Environ. Sci. Technol. 2008, 42, 4863–4869
Comparison of Selection Methods To Deduce Natural Background Levels for Groundwater Units JASPER GRIFFIOEN,* HILDE F. PASSIER, AND JANNEKE KLEIN TNO Geological Survey of The Netherlands, P.O. Box 80015, 3508 TA Utrecht, The Netherlands
Received December 28, 2007. Revised manuscript received April 1, 2008. Accepted April 9, 2008.
Establishment of natural background levels (NBL) for groundwater is commonly performed to serve as reference when assessing the contamination status of groundwater units. We compare various selection methods to establish NBLs using groundwater quality data for four hydrogeologically different areas in the highly populated and developed subcatchment Western River Rhine, The Netherlands: selection of old groundwater (before 1945), of tritium-free groundwater (i.e., infiltrated before 1950), and of groundwater having no agricultural contamination by NO3 and SO4. Differences as well as similarities in percentile values for Cl, NH4, and SO4 concentrations are observed among the selection methods as well as the spatial units, pointing out that selection of the data set is a crucial step in deducing NBLs. The following general points of attention are deduced: (1) reference to composition of recharge water (rain or river infiltrate) is necessary to confirm the statistical outcomes, (2) old analyses are affected by conservation errors after sampling for redox-sensitive solutes and may be obtained by selective sampling, (3) old analyses are the only direct reference for NBLs for groundwater units having only anthropogenically influenced, young groundwater at present, and (4) establishment of a priori percentile values as maximum NBL is not right and confirmation by additional processbased insight in the controls on water composition is necessary.
Regional geohydrological and geochemical differences in the natural conditions of groundwater make it necessary to differentiate between groundwater units. Good groundwater management may require establishing different NBLs for different groundwater units. To be able to determine the NBL easily and rapidly for numerous groundwater units, a relatively simple and robust approach is preferred. Nowadays, natural background levels are hard to establish in many areas of the world due to widespread contamination of young water systems by, e.g., industrialization, urbanization, and agriculture. They have been determined in various ways: by statistical analysis (2, 3), by historic analysis (4), or by selecting a subset of data that is assumed to have the natural composition based on hydrological and geochemical tracers (e.g., see ref 5). The latter method was adopted within the framework of the EU specific targeted research project BRIDGE as a procedure to obtain NBLs and threshold values for groundwater bodies (cf. ref 6). Intercomparison of the various methods was not performed previously, and their associated advantages and limitations are not well realized. The objective of this study is to compare various selection approaches to establish natural background levels for several essentially different groundwater units in a highly developed area with widespread groundwater contamination using existing groundwater composition data. We focus on fresh groundwater within the first tens of meters below the surface in the western Netherlands in an area where brackish and saline paleogroundwater is also present at shallow depth. In the study area, groundwater recharging as precipitation surplus or river bank infiltration is ubiquitous and this recharge water may introduce contamination. The selection approaches were compared for Cl, SO4, and NH4, which are among the so-called Water Framework Directive (WFD) pollutants according to the EU Groundwater Directive (2006/ 118/EC). They naturally occur at low concentrations outside marine influence (or evaporites) and are typically indicator solutes for contamination. These selection methods can be applied for other compounds and groundwater units in highly developed areas where groundwater recharge occurs. The methods are particularly applicable where residence times are several tens of years or more.
Study Area Introduction Knowledge on natural background levels of concentrations in groundwater is important to identify contamination of groundwater by anthropogenic activities (1). Here, natural background levels (NBLs) are defined as concentrations present in water as controlled by natural geogenic, biological, and atmospheric processes. The recent European Union (EU) Groundwater Directive (2006/118/EC) explicitly asks to consider background levels when establishing threshold values for groundwater pollutants. The maximum NBL is an important reference value in the process of determination of the good status of the groundwater, in the context of ecological and human functions and for the protection of the groundwater itself. The maximum NBLs are compared with the composition of shallow, infiltrating groundwater to serve as an early warning of groundwater aquifer contamination, so that counteracting management measures can be taken in time to protect the deeper aquifer. * Corresponding author phone: +31-30-256.4470; fax: +31-30256.4755; e-mail:
[email protected]. 10.1021/es7032586 CCC: $40.75
Published on Web 05/21/2008
2008 American Chemical Society
Geography. The study area is the Western River Rhine subcatchment as it is defined within the implementation of the EU water framework directive. It is the most western subcatchment of the Rhine catchment bordering the North Sea (Figure 1). The subcatchment contains large urban areas and is heavily industrialized (20%). The main land use type in the less dense populated areas is intensive agriculture (58%). Forest and other types of nature comprise a small area (8%) and are situated in the sandy areas along the North Sea or at the eastern fringe. Anthropogenic activities put strong pressures on the environment in this densely populated and strongly developed area. The geomorphology of the Western River Rhine subcatchment generally consists of a polder landscape with an altitude near or below sea level. A coastal dune belt is found in the western part along the North Sea. On the northeastern side the subcatchment is bordered by a fresh lake (IJsselmeer). In the east the deltaic area is bordered by the more elevated terrain of ice-pushed ridges, in the south Pleistocene sandy deposits are found that are associated with the Meuse subcatchment, and in the southwest the estuaries of the Rivers Schelde and Meuse are found together with a series of islands or peninsulas. VOL. 42, NO. 13, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4863
FIGURE 1. Different WFD groundwater bodies considered with the sampling locations for the OXC and tritium methods within subcatchment Western River Rhine (left) and the geoscientific units with sampling locations for the historic method (right). Geology and Hydrogeology. The study area is dominated by the deltaic area of the River Rhine where Holocene deposits lie at the surface. The hydrological basis lies at 200-500 m below the surface (7). A thick sequence of predominantly continental deposits is overlying this basis. The coastal dune belt has several tens of meters of Holocene eolian deposits at the surface, forming a phreatic aquifer. The polder area has a confining layer at the surface that consists of fluvial deposits and peat in the eastern part or a combination of fluvial and marine deposits and peat in the western part. The thickness of this confining layer increases westward from less than 1 to over 50 m. In the elevated terrain in the east, tens of meters thick unconsolidated deposits being icepushed during the Saalien ice age lie at the surface, which form a phreatic aquifer. Below the polder area and the coastal dunes lies a several tens of meters thick aquifer complex, which is composed of Late and Middle Pleistocene fluvial deposits regionally intercalated with Eemian marine sediments or Weichselian fluvioglacial deposits. Fresh to saline paleogroundwater dating back to about 5000 years B.P. is present particularly in both the confining layer and the aquifers below the polder area (8). Exfiltration of saline and brackish groundwater gives rise to undesired salinization of surface water. Hydrodynamics. The study area lies in a temperate, coastal climatic zone. Groundwater recharge of precipitation surplus happens at the phreatic aquifers of the coastal dunes and ice-pushed ridge and in the relatively high-lying polders. River bank infiltration happens toward the lower lying polders from the River Rhine and its deltaic river branches. Regional groundwater exfiltrates at the lowest lying polders and to some extent in a series of polders along the elevated terrain of the coastal dunes or ice-pushed ridge. Groundwater flow patterns are mainly controlled by differences in surface water levels between polders, lakes, canals, and rivers, resulting in 4864
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 13, 2008
a complex system of local to regional groundwater flow systems (Figure 2).
Methods Approach. Three approaches to deduce the natural background levels were intercompared. First, historic analyses before 1945 were retrieved from the geoscientific database of TNO Geological Survey of The Netherlands (www. dinoloket.nl) as agricultural activities and other anthropogenic activities strongly intensified after 1945 (e.g., see refs 9–11). Second, filters (i.e., well screens) having nonrecent groundwater were selected from operational monitoring networks, where nonrecent groundwater is defined as groundwater that does not contain tritium. Groundwater having tritium content above 5 tritium units (TU), when sampled in 1983 as for the data set considered (see below), is typically groundwater that has infiltrated after 1950 (12). This is referred to as “recent groundwater” as opposed to “nonrecent groundwater” that has infiltrated before 1950. Recent groundwater is likely more anthropogenically influenced in developed areas of the world, as the intensification of agriculture, industrialization, etc., and associated contamination of groundwater has grown strongly after World War II (13). This makes tritium content in groundwater a suitable tracer. Last, filters with groundwater having an oxidation capacity (OXC) that refers to nitrate and sulfate concentrations for natural land use were selected from existing monitoring networks (cf. ref 14; see below). The first approach is henceforth referred to as the historic method, the second as the tritium method, and the last as the OXC method. As NBLs serve to primarily identify contamination of recharging fresh groundwater, only fresh groundwater having [Cl] < 200 mg/L was considered according to all three methods. Historic data on the composition of rainwater were obtained from Stuyfzand (15) and Leeflang (16), and those
FIGURE 2. Schematic west-east cross-section of flow systems in the northern part of subcatchment Western River Rhine, The Netherlands. on the River Rhine, from various historic sources. The rainwater composition includes dry deposition because continuously open rain gauges were used in the past. Data Selection. The data set used in both the tritium method and the OXC method consists of groundwater analyses in Western River Rhine from the national and four provincial monitoring networks (cf. ref 17). For the national monitoring network, groundwater quality is measured yearly at about 360 locations at 10 and 25 m below the surface. For the provincial networks, measurements are yearly, once per 2 years or once per 4 years There are 20-65 locations per province and in total about 300 locations in The Netherlands. The monitoring wells also contain two filters at similar depths. In the province of Zuid-Holland there are two additional, frequently sampled filters at 3 and 15 m below the surface. Median values per monitoring network filter were calculated from analyses for the period 1994-2005 before statistics per groundwater unit were calculated. A median value is a robust statistically representative value for time series of groundwater analysis. For the tritium method, filters were selected having a tritium content smaller than 5 TU in 1983 when they were sampled for tritium analysis (12). For the OXC method, both NO3 and SO4 concentrations were considered to exclude anthropogenic influence. The OXC method to exclude samples with anthropogenic influence is for fresh samples ([Cl] e 200 mg/L): samples having [OXC] > 2 mequiv/L are removed with OXC ) 7[SO4] + 5[NO3] where OXC is the oxidation capacity in milliequivalents per liter and [i] is the concentration of species i in millimoles per liter. The values of 7 and 5 refer to the amount of electrons transferred when NO3 gets reduced to N2 or SO4 gets reduced to FeS2, respectively. The use of OXC provides insight when nitrate reduction in association with pyrite oxidation is a relevant process (18): contamination of groundwater by nitrate gets masked when nitrate reduction occurs, but the production of SO4 by pyrite oxidation still results in groundwater having a high content of total dissolved solids and
poor quality. Oxidation of pyrite may be an actual process in marine or fluviatile Holocene deposits (19, 20). The criterion between natural composition and anthropogenic altered composition was set to 2 mequiv/L. This criterion is based on the average N (NO3 and NH4 that nitrifies after precipitation events) and SO4 concentrations in rainwater before 1945 corrected for the evaporative concentration from rainwater to groundwater recharge. The evaporative concentration factor is about 2.5 in The Netherlands as average annual rain is about 750-900 mm and the rainfall excess is 250-350 mm/year (12). Nitrogen saturation in the soil is assumed in this way, which is unlikely at low atmospheric N deposition for natural forest ecosystems (20–22). However, this criterion serves to split off relatively high nitrate groundwater due to nitrate leaching through the soil under high atmospheric deposition in nature areas or manuring in agricultural areas. Table 1 shows the numbers of samples for all three selection approaches, and Figure 1 shows the associated locations. For the selection of groundwater analyses before 1945 we retrieved data from the Dutch groundwater database DINO (www.dinoloket.nl). All old complete analyses (at minimum Na, Ca, Mg, alkalinity, Cl, and SO4 and preferably also NO3, K, Fe, NH4, pH, and Al) were collected and checked for electroneutrality condition. If the absolute value of electron balance was above 10%, analyses were discarded, where electron balance is calculated as the ratio between (sum of cation - sum of anions) and {2 × (sum of cations + sum of anions)}. For a few filters, several controlled analyses before 1945 were available and the median values per filter were calculated. The sampling filters selected for the historic method are unevenly spread across the geographical units, which relevance will be more elaborately discussed in Results and Discussion. Data Handling. The three methods were compared for three aqueous compounds: Cl, SO4, and NH4. These three compounds are so-called WFD pollutants according to the EU Groundwater Directive (2006/118/EC). The nutrients N and phosphate are the most critical compounds for surface VOL. 42, NO. 13, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4865
TABLE 1. Medians and 90th Percentiles (mg/L) as well as the Amount of Samples for the Various Combinations of Groundwater Unit, Compound, and Selection Methoda median or 90th percentile for given unit, method, and amount of sample Holland
Cl-50% Cl-90% NH4-50% NH4-90% SO4-50% SO4-90%
dunes
ice-pushed ridgeb
riverine area
OXC,61
historic,17/16
tritium,65
OXC,99
historic,131/129
tritium,41
OXC,68
historic,70
tritium,50
OXC,23
historic,80
96 A 160 A 16.4 A 35 A 1.0 A 1.0 A
86 A 144 A 3.3 B 10.9 B 4.6 A 137 B
88 A 172 A 17.1 A 38.8 A 1.0 A 0.4 A
48 A 120 A 3.0 A 15.6 A 6.0 A 23.0 A
40 B 79 A 1.0 B 4.7 B 4.5 A 23.6 A
41 AB 102 A 3.7 A 13.4 A 0.6 A 31.9 A
61 A 166 A 3.8 A 22.1 A 2.0 A 18.3 A
31 A 89 B 0.8 B 7.1 B 19.4 B 109 B
39 A 91 AB 1.6 AB 21.0 AB 2.1 AB 55.6 B
57 A 156 A 0.1 A 1.5 A 19.0 A 22.0 A
19 B 106 B 0.1 A 1.3 A 15.9 A 56.5 B
a Letter A refers to reference value including its confidence intervals, B is a different percentile value for another method without overlapping confidence intervals with A, and AB indicates the percentile value with confidence intervals for a third method that overlaps with both A and B. b For tritium method, there were insufficient amounts in this unit.
water quality in The Netherlands. Eutrophication of surface water is a major water quality issue, and surface water concentrations above Dutch environmental threshold values are frequently detected. Unfortunately, it appeared that old analyses are unreliable with respect to PO4, as 457 analyses out of 676 old analyses had a concentration equal to 0. This is unrealistic for the study area, since high natural concentrations are frequently observed presumably due to natural mineralization of sedimentary organic matter (23, 24). Trace metals, which are also of large interest for NBLs, could not be intercompared because their analyses are seldom available before the 1970s. Intercomparison between the data sets was performed by comparing the medians (p50) and 90th percentile values (p90) of data sets with their confidence intervals as p50 and p90 are two characteristic figures for the range in background levels present. Here, the 95% two-sided nonparametric confidence intervals were calculated according to Helsel and Hirsch (25). When the confidence intervals of two percentile values did not overlap, they are said to be different from each other. The medians of a data set are used as the prime characteristic to compare whether the data sets are identical. The p90 is calculated as the reference for maximum NBL. This value may subsequently be used to derive a threshold value as required in implementation of the EU Groundwater Directive (2006/118/EC), as explained by Wendland et al. (6). One may argue that higher percentile values, e.g., 95 or 97.3%, would be more suitable as reference for maximum NBL (cf. refs 1 and 2), but higher percentiles also require larger data sets as the confidence intervals increase strongly with more extreme percentiles away from the median (14). Spatial Units. Intercomparison of the three selection methods was made for four spatial units within the WFD subcatchment Western River Rhine. On the one hand, these units are the WFD groundwater bodies, and on the other hand, they are geogenic units each having a characteristic geological and hydrogeological buildup. The four geographical units are as follows: (1) the Holland area where brackish and saline groundwater is found at shallow depth in addition to fresh groundwater (fresh samples only; OXC and tritium methods) or where Holocene marine deposits are found in the confining layer (historic method); (2) the Holocene coastal dunes without a confining layer (all three methods); (3) the riverine area where only fresh groundwater occurs at shallow depth (OXC and tritium methods) or where the confining layer consists only of Holocene peat and fluvial deposits (historic method); (4) the ice-pushed ridge containing a phreatic aquifer (all three methods). Note that the borders between the Holland and riverine areas differ 20 km at most for the OXC and tritium methods on the one hand and the historic method on the other hand. This is because delineation is based on the groundwater Cl concentration (OXC and tritium methods) or the sedimen4866
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 13, 2008
tological origin of deposits in the confining layer (historic method). For the first two methods, the border is defined as the most western occurrence of fresh groundwater; for the latter method the border is defined as the most inland, eastern occurrence of Holocene, marine deposits in the confining layer (cf. Figure 1).
Results and Discussion Table 1 shows medians and 90th percentiles for the combinations of selection method, compound, and groundwater unit. As mentioned before, the median serves to indicate the prime characteristic of the various data sets and to allow intercomparison among both groundwater units and methods. The p90 may serve directly as a reference value for the maximum NBL. Chloride. The medians for chloride are rather identical for the three selection methods per spatial unit (Table 1). A clear difference between the OXC and historic methods is observed for the ice-pushed ridge. Note that no values for the ice-pushed ridge could be calculated for the tritium method, because these areas are regional infiltration areas where young groundwater is only found in the shallow subsurface. The values are highest for Holland, where brackish and saline groundwater is also found. A continuum in Cl concentration from less than 100 mg to concentrations as high as 5000-10000 mg/L is observed, and the statistics for Cl are directly influenced by the upper selection limit of 200 mg/L. The p90s are quite similar for Holland but differ among the methods for the other areas. Reference to the historical composition of rainwater and River Rhine water is useful here. In the 1930s, Cl concentrations in rainwater sampled tens of kilometers away from the coast were in the range of 2-5 mg/L, and close to the coast 5-30 mg/L due to sea spray. The Cl concentration in the River Rhine was naturally 12 mg/L (26) and in the range of 45-85 mg/L in the 1910s to 1930s. In The Netherlands, the evaporative concentration factor is about 2.5 for plants and trees (12, 27). Taking into account this evaporative concentration factor, the p90s of the historic and tritium methods correspond for the dunes and riverine area to both the maximum historic rainwater concentration of Cl close to the coast and the maximum historic Cl concentration of the River Rhine, but those of the OXC method do not. The OXC method appears to be a less appropriate method to deduce NBLs of chloride for the dunes and the riverine area, because modern groundwater selected by the OXC method may have high Cl concentrations but low NO3 concentrations due to subsurface denitrification as well as SO4 concentrations due to SO4 reduction in the redoxreactive coastal lowlands of The Netherlands (28). The two p90s for the ice-pushed ridge do not at all correspond to historic rainwater composition away from the
FIGURE 3. Cumulative frequency distribution of chloride (left) and sulfate (right) for the riverine area and ice-pushed ridge according to the historical method.
TABLE 2. General Advantages and Limitations of the Three Selection Methods Applied To Deduce Natural Background Levels method OXC historic
tritium
advantage
potential limitation
-present-state analytical and conservation techniques -widely applicable -most direct reference to natural groundwater
-method inappropriate due to reduction of both NO3 and SO4 in groundwater -understanding of recharge mechanism necessary -conservation errors for redoxsensitive compounds -analyses unlikely present for many trace elements -some anthropogenic contamination may be present as well -selective sampling for historic drinking water winning -no old groundwater in several kinds of groundwater units existing -3H analysis necessary
-unambiguous tracer of groundwater age -present-state analytical and conservation techniques
coast whether or not corrected for evaporative concentration. Inclusion of anthropogenically influenced groundwater seems relevant for this area even for the historic analyses. Reference to p50 seems to be more appropriate for the icepushed ridge. Inspection of the cumulative frequency distribution of Cl according to the historic method points out a concave cumulative frequency distribution having an inflection point close to p70 (Figure 3). The concentrations above these inflection points are likely anomalous due to the influence of the groundwater composition by anthropogenic activities. This argues against a priori fixation of percentiles as reference for NBLs (6, 29). An explanation is that many groundwater quality investigations were performed in the past for exploration of drinking water and not for environmental investigation. Selective drilling in the vicinity of urbanized areas was performed where contamination by domestic and agricultural activities is likely. This causes bias of the statistics toward higher p90s when aggregating the analyses at a regional scale. Selective drilling is also evidenced by the clustered occurrence of sampling filters for the historic method (Figure 1). Ammonium. For ammonium, the medians and p90 for the OXC and tritium methods are similar and differ with those of the historic method, except for the ice-pushed ridge, albeit there is overlap for the riverine area in the confidence intervals for the historic and tritium methods (Table 1). The
high absolute values especially in Holland should be noted: values above the Dutch environmental nitrogen standard for surface water of 2.2 mg of N/L are frequently observed. In this area, these concentrations are considered to be natural and originate from mineralization of geologically young sedimentary organic matter. The high NH4 concentrations together with high PO4 concentrations are associated with high CO2 pressures in confined groundwater. A preliminary investigation using groundwater analyses after 1945 shows that the nutrient concentrations PO4 and NH4 are not correlated with Cl in Holland, confirming a geogenic origin. This topic is beyond the scope of the present study. The systematically lower percentiles for the historic method are explained from a sampling artifact. We propose that nitrification of NH4 after sampling gave rise to a systematic underestimation of the NH4 concentration for the old groundwater analyses. Proper analysis for NH4 requires acidification of samples using H2SO4. This was not a common procedure before 1945 (30, 31). The historic method, therefore, appears not to be appropriate for redoxsensitive NH4 and likely also other redox-sensitive solutes. The p90s according to the OXC and tritium methods serve as maximum NBLs. Sulfate. The percentiles of SO4 indicate similarities as well as striking differences among the three methods (Table 1). The results for the dunes are near-identical among the three VOL. 42, NO. 13, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4867
methods except for a lower median in case of the tritium method. The lower median for the tritium method may be ascribed to selective sampling of aged groundwater in the downstream part of groundwater flow systems and a prolonged period of over 50 years for mineralization of sedimentary organic matter with associated SO4 reduction during groundwater flow. The p90s are equal among the three methods, and combined with the natural land use in the dunes, this implies that these values reflect maximum NBLs. For the ice-pushed ridge, the medians for the OXC and historic methods do not differ but the p90s do. In the 1930s, SO4 concentrations in rainwater sampled tens of kilometers away from the coast were in the range of 3.5-8 mg/L (15, 16), and close to the coast they were 4-8 mg/L with early outliers up to 17 mg/L. The median values for SO4 are equal to or lower than the rainwater concentration when evaporative concentration is included, but p90 is certainly not comparable to the rainwater concentration for the historic method. As p90 for the ice-pushed ridge in the historic method is higher than that for the dunes, an additional source for SO4 must have been present. This may be (NH4)2SO4 fertilizer salts, because the sandy soils at the ice-pushed ridge are naturally infertile. The cumulative frequency distribution curve of SO4 for the ice-pushed area bends increasingly upward from p70 on for the historic method (Figure 3). Appropriateness of fixation of the natural SO4 background levels at p90 is questionable for reasons explained before. Notable differences are observed for sulfate in both Holland and the riverine area: both percentiles are highest for the historic method although there is overlap in the confidence intervals with those of the tritium method. In the case of the median value for Holland, this is attributed to analysis of samples using the BaSO4 precipitation-and-weight method (30, 31), which had a higher detection limit than modern analytical methods (commonly up to 5 mg/L versus 0.1-2.0 mg/L). The p90 values of the historic method for Holland and riverine areas point to other sources for SO4 than atmospheric deposition associated with rainwater infiltration or SO4 dissolved in river water associated with river bank infiltration. The sulfate concentration in the River Rhine was naturally 35 mg/L (26), in the range of 35-55 mg/L in the 1910s to 1930s, and up to 100 mg/L in the 1960s to 1970s. Albeit the detection limit was high, systematic analytical errors for groundwater analyses of SO4 are unlikely before 1945 because SO4 does not react after sampling. The cumulative frequency distribution of SO4 in the riverine area shows a clear inflection point slightly above 50 mg/L for the historic method (Figure 3), which coincides with the maximum River Rhine concentration before 1940. Sulfate sources in the soil are oxidation of pyrite due to lowering of the groundwater table by drainage and use of (NH4)SO4 fertilizer salts. The latter is unlikely as the soils in Holland and riverine areas are naturally fertile due to high clay content and marine origin in case of deep polders in Holland. However, the occurrence of acid-sulfate soils was a well-known agricultural management problem in the 1950s in this area, which points out too much drainage of formerly anaerobic soils. Selection of a maximum NBL based on these three methods is hard because of the counteracting impact of subsurface SO4reduction and pyrite-oxidation by drainage. A criterion based on River Rhine composition is more suitable for fresh groundwater: 35 mg/L as a strictly natural concentration or 55 mg/L as seminatural concentration is appropriate for Holland and the riverine areas, where groundwater recharge of both rain and river water happens. Comparison. A process-based approach to deduce NBLs for natural compounds is feasible and preferable as an alternative for fully statistical approaches. The three selection methods presented here (historic, tritium, and OXC methods) each have their advantages and limitations as pointed out 4868
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 13, 2008
in the preceding text and summarized in Table 2. One additional aspect is that analyses of trace elements are unlikely present for the historic method. Groundwater analyses on trace elements were seldomly performed before the 1960s as they were often laborsome, of small practical interest in the past, and confronted with high detection limits. The other two methods can be used for any natural groundwater compound. General intercomparison of the medians and p90s per area reveals that no pattern occurs: they can lie within 10% of each other for one series of three values and they differ frequently more than 50% for other series of three (Table 1). A combination of methods strengthens the establishment of NBLs as mutual confirmation is reached and limitations of individual approaches become clear. Establishment of NBLs using predefined percentile values in selection methods, and in statistical methods as well, is not right and should be confirmed by additional data on the composition of recharge water for compounds that do not have a completely geogenic origin as Cl and SO4.
Acknowledgments Sophie Vermooten and Marie¨lle van Vliet are acknowledged for their contribution in the data selection and handling. The research described was financed by the European Commission within the BRIDGE project under Contract No. 006538 as well as the geoscientific information program of TNO Geological Survey of The Netherlands. The views expressed in this paper are purely those of the writers and may not be regarded in any circumstances as stating an official position of the European Commission.
Literature Cited (1) Langmuir, D. Aqueous environmental geochemistry; Prentice Hall: Upper Saddle River, NJ, 1997. (2) Lee, L.; Helsel, D. Baseline models of trace elements in major aquifers of the United States. Appl. Geochem. 2005, 20, 1560– 1570. (3) Kunkel, R.; Wendland, F.; Hannappel, S.; Voigt, H. J.; Wolter, R. The influence of diffuse pollution on groundwater content patterns for the groundwater bodies of Germany. Water Sci. Technol. 2007, 55, 97-105 (doi: 10.2/86/wst.2007.077). (4) Limbrick, K. J. Baseline nitrate concentration in groundwater of the Chalk in south Dorset, UK. Sci. Total Environ. 2003, 314316, 89–98. (5) Edmunds, W. M.; Carillo-Rivera, J. J.; Cardona, A. Geochemical evolution of groundwater beneath Mexico City. J. Hydrol. 2002, 258, 1–24. (6) Wendland, F.; Berthold, G.; Blum, A.; Elsass, P.; Fritsche, J. G.; Kunkel, R.; Ruediger, W. Derivation of natural background levels and threshold values for groundwater bodies in the Upper Rhine Valley (France, Switzerland and Germany). Desalination 2008, 226, 160–168. (7) Dufour, F. C. Groundwater in the Netherlands. Facts and figures. TNO Netherlands Institute of Applied Geoscience: Delft/Utrecht, The Netherlands, 2000. (8) Post, V. E. A.; Van der Plicht, H.; Meijer, H. A. J. The origin of brackish and saline groundwater in the coastal area of the Netherlands. Neth. J. Geosci./Geolog. Mijnbouw 2003, 82, 133– 147. (9) Mylona, S. Sulphur dioxide emissions in Europe 1880-1991 and their effect of sulphur concentrations and depositions. Tellus 1996, 48B, 662–689. (10) Griffioen, J.; Keijzer, T. Biogeochemical cycles of chloride, nitrogen, sulphate and iron in a phreatic aquifer system in the Netherlands. In Proceedings of the 10th International Symposium on Water-Rock Interaction; Cidu, R., Ed.; Balkema: Lisse, The Netherlands, 2001, 1, 521–524 (11) Broers, H. P.; Van der Grift, B. Regional monitoring of temporal changes in groundwater quality. J. Hydrol. 2004, 296, 192–220. (12) Meinardi, C. R. Groundwater recharge and travel times in the sandy regions of the Netherlands; Ph.D. thesis, Vrije Universiteit, Amsterdam, the Netherlands, 1994. (13) Meybeck, M.; Chapman, D. V.; Helmer, R. Global freshwater quality. A first assessment; Blackwell: Oxford, U.K., 1989. (14) Broers, H. P. Strategies for regional groundwater quality monitoring. Ph.D. thesis, University Utrecht, The Netherlands, 2002.
(15) Stuyfzand, P. J. The composition of rain water along the coast of Holland. KIWA-Report No. SWE-91.010; KIWA: Nieuwegein, The Netherlands, 1991 (in Dutch). (16) Leeflang, K. W. H. The chemical composition of precipitation in The Netherlands. Chem. Weekbl. 1938, 35, 658–664. (in Dutch). (17) Van Duijvenbooden, W. Groundwater quality monitoring in The Netherlands. In Regional groundwater quality; Alley, W. M., Ed.; Van Nostrand Reinhold: New York, 1993; pp 515535. (18) Postma, D.; Boesen, C.; Kristiansen, H.; Larsen, F. Nitrate reduction in an unconfined sandy aquifer: Water chemistry, reduction processes, and geochemical modelling. Water Resour. Res. 1991, 27, 2027–2045. (19) Ritsema, C. J.; Groenenberg, J. E. Pyrite oxidation, carbonate weathering, and gypsum formation in a drained potential acid sulfate soil. Soil Sci. Soc. Am. J. 1993, 57, 968–976. (20) Van Helvoort, P. J.; Griffioen, J.; Hartog, N. The reactivity of heterogeneous riverine unconsolidated sediments: The RhineMeuse delta (the Netherlands). Appl. Geochem. 2007, 22, 27352757 (doi: 10.1016/j.apgeochem.2007.06.016). (21) Gundersen, P.; Emmett, B. A.; Kjonaas, O. J.; Koopmans, C. J.; Tietema, A. Impact of nitrogen deposition on nitrogen cycling in forests: A synthesis of NITREX data. Forest Ecol. Manage. 1998, 101, 37–55. (22) Wright, R. F.; Alewell, C.; Cullen, J. M.; Evans, C. D.; Marchetto, A.; Moldan, F.; Prechtel, A.; Rogara, M. Trends in nitrogen deposition and leaching in acid-sensitive streams in Europe. Hydrol. Earth Syst. Sci. 2001, 5, 299–310.
(23) Griffioen, J. Uptake of Phosphate by Fe-hydroxides during Seepage in Relation to Saline/Fresh Groundwater Displacements. Environ. Sci. Technol. 1994, 28, 675–681. (24) Griffioen, J. Extent of immobilization of phosphate during aeration of nutrient-rich, anoxic groundwater. J. Hydrol. 2006, 320, 359–369. (25) Helsel, D. R.; Hirsch R. M. Statistical methods in water resources; Studies in Environ. Science No. 49; Elsevier: Amsterdam, The Netherlands, 1992. (26) Molt, E. L. Contamination of River Rhine water. Water 1961, 45, 143-148; 160-164 (in Dutch). (27) Gehrels, J. C. Groundwater level fluctuations. Separation of natural from anthropogenic influences and determination of groundwater recharge in the Veluwe, the Netherlands. Ph.D. thesis, Vrije Universiteit, Amsterdam, the Netherlands, 1999. (28) Van den Brink, C.; Frapporti, G.; Griffioen, J.; Zaadnoordijk, W. J. Statistical analysis of anthropogenic versus geochemicalcontrolled differences in groundwater composition in The Netherlands J. Hydrol. 2007, 336, 470-480 (doi: 10.1016/ j.jhydrol.2007.01.024). (29) Reimann, Cl.; Filzmoser, P.; Garrett, R. G. Background and threshold: Critical comparison of methods of determination. Sci. Total Environ. 2005, 346, 1–16. (30) Collins, W. D. Notes on practical water analysis; U.S. Geological Survey Water-Supply Paper 596-H; USGS: Washington, DC, 1928. (31) Meerburg, P. A.; Massink, A. Methods for chemical and bacteriological drinking water research. Noordhoff: Groningen, The Netherlands, 1934 (in Dutch).
ES7032586
VOL. 42, NO. 13, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4869
