Eu risk assessment

Measured levels

• Aquatic compartment (incl. Sediment)

The predicted environmental concentrations (PECs) that are presented in this section are derived from monitoring data frequency distributions of ambient lead exposure concentrations in European surface waters (freshwater, marine) and sediments. Additionally, an overview is given on reported background concentrations of lead in the different environmental compartments.

Methodology
In order to develop a scientifically sound and practicable report, the procedures used in the present report are based on the methods and concepts laid down in the Technical Guidance Document (EC, 2003) for environmental risk assessment in the European Union and on the Combined Monitoring based and Modelling based Priority Setting procedure (COMMPS, 1999).

Monitoring data on the ambient (background and anthropogenic) water concentration of lead for different European surface waters were collected from environmental agencies, scientific literature and extensive databases. A detailed overview of the databases used is given in the report “Probabilistic distribution of lead in European surface waters”.

In general, concentrations in the environment can be affected by a large number of processes that relate to the amount released, the spatial and temporal distributions of the releases, and the results of the action of a large number of transportation and transformation processes on the substance. The likelihood and extent to which these myriad of processes will affect a particular quantity of substance in the environment is essentially random and frequency distributions of exposure concentrations in the environment will therefore most likely be distributed according to a particular model with the log-normal model being the most often observed (Klaine et al.; 1996, Solomon et al., 1996; Solomon and Chappel, 1998).
The following general selection criteria for the collection of monitoring data from surface waters were applied:

To increase the relevance of the monitoring data, only the most recent monitoring data were used for PEC derivation (1999-2004). Older monitoring data were only used if no recent data were available;

All measurements of a specific data set were excluded from further investigation if more than 80% of the measurements fall below the detection limit (COMMPS, 1999);

With respect to the measurements below the detection limit (DL), it was decided to set those entries

Outliers were identified according to the statistical approach proposed in the TGD (EC, 2003), i.e. Log10(Xi) > log10(p.75) + K(log10(p.75) – log10(p.25)) with Xi being the concentration above which a measured concentration may be considered an outlier, pi the value of the i^th percentile of the (non-parametric) distribution and K a scaling factor. A scaling factor K=1.5 is applied, as this value is used in most statistical packages.
Using the statistical computer package @Risk (Palisade Decision Tools) - a computational tool that allows the selection of the best parametric distribution that fits the input data - the distribution that most likely reproduces the monitoring data is identified. The goodness-of-fit tests used for screening the selected distribution are Chi-Square, Kolmogorov-Smirnov and Anderson-Darling. The latter test is mainly focussing on the goodness-of-fit in the tails of the distribution, and is therefore the most appropriate test when 90^th percentiles are considered. Acceptance or rejection of a fitted distribution is dependent on the critical value of that fit. Critical values for the Anderson-Darling test are calculated using Monte-Carlo simulations (Stephens, 19974, 1977), with special cases for the Normal, Exponential, Weibull and Extreme Value distributions. For all other distributions, the critical values are estimated using an “all parameters known” test that is more conservative than the tests for the above-mentioned specific distributions. The reported confidence, α-1, is the probability that the null hypothesis (i.e. the input data are created by the distribution function reported by BestFit) is rejected when it is, in fact, true. When a median value is calculated, the goodness-of-fit test used for screening the selected distribution is Kolmogorov-Smirnov.
Non-parametric distributions were used when no parametric distribution could be fitted significantly (p<0.05) to the data points.
From the produced distributions it is possible to assign probabilities to the likelihood that a measure will exceed a certain value. This principle can also be applied to concentrations of substances in the environment, taking into account, however, that these data are usually censored by the limits of analytical detection. The reasonable worst-case ambient PECs (RWC-ambient PEC) for the different surface waters is computed as 90^th percentiles of the measured lead concentration in the sampled surface waters, which is in agreement with the procedures as described in the TGD (EC, 2003). The principle of deriving the PEC from the computed frequency distribution of exposure concentrations is illustrated in Figure 3.1.9-2. This cumulative distribution function, hereafter called CDF, can be used to estimate the likelihood that a particular concentration of the substance will be exceeded in the environment.



Figure 3.1.9 17 Cumulative distribution function and derivation of the 90^th percentile
Regional RWC-ambient PEC concentrations can be derived as follows for the different water catchments or countries/regions:
RWC-ambient PEC_country = median value of all 90^th percentiles that have been derived for the different sites, rivers/catchments or regions.
It is not always feasible to perform this type of data treatment:

no 90P can be calculated when insufficient data points for a specific site are available: the use of the above-mentioned statistical approach is only possible with 3 or more measurements;

the availability of information on 100 to 1000s different sites/rivers within a specific region: a thorough analysis of such a database is very time-consuming and is therefore not always feasible (e.g. the Swedish data set).
When analysis of monitoring data according to the recommended procedure in the TGD (EC, 2003) can not be performed, a river- or region-specific approach is applied: data for one river, river system or region within a country are grouped and a 90th percentile is calculated.
The TGD (EC, 1996; 2003) states that the regional RWC-ambient PEC should be the mean value of all site/area-specific 90P-values. This approach, however, assumes that none of the data points is affected by any point sources: environmental parameters are considered to be log-normal distributed, in which case the mean and median value of log-transformed monitoring data are the same. However, the effect of point source contamination – often too small to be detected with the conventional outlier-analysis, will stretch the upper part of the log-distribution to the right, resulting in a higher mean but will affect much less the median. In a case where some water samples are influenced by point source contamination, the 50^th percentile (median value) of the fitted distribution is the best estimate for a mean value without any impact of point sources. Therefore both the mean and median are reported in this study. A schematic overview of the followed methodology is presented in Figure 3.1.9-3.


Figure 3.1.9 18 Schematic overview of RWC-ambient PEC-derivation
Background concentrations of Pb in the freshwater environment (water & sediment)
In order to interpret the determined ambient lead concentrations, it is important to evaluate the data in view of background reference concentrations. “True” natural background concentrations can hardly be found in most European surface waters as a result of historical and current anthropogenic input from diffuse sources. This issue was discussed for the EU Water Framework Directive by a group of experts and the following definition was agreed: “The background concentration of target metals in the aquatic ecosystems of a river basin, river sub-basin or river basin management area is that concentration in the present or past corresponding to very low anthropogenic pressure. The methodologies proposed for setting the background concentrations were: (1) trace metal concentrations in groundwater (shallow and/or deep); (2) analysed values for trace metal concentrations in pristine areas (with assurance that river basin is pristine or nearly so) (3) expert judgment (incl. international agreements; river basin commissions) (EAF, 2004). A draft working document discussed further the approach and stated that the first step in this process is to elucidate default background concentrations applicable to a large part of Europe. It was agreed that the most important database is the FOREGS Geochemical Baseline Programme (FGBP) published in March 2004 (http://www.gsf.fi/foregs/geochem/). FOREGS (Forum of European Geological Surveys) Geochemical Baseline Programme sought to provide high quality environmental geochemical baseline data for Europe based on samples of stream water, stream sediment, floodplain sediment, soil, and humus collected all over Europe. High quality and consistency of the obtained data were ensured by using standardised sampling methods and by treating and analysing all samples in the same laboratories. Five random points were selected in each Global Terrestrial Network cell (160*160 km2), one point in each quadrant and one point random in the cell. The points were used to select the five nearest small drainage basins of <100 km2. The sampling sites selected for stream water analyses of dissolved metals were typical of locally unimpacted or slightly impacted areas. As a consequence, the metal concentrations – and lead more specific – that are determined in these samples can be considered as relevant background concentrations. These lead concentrations are fundamentally different from the values that were used for the derivation of a RWC-ambient PEC: the surface waters that were used for the RWC-ambient PEC did not represent pristine areas, but only excluded locations that were directly impacted by local point sources.
The FOREGS-data set is considered to be of high quality: a detailed description of sampling methodology, sampling preparation and analysis is given by Salminen et al. (2005):

running stream water was collected form small, second order drainage basins (<100 km²);

whenever possible, sampling was performed during winter and early spring months, and was avoided during rainy periods and flood events;

a full description of sampling materials and sampling volumes is provided, and all materials were rinsed twice with unfiltered or filtered stream water (depending on the type of water sample);

all potential contaminating factors were reduced during the sampling period (wearing of gloves, no smoking in the area allowed, no hand jewelry was allowed , running vehicles during sampling was prohibited, etc..)

If it was not possible to use non-metal equipment (e.g., spades, sieves), unpainted steel equipment was used for sediment sampling (no aluminium or brass). A composite sample was made from subsamples taken from beds of similar nature (ISO-5667-12, 1995), and minimum amount of sediment sample was 0.5 kg dry wt. A detailed description of the whole sampling procedure is given in Salminen et al. (2005).

The programme resulted in 807 stream water samples and 845 sediment samples spread over Europe. Water samples were analysed by ICP-MS (detection limit 0.005 µg/L), and dissolved lead ranged between <0.005 and 10.6 µg Pb/L, with a 50^th percentile of 0.093 µg Pb/L. Taking into account the high quality of the data set, this value is accepted as a typical background concentration for Pb in European surface waters (EU-regional scale).

Determination of the Pb-content in the collected sediment samples was done by ICP-AES, aqua regia destruction, Detection Limit: 3 mg/kg). Lead concentrations ranged between <3 and 4880 mg/kg dry wt, with a 50^th percentile of 14 mg/kg. Taking into account the high quality of the data set, this value is accepted as a typical background concentration for Pb in European freshwater sediments (EU-regional scale).

Table 3.1.9-6 presents the country-specific background levels of Pb in surface water and sediment that were determined with the raw data from the FOREGS program.

Table 3.1.9 64: Country-specific 10^th, 50^th and 90^th percentiles of Pb-background levels in surface water and sediment.

Country	Surface water (µg Pbdiss/L)			Sediment (mg Pb/kg dry wt)
	10^th percentile	50^th percentile	90^th percentile	10^th percentile	50^th percentile	90^th percentile
Austria	0.016	0.039	0.157	3.1	8.9	25.9
Belgium	0.063	0.093	1.186	17.0	29.8	71.7
Czech R.	0.061	0.094	1.569	19.1	40.7	122.1
Germany	0.012	0.079	0.548	9.0	17.7	40.4
Denmark	0.034	0.145	0.596	2.4	7.6	21.8
Estonia	0.040	0.070	0.148	1.4	4.7	15.9
Spain	0.014	0.059	0.245	6.5	15.7	38.0
Finland	0.048	0.173	0.405	1.3	5.4	22.8
France	0.031	0.134	0.588	7.4	17.7	42.4
Greece	0.015	0.026	0.057	7.1	12.0	18.8
Croatia	0.046	0.064	0.166	9.9	12.8	19.2
Hungary	0.035	0.067	0.127	2.9	8.4	24.3
Ireland	0.099	0.178	0.582	7.4	19.1	78.3
Italy	0.024	0.099	0.793	4.7	18.3	71.0
Lithuania	0.082	0.201	0.829	7.9	10.8	17.7
Latvia	0.100	0.253	0.666	1.2	4.1	14.5
The Netherlands	0.085	0.138	1.253	1.3	6.9	37.4
Norway	0.015	0.052	0.395	2.2	8.2	30.3
Poland	0.034	0.070	0.220	3.0	8.8	25.7
Portugal	0.011	0.099	0.899	13.2	19.4	35.7
Sweden	0.015	0.100	0.656	4.4	12.3	34.5
Slovakia	0.045	0.096	0.277	13.7	19.8	35.4
Slovenia	0.101	0.138	0.539	11.7	16.2	25.1
United Kingdom	0.030	0.129	0.449	14.6	33.8	83.0