deviee. Nevertheless no data have been excluded from the assessment on the basis of this eriterion. 

Random daytime sampling (RDT) Figure 3.9 shows the results of the random daytime samples in all properties, compared to the composite proportional sample. Figure 3.9 indicates that the relationship between RDT and COMP showed considerable variation. In general, RDT seems to overestimate COMP (slope >1), though for some individual properties RDT underestimates COMP. As a result of the variations the correlation coefficient, R, is 0.61. 

In summary, all tested sampling procedures show a linear relationship with the composite proportional sample, albeit that the correlation is poor. RDT and the 30-minute stagnation samples, 30MS1, 30MS2 and 30MSA show the best correlation R = 0.5-0.6). RDT however, generally seems to overestimate COMP (slope > 1), whereas 30MS somewhat rmderestimates COMP (slope > 1). FF clearly shows the poorest relationship to COMP = 0.29 and the slope is 0.57. 

Ratio between the tested protocols and the composite proportional sample... 

The ratio between the result obtained by the tested protocol and the result of the composite proportional sample is another w to erqrress the representativeness of the sampling method. Before showing the results, the approach used is explained ... 

Ratio between PROP and COMP Ideally, the ratio between the lead values obtained by the tested protocol (PROT) and the composite proportional sample (COMP) should be unity, or at least constant over a wide concentration range. To determine the variation in the ratio between PROT and COMP, the average ratio is calculated over both the combined test areas (total, all results) and the individual test areas (A-K). The 95 % prediction range of the average ratio is ... 

The 90-% prediction range of the tested protocols The predietion range of a protocol reflects the accuracy of that protocol in predicting the value of the composite proportional sample. Therefore, the 90-% prediction range is applied to assess the representativeness of the tested protocols. 

Summarizing, 30MSA gives the most aeemate reflection of the weekly average concentration, as determined by composite proportional sampling. 

Even though test areas and properties were selected where the lead was expected to be significant (69 % of the tested properties had lead service pipes and/or lead plumbing), only 44 % of the composite proportional samples exceeded the PV of 10 igl. About 20 % of all composite proportional samples exceeded the intermediate PV of 25 igf[ and 19 % exceeded the value of 50 i. ... 

Flushing the tap before use is an effective measme to reduce considerably the lead concentration at the consmner s tap. However, flushing does not guarantee that the lead level will be below 10 This study shows that in 50% of properties where the composite proportional sample exceeded 10 (xg/1, the fully flushed sample was less than 10 (xg/1. For 70 % of properties where lead exceeded 25 (xg/1, the lead concentration after flushing was less than 25 ig . ... 

---