Journal of Hydrology 261 (2002) 24±47

www.elsevier.com/locate/jhydrol

Use of abstraction regime and knowledge of hydrogeological conditions to control high-¯uoride concentration in abstracted groundwater: San Luis PotosõÂ basin, Mexico J.J. Carrillo-Rivera a,*, A. Cardona b, W.M. Edmunds c a Instituto de GeografõÂa, Universidad Nacional AutoÂnoma de MeÂxico, CU, CoyoacaÂn, 04510, DF, Mexico Posgrado en Ciencias de la Tierra, Universidad Nacional AutoÂnoma de MeÂxico, CU, CoyoacaÂn, 04510, DF, Mexico c The British Geological Survey, Wallingford, OX10 8BB, UK

b

Received 29 September 1999; revised 23 October 2001; accepted 5 November 2001

Abstract Signi®cant amounts of ¯uoride are found in the abstracted groundwater of San Luis PotosõÂ. This groundwater withdrawal induces a cold, low-¯uoride ¯ow as well as deeper thermal ¯uoride-rich ¯ow in various proportions. Flow mixing takes place depending on the abstraction regime, local hydrogeology, and borehole construction design and operation. Fluoride concentrations ( < 3.7 mg l 21) could become higher still, in time and space, if the input of regional ¯uoride-rich water to the abstraction boreholes is enhanced. It is suggested that by controlling the abstraction well-head water temperature at 28±30 8C, a pumped water mixture with a ¯uoride content close to the maximum drinking water standard of 1.5 mg l 21 will be produced. Further, new boreholes and those already operating could take advantage of ¯uoride solubility controls to reduce the F concentration in the abstracted water by considering lithology and borehole construction design in order to regulate groundwater ¯ow conditions. q 2002 Elsevier Science B.V. All rights reserved. Keywords: Fluoride; Hydrogeochemistry; Groundwater; Water±rock interaction; Mexico

1. Introduction Natural contamination due to particular geological environments can be an important factor in limiting available water resources. Groundwater is the major source of potable water supply in arid and semi-arid areas and its availability may be threatened not only by the introduction of contaminants by human activities but also by natural processes. Until recently, the impact of trace elements in the water supply was not considered in the management of groundwater in MeÂxico. However, the contribution of some trace elements (¯uoride, iron, arsenic, lead, cadmium, for * Corresponding author.

example) to natural groundwater pollution and to health related effects may be substantial in many groundwater regions world-wide (Edmunds and Smedley, 1996). Fluoride is a common natural contaminant in groundwater supplies in both industrialised and developing countries. The present work discusses natural contamination by deep groundwater ¯ow in the vicinity of the city of San Luis PotosõÂ (SLP), Mexico (Fig. 1a). A signi®cant amount of low salinity groundwater (200±350 mg l 21, total dissolved solids) with temperature between 33.8 and 40.4 8C and with high concentrations of ¯uoride (2.10±3.65 mg l 21) is pumped by public supply boreholes. The hydrogeology and a de®nition of the regional hydrochemistry of the area are given in

0022-1694/02/$ - see front matter q 2002 Elsevier Science B.V. All rights reserved. PII: S 0022-169 4(01)00566-2

J.J. Carrillo-Rivera et al. / Journal of Hydrology 261 (2002) 24±47

25

Fig. 1. (a) Location of study area showing surface geologic units and area where shallow groundwater contaminated with sewage water ef¯uents have been traced (Adapted from Carrillo-Rivera and Armienta, 1989, and Carrillo-Rivera et al., 1996). (b) Hydrogeological crosssection along a horizontal ¯ow line indicating the local, intermediate and regional ¯ows. The hydraulic conductivity (horizontal) refers to the upper part of the deep aquifer (Tv and TGU materials), and the temperature variation was measured during discharge at borehole head. The location of section is indicated in (a).

Carrillo-Rivera et al. (1996): relevant aspects are presented in the following section. Dental ¯uorosis has recently been recognised in SLP in people with high exposure to naturally occurring ¯uoride in drinking water (Grimaldo et al., 1995). MedellõÂn-MilaÂn et al. (1993) reported some degree of dental ¯uorosis in 84% of the population between 6 and 30 years of age; 34% of children 11±13 years old

had severe ¯uorosis. Only children show severe dental ¯uorosis, as opposed to senior citizens who lack signi®cant effects (Sarabia, 1989). As ¯uoride shares a common source with arsenic, lead and cadmium, the presence of the former is of additional concern as the latter may also be introduced into the water supply (MedellõÂn-MilaÂn et al., 1993). A wide variety of studies is available on various

26

J.J. Carrillo-Rivera et al. / Journal of Hydrology 261 (2002) 24±47

Fig. 1. (continued)

aspects of ¯uoride related to groundwater. Nordstrom and Jenne (1977), Edmunds et al. (1984), Robertson (1986), Zhaoli et al. (1989), Travi and Faye (1992), Nordstrom et al. (1989) and Hitchon (1995), among others, studied ¯uoride as a result of water±rock interaction in various aquifers with different lithologies. From the groundwater contamination point of view, Kilham and Hecky (1973), Handa (1975a,b), Nanyaro et al. (1984), Gaciri and Davies (1993) and Pekdeger et al. (1992) have found that ¯uoride is a natural occurring contaminant due to its observed high concentrations in groundwaters low in calcium and magnesium. From the human health perspective, several researchers have made important contributions in relation to skeletal and dental ¯uorosis such as Teotia et al. (1981), Diesendorf (1986), Tennakone et al. (1988), Dissanayake (1991) and Rajagopal and Tobin (1991); studies on SLP have been reported by MedellõÂn-MilaÂn et al. (1993) and Grimaldo et al. (1995).

With regard to treatment procedures to remove high ¯uoride concentrations from groundwater, two main systems have been experimentally implemented: (i) through sorption onto arti®cial and natural resins (Tejook, 1993; Sarabia, 1989; Ovalle, 1996) and (ii) through the control of ¯uoride solubility by adding gypsum to the arti®cial sand-pack ®lter of a borehole (Appelo and Postma, 1996). Both techniques could treat groundwater; although, the gypsum technique increases sulphate concentration in the abstracted water, and resins are expensive and produce waste disposal problems. No technique has so far been demonstrated to solve the problem at source, before abstraction. The objectives of this paper are: (i) to identify the evolution of high ¯uoride waters in time and space in the area of SLP, (ii) to de®ne the possible solubility controls of ¯uoride concentration in groundwater and (iii) to discuss an in situ technique to control ¯uoride in the groundwater system.

J.J. Carrillo-Rivera et al. / Journal of Hydrology 261 (2002) 24±47

27

Table 1 Average of geochemical analyses of ¯ow systems for shallow and deep groundwater from the San Luis PotosõÂ surface drainage basin. Depth is indicated in metres, temperature in 8C, EC in mmho s cm 21, pH without units, other chemical constituents in mg l 21. ND means no data. Geochemical analyses of seven boreholes from 1962 abstracting regional ¯ow, and intermediate ¯ow water analyses from six boreholes from 1962, and their respective averages, are included as a reference Sample

Depth

Temp

pH

EC

SiO2

HCO3

Cl

SO4

NO3

Na

K

Ca

55

154

13.0

18.0

5.0

56.0

5.0

19.0

5.7

3.2

10.1

11.9

8.8

32.6

Mg

F

Li

B

1.09

3.2

0.19

0.190

10.4

1.73

0.3

0.01

0.031

10.2

25.1

1.94

1.0

0.06

±

0.4

0.01

0.08

Regional ¯ow system (Group I) 266 36.6 7.13 393 Mean a Intermediate ¯ow system (Group II)

210 24.5 6.56 140.2 95.6 Mean a Mixed water (regional and intermediate) (Group III)

72.6

252 26.7 7.11 Mean a Local ¯ow system (Group IV)

260

83

138

11.4

11.8

5.8

45 21.0 6.22 Mean a Regional ¯ow system (1960)

820

70

212

85.4

84.8

29.2

50

23.8

87.1

. 200 38.0 7.30 360 PDC b GLM b . 200 31.0 7.60 370 TDA b . 200 30.5 7.70 350 . 200 37.5 7.40 390 FDC b SRH-1 b . 200 35.0 7.50 340 SRH-8 b . 200 29.0 7.50 400 LF b . 200 30.0 7.80 270 Mean . 200 33.0 7.54 354 Intermediate ¯ow system (1960±1962)

ND ND ND ND ND ND ND ND

159 152 159 157 160 172 143 157

15.0 14.0 15.0 13.0 13.0 17.0 9.0 13.7

24.0 31.0 24.0 29.0 24.0 20.0 1.0 21.9

ND ND ND ND ND ND ND ND

54.0 51.0 46.0 55.0 56.0 52.0 35.0 49.9

7.0 13.0 10.0 6.0 4.0 13.0 5.0 8.3

22.0 21.0 24.0 21.0 19.0 24.0 23.0 22.0

1.00 1.00 1.00 0.50 0.50 1.50 1.00 0.93

ND ND ND ND ND ND ND ND

ND ND ND ND ND ND ND ND

ND ND ND ND ND ND ND ND

MYR b JT b CE c CC c BSR c Uc Mean

ND ND ND ND ND ND ND

5.0 11.0 6.0 7.8 10.0 5.0 7.5

3.0 3.0 5.4 12.8 14.4 6.4 7.5

ND ND ND ND ND ND ND

11.0 16.0 11.5 15.0 14.0 18.5 14.3

11.0 9.0 8.8 11.0 8.6 5.0 8.9

10.0 12.0 7.0 10.5 10.0 4.0 8.9

2.5 3.0 1.8 1.8 1.0 1.6 2.0

ND ND ND ND ND ND ND

ND ND ND ND ND ND ND

ND ND ND ND ND ND ND

a b c

240 254 . 200 . 200 . 200 . 200 . 200

24.5 28.0 23.5 24.0 26.5 26.5 25.5

6.7 7.2 6.8 7.1 7.2 7.2 7.0

150.0 168.0 83.0 110.0 100.0 75.0 114.3

60.0 81.0 59.0 30.0 51.0 51.0 55.3

13.1

Modi®ed after Carrillo-Rivera et al. (1996). IIZA (1962). Stretta and Del Arenal (1960)

2. The study area 2.1. Climate The City of San Luis PotosõÂ is located approximately 400 km north-west of MeÂxico City at the south-eastern end of the Sierra Madre Occidental and is the capital of the state of the same name. The climate is semi-arid with an average rainfall, mainly during the summer months (June±September), of about 400 mm annum 21. The study area is a closed drainage basin (Fig. 1a) lacking any perennial runoff.

The mean annual and summer temperatures are 17 and 21 8C, respectively. The average annual potential evaporation is approximately 2000 mm (1961± 1990). The elevation of the basin ¯oor is between 1850 and 1900 m above datum (mean sea level). 2.2. Hydrogeological setting GeologyÐthe main geological features of this area are typical of a number of similar surface basins in the Sierra Madre Occidental, Central Alluvial Basins (Back et al., 1988) as well as other regions of north-western

28

J.J. Carrillo-Rivera et al. / Journal of Hydrology 261 (2002) 24±47

MeÂxico and south-western US. An average of about 1700 m of Tertiary fractured lava ¯ows, tuff and ignimbrite cover the inferred hydrogeological basement (Cretaceous limestone and calcareous mudstone, and a Tertiary quartz±monzonite intrusion). These volcanic rocks have an abundant glassy matrix of about 80% in volume. Zielinsky (1983) and Robertson (1986) found high ¯uoride in the glassy matrix in rocks of similar age and composition to those in SLP at the northern boundary of the Sierra Madre Occidental. Geochemical analyses of local lava ¯ows reported ¯uorine (.0.2% by weight) and occasionally small topaz crystals partially ®lling fractures and cavities in the rock (TristaÂn, 1986). A clastic sequence (gravel, sand, silt and clay derived from the surrounding Tertiary volcanic rocks) was deposited on top of the volcanic units as basin ®ll (Fig. 1b). Calcareous material resulting from erosion of Cretaceous rocks located to the northeast of the SLP city area is expected to be a source of clastic deposits in the basin ®ll. These sediments are interbedded with pyroclastic material; both are referred to in this paper as Tertiary Granular Undifferentiated (TGU). Borehole logging data indicate the presence of a 50±150 m thick bed of ®ne grained and compactsand (locally referred to as `clay layer') fully enclosed in the TGU. This layer is found under most of the ¯at part of the basin except at the edges and has an area extent of about 300 km 2. X-ray diffractometry of this compact sand shows about 90% by volume of quartz and sanidine, as well as a small fraction of carbonates and clays (halloysite, illite and montmorillonite). Aquifer systemÐthe hydrogeology of the study area is de®ned in detail by Carrillo-Rivera (1992) and Carrillo-Rivera et al. (1996). In this drainage basin, two main aquifers are separated by the low permeability compact-sand layer. The shallow aquifer is in alluvium, uncon®ned, perched on the compact-sand layer; it is recharged by rainfall and untreated sewage derived from agricultural return ¯ow. The depth to the water table is in the order of 5±30 m and mean observed drawdowns of about 4 m are produced by an average abstraction of 10 l s 21. The deep aquifer units have a regional distribution with limits far beyond the boundary of the basin. Within the drainage basin, this aquifer is in both the TGU and fractured volcanic units; it is con®ned over

most of the ¯at part of the basin and has a depth to the potentiometric surface between 60±150 m. Usually, boreholes tapping this aquifer terminate at a depth of 350±450 m in lava ¯ows, tuffs and/or the TGU and may penetrate about 100±200 m of these volcanic units. HydrochemistryÐthe chemical characteristics of the groundwater in the San Luis PotosõÂ drainage basin are different for shallow and deep aquifers. Groundwater in the shallow aquifer is calcium-chloride-bicarbonate in composition. Shallow hand-dug wells in the north-east of SLP city (Fig. 1a) suggest a strong in¯uence of contamination from sewage wastewater. This contamination appears to be evident from high nitrate concentrations (mean 29.2 mg l 21, ranging from 2.3 to 79.9 mg l 21), high bicarbonate (mean 212.0 mg l 21, ranging from 94.4 to 414.8 mg l 21) and high chloride concentrations (mean 85.5 mg l 21, ranging from 12.0 to 225.0 mg l 21) (Group 4, Table 1). Carrillo-Rivera and Armienta (1989) and Carrillo-Rivera (1992) discuss the chemistry of contamination derived from untreated sewage ef¯uents used in an area of about 35 km 2 (Fig. 1a) where agricultural practices have been developed for more than 40 years (Stretta and Del Arenal, 1960). Temperature, lithium, ¯uoride and sodium have been used to identify two main chemical groundwater types in the deep aquifer. High concentrations of these elements and discharge temperatures higher than 33.8 8C are associated with the sodium-bicarbonate waters, ¯owing through the fractured volcanic rocks (Group 1, Table 1). Geothermometry results suggest that the chemistry of this groundwater is controlled by low temperature (70 8C , T , 80 8C) devitri®cation processes affecting the abundant glassy matrix present in the rock. This temperature range for groundwater at depth, was estimated applying quartz, chalcedony, Na±K±Ca and Na/K geothermometers. Results indicate that chalcedony (mean, 79.5 8C) and Na±K±Ca (mean 71.3 8C) geothermometers provided evidence for the last equilibrium temperature for this groundwater and support volcanic glass being the main control of SiO2 dissolution (Carrillo-Rivera et al., 1996). Fluoride and silica are incorporated into groundwater during dissolution of the reactive glassy matrix of aquifer material; lithium, sodium and calcium are mobilised from the glassy matrix by

J.J. Carrillo-Rivera et al. / Journal of Hydrology 261 (2002) 24±47

29

Table 2 Groundwater abstraction, number of active boreholes and water level decline in the San Luis PotosõÂ surface drainage basin from 1962 to 1987. Abstraction rate in m 3s 21, water level decline in m annum 21. ND means no data (Adapted from Carrillo-Rivera, 1992) Year

Abstraction rate (m 3 s 21)

Number of active deep boreholes

Water level decline (m)

1962 1972 1977 1984 1986 1990

0.550 0.778 1.890 2.300 2.630 2.600

82 ND 193 230 265 280

ND 0.9 1.0 ND ND 1.3

ionic exchange with hydrogen derived from carbonic acid dissociation (Cardona, 1990). Magnesium is mainly ®xed in the matrix of the rock and retained by clays formed as a result of the low temperature devitri®cation processes. Petrographic evidence suggests a minor solute contribution from phenocryst (mainly quartz and sanidine) dissolution (Instituto de GeografõÂa, 1999) in the rocks of the SLP area. A second groundwater chemical type is calciumbicarbonate water with well-head temperatures between 23 and 27 8C that circulates through the granular portion of the deep aquifer. This water is characterised by low lithium, ¯uoride and sodium concentrations, and high silica values (Group 2, Table 1). Cardona (1999) explained the chemistry of this groundwater by inverse geochemical modelling using chemical data for local rainfall and reactions involving incongruent dissolution of minerals identi®ed by X-ray diffractometry in the granular material. The main reactions reported involve incongruent dissolution of sanidine, Ca-montmorillonite, Namontmorillonite and illite resulting in the formation of kaolinite; in addition there is cation exchange between calcium and sodium, and carbonate dissolution. Hydraulic characteristicsÐpumping-test analysis results reported by Carrillo-Rivera (1992) indicate average horizontal hydraulic conductivity for the shallow aquifer material to be about 2 £ 10 24 m s 21, and about 1 £ 1024 m s 21 for the TGU, values that correspond with boreholes producing groundwater with low temperature. Fractured volcanic units show low horizontal hydraulic conductivity ( < 0.01 £ 10 24 m s 21) which corresponds with areas where the groundwater temperature is high (Fig. 1b). A high vertical hydraulic conductivity (about 10 22 m s 21) has been de®ned for

some boreholes tapping the fractured volcanic units. A lack of any relationship between the depth of individual boreholes and water temperature has been observed (Cardona, 1990; Carrillo-Rivera, 1992), and no information is available on hydraulic conductivity distribution with depth. The average horizontal hydraulic conductivity of the compact-sand layer was computed to be < 10 29 m s 21, as a statistical mode from 26 pointpiezometer slug-tests using the Hvorslev (1951) method. The slug-tests were carried out along the upper 20 m of the compact-sand layer, in the southern part of SLP City, where values from 1.32 £ 10 26 to 3.5 £ 10212 m s 21 were reported (GeoingenierõÂa Internacional, 1996); horizontal and vertical hydraulic gradients for the same site were 0.006±0.008 and < 1 m m 21, respectively. Runoff collected in San Jose and El Peaje dams provide a seasonal water supply in the order of 0.20 m 3 s 21 for the city of SLP. Consequently, about 95% of the total current consumption (2.60 m 3 s 21) is derived from the deep aquifer. Abstraction causes a corresponding water level decline of < 1.3 m annum. Table 2 shows annual groundwater abstraction rates (Q) and the average water level decline over the period from1962 to 1990. Q1972 (0.78 m 3 s 21) and Q1977 (1.89 m 3 s 21) produced average water level declines of 0.9±1.0 m annum 21, respectively. When Q1972 is compared to Q1990, these data indicate that by increasing the abstraction some 330% (from 0.78 to 2.60 m 3 s 21) there is an increase in rate of water level decline of only 40%. Boreholes tapping the fractured volcanic units of the deep aquifer produce from 0.005 to 0.055 m 3 s 21, while boreholes in the granular material report yields of about 0.003±0.035 m 3 s 21 (Carrillo-Rivera, 1992). In general, the deep aquifer drawdowns vary between 1 and 50 m; however, an

30

J.J. Carrillo-Rivera et al. / Journal of Hydrology 261 (2002) 24±47

average of 20 and 14 m is measured in abstraction boreholes in volcanic material and TGU, respectively. Groundwater ¯ow in the horizontal direction is toward the main abstraction areas located in the centre and south-east of the city of SLP. There are some in¯ows to deep (.200 m) abstraction boreholes derived from the shallow aquifer, which take place through the compactsand layer or at its edge (or due to faulty borehole construction). Computations with a numerical model (Rathod and Rushton, 1991) suggest that boreholes have a maximum input of water from the shallow aquifer of approximately 1±3% of the total obtained yield. This in¯ow is consistent with observed drawdown response with abstraction time (i.e. lack of leaky effect as described by Hantush (1956) in boreholes located in areas where groundwater has a temperature of 23±27 8C (Carrillo-Rivera, 1992). A lowering in the rate of increase of drawdown with time is observed in boreholes where groundwater temperature is higher than 32 8C. A similar effect has been documented in an area with thermal in¯ow (44.1 8C) in South Arabia (Kawecki, 1995). An overall water input from the shallow to the deep aquifer through the compact-sand layer was investigated with a ®nite difference numerical approach using the computer code MODFLOW (McDonald and Harbaugh, 1988). Modelling was carried out under steady state that included differences in hydraulic gradients between shallow and deep aquifers (as in Fig. 1b) as well as variable thickness in the compactsand layer. Homogeneous and anisotropic conditions (10 29 and 10 210 m s 21, for horizontal and vertical hydraulic conductivity, respectively) were considered for the compact-sand layer. The domain integrated 500 cells each 1000 £ 1000 m 2, that included the shallow and compact-sand layer thickness as well as a 1700 m thickness of Tertiary volcanics. General head boundaries were used for both shallow and deep aquifers; recharge was assigned to the shallow aquifer. Assuming a porosity of 0.1, the vertical and horizontal ¯ow velocities obtained within the compact-sand layer were approximately 0.03 and 0.0006 m annum 21, respectively. The 300 £ 10 6 m 2 of the compact-sand allows an in®ltration of about 0.03 m 3 s 21. Water budgetÐa groundwater balance equation,

considering the con®ned area of the developed part of the deep aquifer to a depth equal to that of the abstraction boreholes ( < 400 m), and as applied to available data for October 1986 to September 1987, can be written as follows: Inflow…LIN 1 RVER † 2 outflow…Q 1 LOUT † ˆ change in storage ^ …SDwA†

…1†

where LIN and LOUT are lateral input and lateral output, respectively; RVER the vertical in¯ow; Q the groundwater abstraction; S the storativity; Dw the change in water level elevation and A is the surface area. RRAIN includes in¯ows related to direct vertical recharge from rainfall (RRAIN), in®ltration from the overlying semi-con®ning layer (IOSC) and vertical ¯ow from beneath (IVIB). The available information allows estimates to be made, with some uncertainty involved for all variables in Eq. (1), exception made for (IVIB). Rearranging variables: Q 1 …SDwA† 2 LIN 2 RRAIN 2 IOSC 1 LOUT ˆ IVIB …2† Estimated values, as indicated in Carrillo-Rivera (2000), for related variables are as follows. Q (2.69 m 3 s 21, error ^ 20%) was computed from continuous ¯ow-meter measurement (6% in volume), instant abstraction determination and duration of operation (24%), and electricity consumption per unit of discharge and total energy consumption (70%). S (0.001) was obtained from pumping-test analyses. Dw (21.35 m) is the water level difference for 1989±1990 data. A (300 £ 10 6 m 2) is the surface area of the con®ning bed. LIN (0.15 m 3 s 21) was computed as groundwater ¯ow to the centre of the abstraction area using transmissivity values of 1.9 £ 10 24 ±19.0 £ 10 24 m 2 s 21 and 8.1 £ 10 24 ± 40.0 £ 10 24 m 2 s 21 for the fractured media and granular material, respectively, and an average hydraulic gradient of 0.004. RRAIN is equal to 0.0 m 3 s 21 as con®ned conditions prevail. IOSC (0.2 m 3 s 21) was estimated with MODFLOW. LOUT is considered nil (m 3 s 21) as suggested by the potentiometric head distributions. Solving for IVIB yields an estimate of < 2.3 m 3 s 21, a value that includes uncertainties in Eq. (2). The error of about ^20% in determining the abstraction rate is considered the most important uncertainty. Analyses considering an uncertainty of

J.J. Carrillo-Rivera et al. / Journal of Hydrology 261 (2002) 24±47

31

Table 3 Four examples of possible reactions to reproduce observed mixed groundwater, during calculations Cl 2 was considered as a conservative element, (in parenthesis) water analysis of that particular groundwater ¯ow that was used to make calculations Example

Regional (%)

Intermediate (%)

Local (%)

Main reactions

Uncertainty

Average analyses Borehole 37

33 23 (Borehole 11)

64 74 (Average)

3 3 (Borehole 57)

15 3

Borehole 40

9 (Borehole 3)

85 (Average)

6 (Borehole 58)

Borehole 43

15 (Average)

83 (Average)

2 (Borehole 52)

Dissolution of calcite Dissolution of calcite, sanidine, illite Precipitation of MgMontmorillonite plus ion exchange Dissolution of calcite, sanidine, illite Precipitation of caolinite Dissolution of calcite, sanidine and Ca-Montmorillonite Precipitation of MgMontmorillonite

^20% for Q, leaving other components of the budget as constant, would result in the IVIB estimate ranging between 1.8 and 2.8 m 3 s 21. An independent chemical based estimate for IVIB shows that about 1.82 m 3 s 21 of the 2.6 m 3 s 21 groundwater abstraction has a temperature higher than 33.8 8C (Herrera et al., 1992), which was considered to represent a groundwater input from beneath. Further, the decrease of groundwater production per unit of water level decline, as identi®ed for 1972± 1990 (Table 2), is considered as further support for an additional source of water. Flow systemsÐlocal ¯ow systems have been identi®ed in the shallow aquifer, while intermediate and regional ¯ows were identi®ed in the deep aquifer by Carrillo-Rivera et al. (1996). This ¯ow system de®nition follows that of Wallick and ToÂth (1976), which implies a hierarchy arrangement where local system ¯ows travel at shallower depths than intermediate ¯ow systems, and regional ¯ow systems reach the deepest sections of the groundwater ¯ow domain. The local ¯ow system circulates in the shallow aquifer, having a temperature of 21 8C ^ 1; its average major ion, and some minor and trace elements concentrations are as shown in Table 1, Group 4. Complete geochemical data for local groundwater, as well as for intermediate and regional ¯ows, is given in Carrillo-Rivera et al. (1996). The intermediate groundwater ¯ow system is considered an end-member (Group 2, Table 1) that circulates at shallow depth in granular material of the deep aquifer. Geology and hydrochemistry

10 15

suggest the recharge area of this ¯ow to be runoff generated in the upper reaches of the basin and in®ltrating the TGU material in piedmont area, mainly by the Sierra de San Miguelito (Cardona and CarrilloRivera, 1996). The regional ¯ow system (Group 1, Table 1) is considered as another end-member, which has been observed to originate in fractured volcanic rocks. Both the wide distribution of volcanic rocks far beyond the surface boundary of the basin and the intermediate ¯ow that is found at shallow depth in the study area suggest that the regional ¯ow has its origin beyond the western limit of the drainage basin. Geothermometry results indicate a minimum equilibration temperature, for groundwater from the regional ¯ow system, of about 75 8C ^ 5 which, using local geothermal gradient data implies that this ¯ow travelled to a depth of 1.7 km ^ 0.4. There are no discharge areas for the intermediate and regional ¯ow systems within the drainage basin, and except for local temporary springs, the closest thermal (36 8C) natural outlet, Ojo Caliente, is located some 27 km to the south of the drainage basin divide (Carrillo-Rivera et al., 1992). Mixture of ¯owsÐthe main source of abstracted groundwater in the deep aquifer units comes from the mixing of regional and intermediate ¯ow systems to various degrees producing Group 3 chemistry as shown in Table 1. Hydraulic evidence presented above is consistent with a negligible input of groundwater leaking through the compact-sand layer into the deep aquifer. Considering advection as a major

32

J.J. Carrillo-Rivera et al. / Journal of Hydrology 261 (2002) 24±47

Fig. 2. Conceptual model of mixing among identi®ed groundwater groups: shallow (contaminated water), intermediate and regional ¯ows. Calculated amount of Cl and Li shows the degree of uncertainty considered in the inverse modelling.

control of shallow water moving across the compactsand layer, the computed vertical ¯ow velocity is suf®ciently low ( < 0.028 m annum 21) for contaminated water to be detected in the deep aquifer. The use of sewage water in irrigation practices started some 40 years ago. Should dispersion, or preferential ¯ow, prevail in some parts of the compact-sand layer, this could explain the high nitrate concentrations in Group 3 boreholes 38 and 39 (7.7 and 11.0 mg l 21, respectively), and the high chloride concentrations (27.5 mg l 21) in borehole 48, as reported by Carrillo-Rivera et al. (1996). Results are supported by the limited evidence of contamination of groundwater in groups 1 and 2, an effect that is only traced in borehole 16 tapping intermediate ¯ow with 33.1 mg l 21 of nitrate. Indication of contamination of the mixed waters by the shallow waters would be suggested by samples displaying a deviation from linear mixing relationships between end-members (groups 1 and 2). Few samples show such deviation. However, most boreholes are located away from the area where shallow contaminated groundwater has been de®ned and hence any in®ltration from the shallow system may not be easily identi®ed chemically. Despite these problems a chemical mass balance was attempted to de®ne the overall importance of the proportion of contaminated shallow groundwater in water abstracted from the deep aquifer. The

computed in¯ow is considered to represent the minimum concentration of shallow groundwater entering the deep aquifer, corresponding to that collected by abstraction boreholes during working hours. This mass balance was made based on the following assumptions: (i) contaminated water entering the deep aquifer is represented by the average composition of sampled shallow boreholes; (ii) the compact-sand layer produces a negligible change in the chemical composition of in®ltrated water; (iii) water in¯uenced by contamination is represented by an excess in salinity over the average composition of mixed groundwater (i.e. the difference between theoretical and observed average composition of mixture). Inverse modelling to calculate the relative proportions of mixed water obtained in production boreholes as well as reactions produced during mixing was made with phreeqc (Parkhurst, 1995). No source or sink was considered for chloride, phreeqc assumed this element to be conservative in due calculations. Four examples were analysed to estimate the proportions of in¯ow from local ¯ow to abstraction boreholes where mixed ¯ow has been de®ned (Table 3). The ®rst example included the average geochemical analyses (Table 1) of mean concentration for the regional, intermediate and local ¯ows. Three additional examples for boreholes 37, 40 and 43 were considered; water analyses of their thermal, intermediate and local ¯ow components were represented by water

J.J. Carrillo-Rivera et al. / Journal of Hydrology 261 (2002) 24±47

Fig. 3. Groundwater temperature distribution for the deep aquifer (in 8C) at borehole-head for 1962 (IIZD, 1962).

33

34

J.J. Carrillo-Rivera et al. / Journal of Hydrology 261 (2002) 24±47

Fig. 4. Groundwater temperature difference (in C) for the deep aquifer, between 1962 and 1987.

analyses from nearby boreholes, or by the average values of the geochemical analyses (Table 1) when these analyses were unavailable. Table 3 shows the percentage for the different proportions of thermal,

intermediate and local ¯ow components in the mixture for each example. Individual values used for the geochemical analysis for regional, intermediate and local groundwater are reported in Carrillo-Rivera et al.

J.J. Carrillo-Rivera et al. / Journal of Hydrology 261 (2002) 24±47

35

Fig. 5. Fluoride distribution in groundwater of the deep aquifer (in mg l 21) after several months of continuous abstraction (1987). Number next to borehole symbol corresponds to those in Table 1.

36

J.J. Carrillo-Rivera et al. / Journal of Hydrology 261 (2002) 24±47

Fig. 6. Relation between borehole-head temperature to ¯uoride concentration in groundwater abstracted from the deep aquifer, San Luis PotosõÂ (1987).

(1996). Results are considered to indicate the signi®cance of the input of shallow groundwater to deep abstracted water, which according to obtained results it could vary, depending on location. Bicarbonate, sodium, potassium, calcium, silica, magnesium and pH, were considered as non-conservative. Consequently, reactions deduced to control the chemistry of groundwater from the intermediate ¯ow including illite, montmorillonite, sanidine and calcite dissolution, as well as cation exchange between sodium and calcium, were utilised. The results (Table 3) suggest that average mixed water, as obtained by boreholes, comprises 33, 64 and 3% of regional, intermediate and local ¯ow system waters, respectively. The value of the local ¯ow input of 3% for the average groundwater analyses agrees in principle with the computed percentage for boreholes 37, 40 and 43; of 3, 6 and 2%, respectively. Uncertainty values from 8 to 15% were applied in the modelling; in the ®rst example (Table 3) the model applies the value of uncertainty of 15% affecting only analysis of the local ¯ow, which implies ^12.8 mg l (see Fig. 2). The mass transfer produced by theoretical reactions between TGU material and the mixed water obtained are also shown in Table 3. The reactions column represents a set of feasible reactions produced during mixing to obtain the concentration of water of group 3. Reaction results

are in agreement with saturation indices; however, regarding the saturation index of Ca±Mg-montmorillonite there was no data on aluminium concentration as to assist modelling interpretation (see Section 4.2). As proposed in Fig. 2, a conceptual model for the mixed-water obtained through boreholes considers that the average derived chloride content in the abstracted mixed water has an input of < 3% in volume from the shallow aquifer. The higher average in¯ow from the shallow aquifer and from the intermediate system obtained from the chemical mass balance, as compared with results from hydraulic computations, is believed to be due to uncertainties in estimates of hydraulic parameters, as well as local increases in vertical hydraulic gradient during abstraction. 3. Methods Standard hydrogeochemical procedures were used to further validate the previous hydrogeological model of groundwater ¯ow. Under prevailing hydrogeological conditions two hydraulic conditions were examined: (i) intermediate groundwater ¯ow that travels mainly horizontally at deep borehole

J.J. Carrillo-Rivera et al. / Journal of Hydrology 261 (2002) 24±47

abstraction depth, and (ii) deeper regional in¯ows that reach abstraction areas from beneath. Chemical constituents in groundwater provide valuable information on the history of the ¯ow path the surrounding environment. Consequently, a groundwater sampling programme was made to collect water from boreholes tapping different lithologies, hydrogeological conditions and borehole discharge temperatures. Detailed sampling techniques, (University of Waterloo, 1983), and analytical procedures are discussed in Carrillo-Rivera (1992), Cardona (1990) and Cardona et al. (1993). Chemical analyses were carried out according to procedures, accuracy and precision equivalent to APHA, AWWA, WPCF standards (1989). Analyses were carried out, in duplicate with results reported as an average (Carrillo-Rivera et al., 1996).

4. Results and discussion 4.1. Fluoride and temperature evolution 1962±1987 Groundwater production from the deep aquifer began in the 1940s but abnormal temperature readings (i.e. higher than ambient) were not detected in the abstracted waters, at least up to the early 1960's. A UNESCO groundwater survey reports deep borehole abstraction water temperatures. Two boreholes in La Florida (to the SE of the city of SLP) had temperatures of 30.0 and 31.0 8C, respectively; a third one (Fig. 3) in the south of the city had a temperature of 38 8C (Stretta and Del Arenal, 1960). Two years later, reports indicate the construction of eight additional deep boreholes tapping thermal water with temperatures up to 38 8C (IIZD, 1962). Other boreholes had waters with well-head temperatures lower than those measured at present. In general, a temperature increase in the abstracted groundwater with time has been identi®ed (Fig. 4) suggesting deeper groundwater is being induced to pumping levels. Fig. 3 indicates the location of the ®rst boreholes where thermal water was tapped in 1962. The total groundwater abstraction was the order of 0.55 m 3 s 21. A comparison of water temperatures for 1962 with those of 1987, as reported in Carrillo-Rivera et al. (1996), indicates signi®cant differences as shown in Fig. 4. There is a temperature rise of approximately

37

5±10 8C, with a maximum rise of 15 8C: increase of .5 8C are seen over 60% of the area. Minor increases in temperature occur away from the area of known faults. Major temperature changes are related to areas where faults in the volcanic material have been identi®ed, the in¯uence of thermal water decreases away from fault areas. New thermal areas with increasing groundwater temperatures are also recognised in the north and south-east of the basin. Disregarding that boreholes for the survey of 1960 have been closedown, when Fig. 3 is compared to Fig. 5, high groundwater temperature areas correspond with those of high ¯uoride concentrations. A good correlation between ¯uoride and groundwater temperature (correlation coef®cient r ˆ 0:80) is observed for available data of 1987 (Fig. 6). Fluoride concentrations and temperature are higher in the regional ¯ow water than in water from intermediate ¯ow. The equation that relates groundwater temperature (T ) to its ¯uoride content (F) is shown in Eq. (3): F ˆ …Td25:005†=3:562

…3†

where temperature T is in 8C, and ¯uoride concentration F is in mg l 21. The heterogeneous nature of the aquifer units re¯ects the presence of two end-members. One endmember (in fractured volcanic rocks) has a high ¯uoride content and temperature; in contrast, the other (in TGU) has low ¯uoride concentration and is colder. The greater scatter of the mixed water samples suggests that some mixing may occur outside the borehole, a linear relationship is a reasonable summary of the complete dataset and as an approximation for conservative mixing. The earliest available major element analyses for the regional and intermediate groundwater (IIZD, 1962) indicate a remarkable similarity to analysis carried out in 1987 (Table 1). These data suggest that abstracted thermal water has reached a chemical quasi-steady state. Available data indicate that a fairly constant but minor amount of water from the shallow aquifer is seeping through the compact-sand layer and reaching the abstraction zone of regional and intermediate ¯ows. This is suggested by similar average values of chloride and bicarbonate concentrations for 1962, as compared to 1987 values (Table 1). Consequently, ¯uoride concentrations can be interpolated for 1962, when ¯uoride was not determined,

38

J.J. Carrillo-Rivera et al. / Journal of Hydrology 261 (2002) 24±47

Fig. 7. (a) Fluorite saturation index for regional ¯ow as related to lithium concentration; (b) Calcite saturation index for the same samples as related to lithium concentration; (c) ¯uorite saturation index for same samples as related to calcite saturation index. Equilibrium range for saturation indices is indicated following Jenne et al. (1980).

J.J. Carrillo-Rivera et al. / Journal of Hydrology 261 (2002) 24±47

39

Fig. 7. (continued)

using Eq. (3) and assuming the relationship between temperature and ¯uoride has prevailed for the regional and intermediate ¯ows. IIZD (1962) reports 13 boreholes with temperatures between 28 and 38 8C, abstracting about 0.34 m 23 s 21. Cold-water abstraction was about 0.21 m 23 s 21. An estimate of ¯uoride input to the water supply system may be made for each borehole. Integral instant discharge and the computed ¯uoride concentration yield about 80 kg day 21. Similar computations were made by Herrera et al. (1992) applying a conservative mixing model with lithium using data from 1987; the calculated abstraction of regional and intermediate waters being 1.90 m 3 s 21 and 0.70 m 3 s 21, respectively (see Section 4.3). Fluoride input to the water supply system late in the 1980s was around 530 kg day 21. The non-linear relation between groundwater abstraction for 1962±1987 and its ¯uoride content is due to an increase in the production of thermal as opposed to cold water. Borehole 34, located in the southern part of the surface basin divide, taps fractured media and showed water temperature and electrical conductivity, and ¯uoride and sodium content to increase with abstraction time. The regional input reaches quasi-steady

state conditions, in terms of chemical response, after some 90 days of continuous abstraction (CarrilloRivera et al., 1996). This response was interpreted as thermal water up-coming as a result of density contrasts. The control on the upward migration of thermal water due to an overlying cold water hydraulic head has been postulated by De Marsily (1986) and Leroy et al. (1992). Density (r ) difference between cold (25 8C, r ˆ 997:0 kg m23 ) and thermal waters (75 8C, r ˆ 976:5 kg m23 ) is the same as that between sea and fresh waters. The interface between cold and thermal waters is analogous to that of marine±fresh water interface response, except that in the present case denser water overlies less dense water. Vertical tensional faulting in volcanic units may also be expected to increase hydraulic conductivity and rise of regional groundwater along fracture planes. 4.2. Mineral saturation The most important factors that dominate ¯uoride in groundwater are solubility controls and availability in aquifer minerals. Edmunds et al. (1984) and Nordstrom et al. (1989) found that ¯uorite solubility and

40

J.J. Carrillo-Rivera et al. / Journal of Hydrology 261 (2002) 24±47

Fig. 8. Groundwater ¯ow distribution in the vicinity of boreholes depending on their construction and lithology. Solid line in boreholes indicate screen length; TGU means Tertiary Granular Undifferentiated (basin ®ll); F means fractured media; f indicates a geological fault; KhTGU is horizontal hydraulic conductivity of TGU material; KVf is vertical hydraulic conductivity of fractured media.

carbonate geochemistry control ¯uoride and calcium concentrations in groundwater derived from felsic rocks. Handa (1975a,b) and Robertson (1986) in different geological environments, detected similar controls where higher ¯uoride concentrations correlated with lower calcium content. The abundance control was described by Deutsch et al. (1982) in a basalt aquifer where all samples were under-saturated in respect to ¯uorite. An understanding of controls on ¯uoride was achieved by calculating the saturation indices of Fbearing minerals using WATEQ4F (Ball et al.,

1987). Saturation indices for the thermal waters included well-head temperature. Similar computations were made for the average analysis of thermal water using the minimum equilibrium temperature at depth of 75 8C. Fluorite saturation indices and lithium concentrations were used to analyse solubility controls in the ¯ow system involved. Fig. 7a suggests that concentrations of ¯uoride are controlled by ¯uorite in thermal waters (regional system) independently of residence time as evidenced by lithium concentrations. Differences in residence time (Edmunds and Smedley,

J.J. Carrillo-Rivera et al. / Journal of Hydrology 261 (2002) 24±47

1996) as expressed by a lithium concentration of 0.07 mg l 21, appear to be signi®cant since intermediate water travelling within the basin has concentrations of 0.01±0.02 mg l 21 (Fig. 7a). Nordstrom and Jenne (1977) established that thermal waters in the western USA were in chemical equilibrium with both calcite and ¯uorite. However, the plotting of the mean thermal SLP waters for 75 8C is outside the ¯uorite equilibrium fringe, but inside the calcite equilibrium zone (Fig. 7b), where a range of the ^0.5 units de®nes the equilibrium zone (Jenne et al., 1980) representing errors involved in related ®eld measurements and analytical procedures involved in SI computations. This SI¯uorite suggests that some ¯uoride is lost during the ascent of the thermal water, due to a re-equilibrium of ¯uoride to the new temperature before the water reaches the borehole discharge. The intermediate ¯ow system is under-saturated with respect to ¯uorite. Both end-members establish the mixing path as depicted in Fig. 7a suggesting a control on ¯uoride in the abstracted groundwater; samples which do not fall on the mixing path are considered to re¯ect the effect of local ¯ow conditions in the vicinity of the borehole due to lithology and borehole construction. During the rise of the thermal water to the wellhead, its temperature is lowered from around 75 8C to an average of 36.3 8C, resulting in under-saturation with calcite (Fig. 7b). Intermediate waters in equilibrium with respect to calcite (top, area A) belong to boreholes in TGU located in the vicinity of calcareous rock outcrops. Samples of area A, appear to behave as the end-member proper for the mixing line, producing the observed distribution along this line (areas B, C) depending on calcite availability in formation units. However, equilibrium with respect to calcite for mixed waters in area C, support additional reactions during mixing as described in the previous section. The interrelation of SI¯uorite and SIcalcite (Fig. 7c) emphasises the fact that both calcite and ¯uorite control the dissolved ¯uoride and calcium only in some samples. Thermal waters in equilibrium with ¯uorite and under-saturated with calcite could still augment calcium content and theoretically decrease dissolved F. Calculations made with PHREEQE (Parkhurst et al., 1980) reveal a feasible increment of about 13.0 mg l 21 of calcium to achieve calcite equilibrium at measured temperatures. Maintaining

41

the same ¯uorite saturation index, this calcium increment yielded a 0.7 mg l 21 loss of dissolved ¯uoride. An additional ¯uoride loss could be expected by lowering the temperature (to 25 8C) diminishing the ¯ow velocity of the thermal water and inducing its ¯ow through the TGU before abstraction (Fig. 8b). This temperature decrease allows increased calcium dissolution (2 mg l 21) and ¯uoride precipitation (0.5 mg l 21). Fluorite over-saturated waters are uncommon in natural low-temperature groundwater systems (Nordstrom et al., 1989; Handa, 1975a,b; Robertson, 1986; Edmunds et al., 1984, among others). The present results support the assumed fast re-equilibration in ¯uoride content, when thermal water lowers its temperature from 75 to 36 8C, and suggest that proposed ¯uoride solubility controls are feasible. 4.3. Control of ¯uoride in abstracted groundwater Lithium may be considered as a reliable indicator of residence time along many groundwater ¯ow systems (Edmunds and Smedley, 1996). This element is mainly released from felsic igneous rocks in recharge areas and is prone to remain in solution (Hem, 1970). The strong linear grouping of data in Fig. 7a suggests that in some boreholes there is a mixing trend of regional water enriched in lithium and ¯uoride with dilute intermediate ¯ow water with lower Li and F. The chemistry of thermal water emphasises the role of temperature and residence time in the mobilisation of lithium and ¯uoride, suggesting a common origin for both species. Consequently, the linear relationship detected between lithium and ¯uoride concentrations, is considered to re¯ect mixing in the borehole of different proportions of intermediate and regional water. An interpretation of Fig.7a could be examined using the ¯ow system affected by groundwater abstraction and the conditions developed in the vicinity of the borehole. Five cases have been recognised (Fig. 8), each having lithology and borehole construction control with horizontal ¯ow, vertical upward ¯ow or vertical downward ¯ow. Three controls (a±c) may be considered involving mixture between intermediate and regional ¯ows, with two other cases representing the presence of end-members.

42

J.J. Carrillo-Rivera et al. / Journal of Hydrology 261 (2002) 24±47

Case (a)Ðhorizontal ¯ow towards the borehole dominates in the TGU and fractured media when KhTGU < KVf (Kv, vertical hydraulic conductivity; subscript f, fractured media). This appears to agree with K results of numerical modelling from pumping-test analyses of adjacent boreholes (CarrilloRivera, 1992) that indicate values for KhTGU < 1.0 £ 10 25 ms 21 (standard deviation, s ˆ 0:40 £ 10 25 ms21 ) and KVf < 0.9 £ 10 25 ms 21 …s ˆ 1:0 £ 10 25 ms21 †: In this case vertical in¯ows from the fractured volcanic material are as important those from the TGU material (Fig. 8a). Mixing of regional and intermediate ¯ows occurs in the borehole and no major deviations from the theoretical conservative mixture are expected; additional water±rock reactions are likely to be minimal over this time scale. This behaviour is considered to be representative of boreholes 28, 30, 32, 34, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 48, 49, 50 (Fig. 5). Case (b)Ðthermal water rises to the abstraction zone via the fractured tertiary granular undifferentiated (TGU) in higher proportions than the cold end-member; the borehole only penetrates the TGU material (Fig. 8b). Groundwater mixing occurs in the vicinity of the borehole as KhTGU , KVf (KhTGU < 0.7 £ 10 25 ms 21 …s ˆ 0:6 £ 10 25 † and KVf < 1.9 £ 10 25 ms 21). Slow velocities in the granular material permit reactions with thermal water involving changes, mainly, in major elements such as sodium exchanged for calcium and, to a lesser degree, magnesium (Table 1). However, because the lack of an additional source of minor elements, and in the absence of solubility controls, ¯uoride and lithium are expected to have small deviations from the theoretical conservative mixture region. This situation is considered to occur in the case of boreholes 31 and 35 (Fig. 5) where discharging groundwater has a lower temperature than that expected from Eq. (3), and Fig. 6. Fig. 7a suggests that groundwater from these boreholes had lithium addition or ¯uoride removal. The lower temperature water appears to indicate that groundwater is derived from the TGU material; however, lithium concentration is relatively high, hence ¯uoride seems to have been removed. Analytical results below the mixing line of Fig. 7a are interpreted as ¯uoride removal. Fluorite solubility control fails to explain the observed ¯uoride de®ciency, as mixed waters are under-saturated with respect to this

mineral. Further investigations of ion exchange and sorption-desorption reactions as well as ¯uorapatite solubility control, could help to explain these results. Case (c)Ðthe two sites which plot above the mixing line of Fig. 7a are interpreted as arising from ¯uoride addition to abstracted groundwater. Both samples are in equilibrium with respect to ¯uorite suggesting solubility controls similar to those of regional ¯ow. This involves additional ¯uoride dissolution during or after mixture when water from the intermediate system is induced into the fractured volcanic units due to borehole construction (Fig. 8c). Estimated regional values of KhTGU < 0.7 £ 1025 ms 21 …s ˆ 0:4 £ 10 25 ms21 † and KVf < 1.9 £ 1025 ms 21 …s ˆ 1:8 £ 1025 ms21 † suggest that this hydraulic conductivity difference and borehole construction play a major role in the observed hydrochemistry, resulting in higher concentrations and lower temperatures as seen in boreholes 29 and 47 (Fig. 5). Cases (d) and (e) correspond to groundwater derived from the thermal and cold end-members, respectively. Case (d) is obtained when KhTGU is small compared with KVf, and it is further emphasised when borehole construction taps the fractured volcanic units only. Hydrological properties of aquifer units indicate average values for KhTGU < 0.7 £ 10 25 ms 21 and KVf < 1.9 £ 10 25 ms 21. The regional in¯ows are superimposed on the system and thermal ¯ow prevails (Fig. 8d). This case is observed in boreholes (1, 2, 3, 4, 5, 6, 7, 8, 9, and 11) which penetrate fractured volcanic units and which are located close to major lineaments detected in aerial photographs (Fig. 5). In case (e) water from the intermediate system is obtained when the boreholes only tap the granular material, (Fig. 8e), and the KVf, is too low to support thermal water input. These conditions are suggested by: (i) very low ¯uoride and lithium concentrations in abstracted groundwater to the northeast and east of the city of SLP, (ii) the lack of evidence of major fractures in the aquifer materials and (iii) the presence of TGU material (see Figs. 2 and 3). Case (e) is considered to occur in the case of boreholes 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26 and 27. 4.4. Fluoride management strategy A practical option to alleviate the high ¯uoride in

J.J. Carrillo-Rivera et al. / Journal of Hydrology 261 (2002) 24±47

43

Fig. 9. Relation of ¯uoride to lithium content in groundwater of the deep aquifer.

the water supply to SLP city, is to avoid groundwater abstraction within the high ¯uoride thermal zone (Figs. 3 and 5). Boreholes need to be located in the cold water sector north-east of the city. However, in practice this is not a feasible option. Economic constraints are encountered for shifting the abstraction zone (new boreholes, distribution network, additional energy costs). Also, this area is highly susceptible to contaminants derived from the shallow aquifer. A further limiting factor is that there will still be a limited amount of groundwater available from the intermediate ¯ow system in the long term (CarrilloRivera, 1992). The area has little horizontal natural replenishment (estimated at 0.20±0.80 m 3 s 21 with Darcy's law and by chemical mass balance, respectively). An interpretation of the results indicates two possible strategies to diminish ¯uoride in the abstracted groundwater: mixing and solubility controls. Hydrogeological conditions permitting, the mixing of cold and thermal waters could be an appropriate management procedure to keep ¯uoride levels below drinking water standards. Cold water-thermal water input

should be adjusted to manage mixing proportions. The entrance of cold and thermal water into a borehole relies on several factors including: porous and fractured media hydraulic conductivity, borehole depth and design, well-losses, porous media thickness, abstraction rate, time since abstraction began, drawdown, difference in water density of related ¯ows, and thickness of intermediate ¯ow. The optimal management drawdown for each borehole to attain an acceptable ¯uoride concentration is, however, dif®cult to obtain for existing boreholes where data on borehole construction are often unavailable. A practical alternative is to apply the correlation between ¯uoride and temperature. This ¯uoride control may be implemented in existing boreholes by adjusting the abstraction rate and/or abstraction schedule (i.e. varying induced mixing proportions from each end-member). The cold and thermal water mixture has a F concentration conformable to drinking water standards ( < 1.5 mg l 21) by maintaining its temperature at 30 ^ 2.8 8C. Due to data dispersion …correlation coefficient ˆ 0:8† corresponding expected values of ¯uoride concentration could vary

44

J.J. Carrillo-Rivera et al. / Journal of Hydrology 261 (2002) 24±47 21

by ^0.8 mg l . These deviations from the value predicted from Eq. (3) imply the need to carry out individual tests in each borehole. If new ®eld determinations are not possible, electric conductivity (EC) monitoring could provide an additional ®eld test to identify in¯ow of regional groundwater with relative high salinity. A desirable alternative is to carry out a step drawdown test where ¯uoride and temperature determinations could be made at selected intervals in order to establish a local temperature-¯uoride relationship. At the same time, such a test would permit additional controls on ¯uoride concentrations to be de®ned in the hydrogeological setting where the borehole is located. Carrillo-Rivera et al. (1996) report a parabolic relationship between abstraction time and ¯uoride concentrations, temperature, sodium, and EC values in borehole 34; as time increases, the rate of increase in each of the other parameters decreases. This borehole obtains more thermal water with abstraction time, increasing ¯uoride concentration from 0.4 mg l 21 (at 30.4 8C, t ˆ 1 h) to 0.88 mg l 21 (at 33.6 8C, t ˆ 60 h). Such a response suggests ¯uoride levels could be controlled with abstraction time; however, this borehole is the only source for the town supply and subject to continuous abstraction. Data indicate that after abstraction for < 150 h, ¯uoride and lithium concentrations are 1.2 and 0.11 mg l 21, respectively; water temperature is 34.4 8C and EC changes from 170 to 245 mS cm 21. These results imply a case (a) situation (Fig. 9) where expected ¯uoride values are higher than those de®ned with Eq. (1). The adjustment of water quality to abstraction rate has not yet been applied in the SLP basin, but experience in other areas of similar hydrogeology indicates its feasibility. In the Las Avenidas subbasin (located < 90 km north of Mexico City), standard step drawdown test observations included the continuous monitoring of electrical conductivity and temperature (borehole 6-bis: 300 m depth in fractured volcanic units) and water analyses on selected samples were carried out (Carrillo-Rivera and Cardona, 1998). The results suggested a relation between borehole yield (32.0±70.0 l s 21) and both groundwater salinity (557 mmho s cm 21 ± 2,030 mS cm 21) and temperature (29±39 8C). The desired water quality and salinity (700 mS cm 21)

were obtained using a groundwater abstraction of 40.0 l s 21 (and 30 8C). The reliability of the system of ¯uoride management by mixing procedures over extended periods of time depends on the availability of regional and intermediate ¯ows in suf®cient quantity and of similar quality to those used to derive Eq. (3). Groundwater abstraction has increased about ®ve-fold from 1960 to 1987 in which time the thermal water component has increased. In 1987 the thermal water represented about 70% of the total abstraction rate. The lack of continuous ¯uoride and temperature data prevents a complete ¯uoride concentration evolution with abstraction time being de®ned. However, the water quality of the regional ¯ow in SLP is a blend of rather uniform concentrations of, for example, chloride and sodium, and these values are very different from those of the intermediate system. (see Table 1); their concentrations for 1962 and 1987 have remained effectively constant, regardless of a groundwater abstraction increase from 0.55 to 2.6 m 3 s 21. The large volumes and homogeneity of the fractured unit (Sierra Madre Occidental), and the relative large residence times as proposed by lithium concentrations in the regional ¯ow system suggest that current stress on the regional ¯ow system is expected to have unimportant alteration in terms of its water quality. Any future changes in water quality, including those that could result in groundwater input from the shallow aquifer appear to bear little effect in the proposed ¯uoride management. Vertical in¯ow through the compactsand layer would provide water with a low ¯uoride content ( < 0.4 mg l 21, Table 1) and a high calcium concentration thus may mean that a new regression equation would need to be obtained. The mean value concentrations for chemical components of the intermediate ¯ow system show no evidence of signi®cant change. Current groundwater usage as applied since the middle of the XX century has posed little stress on the obtained inorganic water quality. Actions to avoid organic water quality deterioration from in¯ows across the compactsand layer may be needed if the control on ¯uoride by means of water chemistry through mixing is to be continued. A different mixing procedure could be to tap intermediate and regional water and carry out the blending at ground surface. However, this option needs a long main system to deliver water where at

J.J. Carrillo-Rivera et al. / Journal of Hydrology 261 (2002) 24±47

present is supplied locally by boreholes. The largest input to water supply is that of regional ¯ow (about 70% in volume) as compared to that of the intermediate ¯ow, and the latter is not in similar quantities as that of regional ¯ow to provide with expected needs. Borehole construction and lithology may affect the thermal ¯ow (Fig. 8b) by inducing it through the TGU before it reaches the abstraction zone. Thus water temperature decreases and dissolves suf®cient calcium to reach calcite equilibrium. Calculations indicate a 30% reduction in the thermal water ¯uoride content (if the original SI¯uorite persists) by this process, allowing lesser amounts of cold water be needed to reach the desired ¯uoride concentration in the mixed water. This permits a long term sustainable management of the intermediate ¯ow system. A possible increase in temperature of the TGU material due to the induced ¯ow of thermal water through its matrix is anticipated to be minor as Case (b) boreholes (Fig. 9), have lower water temperature than expected from their lithium contents. The thermal water abstraction through the TGU in these boreholes exhibits a temperature rise of about 3 8C related to the cold water. 5. Conclusions Chemical analyses of thermal waters between 1962 and 1987 show a remarkable similarity in major ion composition; however, there are no ¯uoride analyses available for 1962. Analyses for 1987 gave ¯uoride concentrations which have a linear relationship to groundwater temperature. This correlation permits an estimate to be made of ¯uoride content in abstracted groundwater from 1962. Current groundwater abstraction has increased from 0.550 m 3 s 21, in 1962, to < 2.60 m 3 s 21 for 1987, inducing vertical ¯uoride-rich ¯ows into the centre of the SLP basin. This now affects boreholes located over an area estimated to be ten times larger than in 1962 (Figs. 3 and 5). Saturation indices suggest that ¯uoride concentrations in the regional waters are controlled by ¯uorite solubility irrespective of abstraction groundwater temperature. Geothermometry calculations suggest that the deeper water has a temperature of about 75 8C At this depth the waters are undersaturated with respect to ¯uorite. This is interpreted as an effect

45

of ¯uoride loss, which occurs during the ascent of thermal water to borehole discharge. A natural ¯uoride concentration control may be postulated by increasing calcium and lowering water temperature, which is feasible as the thermal waters are calcite under-saturated; calculations indicate that this mechanism is viable. Since ¯uorite over-saturation is not anticipated, then application of this solubility control is recommended for natural ¯uoride reduction in future groundwater abstraction schemes. New boreholes sites are to be located in areas where TGU material is calcium-rich, and the design of the boreholes will be as indicated in Fig. 8b. Appropriate mixing between thermal and cold waters keeps ¯uoride within acceptable drinking water limits. This is achieved by maintaining borehole-head temperature between 28 and 30 8C. Some existing boreholes currently abstract appropriatelymixed low ¯uoride water. New boreholes should be located, designed, constructed and operated to withdraw thermal waters from the fractured volcanic units via the TGU. Fluoride control could be achieved through the appropriate mixture of cold and thermal ¯ows, and cases have been identi®ed where borehole lithology and its construction design may also control the resulting mixture. A well-head test should be devised which includes the de®nition of both the particular temperature-¯uoride relation and ¯uoride solubility controls. Any efforts to diminish ¯uoride concentrations in pumped groundwater in SLP, taking advantage of the hydrogeological and geochemical controls could prove to be bene®cial. In MeÂxico, at least some 15% of the total population estimated at 100 M, are supplied with thermal ¯uoride rich-water in the semi-arid eastern portion of the Sierra Madre Occidental and therefore the proposed model and remedies could have wider applicability in similar hydrogeological in Central and South America. The application of the proposed mechanisms of ¯uoride attenuation by the hydrogeological environment may be optimal to conventional treatment plants, which yield higher costs and the need for an additional by-product management. Acknowledgements The authors would like to thank The British

46

J.J. Carrillo-Rivera et al. / Journal of Hydrology 261 (2002) 24±47

Council, the National Science and Technology Council of MeÂxico, the Sistema de InvestigacioÂn Miguel Hidalgo, ComisioÂn Nacional del Agua (DireccioÂn TeÂcnica) and INTERAPAS for their partial ®nancial support to perform hydrogeological studies in the San Luis PotosõÂ area. Constructive comments by reviewer, John Tellam, signi®cantly improved the original manuscript. This paper is published with the permission of the Director, British Geological Survey (NERC).

References APHA, AWWA, WPCF, 1989, 1989. Standard Methods for the Examination of Water and Wastewater. 17 edAPHA, AWWA, WPCF, Washington. Appelo, C.A.J., Postma, D., 1996. Geochemistry, Groundwater and Pollution. A.A. Balkema, Rotterdam 536pp.. Back, W., Rosenshein, J.S., Seaber, P.R., 1988. Hydrogeology. The Geology of North America. Geol. Soc. Am. 2, 524. Ball, J.W., Nordstrom, D.K., Zachmann, D.W., 1987. WATEQ4FÐa personal computer translation of geochemical model WATEQ2, with revised database. USGS open ®le report, pp. 87±50. Cardona, B.A., 1990. CaracterizacioÂn ®sico-quõÂmica y origen de los soÂlidos disueltos en el agua subterraÂnea en el valle de San Luis PotosõÂ: su relacioÂn con el sistema de ¯ujo. MSc Thesis, Facultad de IngenierõÂa Civil, Universidad A. de N.L., MeÂxico, 97pp. Cardona, B.A., Carrillo-Rivera, J.J., 1996. Mecanismos naturales de control de ¯uoruro en sistemas regionales de ¯ujo: Sierra Madre Occidental, MeÂxico. 3er Congreso, AsociacioÂn Lainoamericana de HidrologõÂa SubterraÂnea para el Desarrollo. San Luis PotosõÂ, MeÂxico. Cardona, B.A., Carrillo-Rivera, J.J., Armienta, M.A., 1993. ContaminacioÂn y valores de fondo en aguas subterraÂneas de San Luis PotosõÂ. GeofõÂsica Internacional 32-2, 277±286. Carrillo-Rivera, J.J., 1992. The hydrogeology of the San Luis PotosõÂ Area, MeÂxico. MeÂxico. PhD Thesis, University of London, U.K., 203pp. Carrillo-Rivera, J.J., 2000. Application of the groundwater-balance equation to indicate interbasin and vertical ¯ow in two semi-arid drainage basins, MeÂxico. Hydrogeol. J. 8-5, 503±520. Carrillo-Rivera, J.J., Armienta, M.A., 1989. DiferenciacioÂn de la contaminacioÂn inorgaÂnica de las aguas subterraÂneas del valle de la ciudad de San Luis PotosõÂ, MeÂxico. GeofõÂsica Internacional 28-4, 763±783. Carrillo-Rivera, J.J., Cardona, A., 1998. Flujo vertical en la evolucioÂn del agua subterraÂnea extraõÂda: Subcuenca del rõÂo de las avenidas, Hgo, MeÂxico. Proceedings 48. Congreso Latinoamericano de HidrologõÂa SubterraÂnea. ALHSUD. Montevideo, vol. 3, pp. 1323±1342. Carrillo-Rivera, J.J., Clark, D.I., Fritz, P., 1992. Investigating recharge of shallow and paleo-groundwaters in the Villa de

Reyes basin, SLP, MeÂxico with environmental isotopes. Appl. Hydrogeol. 1±4, 35±48. Carrillo-Rivera, J.J., Cardona, A., Moss, D., 1996. Importance of the vertical component of groundwater ¯ow: a hydrochemical approach in the valley of San Luis PotosõÂ, MeÂxico. J. Hydrol. 185, 23±44. De Marsily, G., 1986. Quantitative Hydrogeology for Engineers. Academic Press, New York, p. 440. Deutsch, W.J., Jenne, E.A., Krupta, K.M., 1982. Solubility equilibria in basalt aquifers: The Columbia Plateau, Eastern Washington, US. Chem. Geol. 36, 15±34. Diesendorf, M., 1986. The mystery of declining tooth decay. Nature 322, 125±129. Dissanayake, C.B., 1991. The ¯uoride problem in the groundwater of Sri LankaÐenvironmental management and health. Int. J. Environ. Studies 38, 137±156. Edmunds, W.M., Smedley, P.L., 1996. Groundwater chemistry and health-an overview. Environ. Geochem. Health, Appleton, J.D., Fuge, R., McCall, G.J.H. (Eds.). Geol. Soc. Special Pub. 113, 91±105. Edmunds, W.M., Andrews, J.N., Burgess, W.G., Kay, R.L.F., Lee, D.J., 1984. The evolution of saline and thermal groundwaters in the Carnmenellis granite. Miner. Mag. 48, 407±424. Gaciri, S.J., Davies, T.C., 1993. The occurrence and geochemistry of ¯uoride in some natural waters of Kenya. J. Hydrol. 143, 395±412. Geoingenieria Internacional, 1996. Estudio hidrogeoloÂgico de la contaminacioÂn del agua subterraÂnea en la zona industral de SLP. ComisioÂn Nacional del Agua, contrato GAS-013-96. Grimaldo, M., Borja-Aburto, V.H., RamõÂrez, A.L., Ponce, M., Rosas, M., Diaz-Barriga, F., 1995. Endemic ¯uorosis in San Luis PotosõÂ, MeÂxico. Environ. Res. 68, 25±30. Handa, B.K., 1975aa. Geochemistry and genesis of ¯uoridecontaining groundwaters in India. Ground Water 13, 257±281. Handa, B.K., 1975bb. Geochemistry and genesis of ¯uoridecontaining ground waters in India. Ground Water 13, 275±281. Hantush, M.S., 1956. Analysis of data from pumping test in leaky aquifers. Trans. Am. Geophys. Un. 37, 702±714. Hem, J.D., 1970. Study and interpretation of the chemical characteristics of natural groundwater. Second Edition. USGS Water Supply Paper, 1473, 363pp. Herrera, I., Medina, R., Chargoy, L., Carrillo-Rivera, J.J., 1992. Evaluation of hydrothermal sources that sustain an overexploited aquifer at San Luis PotosõÂ, MeÂxico, IAH. Selected Papers on Aquifer Over-Exploitation 3, 47±60. Hitchon, B., 1995. Fluoride in formation waters. Alberta Basin, Canada. Appl. Geochem. 10, 357±367. Hvorslev, M.J., 1951. Time Lag and Soil Permeability in Groundwater Observations. US Army Corps Engrs. Waterways Exp. Sta. Bull., vol.36. Vicksburg, Mississippi. Instituto de GeografõÂa, 1999. RecopilacioÂn, integracioÂn y ordenamiento de informacioÂn relativa a los sistemas de ¯ujo de agua subterraÂnea en la cuenca de San Luis PotosõÂ y su relacioÂn con la Sierra Madre Occidental. Convenio SGTGAS-001/PRO99. IIZD, 1962. GeohidrologõÂa de ocho pozos de la SecretarõÂa de Recursos HidraÂulicos perforados en la Ciudad de San Luis PotosõÂ.

J.J. Carrillo-Rivera et al. / Journal of Hydrology 261 (2002) 24±47 Instituto de InvestigacioÂn de Zonas DeseÂrticas, Universidad ASLP, 12 pp. Jenne, E.A., Ball, J.W., Vivit, D.V., Barks, J.H., 1980. Geochemical modelling apparent solubility controls on Ba, Zn, Cd, Pb, and F in waters of the Missouri Tri-State mining area. In: Hemphill, D.D. (Ed.). Trace Substances in Environ Health XIV. University of Missouri, Columbia, pp. 353±361. Kawecki, M.W., 1995. Correction for Temperature Effect in the Recovery of a Pumped Well. Ground Water 33 (6), 917±926. Kilham, P., Hecky, R.E., 1973. Fluoride geochemical and ecological signi®cance in East African waters and sediments. Limnol. Oceanogr. 16, 932±945. Leroy, C., Hulin, J.P., Lenormand, R., 1992. Tracer dispersion in strati®ed porous media: in¯uence of transverse dispersion and gravity. J. Contamin. Hydrol. 11-1/2, 51±68. McDonald, M.G., Harbaugh, A.W., 1988. A modular three-dimensional ®nite difference ground water ¯ow model. USGS Techniques of Water Resources Investigations Report, Book 6, Chapter A1. MedellõÂn-MilaÂn, P., Alfaro-De la Torre, M.C., De Lira-SantillaÂn, A.G., Nieto-Ahumada, B., Sarabia-MeleÂndez, I., 1993. Fluoride in drinking water its correlation with parameters of the aquifer and effect on dental health in the City of San Luis PotosõÂ, MeÂxico. Proceedings of Water Quality Technology Conference, American Water Works Association 2, pp. 1011±1024. Nanyaro, J.T., Aswathanarayana, U., Mungore, J.S., Lahermo, P.W., 1984. A geochemical model for the abnormal ¯uoride concentrations in waters in parts of northern Tanzania. J. Afr. Earth Sci. 2, 129±140. Nordstrom, D.K., Jenne, E.A., 1977. Fluoride solubility in selected geothermal waters. Geochim. Cosmochim. Acta 41, 175±188. Nordstrom, D.K., Ball, J.W., Donahoe, R.J., Whittemore, D., 1989. Groundwater chemistry and water±rock interactions at Stripa. Geochim. Cosmochim. Acta 53, 1727±1740. Ovalle, T.J., 1996. Desarrollo de un medio adsorbente a base de carboÂn aluminizado para la remocioÂn de ¯uoruro presente en soluciones acuosas. MSc Thesis. Faculty of Chemical Eng. Universidad ASLP, MeÂxico. 96pp. Parkhurst, D.L., 1995. phreeqcÐa computer program for speciation, reaction path, advective-transport and inverse geochemical calculations. USGS Water-Res. Inv. Rep. 95±4227, 143 pp. Parkhurst, D.L., Thorstenson, D.C., Plummer, N.L., 1980. phreeqcÐa computer program for geochemical calculations. USGS Water-Res. Inv. 96, 210 pp. È zeguÈr, N., Schneider, H.J., 1992. Hydrochemistry of Pekdeger, A. O ¯uoride in shallow aqueous systems of the GoÈlcuÈk area, SW Turkey. In: Kharaka, Y.K., Maest, A.S. (Eds.). Proceedings 7th International Conference on Water±Rock Interaction, Utah, pp. 821±824.

47

Rajagopal, R., Tobin, G., 1991. Fluoride in drinking water: a survey of expert opinion. Environ. Geochem. Health 13, 3±13. Rathod, K.S., Rushton, K.R., 1991. Interpretation of pumping from two-zone layered aquifers using a numerical model. Ground Water 29-4, 499±509. Robertson, F.N., 1986. Occurrence and solubility controls of trace elements in groundwater in alluvial basins. Regional aquifer systems in the US. Am. Water Res. Assoc., 69±80. Sarabia, M.F.I., 1989. Contenido de ¯uoruros en el agua de consumo y sus efectos en el tejido dental, San Luis PotosõÂ, MeÂxico. MSc Thesis. Universidad A. de S.L.P., 187pp. Stretta, E.J.P., Del Arenal, R., 1960. Estudio para el abastecimiento de agua potable para la Ciudad de San Luis PotosõÂ. Applied Sciences Institute, Hydrology Section, UNESCO and Instituto de GeofõÂsica, UNAM Int. Rep., Jan. 40 pp. Tejook, T.K., 1993. De¯uoridation of water supplies. Water Lines 2, 26±27. Tennakone, K., Wickramanayake, S., Fernando, C., 1988. Aluminium contamination from ¯uoride assisted dissolution of metallic aluminium. Environ. Pollut. 2, 133±144. Teotia, S.D.S., Teotia, M., Singh, R.K., 1981. Hydrogeochemical aspects of endemic skeletal ¯uorosis in IndiaÐan epidemiological study. Fluoride 4, 69±74. Travi, Y., Faye, A., 1992. Fluoride in Paleocene aquifers in Senegal. An example of the contamination of a con®ned aquifer by its roof zone, aggravated by intensive exploitation. IAH Selected Papers on Aquifer Over-Exploitation 3, 145±154. TristaÂn, G.M.,1986. EstratigrafõÂa y tectoÂnica del graben de Villa de Reyes, en los estados de San Luis PotosõÂ y Guanajuato, MeÂxico. Inst. Geol., Universidad A. de S.L.P., Foll. Tec., 107pp. University of Waterloo, 1983. Notes of ®eld methods in contaminant hydrogeology. Short Course. Groundwater Research Institute, Department of Earth Sciences, Waterloo, Ontario, Canada, pp. 1±68. Wallick E.I., ToÂth, J., 1976. Methods of regional groundwater ¯ow analysis with suggestions for the use of environmental isotopes. Interpretation of Environmental Isotope and Hydrochemical Data in Groundwater Hydrology. Proceedings of Advanced Group Meeting, Vienna, pp. 37±64. Zhaoli, S., Mi, Z., Minggao, T., 1989. The characteristics of ¯uoride in groundwater of north China and the signi®cance of ¯uoritewater interaction to ¯uorine transportion. Proceedings of 6th International Symposium on Water Rock Interaction. Malvern, UK Balkema, pp. 801±804. Zielinsky, R.A., 1983. Experimental leaching of volcanic glass: implications for evaluation of glassy volcanic rocks as sources of uranium. Uranium in Volcanic and Volcano-Clastic Rocks, Goodell, P.C., Waters, A.C. (Eds.). AAPG Stud. Geol. 13, 1±11.