Synopsis
The new EU Water Framework Directive (Directive 2000/60/EC) implies that hard rock units will need to be considered as aquifers, and to be hydrogeologically and hydrogeochemically characterized. This is not a straightforward task, given that such aquifers are heterogeneous, discontinuous and strongly three-dimensional in the distributions of both hydraulic and hydrochemical properties. The Geological Survey of Norway has recently carried out characterizations of Norway’s hard rock aquifers using simple non-parametric statistical techniques, focusing on median and percentile values for aquifer properties and concentrations of chemical parameters. Such characterizations reveal far larger parameter ranges within given aquifer lithological units than between different units. Indeed, for most lithologies, median well yields and concentrations of major chemical parameters (e.g. pH, major ions) are remarkably similar. For certain elements, however, systematic differences can be detected. For example, elements such as uranium, radon and fluoride are significantly enriched in Precambrian granites and gneisses, and depleted in anorthosites.
Introduction
This paper considers some aspects of the philosophy of siting successful wells in hard rock aquifers. It also suggests means of characterizing such heterogeneous aquifers, both in terms of their hydraulic parameters and their hydrogeochemistry. In this paper, the term ‘hard rock’ aquifer is taken to mean the crystalline metamorphic or igneous rock (basalt, slate, granite, schist, gneiss etc.) underlying large areas of Scandinavia, Scotland and Wales. It is not taken to include deeply weathered basement terrain, with significant saprolite development, such as is typical of much of tropical India or Africa (Lloyd 1999).
Approaches to the science of locating successful water supply boreholes in hard rock aquifers typically focus on the identification of fracture zones or lineaments, by means of field surveys and study of maps, aerial photographs or satellite images. Geophysical techniques (ground penetrating radar, resistivity, electromagnetic and magnetic techniques) are often used to similar ends. Some authors have argued that certain fractures are likely to be more productive than others, depending on their orientation in past or current in-situ stress fields, resulting in Larsson's (1972) ‘Hydrotectonic’ theory (further discussed and criticized by Banks et al. 1994; 1996).
The authors do not wish to suggest that fracture zones never represent groundwater flow pathways, nor do they wish to suggest that geophysics are of little use or that stress orientation is not of significance. The relative lack of success of hydrogeologists in locating productive wells in hard rock terrain (and the continued apparent willingness of customers to use dowsers in preference to hydrogeologists) does, however, indicate that current techniques and theories lack predictive power. In particular, the following points should be noted:
Not all fracture zones are water transmissive. For example, Banks et al. (1992) clearly demonstrated that water leakages to Hvaler subsea tunnel (southern Norway, Fig. 1) were derived from small fractures in relatively massive granite, rather than from major fracture zones and crush zones (which were often ‘tightened’ by smectite-rich fault gouge).
Remote sensing and most geophysical techniques are unable to distinguish between water-transmissive fractures and clay/gouge-filled fractures (both result in lineaments in the terrain, and high conductivity electromagnetic responses).
Most ‘hydrotectonic’ theories are simply too plausible: too many tectonic/stress scenarios can be envisaged to explain the yield (or lack thereof) of a set of boreholes. Such theories have explanatory power, but lack predictive power (Banks et al. 1996).
Water leakages in the Hvaler tunnel (after Banks et al. (1992), reproduced with permission of the Geological Society Publishing House). Note that zones of water leakage and injection grouting do not necessarily coincide with fracture zones as detected by geophysics and in the tunnel itself.
Many hydrogeologists approach well siting in hard rock terrain as a game of chess. By using the right tools and the right strategies, a positive outcome ought to be assured. In reality, this is often a delusion. The resources at the disposal of most groundwater supply investigations are insufficient to adequately characterize the complexity of most hard rock aquifers. The inability of hydrogeologists to predict the hydraulic properties of hard rock aquifers has its roots in the fractured, discontinuous and extremely heterogeneous nature of such aquifers. In a similar manner, the chemical quality of water from a single prospective borehole cannot be adequately predicted, and also has its roots in heterogeneous and discontinuous fracturing and fracture mineralization. As will be seen later in this article, hydrochemical parameters exhibit wide variations within a single lithology, largely reflecting heterogeneity in fracture residence times and hydrodynamics.
The authors would argue that well siting in hard rock terrain is more akin to a game of bingo. The outcome of drilling is largely determined by chance. Yet, hydrogeologists are still required to make prognoses for clients. Likewise, the EU Water Framework Directive (Directive 2000/60/EC) requires the hydrogeological and hydrochemical characterization of hard rock aquifers. It is clearly problematic to approach such characterization in a deterministic manner, but can we approach it in a probabilistic manner? Can we use statistics to assist in decision-making, when drilling for water? Can we statistically characterize hard rock aquifers?
The statistical characterization of well yield from hard rock aquifers
Some form of statistical technique has been used to characterize aquifers in most of the hard-rock-dominated nations of Europe, including Sweden, Finland, the Czech Republic and Norway. Indeed, the IAH hydrogeological map of Europe uses cumulative frequency diagrams for such purposes (e.g. Persson et al. 1985).
Let us take a data set from the borehole archives of the Geological Survey of Norway, for example, 24 boreholes from the island group of Hvaler, in the Precambrian Iddefjord Granite (Table 1). In this data set, yields range from 40001/h to dry (nominal 1 1/h entered into archive). We can calculate an arithmetic mean yield of 7641/h, with a standard deviation of 998 1/h, but is this a meaningful characterization of the Iddefjord Granite of Hvaler? Figure 2 suggests not: parameters such as the arithmetic mean and standard deviation only adequately characterize the central values of normally (or at least symmetrically) distributed data sets. Figure 2 demonstrates the distribution of yields to be highly skewed towards low yields (typical of all hard rock yield distributions). To characterize non-normally distributed data sets, we need to use ‘non-parametric’ statistics. The median value 400 1/h is taken as a central value, while the interquartile range (787 1/h) between the 25th percentile value (138 1/h) and the 75th percentile value (925 1/h) is a non-parametric measure of variance.
Histogram showing the frequency distribution of the borehole yield data set for 24 boreholes in the Iddefjord Granite of Hvaler (Table 1).
Yield (from short-term testing) of 24 water supply boreholes in the Iddefjord Granite of the Norwegian Hvaler Islands. The yield of 1 1/h is probably a nominal figure assigned to a ‘dry’ borehole.
In a sense, these statistics, either as raw numbers or as a cumulative frequency distribution (c.f.d.) curve (Fig. 3), represent a characterization of the Iddefjord Granite aquifer (provided that a great enough data set is available). Although neither boxplots nor c.f.d. diagrams make any specific assumptions about the type or nature of the distribution of yield, it is intriguing that, for many lithologies, the distribution of borehole yields often approximates to log-normality (Banks 1998). In such a case, the distribution can be characterized by only two parameters, the median (which will be identical to the geometric mean) and the standard deviation of the log values. The explanation for this phenomenon lies in the hierarchical evolution of fractures within the lithosphere and the observed lognormal distribution of fracture apertures (Banks et al. 1996).
Cumulative frequency distribution diagram for the borehole yield data set for 24 boreholes in the Iddefjord Granite of Hvaler (Table 1). The median is located at cumulative probability = 0.5, the 25th percentile at cumulative probability = 0.25 etc.
Furthermore, it can be argued that such statistics are meaningful in providing a first-order characterization of the hydraulic properties of an aquifer, given that well yield will be closely related to specific capacity (yield per metre drawdown) and that specific capacity is approximately proportional to aquifer transmissivity. Czech and Swedish hydrogeologists have used borehole yield as a surrogate measure for aquifer transmissivity when producing hydrogeological maps (Jetel & Krasny 1968; Krasny 1975; Carlsson & Carlstedt 1977), while Wright (1997) has used a plot of well yield against specific capacity (the ‘QSC diagram’) to characterize aquifers in Ireland. Banks (1992) has specifically examined the published relationships between well yield and transmissivity and concluded that:
where Ta - apparent transmissivity (m2/s) and F = specific capacity (m3/s/m drawdown)
Application of statistical characterizations to hydrogeological problem solving
Cumulative frequency curves can be constructed for each specific lithology in an area. Although yield distributions are often very similar, some systematic differences can be distinguished (Fig. 4). From such curves, one may offer a client the possibility of making an informed decision as to the chances of obtaining a given yield. For example, the chances of obtaining a yield of 12001/h in the Iddefjord Granite are only around 28%. However, the chances of obtaining a yield of 100 1/h in almost all lithologies is very good, often around 90%. Such low yields are in fact adequate for many domestic properties, summer cottages or small farms, and this statistic helps to explain the apparent relative success of dowsers, who are typically employed by clients with such properties. This statistic also explains why many Scandinavian drillers are able to offer a ‘yield guarantee’ to small clients, absorbing the cost of the small risk of failure into their pricing structure.
Cumulative frequency diagram showing yield distribution curves for Norwegian water boreholes in the Precambrian Iddefjord Granite, Cambro-Silurian metasediments of the Norwegian Caledonian terrain, and Precambrian gneisses. The '1001/h' guideline shows the approximate 10% yield for most lithologies (i.e. 90% of wells yield better than this figure).The 50% and 72% guidelines show the median yield and the 72% yield for the Iddefjord Granite. After Banks & Robins (2002) and based on data from Morland (1997). Reproduced with permission from the Geological Survey of Norway.
Furthermore, yield distributions, once characterized, can be combined to permit an estimation of the chances of obtaining a given total yield from a given number of multiple wells. Banks (1998) used a spreadsheet method of combining frequency curves in order to estimate yield distributions from multiple wells and points out that, for single wells, the 50% likely yield (F50) is given by Y5Q=Mm, where Mm is the median yield of single borehole.
For a non-interfering wellfield of n wells, the total 50% probability yield approaches: Y50(tota\)=nMa, where Ma is the arithmetic mean yield of a single well (n = large). For smaller wellfields of n wells: nMm < F50(total)<nMa.
Gustafson (2002) takes a similar approach and uses equal density sampling of cumulative frequency distribution curves to estimate the likely total yield of the best i boreholes out of a total drilled number b boreholes (assuming that (b — i) boreholes will not be commissioned; Fig. 5). Such approaches are not deterministic, but they do allow a client to make an informed cost- risk-benefit analysis of drilling. The economic aspects of such decision-making are clearly discussed further by Gustafson (2002).
Cumulative plots of total predicted borehole yields from the i best wells out of b drilled. These plots are based on a Swedish data set of (rather good) boreholes, sited on the basis of Very Low Frequency (VLF) geophysical measurements (see Müllern (1980) for a description of the VLF technique), and with a median yield of 36001/h for single wells (note that the median yield of all Swedish hard rock wells, whether sited by VLF methods or not, is only some 6001/h (n = 59 000)). After Gustafson (2002) and reproduced with the permission of the Geological Survey of Norway.
The use of non-parametric statistical presentations also allows us to test hydrogeological hypotheses. For example, Morland (1997) examined the entire groundwater well data set held by the Geological Survey of Norway, to try to establish whether yield distributions differed significantly between differing hard rock lithologies. To this end, he used a presentation termed the ‘boxplot’ (Fig. 6). Here, the central ‘box’ contains the interquartile range of data, with a horizontal line at the median value. The parentheses on either side of the median represent a robust 95% confidence interval on the median. The ‘whiskers’ represent the extraquartile data, with outlying data (as defined by a statistical algorithm) represented by squares. Such boxplots are highly useful tools, as they do not make any assumptions about the nature of the data distribution, nor do they place especial emphasis on extreme low or high data values which may represent experimental error or ‘abnormal’ values (the emphasis is rather on the central mass of data).
Boxplots illustrating the distribution of borehole yields within different lithological subgroups in Norway (after Morland 1997). The numbers on the left vertical axis refer to the lithological group (Table 2), those on the right give the total number of boreholes (#) in each subgroup. The 'box' of the boxplot contains the central interquartile range of data, with a line at the median value (and square parentheses giving the 95% confidence interval on the median). The 'whiskers' show the extraquartile data range, with squares showing extreme outlying data.
Lithological groupings used in connection with Norwegian hydrogeological and hydrochemical data, based on the legend to Sigmond's (1992) bedrock map of Norway at scale 1:3 000 000.
Morland (1997) found that, for a total of 12 757 hard rock boreholes, the median yield was 600 1/h (±17 1/h at a 95% confidence level), the median borehole depth was 56 m (±0.58 m) and the median normalized yield was 12.04 1/h/drilled m (±0.38 1/h/m). Intriguingly, a similar investigation of some 59 000 Swedish hard rock boreholes (Gustafson 2002; Fagerlind 1988) also arrived at an identical median of 6001/h (25th percentile = c. 2301/h, 75th percentile = c. 15001/h, arithmetic mean = 1643 1/h).
Morland (1997) divided the Norwegian data set into subsets according to lithology (Table 2; as defined in Sigmond's (1992) geological map), and produced the boxplots shown in Figure 6. From this diagram, lithologies whose median yield differs significantly from each other can be identified by the fact that the 95% confidence intervals do not overlap. For example, the median yield of lithology 57 (Permian volcanics in the Oslo Rift, 25001/h) is significantly higher than any other lithology, while that of 79 does not differ significantly (at a 95% confidence level) from that of lithology 87. The lowest median yield (lithology 72, 2501/h) is from Caledonian gabbros, diorites and ultramafic rocks. It will be noted that, while the spread of yields within any given lithology is large, the difference in median yield between lithologies is modest. This is presumed to reflect the similar geomechanical properties of many hard rock lithologies.
Morland (1997) also demonstrated an intriguing inverse relationship between median yield and median well depth (Fig. 7), presumably reflecting the fact that successful boreholes are terminated at shallower depths than poorly yielding boreholes, which are drilled deeper in the hope of encountering a water strike (see Ronka (1993) for a similar phenomenon in Finland).
Morland's (1997) data set of Norwegian bedrock boreholes, with median normalized yield (1/h per drilled metre) plotted against lithological subgroup (see Table 2) in descending order from left to right. Also shown are median total yield (1/h per borehole) and median borehole depth (m).
Morland (1997) then divided a section of Precambrian gneissic terrain in western-central Norway into regions of differing post-glacial isostatic uplift. The median normalized yield value for each subregion was calculated, and Figure 8 was produced, appearing to confirm a relationship first tentatively identified by Rohr-Torp (1994), between borehole yield and postglacial neotectonic activity. Henriksen (2003) subsequently attempted to continue this study further east, into the Swedish Precambrian terrain. He failed to find such a strong degree of correlation as Morland (1997), however, although he concludes that isostatic rebound might be able to explain around 9% of the observed variation in borehole yield along some profiles.
Correlation between well yield and total postglacial isostatic uplift; the diagram shows the median yield of boreholes in Norwegian Precambrian rocks per metre drilled depth as a function of annual land uplift. After Morland (1997); Banks & Robins (2002), and reproduced with the permission of the Geological Survey of Norway.
A word of caution
Naturally, the utility of the statistical techniques described above depends on the quality of the data set under consideration. In this context, the following caveats should be noted:
There will often be a tendency towards overestimation of yield distributions, because drillers in most countries tend not to report dry or failed wells to the relevant authorities.
The yields reported to well databases are often short-term yields measured over a period of hours. These yields may well not be representative of long-term sustainable yields (Van Tonder et al. 1998).
Hydrogeochemistry
In the same way that borehole yield varies considerably from borehole to borehole over distances of only a few metres within a given hard rock aquifer, depending on whether a transmissive fracture is encountered, so does the hydrogeochemistry. Different fractures contain groundwater of differing chemistry, depending, amongst other things, on the degree of opportunity for water rock interaction on fracture walls (surface area relative to water volume, residence time) and fracture mineralogies (which may differ from whole-rock mineralogy). However, it is possible to use the same statistical techniques as applied to well yield (above) to quantify the chances of obtaining water of a given quality within a given aquifer, and also to reveal statistically significant differences in the water quality distributions of differing lithological units.
For example, Banks et al. (1998b) systematically analysed water samples from some 1604 Norwegian bedrock groundwater boreholes. They observed:
that bedrock groundwater chemistry, like borehole yield, could not be predicted at a given site in a deterministic manner;
that bedrock groundwaters were hydrochemically more mature than both surface waters and water from superficial drift aquifers, e.g. they contained higher dissolved ionic contents, higher pH and alkalinity etc.;
as regards water type, that bedrock groundwaters were typically of calcium bicarbonate composition, although a significant number could be classed as high pH, sodium bicarbonate waters;
that a weak trend to a more ‘mature’ hydrochemistry could be discerned with increasing borehole depth (Frengstad & Banks 2003);
that within any given lithology there was typically a very wide range of concentrations of most parameters;
that the systematic differences between different lithologies were, for most parameters (including major ions and pH) rather low. In fact, variation in concentration between boreholes within a given lithology was often orders of magnitude greater than the variation in median values between different lithologies.
Figure 9 shows a series of boxplots documenting the distribution of groundwater pH amongst differing lithologies in Norway (see Table 2). The wide range in pH values should be noted, as should the fact that the median pH in almost all lithologies is around 8.2 to 8.3. Banks et al. (1998b, 2000) and Frengstad & Banks (2000) suggested pH increases with progressive carbonate and silicate weathering and that this median pH represents an interval of buffering during calcite saturation and precipitation from the water.
Ca+++HC03 – =CaC03+H+
Boxplots displaying distribution of laboratory-determined pH values in Norwegian hard rock groundwater, sorted according to lithological groups (see Table 2). For comparison, boxplots are displayed on the extreme left showing: the entire data set for all hard rock groundwaters (‘Bedrock’, n - 1604), for groundwaters from superficial drift aquifers (‘Drift’, n = 72) and for n = 1 surface water works (‘Surface water’, n = 1). See Banks et al. (1998b, 2000).
The variation in pH within the various lithologies is ascribed largely to varying degrees of hydrochemical maturity (i.e. water-rock interaction) which would probably depend on factors such as residence time, fracture shape etc., which ultimately depend more on the geomechanical properties of the aquifer than on their detailed lithochemistry. As the chemistry of so many other major ions and trace elements depends ultimately on pH and the carbonate system (Frengstad et al. 2001), the concentrations of the majority of hydrochemical parameters in hard rock aquifers can be ascribed to:
hydrodynamic factors (residence time, fracture shape)
universal hydrochemical equilibria; specifically the calcium carbonate system
and not, to any great extent, to the detailed geochemical or mineralogical composition of the host rock.
A few elements do, however, exhibit a discernible lithological dependence. Prominent among these are uranium and radon. Uranium is highly soluble under oxidizing conditions, while radon is ultimately dependent on the uranium and radium content of the solid phases in contact with water. These parameters are thus dependent to a lesser extent on, for example, water pH and the degree of hydrochemical evolution of the water relative to the carbonate system, and more on the uranium content of the host rock aquifer. Indeed, uranium concentrations in bedrock groundwater in Norway vary across six orders of magnitude (and, along with several other elements, exhibit a distribution approaching log-normality; Frengstad et al. 2000). Uranium (and also radon and fluoride) concentrations can also be seen to be clearly elevated in granites (group 92, in Fig. 10) and depleted in groundwaters from anorthosites (group 93). The apparently elevated concentrations in group 72 are, on detailed inspection, also related to minor granitic inliers within that group (Frengstad et al. 2002). Uranium concentrations of up to 2 mg/1 have been recorded elsewhere in Norway (Morland et al. 1997), while over 14 mg/1 have been found in Finland (Asikainen & Kahlos 1979). Norwegian radon concentrations were found to range up to 31 900 Bq/1 (Banks et al. 2000).
Boxplots displaying distribution of uranium concentrations (determined by ICP-MS, in jig/1) in Norwegian hard rock groundwater, sorted according to lithological groups (see Table 2). For comparison, the extreme right-hand boxplot shows the entire data set for all hard rock groundwaters (‘Bedrock’, n - 476). See Frengstad et al. (2000); Banks et al. (2000). The apparently elevated concentrations in group 72 are, on detailed inspection, also related to minor granitic inliers within that group (Frengstad et al. 2002). Data below the analytical detection limit are plotted at a value of half the detection limit for the purposes of graphical presentation.
Boxplots also enable the comparison of groundwater composition in similar aquifers from different climatic conditions. For example, Banks et al. (1998a) compared groundwater from the granitic Isles of Scilly, with groundwaters from the granitic Hvaler islands of Norway. Huge hydrochemical differences were noted (Fig. 11): the Scilly waters were far less hydrochemically mature (lower pH, alkalinity, silicon, Na/Cl ratio, calcium content etc.), lower in fluoride, uranium and radon, but much richer in nitrate and potassium. The Hvaler waters appeared to be more chemically reducing in nature, with higher iron and manganese concentrations. The differences were ascribed to the following factors:
the Scilly Isles are currently submergent in nature, while the Hvaler Isles are emergent (post-glacial isostatic rebound);
Scilly thus has a long subaerial weathering history, while Hvaler has recently been glacially scoured;
the Hvaler Isles are thus composed of relatively fresh, unweathered granite; Scilly's granite is likely to be more weathered and possibly depleted in uranium and basic minerals along the main groundwater flow pathways;
the Scilly Isles have higher topographic gradients than Hvaler (faster groundwater flow);
the Scilly granite may be more permeable than that of Hvaler;
Scilly is subject to relatively intense floriculture and agriculture, compared to Hvaler; soils are sandy and unfrozen in winter and may be more prone to nitrate leaching.
Boxplots comparing distribution of selected parameters in granitic groundwater from the Hvaler Islands (n = 11) with granitic groundwater from the Isles of Scilly (n = 10). Based on data published by Banks et al. (1998a). Data below the analytical detection limit are plotted at a value of half the detection limit for the purposes of graphical presentation.
Shand & Frengstad (2002) and Frengstad (2002) have continued this study further, comparing a range of Norwegian hard rock lithologies with Welsh Lower Palaeozoic argillite aquifers, the Cornish Permian Carnmenellis Granite and the Scottish Devonian Sandstone aquifers. These studies confirmed the generally higher pH, higher alkalinity nature of the Norwegian groundwaters and their generally lower nitrate concentrations.
Conclusions
Hydrogeological prognosis
How might we restore the self respect of the hydrogeologist in hard rock terrain vis-à-vis the dowser? Should we continue to play our games of hydrogeological chess against granitic opponents, marshalling our resources of geophysics and structural analysis upon a playing board of aerial photos and LandSat images?
Alternatively, should we simply accept that hard rock hydrogeology is a game of bingo, and focus our energies on providing clients with realistic estimates of their chances of obtaining a well of adequate yield and chemistry, via statistical techniques such as those described above?
The authors would argue that the correct analogy is possibly a game of poker. The outcome is largely dependent on chance. However, the astute poker player will employ sensible strategies (i.e. using geophysics or lineament analysis, where appropriate) and will have an excellent understanding of the probabilities of various outcomes (using statistical analysis). The best poker players, not being content to accept the probabilities, will probably also have an ace up their sleeves, with which to cheat Mother Nature. One example of such a hydrogeological ‘ace’ would be the use of explosives or hydraulic fracturing (hydrofraccing). The latter is a technique of creating or opening subsurface fractures by injection of water at pressures of several MPa. During the past years, hydrofraccing techniques have advanced considerably (Less & Anderson 1994; Lloyd 1999), although even three decades ago, studies performed in Sweden suggested that hydrofraccing typically results in a median increase in bedrock borehole yield of around 2001/h, with increases of up to 18001/h (Müllern & Eriksson 1977; Fagerlind 1979; 1982).
Aquifer vulnerability and characterization
The EU Water Framework Directive requires the characterization of hard rock aquifers. The Nordic nations have progressed a long way on the road to this objective by the publication of groundwater atlases and maps (Lahermo et al. 1990; Aastrup et al. 1995) or by the analysis of large databases of hydraulic and hydrochemical information (Morland 1997; Banks et al. 1998b; Frengstad et al. 2000; 2001). The hard rock areas of the UK (including Scotland, Wales, SW England and parts of Northern Ireland) need to complete this task within a relatively short time frame. Currently, the British Geological Survey is contracted to characterize aquifer units to produce ‘aquifer property’ maps of Scotland. This current characterization involves placing each aquifer unit in one of several ‘classes’ based on its productivity (typical long-term well yield), groundwater chemistry (pH, mineralization, redox condition), depth of flow, length of flow pathway, transmissivity, porosity and degree of confinement (Fitzsimons et al. 2004).
Additionally, the vulnerability of Scottish groundwater is being assessed in a GIS environment, using a methodology described by O'Dochartaigh et al. (2005). This methodology considers, however, only the vertical contaminant pathway from the surface to the water table. The vulnerability of this pathway, according to O'Dochartaigh et al.’s methodology, does not depend on the properties or lithology of fractured bedrock but is largely determined by the nature of the soils and superficial deposits. However, once any contamination reaches the water table in a fractured bedrock aquifer, it may migrate laterally down-gradient towards specific receptors such as wells and springs. In this case, one would expect the vulnerability of the receptor to depend to some extent on the hydraulic properties of the fractured aquifer.
It is argued that the analysis of available well yield and hydrochemical data sets using simple non-parametric statistical methods would be a straightforward and inexpensive way of characterizing the hydraulic properties and hydrochemistry of fractured, hard rock aquifers. Authors such as Shand & Frengstad (2002) have already attempted the statistical characterization of the hydrochemistry of some British hard rock aquifers. Available data from Scotland are sparse and possibly somewhat non-representative, however, and those available are presented in publications by Robins (1999) and Shand & Frengstad (2002). It is likely that locally operating drillers may hold additional information on the yields and water chemistry of smaller private and farm wells in the hard rock areas of Scotland. There is a case for the systematic collation of such data, via the development of mutually trusting and beneficial relationships between drillers and regulators, in order to establish larger data sets to support statistical analysis. If this could be achieved, it can be argued that the characterization of a Scottish granite aquifer is as simple as presenting a data table such as Table 3.
Example of the possible characterization of, for example, a granite aquifer using non-parametric statistical analysis - Norwegian Precambrian Granite (Lithological Group 92, see Table 2) (data derived from studies of Morland 1997 and Banks et al. 1998b).
- © 2005 Scottish Journal of Geology
References
In this issue
Use of non-parametric statistics as a tool for the hydraulic and hydrogeochemical characterization of hard rock aquifers
