5 CALCULATION OF EMC VALUES FOR A UK APPLICATION

Transcription

1 5 CALCULATION OF EMC VALUES FOR A UK APPLICATION Section 3 concluded by presenting the method to be used to develop the basin scale urban nonpoint source screening assessment. The approach is based on a volume-concentration method in which runoff volume and a stormwater quality coefficient, the EMC (Event Mean Concentration), demonstrated to be independent of each other, are used to calculate the nonpoint source load. Runoff is quantified separately, using data specific to the geographical area of interest, hence much improved load estimations are possible, so long as adequate concentration coefficients are used. Volume-concentration methods employing EMC statistics have been shown to perform better than regression models (Brown 1987; Pandit, 1997), and can give load estimates comparable to deterministic build-up wash-off models (Chandler, 1994; Charbeneau and Barrett, 1998). These assessments can e performed at a fraction of the cost of more complex methods, making them well suited to screening analysis. Section 4 outlined the approach to estimating runoff volume for application of the EMC volumeconcentration method to UK urbanised river basins. This section describes the derivation of site mean EMC values to address the provisional pollutant list developed in Section 2. This section is a summary of the companion report "The Quality of Urban Stormwater in Britain and Europe: Database and Recommended Values for Strategic Planning Models" (Mitchell, 2001) which should be consulted for details of the development of the EMC stormwater database, the database itself and the appropriate statistical analyses. Recommended site mean EMC values derived from analysis of the stormwater quality database are reproduced below. 5.1 Background and properties of the EMC coefficient The pollutant emission coefficient applied for loads assessment within the volumeconcentration methods is the event mean concentration (EMC) averaged for a catchment (the Site mean EMC). The EMC is the flow weighted mean concentration, which for a single storm event is the mass discharged divided by the volume. This is equivalent to collecting the entire stormwater runoff, completely mixing it and then determining the pollutant concentration. Individual storm EMC values can then be summarised for the catchment as either the arithmetic mean, the flow weighted mean (total load from storm events divided by total discharge volume), or the median of event EMC's. When comparing sites the median EMC is the most appropriate statistic as extreme values have less influence, but when a more accurate estimate of loads is required, these extreme values should be considered, hence the mean is the appropriate statistic (Athayde, 1983). For a large number of storms, the flow weighted mean tends to the arithmetic mean. This section summarises the derivation of emission coefficients that can be used for annual load estimation (in Northern Europe and the UK), consistent with the Water Framework Directive objective of setting emission limits relative to environmental quality standards, hence the analysis is based on catchment mean EMC 's. In practice, a UK or European analysis based on median EMC's would be impractical as very few studies cite the median value. The catchment load based on a series of events (usually over a year), is given by: N L j = Σ V i C ij Eqn. 5-1 i =1 Where L j is the annual load of the pollutant j, N is the number of storm events during the year, C ij is the EMC for pollutant j in event i, and V i is the event runoff for event i. In field studies, 58

2 no correlation is found between the event discharge (V i ) and event EMC (C ij ), an observation noted for a wide range of pollutants in stormwater from the USA, UK and France (Mance and Harman, 1978; Athayde, 1983; Hemain, 1986). The independence between event discharge and EMC means that equation 5-1 can be simplified as: L j = C j Σ V i Eqn. 5-2 Where Σ V i is the annual runoff volume, and C j is the catchment mean EMC. Thus when quantifying the annual pollutant load, annual runoff volume can be quantified independently of the stormwater quality concentration, and the annual load calculated as the product of annual storm runoff and the relevant site EMC value (i.e. the emission coefficient C j ). When an EMC value is required for a catchment, but is not readily available, there are two options: identify the EMC value through stormwater quality monitoring, or derive a value from available records. Monitoring will provide the best possible value, assuming that it is sufficiently intensive and prolonged. However, if the data is used only for preliminary screening, it is unlikely that the monitoring cost is justifiable (Williamson, 1991). This is particularly evident if the screening exercise is to be applied to a large catchment or river basin, where the problem of extrapolating monitored values to unmonitored parts of the catchment would still remain. The alternative is to draw on published EMC values. One approach is to identify a catchment with similar characteristics to the catchment of interest and apply the corresponding EMC value. In practice this is difficult to do as it is often not possible to identify a suitable matching catchment. Alternatively, a literature based central (average) EMC value can be derived from a collation of site mean EMC values, preferably from catchments with broadly similar characteristics. However, it is essential to understand the distribution of the collated data so that the most appropriate measure of central tendency can be derived. The US NURP demonstrated that for most pollutants, EMC's values adhere very well to a log-normal distribution for both (1) event to event variation at an individual site, and importantly (2) for site to site variation in median EMC (Athayde et al., 1983). When a distribution is known to be log-normal, the best estimate of the mean is derived from the log-normal relationship, although when the sample size is very large, the arithmetic mean will be similar. A further advantage of the log-normality relationship is that the expected EMC value at any probability of occurrence (X α ) can be determined from: X α = exp (µ lnx + Z α σ lnx ) Eqn 5-3 Where: µ lnx is the mean of the log transformed data Z α is the standard normal probability σ lnx is the standard deviation of the log transformed data Event to event log-normality has been observed in several European EMC databases, including those for the UK (Mance and Harman, 1978) and France (Hemain, 1986). However, to date, no comprehensive analysis of site to site variability in EMC has been conducted for the UK or Europe. 5.2 Objectives of the EMC analysis The aim of the companion stormwater quality report (Mitchell, 2001) was to identify site mean EMC values suitable for application to diffuse pollution screening models applied within northern Europe, and particularly the UK. The objectives were to: 59

3 (1) Assess the validity of the log-normality relationship for European stormwater 4 EMC's; Log-normality in stormwater event EMC has been reported for both US and European data. However, log-normality in site EMC, reported for the US NURP data, has not been similarly demonstrated for a European data set. Demonstration of log-normality permits better selection of improved site mean EMC values for European screening models, and probabilistic modelling of stormwater loads, as parameters of the distribution can be easily calculated. (2) Identify any significant differences in pollutant EMC between different urban land-uses; The US NURP (Athayde et al., 1983) concluded that there were no discernible differences in pollutant site mean EMC between different urban land uses, and hence recommended site EMC values for general urban land only (urban open land was significantly different). However, Novotny (1992) argued that a priori, there should be differences given that there are different intensities of polluting activities within different land uses, and postulates that the reason why statistically significant differences are not detected is due to the small sample size, relative to the study area (i.e. the entire USA). At this large scale, continental variations in factors such as soil type and underlying geology (which were not accounted for in the NURP statistical analysis), may mask any significant differences in EMC that can be attributed to urban land use. Further investigations of land use specific EMC values are underway for the USA drawing on the NURP data, augmented by data now available from monitoring conducted for local governments seeking permits under the US National Pollutant Discharge Elimination Scheme (NPDES) (Smullen and Cave, 1999). However, no similar exercise has been attempted for European EMC data, hence the report aims to assess whether land use specific EMC values can be recommended for use in Northern Europe. (3) Recommend site EMC values suitable for application in UK and northern European diffuse pollution screening models. Recommended site mean EMC values are currently available for application in the United States, largely drawn form the NURP, However, no comparable attempt has been made to derive values appropriate for application in Europe, hence an appraisal of European stormwater EMC's is warranted. The final objective is therefore to derive stormwater (surface runoff) pollutant concentration values which can be recommended for stormwater screening applications in northern European and UK urban catchments. The values can be applied in the context of preliminary planning, particularly the identification of source areas within river basins that contribute the most significant loads, and where BMP and source control techniques, may be most beneficial. These objectives were addressed through: Development of an EMC stormwater quality database; Analysis of the database to assess the log-normality relationship for UK and northern European EMC data; Statistical analysis of the database to derive appropriate average site EMC values; Statistical analysis to assess differences in EMC attributable to land use Recommendation of average site EMC values for application in the screening model. 5.3 Development of the EMC Database No extensive stormwater quality monitoring programme similar to that conducted in the US (NURP, NPDES requirements) has occurred for Europe, hence pollution management is based on EMC values derived from pollution monitoring conducted within the country of 4 Stormwater is here defined as the runoff from urban surfaces only (measured at catchment outflow or flow at the head of a separate sewer system), discharge that can be addressed by source control techniques. See Arnberg- Nielsen et al., (2000) for a review of CSO EMC values, and Appendix E for a summary. 60

4 application. Stormwater quality monitoring at the national level in Europe has been quite limited and sporadic, leading to calls for greater development of national urban stormwater quality databases (see Osborne and Hutchings, 1990 on the UK; Hemain et al., 1990 on France). Databases have been developed in the 1990's, but they remain poorly developed with regard to the quality of surface water runoff. In the UK, for example, the Urban Pollution Database (WRc, 1994) is used to provide default values for use in models recommended by the Urban Pollution Management Manual (FWR, 1994). However, for surface runoff quality, the database addresses only a small number of pollutants (sediment, oxygen demand, ammonia) for a handful of catchments, and with little land use differentiation. Thus Morris and Crabtree (2000), could only draw on data from six UK catchments with no land use differentiation when recommending default surface water quality values for use in the SIMPOL stormwater planning model. Ideally, each national database would contain sufficient stormwater quality data to permit the derivation of nation specific site EMC values suitable for use in probabilistic modelling. However, urban pollution monitoring is expensive, and relatively little of the necessary monitoring has occurred. Recognising this, Hemain et al., (1990) argued for an international stormwater quality database, containing data on wet weather pollution discharge and catchment characteristics, and which could be used to generate further knowledge on stormwater and its variation. The database developed in this research addresses the quality of urban surface water runoff, and is intended to provide improved site EMC values for use in urban pollution management preliminary assessments. The values recommended in the report are intended for application in northern European countries (and particularly the UK), and are based on UK and northern European data, except where sample sizes are very small. It is hoped that geographically limiting the analysis in this way will go some way to controlling for national variations in climate and catchment characteristics (e.g. drainage and street cleaning practices). The stormwater quality database was built through a review of the literature, principally using the web of science database. Citations in identified sources were also used to extend the database, particularly into the "grey literature" of unpublished material and agency reports, which contains a significant proportion of the stormwater quality data that was found. Many promising references proved to be unsuitable as stormwater quality: was addressed as unit loads or mass per unit solid; was derived from grab samples or were non-flow weighted arithmetic means; related to combined sewers, roofs, and atypical study sites (e.g. "shanty towns" in developing countries; or the data was confidential and unavailable. The literature searching produced 160 separate urban stormwater quality studies which present surface runoff quality as a flow weighted event mean concentration. The studies address 676 separate urban catchments, of which 242 are in Europe. Of these 71 were monitored in the UK, half by the Forth River Purification Board in Scotland (Bayes et al., 1994), and the rest were drawn entirely from countries in Northern Europe with broadly similar climates and type of urban development, including: Denmark (2 sites), Finland (8), France (13), Germany (24), Norway (6), Russia (1), Sweden (18), Switzerland (5) and the Netherlands (3). Of the remainder, 436 catchments relate to North America (data drawn from the US EPA and USGS NURP studies (Athayde et al., 1983; Mustard et al., 1987) with many additional studies conducted as part of the National Pollution Discharge Elimination Scheme). Data from 89 developed country sites in the "rest of the world" were also included, mostly from Australia, New Zealand and Japan. Study details and the stormwater quality database are presented in the companion report (Mitchell, 2001). It is stressed that the database deals only with EMC values appropriate to surface water runoff, with monitoring conducted either before drainage to the below ground sewer system or within a separate sewer. This is considered the most appropriate EMC data when the aim of the analysis is to identify catchments where source control techniques may be most beneficial. 61

5 EMC values for combined sewer systems are available, but the application of these values to unmonitored sites requires much more caution as the variability in site EMC is very large, due to great regional differences in sewer inputs from buildings, and the sewer design and layout. A compilation of stormwater EMC values for wet weather Combined Sewer Overflows (CSO's) is presented in Appendix E of Mitchell (2001). 5.4 Definition of Final Pollutant List The database comprises stormwater quality and catchment characteristic data for 678 separately monitored catchments. Sufficient EMC data was obtained to address 18 of the pollutants identified in the provisional list presented in Section 2. These included physical parameters (suspended sediment, biological oxygen demand, chemical oxygen demand), metals (cadmium, chromium, copper, iron, lead, mercury, nickel, zinc), nutrients (total nitrogen, total kjheldal nitrogen, total phosphorous, soluble phosphorous, ammoniacal-n) and hydrocarbons as total oil and grease. Pollutants relevant to the Water Framework Directive are referred to in Article 1.4 of Annex II, where it is stated that "member states shall.estimate and identify significant diffuse pollution, in particular by substances listed in Annex VIII, from urban, industrial, agricultural and other sources". All of the above parameters are included in Annex VIII, but not all of the Annex VIII substances are addressed in the screening model. This is because Annex VIII substances are either not considered significant in terms of diffuse urban pollution (organohalogen and organotin compounds, endocrine disrupters), or because they are potentially significant, but there is insufficient EMC data in the literature to allow identification of meaningful central tendency values (cyanide, arsenic, biocides). Data on many pollutants is presented in the literature studies examined, but not sufficiently frequently to warrant further analysis. Not all studies address all the pollutants: suspended solids is measured in 508 of the 678 catchments, but mercury is measured in just Data Management Catchments that were included in the database are those that are entirely or part urban, or which are green space within an urban area. Agricultural, forestry and other rural catchments are excluded, as are roofs. Roofs are clearly urban, but for purposes of basin scale screening, it is considered unlikely that adequate roof data (cover and type) can be obtained to merit the treatment of roofs as distinct source areas. Land use characteristics are described in nearly all of the studies, although the quality of the description varies markedly between them. For some catchments, quantitative data is given on land use cover, but for others it is only qualitative (e.g. high density residential, industrial estate etc.). However, this descriptive data is usually sufficient to allow the catchments to be confidently allocated to one or other of the land use categories adopted for the database. Where mixed catchments do occur, they play little part in the analysis. Figure 5-1 illustrates the land use categories used to structure the data, and permit analysis of potential differences in site EMC values between land uses. The categories were selected so as to correspond to land uses which are most easily mapped for British (and European) cities using readily available geographical data (see Section 6). Thus if land use specific EMC values are identified, they can be readily applied to UK/European diffuse pollution screening models. 62

6 Figure 5-1 Land use classification used in Stormwater Quality Database and Preliminary Planning GIS-Model ALL URBAN (N=676) URBAN OPEN (N=27) DEVELOPED URBAN (N=538) MAIN ROADS (N=111) M-WAYS (ADT>30000) (N=66) A & B (ADT<30000) (N=38) INDUSTRIAL and COMMERCIAL (N=138) MIXED (N=103) and NOT STATED (N=47) RESIDENTIAL (N=250) INDUSTRIAL (N=60) COMMERCIAL (N=78) LOW DENSITY (N=32) MEDIUM DENSITY (N=86) HIGH DENSITY (N=41) 1. EMC land-use categories correspond to land-use categories derived from CORINE digital database, and applied within a GIS stormwater quality preliminary planning model. 2. N=100 denotes number of stormwater quality records (catchments) in each category. 3. Roads type defined by Average Daily Traffic (ADT), and ADT-Road type table 2.5 in CIRIA Low and high residential density are the lower and upper quartiles of the density data, and correspond to <14 p/ha and >50p/ha respectively. 2.5 p/dwelling is assumed where density is given in dwellings/ha. 5. Light shading represents usual EMC differentiation by land use (Table 5-2), dark shading addresses additional land uses categories applied to recommended EMC values from this study (Figure 5-3).. 63

7 5.6 Descriptive Statistics The Stormwater Quality report (Mitchell, 2001) provides a brief description of the principal pollutant source in urban catchments, together with an indication of the effects in receiving water, and the range of observed instantaneous urban stormwater quality concentration (for comparison with flow weighted concentrations). Observed acute and chronic toxicological effect concentrations and European water quality standards are given. Descriptive statistics from an analysis of the database are then presented, divided according to the land classification described above, and three geographical regions: the UK, Northern Europe and "Global" (i.e. all data). The European data group contains all the UK data, and the Global region contains all the European and UK data. The Global region is included for comparative purposes, and has been used to provide an indicative EMC value where the sample size in the European data group is considered too small. The descriptive statistics include the maximum and minimum observed site EMC by geographical region, and a table of the following parameters: sample size, site mean EMC and standard error, inter-quartile values (all calculated from log values), and three other measures of central tendency, the median, and the arithmetic and geometric means. Alternative percentile values can be calculated from the data presented in these tables using equation 5-3. Tests for log-normality are also conducted using the powerful Shapiro-Wilk statistic, with the final arbitration of log-normality determined with reference to a graphical Q-Q plot, of the type illustrated in Figure 5-2. T-tests for differences in mean site EMC by land use were conducted, and a commentary on the results presented for each pollutant. Figure 5-2. Example normality plots for site mean EMC data (a) Frequency distribution plot (Oil and Grease) (b) Log-normality plot (Oil and Grease) Frequency Normal Quantile Log Oil and Grease (mg/l) 64

8 (d) Log-normality plot (Cadmium) (d) Log-normality plot (Total P) Normal Quantile Normal Quantile Log COD (mg/l) Log Total P (mg/l) 5.7 Conclusion and Recommenced EMC Values The site EMC data was assessed for its fit to a log-normal relationship. Tests indicated that most pollutants adhered very well to a log-normal relationship for all three geographical regions studied (UK, Northern Europe, Global) where sample size permitted an analysis. Zinc and oxidised nitrogen (NO 2 + NO 3 ) did not adhere strongly to the log-normal relationship, but stratification of the data, by geographical region and by land use category demonstrated that the assumption of log-normality was appropriate for these and all other pollutants. It was concluded therefore, that all pollutants should be subject to further analysis (descriptive statistics, test for differences in means) in the log domain. Tests for differences in site mean EMC were conducted for each pollutants, for each of the three geographical regions in the database. For most pollutants (not those with small sample sizes), there were more significant differences observed between site mean EMC between land use categories for the Northern European data, than for the other geographical regions. This tends to support the view that significant differences by land use in the global data set are masked by confounding factors which vary greatly across such a large geographical range. Conversely, fewer differences are found in the UK set than in the European data, as sample sizes here are low, limiting the power of the difference test. Roads are the major source of TSS, BOD, lead, copper, zinc in all geographical regions, and a principal source of other metals in at least one region (sample size limits the analysis). TSS and metal loads are higher from motorways than other main roads, indicating the role of vehicle density in load generation. Nutrient loadings (Tot. P, Tot N. TKN, NO 2 +NO 3 ) are highest from residential areas, and lowest for industrial and commercial areas. Within the residential category, sites of low density have higher BOD and COD and lower P concentrations in stormwater. No clear and consistent patterns between residential density and site mean EMC are evident for other parameters. Using results of the difference tests, land use categories for which significantly different site mean EMC values are evident are identified for each pollutant. The selection is largely objective, although when there is not a strong and consistent pattern between geographical regions, some judgement is used in deciding whether a land use category merits treatment as a distinct category. In some cases it was possible to define up to 5 distinct land use categories, whilst in others, it was not possible to discern any differences at all (small sample size), hence recommended site mean EMC values relate to the all urban land use class. 65

9 Recommended stormwater quality values are given below, where EMC values are presented for (1) each of three land use classes (Open, Developed Urban, Roads) which are a priori commonly considered to be significantly different to each other, an assumption supported by data from the US (Athahyde, 1983; Strecker et al., 1987), and (2) for each land use category (see Figure 5-1) with a significantly different mean site EMC. Table 5-1. 'Average' stormwater quality from site mean EMC meta-studies GLOBAL USA NORTH EUROPE UK NZ This study 1, 8 TSS mg/l BOD mg/l COD mg/l Cd ug/l Cr ug/l Cu ug/l Fe mg/l Pb ug/l Hg ug/l Ni ug/l Zn ug/l Tot P mg/l Sol P mg/l Tot N mg/l TKN mg/l NO 2 +NO mg/l 0.85 NH 4 -N 0.45 mg/l 0.45 Oil & G 4.6 mg/l Duncan (1999) 2 NURP (1983) 3 NURP Update (1999) This Study 1, Ellis William (1989) 5 -son (1991) Notes: 1. Mean of all land uses (all uses except roads and open in italic); 2. Excludes roads and open land use; 3. From Athayde et al., (1983); 4. As updated using NPDES and USGS data by Smullen and Cave (1999); 5. All values from Separate Sewers Outfalls; 6. Mean values (not all same as the values recommended for New Zealand by the author); 7. NO 3 only. 8. Values in bold have sample size N<10. 66

10 Table 5-1 presents site mean EMC values for the All Urban and Developed Urban categories for the three geographical regions. The data is presented with comparable variables from other meta-studies, and it is apparent that the values are broadly consistent with results of these studies, although there is significant overlap in the sources used. However, it is also evident that site mean EMC varies across a significant range, and that a specific Northern European analysis was warranted. Table 5-2 presents EMC values for the three urban land uses which are most often treated as distinct categories in screening models: Main Roads, Developed Urban and Urban Open. Table 5-3 then presents recommended values considering the tests for differences in site mean EMC by land use. In some cases this results in additional land use categories, and in some cases only one category can be justified. Table 5-2 Site mean EMC values by conventional screening analysis land use classes Pollutant Land use category Mean 1st 3rd Data source 1 TSS Urban Open All (no Europe data) mg/l Developed Urban 77 (45) 32 (18) 190 (115) Europe & (UK) Main Roads 191 (149) 101 (68) 361 (327) Europe (UK) BOD Urban Open All (no Europe data) mg/l Developed Urban 9.6 (8.1) 5.9 (4.9) 15.6 (13.3) Europe & (UK) Main Roads Europe COD Urban Open All (no Europe data) mg/l Developed Urban Europe Main Roads Europe Cd Urban Open All (N=2) ug/l Developed Urban Europe (N=28) Main Roads Europe (N=11) Cr Urban Open All (N=2) ug/l Developed Urban 16.4 (5.4) 6.4 (3.0) 42.2 (9.5) All & (Europe N=14) Main Roads All Cu Urban Open All (N=6) ug/l Developed Urban Europe Main Roads Europe Fe Urban Open All (N=2) mg/l Developed Urban All Main Roads All (N=16) Pb Urban Open All (N=11) ug/l Developed Urban (99.0) 71.4 (47.7) (205.3) Europe & (UK N=12) Main Roads Europe (N=26) Hg Urban Open All (N=1) ug/l Developed Urban All (N=22) Main Roads All (N=2) Ni Urban Open All (N=2) ug/l Developed Urban 26.4 (30.4) 15.0 (18.2) 46.5 (50.6) All & (Europe N=13) Main Roads All (N=7) Zn Urban Open All (N=6) ug/l Developed Urban Europe Main Roads Europe (N=25) 67

11 Pollutant Land use category Mean 1st 3rd Data source 1 Tot P Urban Open All (N=21) mg/l Developed Urban Europe Main Roads Europe (N=9) Sol P Urban Open All (N=9) mg/l Developed Urban All Main Roads All (N=4) Tot N Urban Open All (N=14) mg/l Developed Urban 2.43 (2.20) 1.45 (1.50) 4.09 (3.40) All & (Europe N=24) Main Roads All (N=27) TKN Urban Open All (N=11) mg/l Developed Urban All Main Roads All (=20) NO Urban Open All (N=8) mg/l Developed Urban All (N=8) Main Roads All (N=17) NH 4 -N Urban Open All (n=1) mg/l Developed Urban 0.56 (0.55) 0.30 (0.29) 1.06 (1.05) Europe & (UK) Main Roads Europe (N=4) Oil & Urban Open All (N=1) Grease Developed Urban 5.08 (4.2) 2.12 (1.2) (14.9) All & (Europe N=18) mg/l Main Roads All (N=12) 1. Regions are the UK, Northern Europe and All data (European plus mostly US data). 2. Values in bold have sample size N<10, and should be treated cautiously (use Table 4-3 values or local values preferentially). Sample sizes are also given where N<30. Table 5-3 indicates that is appropriate to divide urban land in to additional land use classes for many pollutants, and that clear differences are commonly observed in stormwater pollutant concentrations between residential and industrial/commercial sites. In several cases significant differences in EMC are evident between commercial and industrial sites. However, lumped commercial/ industrial values are recommended here as the industrial and commercial land use data is rarely separately resolved in widely available European land use databases. Whilst differences in EMC were observed between residential density classes, the pattern of differences were not so profound as to justify recommendation of density specific EMC values. However, in several cases, highly significant differences were observed between Motorways and other main roads, and Table 5-3 indicates where it is appropriate to treat these as separate land use classes. The final choice of site mean EMC value/land-use class will also depend upon the availability of digital land use data to which the values can be applied in a GIS based screening analysis. Note that significant differences in load estimates arise depending on the land use resolution used. Figure 5-3 illustrate difference in annual load for one pollutant, using the EMC data presented in Table 5-2 and 5-3 respectively. For both Tables 5-2 and 5-3, values are based, as far as possible, on the European data. Where sample sizes are very low (< 10) this is not practical, hence the global value is recommended instead. Where sample sizes are sufficient (N>10) UK values are also given. Mean and quartile values are presented, and equation 5-3 can be used to derive alternative percentile values for probabilistic modelling. 68

12 Table 5-3. Recommended site mean EMC values for N. European screening applications Pollutant Land use category Mean 1st 3rd Data source 1 TSS Urban Open All (N=18) mg/l Ind./Comm (33.3) (13.9) (80.0) Europe & (UK N=28) 2 Residential 85.1 (46.9) 37.6 (19.7) (111.6) Europe & (UK N=17) Motorways Europe (N=16) Other Main Roads Europe (N=6) BOD Urban Open All (N=4) mg/l Ind./Comm. 9.9 (9.2) 5.9 (5.5) 16.7 (15.5) Europe & (UK N=26) Residential 8.5 (5.2) 5.1 (3.8) 14.1 (7.0) Europe & (UK N=13) Main Roads Europe (N=11) COD Urban Open All (N=16) mg/l Ind./Comm Europe (N=6) Residential Europe (N=20) Main Roads Europe (N=9) Cd ug/l All Urban Europe Cr ug/l All Urban Europe Cu ug/l Urban Open All (N=6) Developed Urban Europe Main Roads Europe (N=21) Fe mg/l All Urban All Pb ug/l 3 Urban Open All (N=11) Ind./Comm Europe (N=11) Residential Europe (N=24) Motorways Europe (N=14) A-Roads Europe (N=10) Hg ug/l All Urban All (N=25) Ni ug/l Urban Open All (N=2) Developed Urban Europe (N=13) Zn ug/l Urban Open All (N=8) Ind./Comm Europe (N=13) Residential Europe (N=25) Motorways Europe (N=25) Other Main Roads Europe (N=10) Tot P Urban Open All (N=21) mg/l Ind./Comm (0.52) 0.16 (0.40) 0.54 (0.70) All & (Europe N=6) Residential 0.41 (0.36) 0.24 (0.20) 0.72 (0.60) All & (Europe N=18) Motorways All Other Main Roads All Sol P Urban Open All (N=9) mg/l Ind./Comm All (N=24) Residential All Main Roads All (N=4) Tot N Urban Open All (N=14) mg/l Ind./Comm All Residential All Main Roads All 69

13 Pollutant Land use category Mean 1st 3rd Data source 1 TKN Urban Open All (N=11) mg/l Ind./Comm All Residential All Main Roads All (N=20) NO Urban Open All (N=8) mg/l Ind./Comm All Residential All Main Roads All (N=17) NH 4 -N Urban Open All (N=1) mg/l Developed Urban 0.56 (0.55) 0.30 (0.29) 1.06 (1.05) Europe & (UK) Oil & G. Urban Open All (N=1) mg/l Developed Urban Europe (N=18) 1. Regions are the UK, Northern Europe and All data (European plus mostly US data). 2. Values in bold have sample size N<10, and should be treated cautiously. Sample sizes are also given where N<30; 3. Due to the reduction in use of leaded petrol, the first quartile value is recommended for all land uses. See Table 5-4 for lower and upper percentile values; 4. It is recommended that EMC values are selected from the same geographical regions, for reasons of consistency, where two or more land uses are addressed for a pollutant. Table 5-4 Recommended site mean EMC values for lead Pollutant Land use category Central value 1 (25 % tile) 10 % tile 50 % tile Data source Pb ug/l Urban Open All (N=11) Ind./Comm Europe (N=11) Residential Europe (N=24) Motorways Europe (N=14) A-Roads Europe (N=10) 1. Recommended central tendency value based on 25 percentile (considers historical decline in site mean EMC in response to introduction of unleaded petrol (see Mitchell, 2001, section 3.9). 70