Contract n° SSPI-2004-006538
BR B RI ID D GE G E
Ba B ac ck kg gr ro ou un nd d c cR Ri it te er ri ia a f fo or r t th he e I ID De en nt ti if fi ic ca at ti io on n o of f G Gr ro ou un nd dw w at a te er r th t hr rE Es sh ho ol ld ds s
Research for Policy Support WP3
D17: Report on the integrated data aggregation methodology
Due date of deliverable: 31 September 2006 Actual submission date: 25 October 2006
The deliverable authors are responsible for the content
Start date of the project : 1 January 2005 Duration : 2 years
AUTHORS: Andreas Scheidleder, Steffen Uhlig (quo data), Claudia Schramm AFFILIATION: UBA
ADDRESS: Spittelauer Lände 5, A-1090 Wien, Austria TEL.: +43-(0)1-31304 3541
EMAIL: andreas.scheidleder@umweltbundesamt.at FURTHER AUTHORS: Ariane Blum, Jan Bronders
Revision [1]
Project co-funded by the European Commission within the Sixth Framework Programme (2002-2006) Dissemination level
PU Public
Content
1 Introduction ... 3
2 Requirements of the legislation ... 4
2.1 Water Framework Directive...4
2.2 Groundwater Daughter Directive – Common position adopted by the Council (23 January 2006) ...4
3 Discussion of methods of data aggregation ... 6
3.1 Methods in use or under investigation in Member States ...6
3.2 Description and statistical assessment of candidate methods for data aggregation...6
Estimation parameters ... 7
Estimation methods (aggregation methods)... 8
4 Proposed aggregation methods and procedure ... 12
4.1 Legal requirements ... 12
Need for data aggregation ...12
Extent of exceedance ...12
Location of exceedance ...12
Level of aggregated concentration ...13
Confidence in the assessment ...13
4.2 Proposed aggregation methods ... 13
Extent of exceedance ...13
Location of exceedance ...14
Level of aggregated concentration ...14
Confidence in the assessment ...14
Preliminary network check – Division of sub-bodies...14
4.3 Proposed aggregation 3-step procedure ... 15
Step 1 – any exceedance? ...15
Step 2 – network homogeneous ...15
Step 3 – sub-body consideration ...16
4.4 Thoughts on the assessment of chemical status – combination of results ... 17
5 Virtual case study – demonstration of methodology ... 18
5.1 Characterisation of test GW-body ... 18
5.2 Test calculations of aggregation methods ... 19
Preliminary Network check ...19
Step 1 – any exceedance? ...19
Step 2 – network homogeneous ...19
Step 3 – sub-body consideration ...20
Location of exceedance ...21
6 References ... 22
1 Introduction
In parallel to the combined approach for the derivation of environmental threshold values (Act. 3.4) an integrated data aggregation methodology, taking into account the regulations of the WFD and the draft Groundwater Daughter Directive, had to be developed. Corresponding to the data aggregation methodology the requirements on the network design and sampling strategies will be defined and matched with Act. 3.2. The applicability of the aggregation methodology was demonstrated by test runs for a ‘virtual’ case study representing the variance of hydrogeological environments and possible receptors like associated surface waters or dependent terrestrial ecosystems.
2 Requirements of the legislation
2.1 Water Framework Directive
Provisions for data aggregation in the WFD (Annex V Section 2.4.5) are as follows:
In assessing status, the results of individual monitoring points within a groundwater body shall be aggregated for the body as a whole. Without prejudice to the Directives concerned, for good status to be achieved for a groundwater body, for those chemical parameters for which environmental quality standards have been set in Community legislation:
– the mean value of the results of monitoring at each point in the groundwater body or group of bodies shall be calculated; and
– in accordance with Article 17 these mean values shall be used to demonstrate compliance with good groundwater chemical status.
2.2 Groundwater Daughter Directive – Common position adopted by the Council (23 January 2006)
The procedure for assessing groundwater chemical status is laid down in Article 4.
2. A body or a group of bodies of groundwater shall be considered to be of good chemical status when:
– (a) the values for the groundwater quality standards listed in Annex I and the relevant threshold values established in accordance with Article 3 and Annex II are not exceeded at any monitoring point in that body or group of bodies of groundwater; or
– (b) the value for a groundwater quality standard or threshold value is exceeded at one or more monitoring points but an appropriate investigation in accordance with Annex III confirms that:
(i) on the basis of the assessment referred to in paragraph 3 of Annex III, the
concentrations of pollutants exceeding the groundwater quality standards or threshold values are not considered to present a significant environmental risk, taking into account, where appropriate, the extent of the body of groundwater which is affected;
(ii) the other conditions for good groundwater chemical status set out in Table 2.3.2 in Annex V to Directive 2000/60/EC are being met, in accordance with paragraph 4 of Annex III to this Directive;
(iii) where appropriate, the requirements of Article 7(3) of Directive 2000/60/EC are being met, in accordance with paragraph 4 of Annex III to this Directive;
(iv) the ability of the body of groundwater or of any of the bodies in the group of bodies of groundwater to support human uses has not been significantly impaired by pollution.
The assessment procedure is furthermore specified in Annex III:
1. The assessment procedure for determining the chemical status of a body or a group of bodies of groundwater will be carried out in relation to all bodies or groups of bodies of groundwater characterised as being at risk and in relation to each of the pollutants which contribute to the body or group of bodies of groundwater being so characterised.
2. In undertaking any investigations referred to in Article 4(2)(b), Member States will take into account:
(a) the information collected as part of the characterisation to be carried out in
accordance with Article 5 of Directive 2000/60/EC and with Sections 2.1, 2.2 and 2.3 of Annex II thereto;
(b) the results of the groundwater monitoring network obtained in accordance with Section 2.4 of Annex V to Directive 2000/60/EC; and
(c) any other relevant information including a comparison of the annual arithmetic mean concentration of the relevant pollutants at a monitoring point with the groundwater quality standards set out in Annex I and the threshold values set by Member States in accordance with Article 3 and Annex II.
3. For the purposes of investigating whether the conditions for good groundwater chemical status referred to in Article 4(2)(b)(i) and (iv) are met, Member States will, where relevant and necessary, and on the basis of appropriate aggregations of the monitoring results, supported where necessary by concentration estimations based on a conceptual model of the body or group of bodies of groundwater, estimate the extent of the body of groundwater having an annual arithmetic mean concentration of a pollutant higher than a groundwater quality standard or a threshold value.
4. For the purposes of investigating whether the conditions for good groundwater chemical status referred to in Article 4(2)(b)(ii) and (iii) are met, Member States will, where relevant and necessary, and on the basis of relevant monitoring results and of a suitable conceptual model of the body of groundwater, assess:
(a) the amounts and the concentrations of the pollutants being, or likely to be, transferred from the body of groundwater to the associated surface waters or directly dependent terrestrial ecosystems;
(b) the likely impact of the amounts and concentrations of the pollutants transferred to the associated surface waters and directly dependent terrestrial ecosystems;
(c) the extent of any saline or other intrusions into the body of groundwater; and
(d) the risk from pollutants in the body of groundwater to the quality of water abstracted, or intended to be abstracted, from the body of groundwater for human consumption.
3 Discussion of methods of data aggregation
In parallel to the combined approach for the derivation of environmental threshold values (Act. 3.4 in WP3) an integrated data aggregation methodology, taking into account the
regulations of the WFD and the Common position adopted by the Council 12062/1/2005, was developed.
In a first step methods already in use and methods currently under investigation in different MS were listed. The results of CIS WG 2.81 provided an important basis for the document in hand. Subsequently, the methods are described. Corresponding to the data aggregation methodology the requirements on the network design and sampling strategies were defined and matched with Act. 3.2. Several scenarios for data aggregation on the spatial base of GW-bodies were examined. A “virtual case study” was calculated and the results of different methods of data aggregation were compared. Out of these investigations a stepwise
approach for data aggregation on the spatial base of GW-bodies or groups of GW-bodies is proposed.
According to the draft GWD Annex III 2(c) the annual arithmetic mean concentration of the relevant pollutant at each monitoring point is the basis for the aggregation on the level of a GW-body or a group of GW-bodies.
3.1 Methods in use or under investigation in Member States
A survey within WG 2.8 showed that the following data aggregation methods are applied by Member States on the level of GW-bodies for assessing the chemical status:
- median,
- arithmetic mean,
- weighted arithmetic mean,
- mean based on the log-normal distribution,
- percentage,
- maximum and minimum values.
It should be noted that maximum and minimum values were considered as accompanying parameters, but not as parameters reflecting the overall chemical status of a GW-body.
Furthermore, the treatment of measurements below the limit of quantification (LOQ), the limit of determination (LOD) and the treatment of unequally distributed sites was not specifically addressed by the Member States.
3.2 Description and statistical assessment of candidate methods for data aggregation
In order to understand the differences and the pros and cons of the different aggregation methods, several estimation parameters are considered within this section. These parameters are unknown but may be interpreted as the true parameter which has to be estimated by the aggregation method. Ideally, the result of the aggregation and the
estimation parameter are equal, but in practise there is always a difference, i.e. an estimation error. This is due to the limited number of monitoring points.
1 Technical Report No 1: The EU Water Framework Directive: Statistical aspects of the identification of groundwater pollution trends, and aggregation of monitoring results. 2001
Estimation parameters
The following different estimation parameters may be considered:
1. What is the percentage (extent) of the GW-body (volume) that is not exceeding a quality standard or TV?
2. What is the true average concentration in a GW-body (average over the volume)?
3. What is the geometric mean in the GW-body of the annual mean concentrations throughout the GW-body? (= antilog of the average of the logged annual mean concentrations)?
4. What is the true concentration value that is not exceeded by the annual mean of x % (50 %, 70 %, 90 %) of the monitoring points in a GW-body?
5. What is the maximum annual mean concentration in the GW-body?
These estimation parameters can be derived from the distribution function F(x)
F(x) = (volume of the GW-body where concentration is below x) / (total volume of GW-body) F(x) can be used to calculate the ratio share of the GW-body below any concentration x. Let x=TV=threshold: F(TV) represents the share (extent) of the GW-body at which the
concentration is below the threshold value TV, i.e. F(TV) represents the percentage of the volume which is not exceeding a TV. Not only the percentage volume not exceeding, but also mean, median, 90 percentile and geometric mean can be derived from F(x) mathematically or graphically, see Figure 1.
Apparently the different estimation parameters may be interpreted as different aspects of the distribution function F(x). It is a political question, which aspects and which estimation
parameters shall be considered.
The percentage volume of the GW-body which is exceeding a standard or TV reflects the extent of pollution. However, it does not reflect the pollution in terms of concentration.
Percentiles reflect the overall status of the GW-body fair to good if the variation coefficient of monitoring point mean values is not too high but they do neither reflect outliers nor the impact of uneven distribution of pollution caused by local or diffuse sources, observed at some monitoring points in a GW-body which show higher concentrations than the rest of the GW-body.
The mean reflects the overall status of the GW-body very well if there is no extreme local pollution. The geometric mean is less often in use. It is very sensitive to small concentrations and less sensitive to higher concentrations.
Figure 1: Distribution function for nitrate in a GW-body. Read: 50 mg/L is not exceeded in approx. 45% of the GW-body. Or: in 70% of the GW-body the concentration is below 70 mg/L.
Nitrate distribution of annual mean concentration throughout the GW-body (Marchfeld, 2005)
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
mg/L
Estimation methods (aggregation methods)
In the following the aggregation methods are described and discussed in relation to the corresponding estimation parameter. For some of these aggregation methods also the statistical confidence intervals are available. These are referred to as confidence limit (CL).
Percentage of sites / volume ration (percentage of volume) not exceeding a standard or TV
A straightforward measure is the percentage of monitoring points in a GW-body not exceeding standards or TVs. Instead of using the un-weighted percentage of sites the weighted percentage of points could be applied, where the different sites are weighted with regard to the assigned volume or area in the GW-body.
A third option is to replace the yes/no (1/0) response for each monitoring point by a statistical estimate for the percentage of the area assigned to this monitoring point, which is not
exceeding standards or TVs. This measure is referred to as volume ratio of non
exceedance. If the annual mean concentration is far below a standard or TV, this percentage will be close to zero, whereas for values clearly above a standard or TV the percentage will be close to 1. If the annual mean concentration equals a standard or TV, the percentage of the area assigned to the monitoring point will be 0.5.
The calculation of weights of the annual mean concentrations for the calculation of the volume ratio is derived from the normal distribution with a standard deviation of 25% which means that for values which are 2 times the standard deviation above a standard or TV the weight is nearly 1 (i.e. 0.98) (Example see Figure 3).
Both the un-weighted and the weighted percentage of sites not exceeding a standard or TV provide an estimate of the percentage of the GW-body not exceeding a standard or TV. The un-weighted percentage may be biased in case of inhomogeneous network or unequal share of each monitoring point. The weighted percentage corrects for this bias. However, both percentages are not very precise with regard to random error: in some years the percentage
will be too small, in others too large. This is especially true when the number of monitoring points is small and the annual mean concentrations are close to a standard or TV. This is demonstrated in the following example. The following Figure 2 represents the outcome of two aggregation methods for estimating the percentage of a GW-body not exceeding a standard or TV based on a time series of 1993–2005 and three monitoring points within a GW-body. In this example the standard or TV was selected to be 50. One monitoring point shows values far beyond 50 and two monitoring points show values around the standard or TV of 50; in some years the measured values are slightly exceeding 50 in others they are slightly below.
It is apparent that due to this random fluctuation the result of the percentage of sites not exceeding (violet line) is varying considerably, whereas the volume ratio (blue line) is apparently less affected by random effects.
Figure 2: Two aggregation methods for calculating the percentage of monitoring points not exceeding a quality standard or TV.
0 10 20 30 40 50 60 70
1 2 3 4 5 6 7 8 9 10 11 12 13
Year
%
Volume Ratio not exceeding
Percentage of Sites not exceeding
Time series of 3 monitoring points
year 1 2 3 4 5 6 7 8 9 10 11 12 13
Point 1 1.8 1.0 1.8 3.3 5.0 4.1 3.6 3.3 3.3 2.7 2.7 2.0 1.7 Point 2 44.2 46.3 48.4 53.9 55.4 51.1 46.8 48.6 44.6 49.7 47.7 51.0 39.8 Point 3 40.8 40.6 42.5 61.2 46.1 54.2 47.1 50.5 52.9 51.9 54.0 57.2 57.6
Figure 3: Weights of annual mean values of monitoring points for volume ratio calculation.
Weighting of Volume Ratio SD=25% and TV=50mg/L
0%
20%
40%
60%
80%
100%
10 20 30 40 50 60 70 80 90
mg/L
Weight
Percentiles (50 (=Median), 70 or 90 Percentile)
The percentiles or quantiles (50 %, 70 %, 90 %) of the annual arithmetic mean
concentrations of monitoring points represent the annual mean concentration which is not exceeded at more than x % (50 %, 70 %, 90 %) of the monitoring points of the area (=GW- body).
The estimation is good in case of evenly distributed sites, but may be misleading in case of unevenly distributed sites. However it is possible to derive weighted percentiles in order to correct for different sub-bodies with different site density.
Methodologies of percentiles would require at least 10 sites for statistically sound results. In several countries small GW-bodies exist, which are monitored by less than 10 sites and cannot be grouped due to different hydrogeological conditions.
Arithmetic Mean (AM) and Confidence Limit (CLAM)
The arithmetic mean of the annual arithmetic mean concentrations over all monitoring points sampled represents simply the average over these sites.
Confidence limits (CL) for the mean are an interval estimate for the mean. Instead of a single estimate for the mean, a confidence interval generates a lower and upper limit for the mean.
The interval estimate gives an indication of how much uncertainty there is in the estimate of the true mean. The narrower the interval, the more precise is the estimate.
The upper confidence limit depends on the variability of the concentration level within the GW-body and on the number of monitoring points. The CL decreases with an increasing number of monitoring points within the GW-body or a decreasing variability of concentration levels. The use of the CL allows for reducing the number of monitoring points in GW-bodies with levels far below any standard or TV and enforces a higher number of monitoring points in GW-bodies with levels close to a standard or TV.
The arithmetic mean reflects the overall status of the GW-body very well in case of evenly distributed sites. The outlier sensitivity of the arithmetic mean is poor and its reflection of the impact of uneven distribution of pollution caused by local or diffuse sources, observed at some points in a GW-body which show higher concentrations than the rest of the points in a GW-body, is fair to good. The replacement of values below LOQ by substitute values may introduce some bias. All measurements are assumed to be stochastically independent and identically distributed.
Compliance with good groundwater chemical status at a given level of confidence can be demonstrated with a statistical test for the null hypothesis
H0: "GW-body is not in good status, i.e. true mean level exceeding a standard or TV"
and the alternative hypothesis
H1: "GW-body is in good status, i.e. true mean level below a standard or TV"
H1 may be considered as statistically proven at significance level alpha/2, if the
corresponding upper CL at confidence level (1-alpha) (e.g. 95 %) is below the limit. Value alpha denotes the probability of making a wrong decision for a good status (although the true, unknown mean exceeds a standard or TV); alpha might vary for different parameters.
Weighted Arithmetic Mean (wAM) and Confidence Limit (CLwAM)
The weighted arithmetic mean (wAM) is another estimator for the average annual mean concentration and is introduced for GW-bodies which can be divided into sub-bodies. It takes into account the share of the sub-bodies and the corresponding arithmetic means.
In order to be consistent with the AM, the calculation of the CLwAM is performed under the same model assumption as with CLAM, i.e. all measurements are assumed to be
stochastically independent and identically distributed. Hence it is assumed that there is no spatial correlation at all, and under this assumption calculation leads to the approximate upper confidence limit. If the sites are evenly distributed, calculation shows that CLwAM = CLAM.
Kriging Mean (KM) and Confidence Limit (CLKM)
The Kriging mean represents the average concentration in the area of a GW-body and takes regard of a heterogeneous distribution of monitoring points.
Both Kriging mean (KM) and arithmetic mean (AM) are methods that can formally be written w1y1 + w2y2 + ... + wnyn
where yi denotes the level of monitoring point i and wi the corresponding weight. For the AM the weights of sites are equal, wi = 1/n. For the KM the weights can be quite different:
monitoring points representing large areas have higher weights than monitoring points representing small areas. For evenly distributed sites these areas and hence the corresponding weights are similar.
The Kriging analysis can also be used for calculating the upper confidence limit for the Kriging mean.
Kriging reflects the spatial structure of the GW-quality data and also to a certain extent the impact of factors affecting the concentration level within an area as there are e.g. land use, hydrogeological conditions, etc. if these are spatially correlated. Anyway, the extension of the model to explicitly include hydrogeological information, etc. requires a much more
complicated statistical algorithm.
Mean based on the log-normal distribution
The mean based on the log-normal distribution equals the geometric mean. This is equivalent with the antilog of the average of the logged annual mean concentrations.
The estimation is less sensitive with regard to outliers, but much more sensitive with regard to the treatment of measurements below the LOQ.
Minimum, Maximum
Minimum and maximum reflect the impact of uneven distribution of pollution caused by local or diffuse sources, the overall status of the GW-body is poorly reflected. Furthermore, the method is highly outlier sensitive.
4 Proposed aggregation methods and procedure
4.1 Legal requirements
Within the assessment of the chemical status of the groundwater the WFD and the draft GWD request for the consideration of the following elements:
- the extent of an exceedance,
- the location of the exceedance,
- the level of the concentration, and
- the confidence in the assessment.
The legal basis of the listed elements is detailed below and all the elements listed are addressed by the proposed aggregation methodology.
Need for data aggregation
Basis of the assessment whether a GW-body or a group of GW-bodies is in good chemical status is the comparison of the annual arithmetic mean concentration of the relevant
pollutants at a monitoring point with the groundwater quality standards set out in Annex I and the threshold values set by Member States in accordance with Article 3 and Annex II. [Annex III 2(c)]
The need for data aggregation on the level of a GW-body or parts within a GW-body is mentioned in Annex III 3 …Member States will, where relevant and necessary, and on the basis of appropriate aggregations of the monitoring results, …, estimate the extent of the body of groundwater having an annual arithmetic mean concentration of a pollutant higher than a groundwater quality standard or a threshold value. [Annex III 3]
Extent of exceedance
According to Article 4 (2)(a) a GW-body is in good status when a quality standard and relevant threshold values is not exceeded at any monitoring point, i.e. the extent of exceedances is 0.
In case of any exceedance at one or more monitoring points further investigation is needed.
Article 4 (2)(b)(i) of the draft GWD requests where appropriate, the extent of the body of groundwater which is affected to be taken into account when assessing whether the
exceedance of a quality standard or a TV at one or more monitoring points does not present a significant environmental risk and classifies a GW-body to be considered in good status.
Location of exceedance
The consideration of the extent of the body of groundwater which is affected should be supported by the consideration of the location of the extent of a GW-body which is affected.
The location of the extent of an exceedance is relevant for assessing according to Annex III 4 GWD the protection of dependent terrestrial and aquatic ecosystems by:
(a) the amounts and the concentrations of the pollutants being, or likely to be,
transferred from the body of groundwater to the associated surface waters or directly dependent terrestrial ecosystems; [Annex III 4 (a)]
(b) the likely impact of the amounts and concentrations of the pollutants transferred to the associated surface waters and directly dependent terrestrial ecosystems; [Annex III 4 (b)]
Furthermore, the consideration of the location of an extent of an exceedance is linked to the GWD requirement of Article 4 (2)(b)(iv) and referring to Article 7(3) WFD where it is laid down that an exceedance does to not significantly impair the ability of a GW-body to support
human uses by assessing according to Annex III 4:
(d) the risk from pollutants in the body of groundwater to the quality of water abstracted, or intended to be abstracted, from the body of groundwater for human consumption. [Annex III 4 (c)]
Level of aggregated concentration
In Annex III 3 of GWD the level of the concentration of a pollutant is requested to be considered in the assessment whether an identified extent of exceedances poses a significant environmental risk resulting in the whole GW-body to be assigned as in poor status.
…Member States will, where relevant and necessary, and on the basis of appropriate aggregations of the monitoring results, …, estimate the extent of the body of groundwater having an annual arithmetic mean concentration of a pollutant higher than a groundwater quality standard or a threshold value. [Annex III 3]
Confidence in the assessment
Finally, according to Annex V 2.4.1 of WFD …Estimates of the level of confidence and precision of the results provided by the monitoring programmes shall be given in the plan. In order to address this requirement properly it is proposed to consider the confidence of the assessment of monitoring results within the proposed aggregation methodology. This means that to the estimated level of aggregated concentrations of substances the analytical
uncertainty, the uncertainty due to the monitoring network and the uncertainty due to the variation of concentrations within a GW-body or a group of GW-bodies has to be added.
4.2 Proposed aggregation methods
This overview describes aggregation methods for the level (percentiles, arithmetic mean, weighted arithmetic mean and Kriging mean), for the extent (percentages of sites / volume ratio not exceeding a quality standard or TV), and for the confidence limit of the respective parameters.
Basis of the assessment is … the annual arithmetic mean concentration of the relevant pollutants at a monitoring point… [Annex III 2(c)].
Extent of exceedance
The estimation of the extent of exceedances should be performed with simple weighted or un-weighted percentages of the monitoring points exceeding a quality standard or TV by their annual arithmetic mean concentrations. However, the use of the volume ratio may be especially useful when the number of monitoring points is small.
Alternatively, if there is significant spatial correlation, the extent can also be estimated by means of Kriging techniques.
Location of exceedance
The consideration of the location of exceedances is rather relevant for the compliance regime than for the aggregation methodology. Kriging can especially be very useful for visualization of spatial groundwater quality and the location of exceedances.
Level of aggregated concentration
Regarding the proper estimation of the level of aggregated concentrations it has already been mentioned that percentiles have some disadvantages when the level shall be
described. They do not reflect properly the impact of point source pollution, the calculation of confidence limits is difficult, there are lower limits with regard to the number of monitoring points, and finally, they are not well-defined in case of a small number of monitoring points as different software packages deliver different results.
Aggregation methods focusing on the mean concentration in the GW-body appear to be more suitable for estimating the level of concentrations in a GW-body. The arithmetic mean however should be used only when the distribution of monitoring points is homogenous.
Alternatively the weighted arithmetic mean based on sub-bodies or assigning each monitoring point with its individual weight might be used. The Kriging mean can be calculated when Kriging is possible, i.e. when there is spatial correlation in the GW-body.
Confidence in the assessment
The above proposed methods of data aggregation also provide confidence intervals, which describe the uncertainty.
In case of 2–5 monitoring points per GW-body the confidence is considered by adding a variable factor to the arithmetic mean. This factor is depending on the level of the threshold value and increases with decreasing levels. The underlying standard deviation is based on a component for the environmental variability (25%) and a component for the analytical error.
The latter is derived from the Horwitz curve, which describes a loglinear relationship between standard error and content to be analysed. The Horwitz curve has been derived empirically from a large number of interlaboratory tests for many different substances. Although the standard deviation obtained from this curve may be e.g. three times higher or only 1/3 of the true analytical error, it is widely accepted as a first order approximation.
In case of more than 5 monitoring points per GW-body the confidence is considered by adding the Upper Confidence Limit of the (weighted) arithmetic mean to the (weighted) arithmetic mean or by adding the Upper Confidence Limit of the (weighted) Kriging mean to the (weighted) Kriging mean (two-sided confidence level: 95 %).
Preliminary network check – Division of sub-bodies
According to the GW1 monitoring guidance the selection of monitoring points within a monitoring network should be based on three main factors:
- the conceptual model(s) including assessment of the hydrological, hydrogeological and hydrochemical characteristics of the body of groundwater including characteristic travel times, distribution of different types of land uses (e.g. settlement, industry, forest, pasture/farm land), pathway susceptibility, receptor sensitivity and existing quality data;
- assessment of risk and the level of confidence in the assessment; including the distribution of key pressures and;
- practical considerations relating to the suitability of individual sampling points. Sites need to be easily accessed, secure and be able to provide long-term access agreements.
For the arithmetic mean, spatial representativity of monitoring points is a prerequisite for data aggregation and should be ensured to allow for sound statistical assessment. This means that the monitoring points should be evenly distributed in a homogenous GW-body or each point represents equal share of the GW-body in a GW-body which is non-homogenous with regard to hydrogeology. The even distribution of monitoring points might be assessed by the representativity index which was elaborated within WG 2.8.
Homogeneity implies furthermore that there is no local accumulation of sites.
Representativity with regard to anthropogenic and natural factors is also regarded as
important. Anyway, GW-bodies are from a hydrogeological point of view heterogeneous and only seldom homogeneous. Therefore it might be useful to subdivide a GW-body into sub- bodies where the requirements of even distribution of monitoring points is fulfilled within each sub-body or where each point represents equal share of the sub-body.
4.3 Proposed aggregation 3-step procedure
Based on the requirements of the WFD and the draft GWD and based on the prerequisites and restrictions of the selected statistical methods a stepwise approach is proposed for assessing whether a GW-body is in good chemical status combining the elements extent, level and confidence. The element location of exceedance is relevant within the status assessment procedure and not part of the aggregation methodology.
Step 1 – any exceedance?
Maximum of annual arithmetic mean concentration per monitoring point < standard or TV good chemical status for the GW-body and the considered parameter. No further
investigation and assessment is needed.
Otherwise proceed with step 2.
Step 2 – network homogeneous
If the monitoring points are evenly distributed in a homogenous GW-body or if in a GW-body which is non-homogenous with regard to hydrogeology each site represents an equal share of the GW-body:
Estimation of the EXTENT by the percentage of points or by the volume ratio.
Estimation of the LEVEL by the arithmetic mean for the whole GW-body or the group of GW-bodies.
Considering the CONFIDENCE of the level of concentration by - depending on the number of monitoring points - adding a variable factor respectively by adding the Upper Confidence Limit of the arithmetic mean to the arithmetic mean.
Alternatively, especially if the distribution of sites is not fully satisfying with regard to uniformity, the spatial correlation within a GW-body can be assessed by using variogram techniques. If there is significant spatial correlation, the EXTENT of the GW-body exceeding a standard or TV as well as the LEVEL of concentration and the CONFIDENCE can also be estimated by means of Kriging techniques. Kriging can be very useful for visualization of spatial groundwater quality and the location of exceedances.
Step 3 – sub-body consideration
If the monitoring points are unevenly distributed in a homogenous GW-body or if in a GW- body which is non-homogenous with regard to hydrogeology each site represents an unequal share of the GW-body:
Delineation of sub-bodies where the monitoring points are evenly distributed within a homogenous sub-body or where each site represents an equal share of the sub-body if the sub-body is non-homogenous with regard to hydrogeology.
Performing the assessments of EXTENT and LEVEL according to step 2 for each sub-body.
Aggregation of results of the sub-body assessments to the level of the GW-body by weighting the single results accordingly.
- Estimation of the extent by the percentage of points or by the volume ratio for each sub- body and calculation of the weighted EXTENT of exceedance by the weighted
percentage of points or by the weighted volume ratio for the whole GW-body
- Estimation of the level by the arithmetic mean for each sub-body and calculation of the weighted LEVEL by the weighted arithmetic mean for the whole GW-body
Considering the CONFIDENCE of the weighted level of concentration by - depending on the number of monitoring points - adding a variable factor respectively by adding the Upper Confidence Limit of the weighted arithmetic mean to the weighted arithmetic mean.
In case the Kriging approach is applied, the weighted extent, the weighted level and the weighted confidence can be calculated respectively.
4.4 Thoughts on the assessment of chemical status – combination of results
As already mentioned, the assessment of the chemical status of the groundwater comprises the consideration of the following elements:
- the extent of an exceedance,
- the location of the exceedance,
- the level of the concentration, and
- the confidence in the assessment.
All the listed elements have to be taken into account when determining the chemical status of a GW-body or a group of GW-bodies.
The assessment of the chemical status goes far beyond the objectives of this document.
Although some thoughts on the assessment arose when discussing the methodologies for the above listed elements especially on how the single elements could be combined in order to consider a GW-body to be in good chemical status. The combination would comprise the elements of data aggregation (extent, level and confidence) together with the location of exceedance.
In the light of the proposed methodologies the extent, the level and the confidence could be addressed by a matrix approach wherein the checking criteria for the extent could be defined individually for each parameter or for groups of substances and/or for each GW-body. The checking criteria for level and confidence could be the standard or threshold value.
There might be areas within the matrix where the determination of the chemical status would be rather clear, where the extent, the level and the confidence results are clearly indicating that a GW-body is in good chemical status or not (e.g. the extent of exceedances is 0% or 100%). But, there are areas within this matrix, where this determination is not that
unambiguous (e.g. the yellow area in the matrix). And within this area the location of the exceedance could be the main element for a clear determination of the chemical status of the GW-body or a group of GW-bodies as it takes into regard the location of the exceedance in respect to the protection of dependent terrestrial and aquatic ecosystems and the protection of abstraction of groundwater for human consumption.
Figure 4: Matrix approach for the determination of the chemical status of a GW-body.
E X T E N T
LEVEL & CONFIDENCE
LOCATION
5 Virtual case study – demonstration of methodology
The different proposed aggregation methodologies for extent, level and confidence have been applied to a test GW-body “GWB1” for demonstration purposes.
The calculation comprises
- a preliminary network check
- calculation of the annual mean per sampling site
- assessment of the spatial correlation within the GW-body by using variogram techniques (Kriging)
- delineation of two sub-bodies
5.1 Characterisation of test GW-body
Table 1 gives a rough characterisation of the GW-body and the distribution of the 73 monitoring points within the GW-body and Figure 5 demonstrates the concentrations of the 73 sampling sites. The measured concentrations range from the limit of detection or
quantification up to more than 200 mg/l nitrate.
Table 1: Characterisation of the test GW-body “GWB1” and distribution of monitoring points
Parameter Nitrate
Area: 942 km²
Minimum Concentration 0,0 mg/l Maximum Concentration 206,0 mg/l
No. of monitoring points 73
Figure 5: Distribution of concentrations over the time series
Measured values
2005 2004
2003 2002
2001 2000
1999 1998
1997 200
150
100
50
0
Body & points
5.2 Test calculations of aggregation methods
The proposed methodologies were applied to the test GW-body. The results of the single steps are presented in a common chapter in order to compare.
Preliminary Network check
In a first step the groundwater monitoring network was checked.
The monitoring network is considered representative with regard to anthropogenic and natural factors. There are no effects of abstraction on groundwater quantity, on groundwater quality and on dependent ecosystems, and there is no artificial recharge or groundwater induced pollutant pressures on dependent ecosystems.
The GW-body is at risk because of nitrate from diffuse pollution sources (agricultural activities).
Step 1 – any exceedance?
In a first step all the annual arithmetic mean concentrations of each of the 73 monitoring points was calculated for 2004 and 2005.
If none of the annual arithmetic mean concentrations per monitoring point exceeds the standard, the GW-body is determined as being in good chemical status for the considered parameter. Else calculations have to be proceeded with Step 2.
In 2004: 46 monitoring points exceed the standard of 50 mg/l nitrate.
In 2005: 41 monitoring points exceed the standard of 50 mg/l nitrate.
The calculation has to be proceeded with Step 2.
Step 2 – network homogeneous
Calculation of extent, level and confidence on the level of GW-bodies if the monitoring points are evenly distributed in a homogenous GW-body or if in a GW-body which is non-
homogenous with regard to hydrogeology each site represents equal share of the GW-body.
Because of an accumulation of sampling sites in the western part of the test GW-body, the proposed methods for extent and level on the basis of percentage of sites and arithmetic mean can not be calculated.
But alternatively, the spatial correlation within the GW-body was assessed by using variogram techniques. The extent of the GW-body exceeding the standard as well as the level of concentration and the confidence were estimated by means of Kriging techniques.
According to the proposal this proceeding may be useful if the distribution of sites is not fully satisfying with regard to uniformity.
Table 2: Results of the calculation of extent, level and confidence by Kriging
Year Extent
Kriging
Level Kriging Mean
95 % Upper Confidence limit of Kriging mean
2004 59.0 65.6
2005 54.6 % 56.8 63.2
Step 3 – sub-body consideration
The monitoring points are not evenly distributed within the homogeneous GW-body. There is a local accumulation of sites in the western part of the GW-body. For this reason two sub- bodies were delineated:
Table 3: Characterisation of sub-bodies
No. of monitoring points Area in km² Area in %
GW-body GWB 1 73 942 100%
- Sub-body 1 30 141 15%
- Sub-body 2 43 801 85%
The calculations for the extent and level are performed for each sub-body and then aggregated to the level of the GW-body by weighting the sub-body results.
The extent of exceedances was calculated by percentages of the monitoring points and by the volume ratio approach.
Table 4: Calculation of extent of exceedance
Extent No. of
monitoring points
Area in
km² Area in %
Volume ratio
Percentage of monitoring points
Sub-body 1 30 141 15% 48.3% 56.7%
Sub-body 2 43 801 85% 55.4% 55.8%
GW-body
weighted 73 942 100% 54.3% 55.9%
The extent is almost the same for the percentage of site calculation and the volume ratio calculation. But as in sub-body 1 many monitoring points exceed the standard only by few mg/l, the volume-ratio extent is considerably lower.
Table 5: Calculation of level and confidence
Level Confidence
Median 70% 90% Arithmetic
mean Kriging Arithmetic
mean Kriging
Sub-body 1 52.8 59.0 89.3 48.6
Sub-body 2 58.1 79.1 119.0 58.9
GW-body
weighted 57.4 66.8
Whole GW-body 55.3 70.3 109.6 54.7 56.8 63.0 63.2
The network has a higher density of monitoring points in sub-body 1, but the nitrate level is lower than in sub-body 2. Therefore both the weighted arithmetic mean and the Kriging Mean for the whole body are higher than the arithmetic mean for the whole GW-body.
Location of exceedance
Kriging, when correctly applied, is especially useful for the visualization of spatial
groundwater quality. It is possible to visualise the location of exceedances which is relevant for assessing the protection of dependent terrestrial and aquatic ecosystems and of human uses.
Figure 6: Kriging Surface of the nitrate concentration. Blue colour stands for very low concentration
6 References
Technical Report No. 1: The EU Water Framework Directive: statistical aspects of the identification of groundwater pollution trends, and aggregation of monitoring results. ISBN:
92-894-5639-6.
http://forum.europa.eu.int/Public/irc/env/wfd/library?l=/framework_directive/guidance_docum ents/technicalsreportsnos1sgw/_EN_1.0_&a=i
Technical report on groundwater monitoring as discussed at the workshop of 25th June 2004.
http://forum.europa.eu.int/Public/irc/env/wfd/library?l=/framework_directive/thematic_docume nts/o_-_groundwater/groundwater_working/groundwater_monitoring/_EN_1.0_&a=i
Monitoring Guidance for Groundwater. Version 10.1. WG C Groundwater.
http://forum.europa.eu.int/Members/irc/env/wfd/library?l=/working_groups/new_groundwater/
drafting_monitoring/gw1-guidance_draft-v10do/_EN_1.0_&a=i
STATUS, CONFIDENTIALITY AND ACCESSIBILITY
Status Confidentiality Accessibility
S0 Approved/Released x PU Public x Work-space x
S1 Reviewed PP Restricted to other programme participants
(including the Commission Services) Internet
S2 Pending for review RE
Restricted to a group specified by the consortium (including the Commission Services)
Paper x
S3 Draft for comments CO
Confidential, only for members of the consortium (including the Commission Services)
