Stofkoncentrationer i regnbetingede udledninger fra fællessystemer 3. Structure and quality of measurement programsMany investigations have been performed regarding the determination of pollutant discharges from combined sewers but only few have been reported in scientific journals and international conferences making it difficult to access the data. In most cases it is not possible to get insight into how and where exactly the measurements were carried out because the measurement conditions are poorly described even in the original national or regional project reports. Aim of investigations The procurement of data leading to a better understanding of the pollution sources and of the physical, chemical, and biological processes involved has a number of aims (Geiger, 1986a):
In order to be able to make sound conclusions regarding the issues above it is of paramount importance to choose field study site carefully. The phenomena of main interest need to occur in the catchment with a sufficient frequency. Large variation of pollutant concentrations In general, measured pollutant concentrations vary considerable between events and from catchment to catchment. A very important part of the variation in results from measurement campaigns stems from the uncertainty arising from how and where in a catchment measurements are carried out, how analyses of water samples are performed and how subsequent data analysis is carried out. These issues are discusses in more detail below. Location of sampling place Some studies are performed on the spill water itself whereas measurements in others studies are taken in-sewer or inside CSO chambers. The latter is e.g. the case for Danish measurements (Miljøstyrelsen, 1990a,b) and these should thus not be compared directly with measurements on spill water. However, in many investigations such information is not available and it may even sometimes be doubtful to distinguish whether the measurements are conducted in combined or separated sewer systems. Hydrographs and pollutographs To identify hydrographs and pollutographs the sampling resolution has to be high enough to avoid aliasing. As the variability in the studied processes is high the result is often too many samples making it impossible to analyze all samples in the laboratory. Thus a compromise is often unavoidable. Furthermore, the correct time synchronization between measurements of different entities is necessary. The accuracy of the different measurements should also be comparable. There is e.g. no reason to do very accurate pollutant transport measurements if the hydraulic data accuracy is very poor. Number of events and data treatment Another concern is the number of events monitored and how the data treatment is conducted. This of course depends on what is being investigated. For example, a limited number of EMC's may be sufficient to obtain a good estimate of SMC for a location. However, if extreme statistical properties of e.g. CSO events are sought a high number of events is needed. Sometimes conclusions about the characteristics of extreme events are based on a limited sample of relatively small events. This is, of course, an extrapolation, which should be done with the uttermost care. Calculation of EMC and SMC A special problem is that it is mostly not clear how EMC's and SMC's have been calculated. For sampling programs that are based on flow-weighted techniques, the EMC should be taken as the flow-weighted mean concentration. In studies employing sequential discrete sampling, the EMC should be taken as the area under the loading curve (loadograph) divided by the area under the flow rate curve (hydrograph). Even these simple calculations may sometimes be misunderstood. The SMC should ideally be taken as the arithmetic mean and should not be confused with the median which is mostly higher than the mean due to the skewness of the data. Especially for small samples the SMC it is sometimes calculated as a volume-weighted mean concentration, and sometimes very small events are discarded from the sample analyzed because they have unexpectedly high pollutant concentrations. Uncertainty from measurement and analyses Last but not least come the inaccuracies originating from the uncertainties in the measuring and analysis methods. A typical investigation may consist of the following:
Rainfall measurements The uncertainty on rain gauge measurements using a traditional tipping bucket gauge is estimated to be 5-10% of the total rain depth (Maksimovic and Radojkovic, 1986). It is important not to introduce a time lag between rainfall measurements and flow measurements on an average basis by taking into account the time of concentration. On top of this comes the effect of spatially distributed rainfall which introduces uncertainty of runoff volumes predicted from rainfall measurements, depending on the catchment size (Arnbjerg-Nielsen, 1996). Although these uncertainties may be significant when estimating pollution runoff loads from rainfall measurements they are not important when calculating EMC's and SMC's based on measurements of flow and pollution concentrations. Discharge measurements The quality of flow measurements depends on the method applied. Traditionally measurements of water depth over weir structures have been used. Presently, the combination of a pressure sensor and a Doppler velocimeter is often used. The newest development is electromagnetic flowmeters, which introduces no obstacles into the flow. Beside these principles a lot of sensors exists based on potentiometers, ultra sound etc. Common for all the measuring devises is that the commercial producers usually quite optimistically estimate the obtainable precision. Dependent on the type of sensor system uncertainties varies from a few percent of the flow rate up-till app. 50 % (Maksimovic and Radojkovic, 1986). Water samples Besides rainfall and run-off measurements the intricate issue of taking water samples is important. Some of the investigations presented in chapter 4 are based on grab samples, others on automatic aqua samplers using either vauum pumps or peristaltic pumps. How large discrepancies this introduces is difficult to assess, but the result of using the different approaches is not the same. The next question is where in the cross-sectional flow to take the sample. Many agree that the only feasible way to sample is to use a flexible tube pointing downstream placed in the flow well above any deposits (Balmforth et al., 1995). Other investigations have tried to find an influence of suction velocity, probe size etc. where variations have been shown to be less than those introduced by e.g. the subsequent laboratory analysis (Schlütter and Schaarup-Jensen, 1998). Others have, however, expressed the importance of the intake velocity (Kleijwegt, 1992). It is important to realise that using an automatic aqua sampler will never depict the mass transport of near bed solids or bed load in general. Geiger (1984) recommends homogenising a part of the flow before sampling takes place. Analytical procedures When water samples are analyzed the analytical procedure introduces additional uncertainty. If for instance one large sample is subdivided and analyzed for total suspended solids (TSS) a coefficient of variation as high as CV=20 % may be expected (Schlütter and Schaarup-Jensen, 1998). This uncertainty obviously depends on which parameter is being determined. Furthermore, the total variance in results due to measurement uncertainty accumulates from numerous sources. In the case of TSS, uncertainty originates from e.g. representativeness of the sample, decomposition of the sample before analysis, analytical procedure, numerical rounding errors, etc. The problem is that it is impossible to track the propagation of uncertainty and that the size of the individual contributions to the uncertainty often is unknown. The human factor Field studies at multiple sites also introduce in-between site discrepancies due to the fact that a number of different people are carrying out the measuring campaigns and analyses of samples and data. The considerations mentioned in this paragraph serve as background for the descriptions of measuring campaigns for EMC/SMC-values presented in chapter 4. It is thus proper to review the rendered data with some reservation.
|
|||