OR/16/033 Assessment of the 2012 designation process for groundwater

Stuart, M E, Ward, R S, Ascott, M, and Hart A J1. 2016. Regulatory practice and transport modelling for nitrate pollution in groundwater. British Geological Survey Internal Report, OR/16/033. 1 Environment Agency

Methodology

The NVZ assessment methodology used for the 2012 review of groundwater quality in England and Wales adopted a weight of evidence approach (Environment Agency, 2012). It combined observed data from monitoring with data on agricultural land use calculated using a national-scale nitrate leaching model. The methodology was reviewed by an external panel made up of independent technical advisors (including BGS) and representatives from affected bodies including all the farming unions and the Country Land and Business Association (CLA). The method was agreed by the whole panel at the final meeting. The results of the review were used to identify new groundwater NVZs. The methodology comprised six main steps:

Step 1 — Identification of groundwater boreholes for analysis and statistical analysis of groundwater quality monitoring data

Monitoring data
The method used both Agency and water company data. The Agency network comprises 15 105 points, including 592 where data were supplied by water companies, plus an additional 1,132 water company monitoring points provided for this exercise. Data from blended and treated sources were excluded. Sampling points included boreholes, springs, wells & adits, pits and landfill sites. Data were cleaned to remove zero values (assumed to be analytical errors), outliers (using a multiple outlier test (Ellis, 1998)), values > 200 mg l^-1, and duplicates within any day. Values below the limit of quantification, ‘<’ values, were treated by dividing the recorded value by 2. Data from samples collected between 1980 and 2009 were included, whereas older data and points with fewer than 6 values were excluded. This left a total of 3839 sites in the final dataset.

Status and trend
All groundwater monitoring points with sufficient data were analysed to determine whether:

• the 95 percentile of the measured nitrate concentrations exceeded 50 mg NO₃ l^-1 in 2010; and;
• the 95 percentile of the nitrate concentrations was likely to exceed 50 mg NO₃ l^-1 in 2027.

If the current or future 95 percentile nitrate concentration exceeded 50 mg NO₃ l^-1 with at least 95% confidence, it was deemed to have failed the statistical test. In practice, this means testing whether the lower 90% confidence interval on the 95 percentile exceeds 50 mg NO₃ l^-1. This approach is precautionary and is required to offer protection against the high uncertainty caused by very limited monitoring data at many groundwater monitoring points.

The year 2027 was chosen as the future time horizon because it (i) is consistent with the approach used in the 2008 review, (ii) allows a sufficient period of time for measures to take effect, and (iii) ties in with the Water Framework Directive river basin planning cycle.

As in the 2008 review, most sites had insufficient data to estimate the 95 percentile concentration. For these sites, the mean concentration was calculated instead and an empirical conversion factor was applied to convert the mean concentration to a 95 percentile concentration. The data were analysed statistically with the method version depending on data availability: Weibull procedure for current status (2010) where there were 24 samples between 2004 and 2009, and AntB or the simpler AntC tools for current status and trends for 2027 for fewer samples. Where there were insufficient data for the Ant tools a 45 mg NO₃ l^-1 threshold was used instead on the basis that the mean is commonly 5 mg NO₃ l^-1 lower than the lower 90% confidence on the 95 percentile.

The variable total inorganic nitrogen (TIN) was calculated for each monitoring point where:
TIN = (total oxidised N (TON) + ammoniacal N) or (nitrate-N + nitrite-N + ammoniacal N)

Step 2 — Identifying pollution in areas between boreholes

Step 1 only enables the assessment of groundwater nitrate concentrations at specific points within aquifers (the exact area of land represented by a monitoring point will depend on the volume and depth of the abstraction and geology). In order to estimate nitrate concentrations across the country and use the methodology at a 1 km² resolution, a statistical interpolation technique (kriging) was used to understand spatial patterns in the groundwater dataset and, hence, estimate nitrate concentrations at unmonitored locations. Kriging was used to assess current and future predicted (to 2027) nitrate concentrations for all areas of groundwater by quantifying the spatial correlation between pairs of measurements over the whole dataset and characterising the relationship between pairs of measurements with different degrees of separation. This relationship is then applied to estimate the measured variable at unmonitored locations from the values observed at surrounding locations (Figure 4.1).

Figure 4.1    Mean predicted groundwater total inorganic nitrogen concentrations in: a) 2010 and a) 2027 as interpolated by ordinary kriging (from Environment Agency, 2012).

The kriging results were used to inform the designation process in a number of ways:

• to estimate the probability that nitrate concentrations exceed the set threshold; this assessment of confidence in the data mirrors that used for surface waters;
• to help delimit the extent of the contamination in aquifers;
• to identify monitoring points with anomalously high or low nitrate concentrations; these may then be screened out if they are deemed to be unrepresentative of the water bodies (e.g. if they are influenced by a local point source discharge).

Kriging has limitations and the method used here took no account of geological formations or changes in geology. The results were reviewed in consultation with local staff e.g. in Step 5.

Step 3 — Modelling assessment of nitrate leaching to groundwater

Agriculture
In addition to the analysis of monitored water quality data, nitrate leaching from agricultural land was calculated from farm census returns to Defra. The data were processed by ADAS using the NEAP-N national-scale nitrate leaching algorithms. This approach considers a single maximum potential nitrogen loss coefficient for individual crop and livestock types, modified by spatially distributed information on soil type and hydrologically effective rainfall (HER). It contains data related to average annual soil drainage, nitrate flux and concentrations from diffuse sources at a 1 km² resolution. It uses average climate conditions (1971–2000) and data on agricultural land use based on the 2010 Defra Agricultural Census. The model does not include any point source contributions.

Nitrate losses for crops and animals were then aggregated to provide estimates of leaching for three land use classes: managed arable crops, managed grassland and rough grazing. Total values for agricultural nitrate were calculated by combining losses for the three categories on a 1 km² scale. To convert the load figures into concentrations, losses were standardised by dividing by HER for the 1 km² cell. Due to the uncertainty in the data coverage, each 1 km² cell was averaged with its direct neighbouring cells, where available. The resulting agricultural nitrate concentration leaching to groundwater is shown in Figure 4.2.

Figure 4.2    Agricultural N loading for 2010 as TIN estimated using the ADAS NEAP-N model (from Environment Agency, 2012).

The principal purposes of using the land use data were to:

• identify the significance of the agricultural contribution to any nitrate pollution identified;
• provide further confidence in the conclusions of the statistical analysis of monitoring data;
• minimise the possibility that the borehole fails due to historic landuse because of long travel times for nitrate at the surface to reach deep groundwater.

Urban leaching
Nitrate leaching from urban land areas was calculated according to the component model of Lerner (2000), whereby nitrate losses to groundwater are expressed as export coefficients per hectare of each urban land cover type. The model was integrated with land cover data from the CORINE 2000 and the ADAS National Land Use databases. The model identifies 14 components of runoff, although only parks and gardens, recreational grassland, construction activities, and industrial and commercial units were estimated from CORINE, and leaking sewers and water mains from population density. These data were then combined to give a total nitrate load per 1 km² from urban sources and divided by the HER for that grid square to give a measure of concentration (Figure 4.3).

Figure 4.3    Estimated urban N leaching as TIN using model of Lerner 2000 (from Environment Agency, 2012).

Atmospheric deposition
Data on nitrate loads from atmospheric deposition were also needed to estimate nitrate leaching to groundwaters, but are not normally included within losses from agriculture in the NEAP-N leaching model. The dataset is based on spatial interpolation of monitoring data then input to the MAGPIE soil leaching model. These data were then input into the NEAP-N model. The methodology used for calculating nitrate from atmospheric deposition was the same as that used in the 2008 review.

Step 4 — Combining the evidence from monitoring and modelling

A GIS-based risk model with a weighted scoring system was used to determine potential groundwater NVZs in 2012 by overlaying the results from Step 2) with those from Step 3). For every 1 km square, the risk model assesses the confidence with which it could be determined that the nitrate concentration in the groundwater exceeded, or is likely to exceed, 50 mg/l and that the source of nitrate includes agriculture. This largely reproduced the results from the model used in the 2008 review. The risk model combines lines of evidence data plus local evidence from Environment Agency specialists. Areas of high risk are shown as a potential groundwater NVZ.

The risk model consists of eight components (Table 4.1). Three components describe pressures and are mainly derived from modelled inputs of nitrate data where the higher the pressure, the greater the risk that groundwater nitrate concentrations will exceed 50 mg NO₃ l^-1. The other five components describe the observed nitrate and draw upon a combination of water quality monitoring data and local Environment Agency evidence.

Table 4.1    Components of the GIS risk model (from Environment Agency, 2012)

Risk	Factors
Pressure	1. Agricultural nitrate leaching from the NEAP-N model (National) Score: <25 = 0, 25–50 = 1, >50 = 2, Weighting = 3
	2. Urban nitrate leaching from the Lerner model (National) Score: <25 = 0, 25–50 = 1, >50 = 2, Weighting = -2
	3. Denitrification or mixing lower the nitrate input from agriculture to groundwater (Area) Score: good evidence = 2, some evidence = 1, no evidence = 0, Weighting = -1
Observed	1. Kriged current groundwater nitrate concentration (National) Score: <25 = 0, 25–50 = 1, >50=2, Weighting = 3
	2. Kriged future (2027) groundwater nitrate concentration (National) Score: <25 = 0, 25–50 = 1, >50 = 2, Weighting = 2
	3. Monitored nitrate is representative of point source pollution (Area) Score: good evidence = 2, some evidence = 1, no evidence = 0, Weighting = -5
	4. Monitored nitrate is unrepresentative of real groundwater nitrate concentrations (Area) Score: yes good evidence = 2, yes some evidence = 1, no evidence = 0, no some evidence = -1, no good evidence = -2, Weighting = 3
	5. Surface water — groundwater interactions identify that surface water quality is a reasonable indicator of groundwater quality (Area) Score: good evidence = 2, some evidence = 1, no evidence = 0, Weighting = 1

Four of the components were derived using national datasets; nitrate monitoring data were interpolated to produce national maps of current (2010) and future (2027) groundwater nitrate concentration and agricultural and urban nitrate leaching were estimated from land use. The other four components were derived using the professional judgement of local area Agency staff.

Weightings were designed to give the greatest importance to groundwater monitoring data and secondary importance to agricultural nitrate loss data derived from the NEAP-N model. The model had the flexibility built-in to incorporate the understanding of local Environment Agency hydrogeologists but scores were set using national lines of evidence. Each component was given a score (positive scores increase the overall risk and negative scores decrease the overall risk) and weightings were applied to these scores. The weighted scores were then combined to yield an overall risk score indicating the strength of evidence that the groundwater was polluted by nitrate from agricultural sources.

For a situation where observed nitrate is unrealistically low and <50 mg NO₃ l^-1 (observed risk 4, Table 4.1) there are three possible reasons:

• nitrate pollution has not passed through the unsaturated zone; it is on its way but has not yet been detected by monitoring;
• deep abstractions sample older, cleaner water that is not representative of current nitrate pressure; due to the depth of monitoring the data are unrepresentative;
• uncertainty in predicted nitrate values caused either by short duration of monitoring or a significant variation in the dataset.

The evidence can include:

• a nearby water company source has been abandoned due to high nitrate;
• an unsaturated zone of >30 m thickness is delaying nitrate measurement;
• an aquifer that is layered is layered or sampling is at depth.

The scores for the pressure risk components are summed to provide an intermediate positive pressure risk score. Since the urban loading score is associated with a negative weight (-2), where urban loading outweighs the agricultural load, the intermediate pressure score could be negative. The methodology indicates that the pressure risk score has to be set to zero as a minimum and therefore pressure risk scores have been set to zero if negative pressures were found. The scores for observed risk components are also summed to provide an intermediate positive observed risk score.

If the risk that groundwater nitrate concentration is exceeding 50 mg NO₃ l^-1 and agriculture is the source, the score will be higher than 8. This will lead to potential groundwater NVZ designations. A medium score ranges from 8 to 3 and shows that either the monitoring or modelling assessments exceeded or were likely to exceed 50 mg NO₃ l^-1. These areas are likely to be included in potential designation areas around high risk areas dependent on the hydrogeological setting. A low score is lower than 3 where both the monitoring and modelling assessments show that nitrate concentrations were not likely to exceed 50 mg NO₃ l^-1. These are generally not considered for designation and any low risk areas that are repeatedly shown to be so may be considered for removal from designation.

All areas scoring higher than 8 and currently not in NVZ areas have been highlighted and presented at the workshops to understand local factors that could affect the final score for these areas. The final risk score map for England and Wales is presented in Figure 4.4.

Figure 4.4    Risk score map showing national lines of evidence prior to local Environment Agency workshops (from Environment Agency, 2012).

Step 5 — Ground-truthing the draft designations

Regional workshops were held to allow local EA staff to comment on the preliminary results of the review. The workshops focussed on where there was less certainty in the results and were attended by observers from external stakeholder groups.

Step 6 — Identifying land draining to polluted waters

Land that is directly above a polluted groundwater does not necessarily drain into it. With professional judgement, the following physical and hydrogeological boundaries that allow major aquifers to be split into groundwater bodies and delineate the catchments of the polluted groundwaters have been used including:

• solid or drift geology (1:50 000): geological boundaries such as faults and geological contacts; high permeability drift outcrops; low permeability drift outcrops;
• risk of solution features (1:50 000);
• surface water outflow features; surface water catchment boundaries; rivers, acting as groundwater catchment divides; lakes; coastlines;
• groundwater level contours and flow lines;
• urban areas.

Before making a final determination of the land draining to a polluted groundwater, further checks were undertaken for those waters which had been identified by limited monitoring data to confirm that:

• the monitoring point data that were exhibiting high nitrate concentrations were robust (data period and representativeness);
• the land draining to monitoring points had an agricultural loading.

These checks were made to prevent any unsupportable designations being made as a result of the greater weight given by the risk model to monitoring data. If the land failed either of these checks it was not considered further for designation as an NVZ during this cycle.

Workshops were held across England and Wales with Environment Agency groundwater specialists to understand factors that could affect the final risk scores. All potentially new NVZ areas were reviewed. The following questions were raised during the meetings to help understand if any local conditions might impact on the final scores.

• Is the monitoring point location correctly recorded? Is it representing the correct groundwater body? Is the monitoring point representative of water quality?
• Are there any issues with the land use characterisation?
• Has the weighting methodology been modified from the initial assessment
• Yes, which lines of evidence have been modified?
• Are there any issues caused by the underlying geology or drift geology?
• Has the recharge area been correctly delineated? Will these changes result in a change in designation? Will this be an increase or decrease in the size of the area to be designated?
• Is there any significant point source influence?
• Is there any evidence of denitrification?
• Is there any inconsistency between the kriging results and the loading from land use? Is there any inconsistency between the kriging results and the overall score?
• Has the recharge area been correctly delineated?
• Will these changes result in a change in designation? If the monitoring point cannot be used to represent groundwater quality, can surface water be used as an alternative?

Designated NVZ areas

The total area identified as draining to polluted groundwater was approximately is 32 858 km² (32 599 km² in England and 259 km² in Wales) compared to 32047 km² identified in the 2008 review (31821 km² in England and 226 km² in Wales). The new area that could be designated for groundwater is 810.4 km² (777.7 for England and 32.7 km² for Wales). The total new area identified for designation was is 2150 km² (2110 km² in England and 40 km² in Wales). This represented a 1.6% increase in the area designated in England and a 0.2% increase in the area designated in Wales.

There were also some areas of existing NVZ that previously met the criteria for a polluted water but passed the 2012 review criteria. The total area concerned was approximately 18 402 km². A total area of approximately 5550 km² (all in England) was removed from designation based on improved evidence, showing good and sustained water quality and a low risk of future deterioration. A map of the proposed groundwater NVZs is shown in Figure 4.5.

Figure 4.5    2008 and new proposed groundwater NVZ areas for 2012 (from Environment Agency, 2012).

Appeals process

The First-tier Tribunal (Environment) was set up in 2010 but only in 2013 has it handled a significant work-load. Previously appeals were heard by panels appointed by the Secretary of State. Appeals were heard during 2013, and according to the Tribunal, there were 455 appeals in all, with 38% (172) allowed, and 13% (57) part-allowed. A small number (11) were dealt with by a Consent Order. 31% (142) appeals were dismissed, and the remainder were either struck out or withdrawn. A number of different appeals may relate to the same NVZ. The change to the appeals process from panels to tribunals resulted in a larger percentage being successful. This was due to a greater emphasis on proving that water in a particular location was polluted. For groundwater this was primarily due to the distribution of monitoring points in areas of complex geology.