Idea Transcript
3
DESCRIPTIVE STATISTICS
For each pollutant a brief description of the principal source in urban catchments is given, 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 using the data presented in these tables using equation 3. Appendix C presents further parameters descriptive of the EMC data distribution, including the standard deviation, skewness and kurtosis. The log-normality of the distribution is also assessed using the Shapiro-Wilk test (high values indicate log-normality) a powerful overall test for normality, and one recommended for sample sizes up to 5000, for which the mean or variance of the hypothesised distribution are not specified in advance (Royston, 1992). The Shapiro-Wilk test was not conducted for sample sizes < 10, as the power of the test was too low (i.e. possibility of rejecting the null hypothesis when it should be accepted). The significance level p of the Shapiro-Wilk statistic is also presented. The smaller the value of p, the more untenable the null hypothesis that there is no log-normality, that is, small values of p indicate that the data is not log-normally distributed. The Shapiro-Wilk test may indicate that a distribution is not normal, when examination of the frequency distribution and log-normality graphical plots strongly suggests normality. This is because the Shapiro-Wilk test is very strongly influenced by outliers. Removal of one or two outliers from a sample size of several hundred is sufficient to indicate normality by the Shapiro-Wilk test. Outliers in the site EMC data were examined using box plots (not shown), but only a few values were excluded. These were values that were considered to be a recording or typographic error, and several which were not considered to belong to the population (e.g. EMC values affected by a volcanic eruption in Washington). The normal probability (Q-Q) plot is considered the single most valuable aid to diagnosis of normality (BBN, 1996), and was used as the final arbiter of log-normality. T-tests for differences in mean site EMC by land use were conducted, and a commentary on the results is presented in section 3 for each pollutant. Final recommended values are given in section 4, 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 2-1) with a significantly different mean site EMC, as identified from the analysis presented here.
21
3.1
Total Suspended Solids
Sediments are an important mechanism for the transport of other pollutants which may cause problems related to toxicity, eutrophication, and suitability for recreational or potable use. Particle size is important as half the pollutant load may be transported bound to particles of 99.9%) with residential and highway land uses, for all three geographical regions, indicating that it is appropriate to treat this as a separate land use class for EMC purposes at the group level. No clear pattern of differences in EMC is discernible between residential density groups. Highly significant differences are found with the Ind./Comm. group in the All data (p > 99%), and with the industrial group in the European data. Further differences are found between different residential groups and the industrial and commercial groups for all geographical regions. Significant differences are found with both A roads and motorways in all geographical areas, including the UK (p > 95%) and Europe (p > 99%). It is concluded it is appropriate to treat residential land use as a distinct land use class at the group level. There are no significant differences between motorways and A-roads. Motorways have significantly higher EMC values than most other land uses, including the ind./comm. and residential groups, for all geographical regions, but this pattern is not generally repeated with the A-roads. It is considered appropriate to treat motorways as a distinct land use re EMC. Open (Global value) Residential, Industrial/Commercial Motorways, A-roads
23
The sample size is large for TSS, particularly for the European and Global regions, hence there is confidence that the detected differences in EMC by land use are real. Values could be drawn from either the UK or European data sets. However, for applications where multiple pollutants are of interest, geographical consistency is important. For most other pollutants it is not possible to recommend UK specific values due to limited sample sizes. Therefore, to maximise consistency it is recommended that values be selected wherever possible from the European region. Recommended values are presented in section 4.
3.2
Biochemical Oxygen Demand
Biodegradable organic materials, whether natural or synthetic can enter water in solution or suspended in runoff. Sources include decaying plant and animal matter, animal excreta, litter, food wastes and hydrocarbons. In the process of decomposition (oxidation), organic materials exert an oxygen demand, either chemically or biologically mediated, which can cause dissolved oxygen (DO) in receiving water to fall to levels at which aquatic life cannot be maintained. The biochemical oxygen demand (BOD) is the amount of oxygen used in the metabolism of biodegradable organics, and usually determined using a standard five day test (BOD5) involving the use of micro-organisms. Altogether 248 BOD5 records were obtained, of which 81 were for Northern Europe and 43 for the UK. A reasonable spread of values is found amongst land use classes at the group level by geographical region, although BOD5 data for UK highways is limited to a single observation. The values range from 2 mg/l for an urban open area in the USA (Athayde, 1983) to 146 mg/l for a developed urban area in Detroit, Michigan (Palmer, 1950). The European data has low value of 3 mg/l for an industrial estate in Scotland (Bayes et al., 1994) to 45 mg/l for a medium density residential district in Aix-en Provence, Paris (Deutsch and Hemain, 1984). The UK maximum of 36 mg/l was from an Edinburgh airport outlet (Bayes et al., 1994). Instantaneous concentrations of BOD5 range from 1 to 7700 mg/l, and stormwater can approach the value of untreated domestic wastewater (Makepeace et al., 1995).
Figure 3-2. Normality plots for Biological Oxygen Demand (a) Frequency distribution (Global data)
(b) Log-normality plot (Global data)
90 4
80 3
Frequency
60 50 40 30
Normal Quantile
70
2 1 0 -1
20 -2
10 0
-3 0.25
0.75
1.25 Log BOD (mg/l)
24
1.75
2.25
(d) Log-normality plot (UK data)
3
3
2
2 Normal Quantile
Normal Quantile
(c) Log-normality plot (N. Europe data)
1
0
0
-1
-1
-2 0.40
1
0.60 0.80
1.00
1.20
1.40 1.60
1.80
-2 0.40 0.60
0.80 1.00
1.20
1.40
1.60
1.80
Log BOD (mg/l)
Log BOD (mg/l)
The BOD 5 values from all regions adheres to a log-normal distribution, as indicated by the frequency distribution and probability plots (Figure 3-2). Appendix C presents the normality statistics of skewness, kurtosis and the robust Shapiro-Wilk statistic. The latter indicates that the assumption of normality is not significant (p < 95%), but tests on the stratified data indicates that the assumption of normality is acceptable. Measures of centrality (arithmetic mean, geometric mean, median) are presented in Appendix D, along with quartile values. Alternative percentile values can be calculated using the log-transformed data given in Appendix C, and equation 3 in section 1. T-tests for differences in mean site EMC were performed between land uses at both group and sub-group level, and for all three geographical regions (UK, Northern Europe, and the Global data set). A summary of land use differences is shown below, including recommendations for the most resolved land use classification for BOD EMC application in the UK and Europe.
Industrial & Commercial
•
Residential
•
•
•
•
•
No differences are found between industrial and commercial EMC values. At the group level, significant differences are found with residential use (UK data, p > 95%) and with all categories of highway (p > 99% for European data), hence it is concluded that it is appropriate to treat this as a separate land use at the group level for EMC purposes. No difference in EMC is discernible between residential density groups. Highly significant differences are found with the Ind./Comm. group in the UK data (p > 95%), with further differences between different residential groups and the industrial and commercial groups in the Global data. Significant differences are found with highways at the group and subgroup levels and also motorways in all geographical areas, and particularly in the UK (p > 95%) and European (p > 99%) data. It is concluded that it is appropriate to treat residential land use as a distinct group at the aggregate level.
25
Highways
• •
Selected land-use categories
• • •
There are no significant differences between motorways and A-roads. Highways have significantly (better than p > 95%) higher EMC values than most other land uses including ind./comm. and residential use for European data, but there is insufficient data to identify any UK differences. This pattern is repeated with the European A-roads. Open (Global) Residential, Industrial/Commercial Main roads
Recommended values for the selected land-uses are presented in section 4. Note that the European values are slightly higher than that for the UK, probably attributable to the larger sample size rather than any real effect. It is also noted that both the UK and European EMC values are consistent with the BOD5 values recommended by Morris and Crabtree (2000) for use in the preliminary planning procedures (e.g. SIMPOL model) of the UK Urban Pollution Management Manual (FWR, 1994).
3.3
Chemical Oxygen Demand
Like BOD, Chemical oxygen demand (COD) is a measure of the oxygen demanded by decomposing substances, and is a measure of the organic content of the water. Unlike the BOD test, which uses micro-organisms, COD is determined chemically, using a strong oxidising agent (potassium dichromate) with the process assisted by a silver sulphate catalyst. The COD test is rapid, about 3 hours, but cannot be used to compare biologically oxidizable from inert organic matter. A total of 336 COD EMC records were obtained, of which 49 were from Northern Europe, but only 4 from the UK. A good spread of values by land use class is found in the European and Global data sets. The global range is from 5 mg/l for an urban district of Zhuhai, China (Zhen-Ren et al., 1993) to 1031 mg/l for a commercial district of Burnaby, Canada (Hall and Anderson, 1988). The European values range from a low of 23mg/l for a residential area of Leystad, Netherlands (Unnk and Van de Ven, 1987) to 510 mg/l for an A-road in the Lake Padderunduann area of Norway (Gjessing et al., 1984). Figure 3-3. Normality plots for Chemical Oxygen Demand (a) Frequency distribution (Global data)
(b) Log-normality plot (Global data) 4
100 3
90
2
Frequency
70 60 50 40
Normal Quantile
80
30
1 0 -1 -2
20 -3
10
-4
0
0.5
1.0
1.5
2.0
Log COD (mgl)
26
2.5
3.0
(c) Frequency distribution plot (N. Europe) 14
3
12
2 Normal Quantile
10 Frequency
(d) Log-normality plot (N. Europe)
8 6
1 0 -1
4 -2
2
-3
0
1.2
1.4
1.6
1.8
2.0
2.2
2.4
2.6
2.8
Log COD (mg/l)
COD adheres very well to a log-normal distribution, as indicated by the frequency distribution and probability plots (Figure 3-3). Appendix C presents the normality statistics of skewness and kurtosis and the Shapiro-Wilk test results, which indicate that the assumption of log-normality cannot be accepted (p < 95%) for the all urban data, but that the assumption is valid for the stratified and European data. Measures of centrality are presented in Appendix D along with quartile values. Alternative percentile values can be calculated using the descriptive log-transformed data in Appendix C and equation 3 from section 1. Industrial & Commercial
• • •
•
Residential
• •
• •
Highways
• •
•
No UK data and limited European data limit the analysis. No differences were found between industrial and commercial values. When treated as a single (aggregate) group, significant differences are found with residential use (Europe, p > 95%). Differences are also found between the industrial and commercial groups and some residential and highway categories. It is concluded that it is appropriate to treat this as a distinct land use for EMC purposes at the group level. No differences in EMC are discernible between residential density groups. Residential EMC is significantly lower than the Ind./Comm. group in the European data (p > 95%), with additional similar differences at the subgroup level The residential EMC is significantly lower than that of highways in the European group (p >99%). It is concluded that it is appropriate to treat residential land use as a distinct use at the group level. Motorways have a significantly higher EMC than that for A-roads in the Global region (p >95%), but not the other regions (where N is small). A-road values are consistently and significantly lower (p > 95%) than most other land uses, including motorways in the All group, but not in the other regional groups, presumably due to a small sample size. It is concluded that it is appropriate to treat A roads distinctly from motorways if using the larger All data set, but that this distinction cannot be made for the other regions where N is too small.
27
Selected land use categories
• • •
Open (from Global data set) Residential, Industrial/Commercial Highways (if Europe) else Motorways and A-roads (if Global data set).
Results for differences in mean site EMC values by land use are summarised below, with recommendations for the most resolved land use classification for COD EMC application in the UK and Europe. Recommended COD EMC values are presented in section 4.
3.4
Cadmium
The principal sources of cadmium are combustion, including that of lubricating oil, wear of tyres and brake pads, and industrial emissions, especially from electroplating works. It is used to cover iron products (sheet iron, nuts and bolts) to prevent corrosion, although these do eventually corrode and release cadmium. Cadmium is used extensively in the manufacture of batteries, paints, and plastics, and is found in agricultural use of sludge, fertilisers and pesticides (Makepeace et al., 1995). Cadmium is highly toxic, and standards are set for fisheries protection and drinking water. Its toxicity is affected by hardness, pH, water temperature and the presence of organic compounds. In stormwater, Cd is largely associated with dissolved solids (Makepeace et al., 1995). A total of 141 Cd EMC records were obtained, of which 39 were from Northern Europe, but only 5 from the UK. A good spread of values by land use class is found in the Global data sets, but there is only one observation from an industrial/commercial site for Europe. The Global values range from 0.2 ug/l for a developed urban site in Sacremento, California (Bumgardner, 1994) to 63 ug/l for a parking lot in Syracuse, New York (Owe et al., 1982). The European range is from a low of 0.5 ug/l for a developed urban area in Viborg, Denmark (Arnberg-Nielsen et al., 1987) to 9 ug/l for a developed urban area in Strassenabflusse, in the German Rhine valley (Muschack, 1989).
Figure 3-4 Normality plots for Cadmium (a) Frequency distribution (Global data)
(b) Log-normality plot (Global data) 3
40
2
Frequency
30 25 20 15
Normal Quantile
35
1
0
-1
10
-2 5 0
-3 -0.8
-0.3
0.3
0.8
Log Cd (ug/l)
28
1.3
1.8
(c) Frequency distribution plot (N. Europe) 14
3
12
2 Normal Quantile
10 Frequency
(d) Log-normality plot (N. Europe)
8 6
1 0 -1
4 -2
2
-3
0
1.2
1.4
1.6
1.8
2.0
2.2
2.4
2.6
2.8
Log COD (mg/l)
Cadmium adheres well to a log-normal distribution, as indicated by the frequency distribution and probability plots (Figure 3-4). Appendix C presents the normality statistics of skewness and kurtosis and the Shapiro-Wilk test results, which indicate that the assumption of lognormality cannot be accepted (p < 95%) for the all urban data, but that the assumption is valid for the stratified and European data. Measures of centrality (arithmetic mean, geometric mean, median) are presented in Appendix D, along with quartile values. Additional percentile values can be calculated using descriptive (log-transformed) data given in Appendix C, and using equation 3 from section 1. Results of t-tests for differences in mean site EMC values by land use are summarised below, including recommendations for the most resolved land use classification for Cd EMC application in the UK and Europe. Industrial & Commercial
•
•
Residential
• •
•
Highways
Selected land-use categories
•
No differences were found between industrial and commercial values. Significant differences are found with residential land use, but these differences are not consistent between regions, suggesting that the differences are due to the small sample size. It is concluded that industrial and commercial land use cannot be treated as a distinct group. No differences in EMC are discernible between residential density groups. Only one significant difference occurs between a residential land use and another land use (low density residential > A-roads in the Global region), but this is based on a sample size of N=3. Residential land use cannot be treated as a distinct group.
•
Motorways have a consistently higher EMC than for A roads in all regions, but these differences are not significant, except in the European data, and only then when applying a one tailed t-test (p >90%). It is concluded that highways cannot be treated as a distinct group.
•
All Urban
29
Few differences in Cadmium EMC between land uses were found, although it is likely that major roads are significant sources. The small sample size hampers a more definitive analysis. Significant differences are found between several land uses and the urban open category, but it is noted that the open EMC value appears much higher than might a priori be expected given the known sources of CD in the urban environment. The mean EMC value is based on only two observations both non-European and both drawn from a secondary source, with the original unpublished sources unavailable. It is therefore considered more appropriate to treat the Open EMC as a missing value. It is concluded that there are no significant land use differences, and that the EMC value should be drawn from the European "All Urban" group, where the sample size is considered adequate (N=28). Recommended Cd EMC values are presented in section 4.
3.5
Chromium
The main sources of Cr are corrosion of welded plating metal, wear of bearings and bushes in engines, dyes, paints, ceramics, paper, heating and cooling coils, fire sprinkler systems, pesticides and fertilisers. In its hexavalent form Cr6+ is soluble, mobile and can be stable for long periods in water of low organic content. Its trivalent form CR3+ has a tendency to form stable complexes, notably with chromium hydroxide. Acute toxicity ranges from 2ug/l for algae, where it bio-accumulates readily, to 265ug/l for rainbow trout fry, whilst chronic toxicity ranges from 3ug/l for invertebrates to 73 ug/l for trout fry (Makepeace et al ., 1995). In stormwater it is predominately associated with suspended sediments, and is more toxic in soft water. A total of 112 Cr EMC records were obtained, of which 19 were from Northern Europe, and just 4 from the UK. A good spread of values by land use class is found in the Global data sets, but the small sample size for the other regions limits the analysis. The Global values ranged from 0.5 ug/l for an urban open area in Burnaby, Canada (Hall and Anderson, 1988) to 580 ug/l for an industrial area in Sydney, Australia (GH and D et al., 1989). The European data has a similar minimum value, and a maximum 140ug/l from the A38 urban motorway in Birmingham, UK (Hedley and Lockley, 1975). Figure 3-5. Normality plots for Chromium (b) Log-normality plot (Global data)
60
3
50
2
40 30 20
Normal Quantile
Frequency
(a) Frequency distribution (Global data)
1 0 -1 -2
10
-3 -0.5
0
0.0
0.5
1.0
1.5
Log Cr (ug/l)
30
2.0
2.5
3.0
(c) Frequency distribution plot (N. Europe)
(d) Log-normality plot (N. Europe) 3
10 9
2
8 Normal Quantile
Frequency
7 6 5 4 3
1
0
-1
2 1
-2
0
0.0
0.5
1.0
1.5
2.0
2.5
Log Cr (ug/l)
Total chromium adheres well to a log-normal distribution, particularly at the land use level in the global data set where the sample size permits stratification for application of the ShapiroWilk test. Frequency distribution and probability plots are shown in Figure 3-5, whilst Appendix C presents the normality statistics, and Shapiro-Wilk test result. Measures of centrality (arithmetic mean, geometric mean, median) are presented in Appendix D, along with quartile values. Additional percentile values can be calculated using the descriptive logtransformed, data given in Appendix C, using the equation 3 from section 1. Results of t-tests for differences in mean EMC values by land use are summarised below, including recommendations for the most resolved land use classification for total Cr EMC application in the UK and Europe. Recommended Cd EMC values are presented in section 4. Industrial & Commercial
• •
•
Residential
•
•
•
Highways
• •
No differences were found between industrial and commercial values. At the group level ind./comm. is significantly lower than both residential and highway groups in the European data. However, N is very small, and this pattern is not repeated in the larger Global group. It is concluded that differences may occur, but that the sample size precludes a more powerful analysis, hence industrial and commercial land use cannot be treated as a distinct group. No significant consistent differences in EMC are discernible between residential density groups, although it is notable that EMC values are inversely related to residential density in both European and All groups. Residential land has a significantly lower (p > 95%) EMC value than Roads in the European data, but there are no differences in the Global data. It is concluded that residential land use cannot be treated as a distinct group. No difference in EMC by road type is found in the Global data. There is no A-road data for the other regions. Roads have a significantly (p >95%) higher value than residential use, and also the general Developed Urban category for the European data. This pattern is not repeated in the Global data, and is attributed to one exceptionally high value from the A38 urban motorway in the UK.
31
•
Selected land use categories
•
It is concluded that there is insufficient data to support the treatment of highways as a distinct group. All Urban
No consistently significant differences in total chromium EMC between land uses were found. The open EMC value is based on only two observations, both non-European and drawn from a secondary source, with the original unpublished sources unavailable. It is therefore considered more appropriate to treat the open EMC as a missing value. It is concluded that there are no significant differences in EMC by land use, and that the EMC value should be drawn from the "All Urban" category. The European EMC value for All Urban category is approximately half that of the All data set, but the sample size of 19 is considered sufficient to select the European value for European applications.
3.6
Total Copper
The principle sources of copper in the urban environment include wear of vehicle parts (bushes, tyres, brake linings), combustion of lubricating oils, corrosion of building materials (roofs, pipes) and industrial wastes, particularly those from electroplating works. It is also widely used in algaecides, fungicides and pesticides. As with all metals, the environmental mobility and bio-availability is dependent upon its concentration in solution. As most studies report total metal concentration (i.e. in solution + particulate phase), it is worth noting that c. 20-40% of total copper in urban runoff occurs in the soluble phase (Luker and Montague, 1994). Instantaneous stormwater concentrations range from 0.06 to 1410 ug/l. Copper is the major aquatic toxic metal in storm water, and its toxicity, which varies with hardness, begins at 17ug/l for invertebrates (D. magna) (Makepeace et al., 1995). Toxicity varies with hardness. The EC standard for fisheries is 40 ug/l and 50ug/l for abstraction for a normal water treatment works, guidelines which are often exceeded. A total of 272 Cu EMC records were obtained, of which 65 were from Northern Europe, and 13 from the UK. A good spread of values by land use class is found in the Global and European data sets, but the small sample size for the UK limits the land use analysis for that region. The values ranged from 2 ug/l for a residential site in Scotland (Heal, 1999) to 1270 ug/l for an urban area of New Jersey (Wilber et al., 1980). The maximum European value of 865 ug/l was recorded for an urban district of Stockholm (Palmgren and Bennerstedt, 1984). One observation for Cu had a particularly high value of 7033ug/l. This value was derived from Duncan (1999), and is for an A-road in Washington, USA. Duncan notes that the observation is affected by aerial deposition from the Mt. St. Helens eruption, hence this value is considered as exceptional and dropped from the analysis, along with all the other Mt. St. Helens affected observations. The remaining observations adhere very well to a log-normal distribution for all geographical regions, as indicated by the frequency distribution and probability plots (Figure 3-6). These observations are confirmed by the normality statistics in Appendix C, showing strong log-normality by land use category. Measures of centrality are presented in Appendix D, along with quartile values. Alternative percentile values can be calculated using the mean, count and standard deviation values (logtransformed) given in Appendix C using equation 3 from section 1.
32
Figure 3-6. Normality plots for Copper (a) Frequency distribution (Global data)
(b) Log-normality plot (Global data) 4
70
3 60
Normal Quantile
2 Frequency
50 40 30
1 0
20
-1
10
-2
0
-3 0.3
0.8
1.3
1.8
2.3
2.8
3.3
2.5
3
Log Cu (ug/l)
(c) Log-normality plot (N. Europe)
(d) Log-normality plot (UK)
3
2
2 Normal Quantile
Normal Quantile
1 1 0 -1
0
-1 -2 -3
-2 0
0.5
1
1.5
2
2.5
3
0
Log Cu (ug/l)
0.5
1
1.5
2
Log Cu (ug/l)
Results of t-tests for differences in Cu EMC values by land use are summarised below, with recommendations for the most resolved land use classification for Cu EMC application in the UK and Europe. Recommended Cu EMC values are presented in section 4.
Industrial & Commercial
•
•
A significant difference is found between commercial and industrial land uses in the European data (but only 5 df), but not in the UK (no data) or Global data sets. At the group level, the EMC is significantly lower than that of highways in the UK group (df=4), but not for the other regions.
33
•
•
Residential
• •
•
Highways
• •
•
Selected LU categories
• • •
There are numerous significant differences between industrial and commercial land uses and other land use categories in the European data, and to a lesser extent in the UK data, but none in the Global data. The observed differences are attributed to small sample sizes in the industrial (N=2) and commercial (N=5) categories. It is concluded that ind./comm. cannot be treated as a distinct group. No differences in EMC are discernible between residential density groups. No differences with other land uses are observed for UK data (N=4), but both the European and Global data display significantly lower EMC than that for highways. It is concluded that it is inappropriate to treat residential land use as a distinct group. There are no significant differences between motorways and A-roads. Highways have significantly (better than p > 95%) higher EMC values than most other land uses including ind./comm. and residential use for European and to a lesser extent the UK and Global data. It is concluded that highways should be treated as a separate category. Open (Global data). Developed Urban Main Roads
No significant differences were observed with the urban open data, but the power of the test is low due to a small sample size (N=6 for the Global Region). The NURP study (Athayde et al., 1983) did find a consistent and significant difference between Cu EMC values for open and urban land use, hence it is considered appropriate to recommend that open land is treated as a distinct category. Values for developed urban and highways should be drawn from the European region, where differences are more significant, but where values are broadly consistent with the Global data region. Recommended values are presented in section 4.
3.7
Iron
Sources of iron include corrosion from vehicles and other steel, combustion of coal and coke, iron and steel industry emissions and landfill leachate. Makepeace et al., (1995) report that the acute toxicity of Fe is in the range 0.32 mg/l for mayfly to 16 mg/l for other invertebrates and fish, but that the addition of Fe to lead, copper and zinc reduces the overall toxicity of stormwater. Instantaneous stormwater concentrations range from 0.08 to 440mg/l, and drinking water standards from 0.05 to 0.3 mg/l. Iron in water is generally in the trivalent (ferric) state, rather than the divalent (ferrous state), and is associated mainly with suspended solids, where sediment concentrations range from 1.4 to 128 mg/g. A total of 77 Fe EMC records were obtained, of which only 7 were from Northern Europe, and just one from the UK. A good spread of values by land use class is found in the Global data set at the land use group level, but the small sample size limits the analysis for other regions. Global values ranged from 0.22 mg/l for a residential district of Singapore (Chiu, 1993), to a value of 66.17 mg/l for the A38 Aston expressway, an urban motorway in Birmingham, UK. Although this value was a far outlier, a check of the source (Hedley and Lockley, 1975), indicated that there was no reason to exclude it from the analysis. The minimum European value was 0.76 mg/l for a mixed development in Lelystad, Netherlands (Van den Berg, 1977). 34
Figure 3-7. Normality plots for Iron (a) Frequency distribution (Global data)
(b) Log-normality plot (Global data) 3
20 18
2
Normal Quantile
16
Frequency
14 12 10 8
1
0
-1
6 4
-2
2 -3 -0.8
0
-0.2
0.5
1.1
1.7
Log Fe (mg/l)
(c) Frequency distribution plot (N. Europe)
(d) Log-normality plot (N. Europe)
3
2
Normal Quantile
1
Frequency
2
1
0
-1
0
-2 -0.5
0
0.5
1
1.5
2
Log Fe (mg/l)
The Fe observations adhere reasonably well to a log-normal distribution, as indicated by the frequency distribution and probability plots (Figure 3-7), and the tests for normality in Appendix C. Measures of centrality are presented in Appendix D, along with quartile values. Alternative percentile values can be calculated using the descriptive log-transformed data given in Appendix C, using equation 3 from section 1. Results of t-tests for differences in Fe EMC values by land use are summarised below, including recommendations for the most resolved land use classification for Fe EMC application in the UK and Europe. Industrial & Commercial
•
No significant differences occur between industrial and commercial land or between these uses and any other. N is very small in both UK and European data. Ind./Comm. cannot be treated as a distinct group.
35
Residential
•
No significant differences occur between residential density groups, or between these uses and any other. N is very small in both UK and European data. Residential use cannot be treated as a distinct group.
Roads
•
Most of the high observations observed for Iron are associated with highways. However, there are no significant differences between motorways and A-roads, or between highways and other land uses. It is concluded that highways cannot be treated as a separate category.
Selected land-use categories
•
All Urban
Whilst there are significant differences between the Open category and several others (inc. Ind./Comm., Residential) the power of the tests of difference is low as there are only two observations for the Open category. EMC values cannot clearly be identified for any distinct urban land use, hence the All Urban value is recommended. As the sample size for the European data is very small (N=7), the Global data set value is recommended. Recommended Fe EMC values are presented in section 4.
3.8
Total Lead
Vehicle exhaust emissions are the principle source of Pb in urban stormwater, derived from atmospheric deposition. Tyre abrasion is also a significant vehicular source. Other sources include lead pipes, plastic guttering, lead roofs and flashing, and as an additive in paints and stains, although these are being eliminated. Approximately 1-10% of Pb in urban runoff exists in the soluble phase, and the remainder is associated with suspended sediment. Lead bioaccumulates in aquatic organisms, benthic bacteria, plants, invertebrates and fish, and exerts acute and chronic toxic effects. Acute toxicity of Pb report 96-h LC50 values of 1.17 to 1.47 mg/l, depending on water hardness and pH. Synergistic effects are increased in the presence of Cu or Zn, but reduced by Fe. The chronic toxicity no effect level for rainbow trout ranges between 0.0072 and 0.36 mg/l (Makepeace et al., 1995). The EC guidelines for potable water abstraction is 0.05mg/l, with lower values for aquatic guidelines (e.g. 0.001 mg/l in Canada). Instantaneous stormwater concentrations range from 0.00057 to 26 mg/l. A total of 402 Pb EMC records were obtained, of which 81 were from Northern Europe, and 20 from the UK. A good spread of values by land use class is found. The Global values ranged from 3 ug/l for a motorway in Austin, Texas (Barret and Malina, 1998), to 4680 ug/l for a site addressed by the US EPA NURP (Mustard et al., 1987). The European values ranged from 12 ug/l for a highway in Floda, Sweden (Malmqvist and Hard, 1981) to 2410 ug/l for the A38 motorway in Birmingham, UK (Hedley and Lockley, 1975). The UK minimum was 19 ug/l for an industrial estate on the Forth river, Scotland (Bayes et al., 1994). Several very low values were excluded from the analysis, including those of: Soderlund and Lehtinen (1972) for a Swedish highway (probable printing error in units); Baird et al., (1996) for a residential and commercial runoff in Texas (dissolved Pb, not total Pb); and Heal (1999) influent to a BMP from a residential site (based on only 3 samples hence probably not flow weighted). The Pb observations adhere well to a log-normal distribution (Figure 3-8), however, note the Shapiro-Wilk statistics (Appendix C) which illustrate how the assumption of log-normality becomes progressively stronger as the data is stratified by geographical region, and by land use class.
36
Figure 3-8. Normality plots for Lead (a) Frequency distribution (All Global data)
(b) Log-normality plot (Global) 4
120 3
100 Normal Quantile
2
Frequency
80 60 40
1 0 -1 -2
20
-3
0
-4 0.3
0.8
1.3
1.8
2.3
2.8
3.3
3.8
Log Pb (ug/l)
(c) Log-normality plot (N. Europe) 1.1.1
(d) Log-normality plot (UK) 1.1.1
3
3 2
Normal Quantile
Normal Quantile
2 1
0
-1
1
0
-1
-2
-3
-2 1.0
1.5
2.0
2.5
3.0
1
Log Pb (ug/l)
1.5
2
2.5
3
3.5
Log Pb (ug/l)
Measures of centrality are presented in Appendix D, along with quartile values. Additional percentile values can be calculated using the mean, count and standard deviation values (logtransformed) given in Appendix C, using the equation 3 from section 1. Results of t-tests for differences in Pb EMC values by land use are summarised below, with recommendations for the most resolved land use classification for Fe EMC application in the UK and Europe. Industrial & Commercial
• • •
There is no UK data, but the commercial land use has a higher EMC in both the European (P > 99.9%, 9 df) and Global (not significant) regions. The group EMC is significantly lower than that for most land uses (inc. residential and highways) for all geographical regions. It is concluded that it is appropriate to treat this as distinct land use at both the group and sub-group level.
37
Residential
• •
•
•
Roads
•
•
Selected LU categories
• • •
No differences in EMC are discernible between residential density groups. The residential EMC is significantly higher than Ind./Comm. land use and significantly lower than that for highways, as judged by both group and sub groups in the European and Global regions (insufficient UK data). It is concluded it is appropriate to treat residential land use as a distinct group at the aggregate level. Motorways have a significantly higher EMC than A- roads in both the UK (p > 90%) and Global (p > 99%) regions. Highways have significantly (better than p > 95%) higher EMC values than most other land uses including ind./comm. and residential use. This is apparent for all three geographical regions. It is concluded that it is appropriate to treat roads as a distinct group at the sub-group level. Open (from Global), Residential, Industrial/Commercial Motorways, A-roads
The variation in lead EMC between land use categories is assumed to be a function of vehicle density, with motorways experiencing more traffic than A-roads, and commercial areas more traffic than industrial areas. It is noted however, that since the introduction of unleaded petrol in 1973 lead emissions have declined significantly. Figure 3-9 illustrates this decline for both roads and developed urban land. For the period 1995-9, the EMC is 17% of the long term average (1970-99) for highways, and 30% for developed urban land. It is assumed that that the lower quartile EMC values are therefore the best estimate of current EMC values. The recommended lower quartile Pb EMC values are presented in section 4.
Figure 3-9. Long term change in Pb site mean EMC (a) Developed urban land (N=304)
(b) Highways (N=87)
Pb EMC (ug/l)
400 350
1200 1000
300 250
800
200 150 100
600 400 200
50 0
0 1970-4 1975-9 1980-4 1985-9 1990-4 1995-9
38
1975-9
1980-4
1985-9
1990-4
1995-9
3.9
Total Mercury
Sources of mercury include coal combustion, paint, dental amalgam and the chlor-alkali industry. Mercury is toxic at very low concentrations (e.g. 0.04ug/l chronic toxicity to D. pulex, and 0.52 ug/l for chronic toxicity to trout, both as methyl mercury). Methyl mercury is the most toxic form, and bio-accumulates more readily than inorganic mercury. In urban stormwater, instantaneous concentrations range from 0.012 ug/l to 2.4 ug/l, levels which exceed drinking water and aquatic guidelines (Makepeace et al., 1995). A total of 25 Hg EMC records were obtained, of which 8 were from Northern Europe and none from the UK. In the Global data set, only half of the given values could be allocated to a land use other than the most general built category of developed urban. Minimum values were 0.03 ug/l in the Global set, a composite EMC from twelve towns in the Canadian Great Lakes (Marsalek and Schroeter, 1988) and 0.1 ug/l in Europe, for a mixed developed area in Goteburg, Sweden (Horkeby and Malmquist, 1977). The highest value of 5.27 ug/l was recorded for a residential district in Aix-en-Provence, France (Deutsch and Hemain, 1984). The Hg observations adhere well to a log-normal distribution (Figure 3-10), although this is only apparent in the Global data set as the sample size is too small for the other regions. Appendix C presents the normality statistics, and measures of centrality are presented in Appendix D along with quartile values. Alternative percentile values can be calculated using the descriptive data in Appendix C, using equation 3 from section 1. Figure 3-10. Normality plots for Mercury (a) Frequency distribution plot (Global data)
(b) Log-normality plot (Global data) 3
9 8
2 Normal Quantile
Frequency
7 6 5 4 3 2
1 0 -1 -2
1 -3
0
-2
-1.5
-1
-0.5
0
0.5
1
Log Hg (ug/l)
Results of t-tests for differences in Hg EMC values by land use are summarised below, including recommendations for the most resolved land use classification for Fe EMC application in the UK and Europe. Industrial & Commercial
•
No significant differences occur between industrial and commercial land or between these uses and any other. There is no UK data and little European data, hence Ind./Comm. cannot be treated as a distinct group.
Residential
•
No significant differences occur between residential density groups, or between these uses and any other. There is no UK data and little European data and the residential use cannot be treated as a distinct group.
39
Highways
•
There are no differences between motorways and A-roads, or between highways and any other land use, hence a separate category is not tenable.
Selected land use categories
•
All Urban.
There is only 1 observation of a Hg EMC for Open urban land use (non-European) and this is considered inadequate to define a separate category, given that the principal source of Hg is likely to be atmospheric deposition. As EMC values cannot clearly be identified for any distinct urban land use, the "All Urban" value is recommended. As the sample size for the European data is small, and dominated by two very high values from a Paris residential district (>5 ug/l), the Global data set value is recommended (see section 4).
3.10 Total Nickel Sources of Nickel in stormwater include corrosion of welded metal plating, wear of bearings and bushings and other moving engine parts, electroplating and alloy manufacture and food production. Although an essential element, Nickel is toxic to aquatic life at levels found in urban stormwater, and its toxicity increases when water is soft. Chronic toxicity ranges from 14.8 ug/l for D. magna, to 530 ug/l for fathead minnows in hard water. Nickel in stormwater is generally associated with suspended sediments and organic matter. Instantaneous stormwater concentrations range from 0.001 to 49 mg/l, whilst environmental and health guidelines range from 0.0083 to 1 mg/l (Makepeace et al., 1995). A total of 80 Ni EMC records were obtained, of which 16 were from Northern Europe and only four from the UK. In the largest (Global) data set, only half of the given values could be allocated to a land use other than the most general built category of developed urban. The Global minimum is 3 ug/l for a commercial areas in Fort Worth, Texas (Baird et al., 1996), and a European minimum of 12.97 ug/l for Maurepas, a high density residential area in Paris (Deutsch and Hemain, 1984). The highest value was recorded from an industrial estate on the Forth river in Scotland (Bayes et al., 1994). Figure 3-11. Normality plots for Nickel (a) Frequency distribution plot (Global data)
(b) Log-normality plot (Global data)
1.1.1
3
30 2
20 15 10
Normal Quantile
Frequency
25 1
0
-1
-2
5
-3 0.25 0.5 0.75 1.0 1.25 1.5 1.75 2.0 2.25 2.5
0
Log Ni (ug/l)
40
The Ni observations adhere very well to a log-normal distribution (Figure 3-11), including for the small European data set. Appendix C presents the normality statistics, whilst measures of centrality are presented in Appendix D, along with quartile values. Alternative percentile values can be calculated using the log-transformed descriptive statistics given in Appendix C, using equation 3 from section 1. Results of t-tests for differences in Ni EMC values by land use are summarised below, including recommendations for the most resolved land use classification for Ni EMC application in the UK and Europe. Industrial & Commercial
• • •
No significant differences occur between industrial and commercial land use. The EMC for commercial land use is significantly higher than that for residential land in the European set (N=2), but not in the Global set. The ind./comm. group cannot be treated as a distinct group.
Residential
•
The low residential density group (Global region) has a significantly higher EMC than the high residential group (P > 95%) but the sample size is small (N=2). Residential use cannot be treated as a distinct group.
Highways
•
The highest observations for Ni are associated with highways, but the small sample size precludes identification of any statistically significant differences, hence highways cannot be treated as a distinct group.
Selected land use categories
• •
Urban Open (Global) Developed Urban
EMC values cannot clearly be identified for any distinct urban land use. The low mean value for the Open land use appears reasonable, considering the principal sources of nickel, but is based on only two observations hence caution is recommended. The sample size for the European Developed Urban group (N=13), is considered sufficient to use the European value, particularly as it is broadly consistent with the Global region value. Recommended values are presented in section 4.
3.11 Total Zinc Sources of Zinc includes wear of tyres and brake pads, combustion of lubricating oil, corrosion of building materials and metal objects such as galvanised roof panels. Zinc is less toxic to aquatic life than copper or lead, but it bioaccumulates easily and is more toxic at high concentrations when in soft water. Chronic toxicity is affected by pH. Zinc is mostly associated with dissolved solids but will adsorb to suspended sediment and especially colloidal particles. Instantaneous concentrations in stormwater range from 0.0007 to 22 mg/l (Makepeace et al., 1995). EC guidelines are 0.3 mg/l for salmonid river, 1 mg/l for cyprinid rivers and 3-5 mg/l for potable water abstraction, depending on treatment works capability. A total of 344 Zn EMC records were obtained, of which 76 were from Northern Europe and 24 from the UK. The global values ranged from 5.09 ug/l for a commercial area in the US EPA NURP (Mustard et al., 1987), to 5800 ug/l for an industrial site in Melbourne, Australia (GH and D et al., 1987). The European data ranged from 16.6 ug/l for a road in South Oxhey, Watford, UK (Hamilton et al., 1987) to 3550 ug/l for the A38 urban expressway in Birmingham, UK (Hedley and Lockley, 1975).
41
The Zn observations adhere reasonably well to a log-normal distribution (Figure 3-12), including for the small European data set. The Shapiro-Wilk statistics indicate that lognormality improves as the data is stratified. Appendix C presents the normality statistics, whilst measures of centrality are presented in Appendix D, along with quartile values. Alternative percentile values can be calculated using the log-transformed descriptive statistics given in Appendix C, using equation 3 from section 1. Figure 3-12. Normality plots for Zinc (a) Frequency distribution plot (Global data)
(b) Log-normality plot (Global data) 4
100
3
90
2 Normal Quantile
80
Frequency
70 60 50 40
1 0 -1
30
-2
20
-3
10
-4
0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
Log Zn (ug/l)
(c) Log-normality plot (N. Europe)
(d) Log-normality plot (UK)
3
3 2
1
Normal Quantile
Normal Quantile
2
0
-1
-2
1 0 -1 -2
-3
-3 1.0
1.5
2.0
2.5
3.0
3.5
1.0
Log Zn (ug/l)
1.5
2.0
2.5
3.0
3.5
4.0
Log Zn (ug/l)
Results of t-tests for differences in Zn EMC values by land use are presented in Appendix E. A summary of land use differences is shown below, including recommendations for the most resolved land use classification for Zn EMC application in the UK and Europe.
42
Industrial & Commercial
•
• •
Residential
•
• •
Highways
•
• •
Selected LU categories
• • •
The European commercial EMC value is significantly higher than that for industrial land use (P > 99.9%, 11 df), a pattern not repeated for the UK (no data) or Global regions. The group level EMC value is significantly less than that for residential and highways in all geographical regions. It is concluded that it is appropriate to treat this as distinct land use at the group level. The high density residential land use group has a significantly higher EMC value in both the European and All regions (p > 95%), but not the UK (no data). There are significant differences in EMC value between residential and other land uses at the sub-group level, but not at the group level. It is concluded that it is appropriate to treat residential land use as a distinct group at the aggregate level, due to the observed sub-group differences, and because the very clear differences are observed for the non-residential developed urban values (i.e. the Comm./Ind. values). Motorways have a significantly higher EMC than A-roads in the Global data set (p > 90%) and are consistently higher for the UK and European regions. Motorways have consistently and significantly higher EMC values across all land uses (except open where N is small) for all geographical regions. It is concluded that it is appropriate to treat motorways as a distinct group. Open (Global), Residential, Industrial/Commercial Motorways, A-roads
It is recommended that values from the European data set (N=76) are used in preference to the Global data set (N=344), for European applications. There is generally good agreement between these data sets, and to a lesser extent the smaller (N=24) UK data set. Recommended EMC values are given in section 4. 3.12 Total Phosphorous Total phosphorous is the sum of particulate phosphorous and dissolved or soluble phosphorous. Both of these fractions can be reactive, acid-hydrolysable or organically bound depending on chemical availability (Duncan, 1999). The reactive phosphorous is readily available but powerful oxidising agents are required to release organic phosphorous. Sources of phosphorous include atmospheric deposition, leaf litter, fertilisers, industrial wastes (chemical, food and building materials) and detergents. Phosphorous is an essential nutrient, and where its supply is limiting it may restrict growth. If availability increases excessively plant and algae growth may proceed rapidly causing eutrophication and oxygen depletion. EC guide levels for phosphorous (P 2 O5) are 0.4 mg/l for fisheries and 0.4 to 0.7mg/l for water abstraction depending upon treatment capability. Makepeace et al., (1995) report instantaneous concentration of 0.01 to 7.3 mg/l. A total of 403 P EMC records were obtained, of which 45 were from Northern Europe and just 2 from the UK. The global values ranged from 0.01 mg/l for a developed urban area in Sydney (Sharpin 1993) to 5.5 mg/l for an urban district in South Korea (Yu et al., 1988).
43
European values ranged from 0.08 mg/l for an urban highway in Stockholm (Soderlund and Lehtinen, 1972) to 2 mg/l for a high density residential district of Paris (Deutsch and Hemain, 1984). The P observations adhere very well to a log-normal distribution (Figure 3-13). Appendix C presents normality statistics, while measures of centrality are presented in Appendix D, along with quartile values. Alternative percentile values can be calculated using the log-transformed descriptive statistics given in Appendix C, and equation 3 from section 1. Figure 3-13. Normality plots for Total Phosphorous (a) Frequency distribution plot (Global data)
(b) Log-normality plot (Global data) 4
140 3
120 Normal Quantile
2
Frequency
100 80 60
1 0 -1
40
-2
20
-3 -4 -2.0
0
-1.6
-1.2
-0.8
-0.4
0.0
0.4
Log Total P (mg/l)
(c) Frequency distribution plot (N. Europe)
(d) Log-normality plot (N. Europe) 3
14 12
Frequency
8 6
Normal Quantile
2
10
1
0
4 -1
2 -2 -1.25
0
-1
-0.75 -0.5
-0.25
0
0.25
0.5
Log Total P (mg/l)
Results of t-tests for differences in P EMC values by land use are summarised below, with recommendations for the most resolved land use classification for P EMC application in the UK and Europe.
44
Industrial & Commercial
•
•
•
Residential
•
• •
Highways
• • •
Selected land use categories
• • •
The industrial EMC value is significantly higher than that for commercial land use in the Global region (P > 99.9%, 69 df), a pattern not repeated for the other regions (insufficient data). The group EMC value is greater than that for highways in both the European (p >99%, df 13) and Global regions (significant at the subgroup level only). It is significantly lower than that for residential land use in the Global region (p > 99%), but not in the European region (N=7). It is appropriate to treat industrial/commercial land as a distinct use at the group level, particularly with respect to the Global region. There are no significant differences between EMC amongst the residential density sub-groups, although from the limited available data it appears that higher values are associated with more dense residential use. The residential EMC is significantly greater than that for highways in both the European (p >95%) and Global regions (p > 99%). It is concluded that it is appropriate to treat residential land as a distinct use at the group level, particularly with respect to the Global region. There are no significant differences between motorways and A-roads. Highways and motorways exhibit significantly lower values across most urban land uses, in both the European and Global regions. It is concluded that it is appropriate to treat motorways as a distinct group. Open (Global), Residential, Industrial/Commercial (if Global), Else Developed Urban Motorways, A-roads (Global), else Highways
Differences between land use specific values are generally small, but they are often significant, particularly in the Global region where the sample size is very large (N=403, compared to just N=45 for Europe). It is recommended that values from the Global data set are used in preference to the Europe data set. This is counter to the general advice to use European values for a European application. However, the consistency, in terms of range and distribution of values between these regions means that the Global region values are considered more reliable due to the very much larger sample size, and offers the advantage of a greater sub-division by land use, desirable for hazard mapping. Recommended EMC values are presented in section 4.
3.13 Soluble Phosphorous Details of the nature and sources of phosphorous and its compounds is given above under Total Phosphorous. Particularly significant sources of soluble phosphorous in cold northern hemisphere climates include leaf fall (Kluesner and Lee, 1974) and large roadside gully pots, common in European cities, which can act as anaerobic digesters between storms (Mance and Harman, 1978). Instantaneous soluble phosphorous concentrations in urban stormwater range from 0.0381 to 3.52 mg/l (Makepeace et al., 1995). A total of 108 Soluble P EMC records were obtained, of which only 9 were from Northern Europe and 5 from the UK. The global values ranged from 0.05 mg/l for a high density residential area in the US (Mustard et al., 1987) to 1.79 mg/l for an industrial site in Korea (Yu et al., 1988). European values ranged from 0.32 mg/l from a bus station to 1.32 mg/l for a
45
shopping precinct, both commercial sites in the South Oxhey catchment, North London (Harrop, 1984). Figure 3-14. Normality plots for Total Phosphorous (a) Frequency distribution plot (Global data)
(b) Log-normality plot (Global data) 3
30
2 25
Normal Quantile
Frequency
1 20 15 10
0 -1 -2
5
-3
0
-4 -2.5
-2.0
-1.5
-1.0
-0.5
0.0
0.5
Log Soluble P (mg/l)
The Soluble P observations adhere very well to a log-normal distribution (Figure 3-14). Appendix C presents the normality statistics, and Appendix D the measures of centrality, along with quartile values. Alternative percentile values can be calculated using the logtransformed descriptive data in Appendix C, using equation 3 from section 1. Results of t-tests for differences in soluble P mean site EMC by land use are summarised below, with recommendations for the most resolved land use classification for P EMC application in the UK and Europe. Industrial & Commercial
•
•
Residential
•
•
The industrial EMC value is significantly higher than that for commercial land use in the Global region (P > 90%, 22 df), a pattern not repeated for the other regions (insufficient data). The group EMC value is significantly lower than the high density residential land use (p >90%), and commercial land use is significantly lower (p > 95%) than residential land use, suggesting that commercial and industrial land use contributes less Soluble P than residential areas. Higher residential density is associated with a higher Soluble P EMC, and there are significant differences (p>95%) between the high residential and medium and low density categories. The group residential EMC is significantly higher than that for open, commercial land uses, but there are no differences with highways.
Highways
•
There are no significant differences between motorways and A-roads, and only one significant difference with any other land use (motorways deliver significantly less Soluble P than high density residential areas).
Selected LU categories
•
Open (Global), Residential, Industrial/Commercial (Global, if Hazard mapping the priority), Else Developed Urban (Global, if Load estimation the priority) Highways
• •
46
Sample sizes for both the European and UK data sets are too small ( 95%), although this pattern is not repeated for the other regions where the sample size is small. At the aggregate level, the ind./comm land use (All) has a significantly lower EMC than that for residential land use (p > 99.9%), and highways (p > 95%) at both the aggregate and sub-group level. It is appropriate to treat this as distinct land use at the group level. Residential EMC increases with density, although not significantly. At the group level, the residential EMC is higher than open land use (p > 95%), and ind./comm. use (p > 99.9%), but there are no differences with highways. It is appropriate to treat this as distinct land use at the group level.
48
Highways
• •
•
Selected land use categories
• • •
Motorways have a lower EMC than A-roads, but the difference is not significant Highways have a significantly higher EMC than the ind./comm. land use, and a significantly lower EMC than that of medium density residential development. Highways can tentatively be treated as a distinct class at the group level. Open (Global), Residential, Industrial/Commercial (Global), Highways (Global).
Small sample sizes for the UK and European regions limits the above analysis to the Global region. However, there is broad agreement between EMC values in the Global region (N=270) and the European region with its much smaller sample size (N=31). It is therefore recommended that values from the Global data set are used, especially when hazard mapping, rather than loads estimation, is the priority. Recommended EMC values for total nitrogen are presented in section 4.
3.15 Total Kjheldal Nitrogen TKN is the sum of organic nitrogen and ammoniacal nitrogen. Further details of nitrogen speciation are presented under total nitrogen. Makepeace et al., (1995) report instantaneous stormwater TKN concentrations of 0.32 to 16 mg/l. A total of 187 TKN EMC records were obtained, of which just six were from Northern Europe and only one from the UK. Global values ranged from 0.34 mg/l for an urban open area of Lake George, New York (Athayde, 1983), to 11.02 mg/l for a high density residential district in the USA (Mustard et al., 1987). The European values range from 1 mg/l for a high density residential district in Lelystad, Netherlands (Uunk and van den Ven, 1987), to 6.54 mg/l for a high density residential district in Les Ulis, Paris (Deutsch and Hemain, 1984). Figure 3-16. Normality plots for Total Kjheldal Nitrogen (b) Log-normality plot (Global data)
60
3
50
2
40 30 20
Normal Quantile
Frequency
(a) Frequency distribution plot (Global data)
1 0 -1 -2
10
-3 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2
0
Log TKN (mg/l)
49
The TKN observations adhere very well to a log-normal distribution (Figure 3-16), although the analysis is restricted to the Global data set. Appendix C presents the normality statistics, including the Shapiro-Wilk statistic which illustrate very strong log-normality by land use class. Measures of centrality are presented in Appendix D, along with quartile values. Alternative percentile values can be calculated using the log-transformed descriptive data given in Appendix C, using equation 3 from section 1. Results of t-tests for differences in TKN site mean EMC values by land use are summarised below, with recommendations for the most resolved land use classification for TKN EMC application in the UK and Europe. Industrial & Commercial
• •
•
Residential
•
•
•
Roads
• •
•
Selected LU categories
• • •
There is no difference between industrial and commercial land uses. At the group level, the ind./comm. land use (Global) has a significantly lower EMC than that for residential use (p > 99.9%), and highways (p > 95%) with further similar significant differences at the sub-group level. It is appropriate to treat this as distinct land use at the group level. Residential EMC increases with density. High density residential use has a significantly higher EMC than both medium (p > 95%) and low (p > 90%) density residential use. At the aggregate level, the residential EMC is significantly higher than open land use (p > 99%), and ind./comm. use (p > 99.9%). The only significant difference with highways is with the high density residential group, which is significantly greater than highways at the group (p >99%) and sub-group (p>95%) levels. It is appropriate to treat residential land use as a distinct group at the group level, and tentatively at the density dependent sub-group level. Motorways have a higher EMC than A-roads, but this difference is not significant Highways have a significantly greater EMC than the open (p > 95%) and ind./comm. (p > 95%) land uses, and a significantly lower EMC than the high density residential sub-group (see above). It is appropriate to treat highways as distinct land use at the group level. Open (Global), Residential, Industrial/Commercial (Global), Highways (Global).
Small sample sizes for the UK (N=1) and European (N=6) regions limits the analysis to the Global region. With these very small samples, it is not possible to draw any firm conclusion on the comparison of the European and Global region data, other than to note that the six European values are all within the range of observed values in the Global region, hence there is no reason to assume that it is inappropriate to use the Global region data in a UK/Europe application. Recommended TKN EMC values are presented in section 4.
50
3.16 Oxidised Nitrogen (Nitrate plus Nitrite) Nitrate (NO 3 ) and Nitrite (NO2 ) are oxidised forms of nitrogen (see total nitrogen). Oxidised nitrogen can contribute to eutrophication of surface waters and is associated with an illness of infants, methemoglobinemia, through consumption of potable water. Makepeace et al., (1995) report instantaneous stormwater concentrations for nitrate of 0.01 to 12 mg/l, and 0.02 to 1.49 for nitrite. EC limits for nitrites are 0.01 mg/l for salmonid fisheries and 0.03 mg/l for cyprinid fisheries, and 50 mg/l for nitrates where water is abstracted for potable supply. A total of 155 oxidised nitrogen EMC records were obtained, of which 11 were from Northern Europe and six from the UK. There is little differentiation in the stormwater literature between forms of oxidised nitrogen, with most studies report values as NO2 + NO3 . The global values ranged from 0.1 mg/l for a high density residential area in the USA (Mustard et al., 1987) to 11.5 mg/l for a mixed use urban area in Baltimore (Athayde, et al., 1983). The European values ranged from 0.45 mg/l for a commercial site, the South Oxhey bus station in North London (Harrop, 1984) to 5.72 mg/l for a high density residential area in Les Ulis, Paris (Deutsch and Hemain, 1984). The observations adhere poorly to a log-normal distribution (Figure 3-17), and the ShapiroWilk statistic confirms that the assumption of log-normality must be rejected. The data exhibits a high number of outliers at both ends of the scale. These were checked but cannot be attributed to recording errors or uncharacteristic conditions, and have not been removed. Nonnormality can occur when two sets of normally distributed data with different parameters are considered collectively. This could be the case here if there is inconsistency between studies as to the definition of oxidised nitrogen - most studies explicitly specify NO2 +NO3 , but several do not, and may refer simply to one compound or another. Stratification of the data certainly improves the Shapiro-Wilk statistic, and log-normality is improved in the small European data set, and in the Global data at all land use levels except the residential group, which exhibits a bi-modal distribution. It is concluded that it is appropriate to derive values from the stratified data, particularly given the strong lognormality of the other nitrogenous compounds, and the need for consistency between pollutants. However, particular caution should be exercised when using percentile values for the residential land use. Figure 3-17. Normality plots for Oxidised Nitrogen (Nitrate + Nitrite) (a) Frequency distribution plot (Global data)
(b) Log-normality plot (Global data) 4
70
3 60
Frequency
40 30
Normal Quantile
2 50
1 0 -1
20
-2
10
-3 -1.0 -0.8 -0.5 -0.3 0.0
0
0.3 0.5 0.8 1.0 1.3
Log NO2+NO3 (mg/l)
51
(c) Log-normality plot (All Europe)
(d) Log-normality plot (Ind./Comm., Europe)
2
3 2 Normal Quantile
Normal Quantile
1
0
1 0 -1
-1 -2 -2 -0.5
-3 -0.3
0.0
0.3
0.5
0.8
1.0
-1
-0.8
Log NO2+NO3 (mg/l)
-0.6
-0.4
-0.2
0
0.2
0.4
Log NO2+NO3
(e) Log-normality plot (Residential, Europe)
(f) Log-normality plot (Highways, Europe)
3 2 2
1
Normal Quantile
Normal Quantile
1
0
-1
0
-1
-2
-2
-3 -1.0 -0.8 -0.5 -0.3 0.0
0.3 0.5 0.8 1.0 1.3
Log NO2+NO3
-3 -0.5
-0.4
-0.3
-0.2
-0.1
0.0
0.1
0.2
Log NO2+NO3
Measures of centrality are presented in Appendix C, along with quartile values. Alternative percentile values can be calculated using the log-transformed descriptive statistics given in Appendix D, using equation 3 from section 1. T-tests for differences in organic nitrogen mean site EMC values by land use are summarised below, with recommendations for the most resolved land use classification for organic nitrogen EMC application in the UK and Europe. Industrial & Commercial
•
•
•
Industrial land use has a significantly higher EMC (p > 95%) than that for commercial land use (Global category only due to small sample size in UK/European regions). At the group level, the ind./comm. land use has a significantly lower EMC than that for residential land use (p > 95%), and highways (p > 90%) with further similar significant differences at the sub-group level. It is appropriate to treat this as distinct land use at the group level.
52
Residential
•
•
•
Highways
• •
•
Selected land use categories
• • •
Residential EMC increases with density, and high density residential use has a higher EMC than both medium (p > 95%) and low (p > 95%) density residential land use. At the group level, the residential EMC is higher than ind./comm. use (p > 95%). Differences with highways are density dependent: the low density residential use has lower EMC (p > 95%) and the high density sub-group a higher EMC (p >99%). It is concluded that it is appropriate to treat residential land use as a distinct use at the group level, although the most obvious differences are density dependent. There is no difference in EMC between road types. Highways have a greater EMC than the ind./comm. land use (p > 95%) land uses. Differences with the residential group occur at the sub-group not group level (see above). It is tentatively concluded that highways can be treated as a distinct group at the group level. Open (Global), Residential, Industrial/Commercial (Global) Highways (Global).
Small sample sizes for the UK and European regions limits the analysis to the Global region. With these very small samples, it is not possible to draw any firm conclusion on the comparison of the European and Global region data, other than to note that the UK and European values are all within the range of observed values in the Global region. There is very good agreement between the composite (i.e. developed urban) values for the UK and Global regions, thus it is considered appropriate to use land use specific values from the Global region in a UK application. Note also that observed differences between land uses are similar and comparable between the nitrogenous compounds. Highest EMC values are found in residential developments, particularly high density areas, followed by highways, with ind./comm. areas and urban open land having the consistently lowest values. Recommended TKN EMC values are presented in section 4.
3.17 Ammonia Ammonia is very toxic to aquatic organisms, with acute toxicity experienced by salmonids from 0.083 mg/l, and 1.1 mg/l for cyprinids. Chronic toxicity has been reported for salmonids as low as 0,0017 mg/l (Makepeace et al., 1995). EC Fisheries standards for ammonia are 1 mg/l for total ammonium (NH4 ) and 0.025 mg/l for unionised ammonia (NH3). A total of 84 ammoniacal-N EMC records were obtained, of which 43 were from Northern Europe and 38 from the UK. The global values ranged from 0.04 mg/l for the Chartwell residential district, Auckland (Williamson, 1986) to 5.5 mg/l from the Houston Industrial estate, Scotland (Bayes et al., 1994). The European minimum was also recorded by Bayes et al., (1994) at Red burn, an industrial site in the Forth river catchment, Scotland. Examination of one far outlier, with an exceptionally high value of 20 mg/l led to the removal of all of the ammoniacal-N observations from one commercial site in Scotland (Bayes et al., 1994). At this site, Edinburgh airport, urea is applied as a de-icer to the runway, accounting
53
for unusually high ammonia values. All ammonia values from this site were dropped from the analysis. The remaining observations adhere very well to a log-normal distribution for all geographical regions (Figure 3-18). Appendix C presents the normality statistics, while measures of centrality are presented in Appendix D, along with quartile values. Alternative percentile values can be calculated using log-transformed descriptive statistics given in Appendix C, using equation 3 from section 1.
Figure 3-18. Normality plots for Ammonia (a) Frequency distribution plot (Global data)
(b) Log-normality plot (Global data) 3
25 2
15
10
Normal Quantile
Frequency
20 1
0
-1
5
-2
0
-3 -1.5 -1.3 -1.0 -0.8 -0.5 -0.3 0.0 0.3 0.5 0.8 Log NH3 (mg/l)
(c) Frequency distribution plot (N. Europe) 3
10
2
6 4
Normal Quantile
12
8 Frequency
(d) Log-normality plot (N. Europe)
1 0 -1
2
-2
0
-3 -1.0
-0.8
-0.5
-0.3
0.0
0.3
0.5
0.8
Log NH4-N (mg/l)
Results of t-tests for differences in ammoniacal-N mean site EMC values by land use are summarised below, including recommendations for the most resolved land use classification for ammoniacal-N EMC application in the UK and Europe.
54
Industrial & Commercial
•
•
There are no significant differences in EMC between industrial and commercial land uses. The only significant difference with any other land use is with the low density residential sub-group, which is lower (p > 90%). There is no evidence to suggest that ind./comm. land should be treated as a distinct group.
Residential
•
The low density residential sub-group has a lower EMC than the high density sub-group (p > 95%). There is a general increase in EMC value with residential density, as observed for all the other nitrogenous compounds. There are no significant differences at the group level.
Highways
•
There are no significant differences between motorways and A-roads, and only one significant difference with any other land use (low density residential use has a significantly lower value than highways (P >90%).
Selected LU categories
•
Open (Global) Developed Urban
•
Sample sizes for the European and UK data sets are similar, and approximately half that of the Global region data. The only significant land use differences occur with the low density residential sub-group, which has a low EMC. However the level of significance is not great, and the sample size small (N=4). It is recommended that non-land use specific values are used due to a general lack of observed differences. The European Developed Urban EMC value should be applied for European applications, and the Global EMC value for open land. Caution is required with the open EMC value as there is only one observation.
3.18 Oil and Grease The oil and grease values must be considered as indicative only, as the database covers the determination of oil and grease (any material soluble in an organic extraction solvent) using a variety of solvents. Makepeace et al., (1995) note that oil and grease is a composite of thousands of hydrocarbon species, each with different properties and toxicity's. Oil and grease sources include operation of vehicles and machinery, food processing and preparation, and natural compounds leached from leaf litter. A total of 78 oil and grease EMC records were obtained, of which 24 were from Northern Europe and ten from the UK. The global values ranged from a low of 0.14 mg/l for Les Marais, a high density residential district in Paris (Gromaire-Metz et al., 1999) to a high of 200 mg/l for a stretch of urban motorway in London (Bliss et al., 1979). The European minimum was recorded by Ellis (1989) for a site in the UK (not specified). The Oil and grease observations adhere very well to a log-normal distribution for all geographical regions (Figure 3-19). Appendix C presents the normality statistics, while measures of centrality are presented in Appendix D along with quartile values. Alternative percentile values can be calculated using log-transformed descriptive statistics given in Appendix C, using equation 3 from section 1.
55
Figure 3-19. Normality plots for Oil and Grease (a) Frequency distribution plot (Global data) 3
25
2 Normal Quantile
30
20 Frequency
(b) Log-normality plot (Global data)
15 10
1 0 -1
5
-2
0
-3 -1
-0.5
0
0.5
1
1.5
2
2.5
Log Oil and Grease (mg/l)
(c) Frequency distribution plot (N. Europe)
(d) Log-normality plot (N. Europe)
7
3
6
2 Normal Quantile
Frequency
5 4 3 2
1 0 -1
1
-2
0
-3 -1.0
-0.5
0.0
0.5
1.0
1.5
2.0
2.5
Log O&G (mg/l)
T-tests for differences in oil and grease EMC values by land use are presented in Appendix E. A summary of land use differences is shown below, including recommendations for the most resolved land use classification for oil and grease EMC application in the UK and Europe. Industrial & Commercial
•
•
Residential
•
There are no significant differences in EMC between industrial and commercial land uses. The only significant difference with any other land use is with the A-roads, which have a significantly lower (p > 90%) EMC. There is no evidence to support the treatment of comm./ind. land as a distinct group. There are no significant differences in EMC between residential density groups, or between and residential group and any other land use, hence residential land should not be treated as a distinct group.
56
Highways
•
There are no significant differences between motorways and A-roads, whilst A-roads have a significantly lower EMC value than industrial and commercial land uses, however, the sample very small (N=2).
Selected land use categories
•
Open (from LU-I) Developed Urban
•
Sample sizes for the European and UK data sets are small, and only moderate for the Global region data, limiting the analysis. It is recommended that non-land use specific values are used due to a general lack of observed differences. The European developed urban EMC value should be applied for European applications, and the All EMC value for open land, although caution is required as there is only one observation.
57