Dissolved uranium, radium and radon evolution in the Continental Intercalaire aquifer, Algeria and Tunisia

Natural, dissolved 238 U-series radionuclides (U, 226 Ra, 222 Rn) and activity ratios (A.R.s: 234 U/ 238 U; 228 Ra/ 226 Ra) in Continental Intercalaire (CI) groundwaters and limited samples from the overlying Complexe Terminal (CT) aquifers of Algeria and Tunisia are discussed alongside core measurements for U/Th (and K) in the contexts of radiological water quality, geochemical controls in the aquifer, and water residence times. A redox barrier is characterised downgradient in the Algerian CI for which a trend of increasing 234 U/ 238 U A.R.s with decreasing U-contents due to recoil-dominated 234 U solution under reducing conditions allows residence time modelling ~500 ka for the highest enhanced A.R. ¼ 3.17. Geochemical modelling therefore identi ﬁ es waters towards the centre of the Grand Erg Oriental basin as palaeowaters in line with reported 14 C and 36 Cl ages. A similar 234 U/ 238 U trend is evidenced in a few of the Tunisian CI waters. The paleoage status of these waters is af ﬁ rmed by both noble gas recharge temperatures and simple modelling of dissolved, radiogenic 4 He-contents both for sampled Algerian and Tunisian CI and CT waters. For the regions studied these waters therefore should be regarded as “ fossil ” waters and treated effectively as a non-renewable resource.


Introduction
Low-level activity, naturally-occurring uranium-and thoriumseries radionuclides and their isotopes in groundwaters can give insight into reductioneoxidation (redox) and geochemical controls, watererock interactions, aquifer mixing, and subsurface residence times of sampled waters in aquifer systems (Andrews, 1991;Bonotto, 2004;Porcelli, 2008). They also have significance for health in terms of their alpha (a-)radioactivity, and many countries have adopted regulatory standards for water use (cf. Chau et al., 2011).
In semi-arid and arid zones in Algeria, groundwater is the principal source of drinking water; and in Tunisia in 2010 the Ministry of Agriculture developed a strategic survey for sustainable water usage by 2050 which could include the use of non-traditional sources of water like desalination of seawater or salty groundwater.
The Continental Intercalaire (CI) formation in North Africa hosts an extensive, regional, internally-drained (endorheic), sedimentary aquifer which underlies Algeria, Tunisia, and Libya. Castany (1981) originally emphasised the deep-basin nature of this aquifer system of the northern Sahara such that "development and management of water stored in aquifer … … [is] 'groundwater mining'". Puri et al. (2006), Mamou et al. (2006) and Edmunds (2012) have asserted that in the Saharan basins the water resource often can be shown to be 'fossil' or palaeowater, almost entirely non-renewable in terms of both their water resources management and in International Law Eckstein, 2003, 2005). Semi-arid/arid regions like Libya and Algeria are heavily dependent on groundwater as their only water resource; coordinated management of this transboundary, shared water resource led to the creation of a Consultation Mechanism Unit for the North Sahara Aquifer System e SASS (Syst eme Aquifere du Sahara Septentrional) in July 1999 (UNESCO, 2010).
Previous attempts to date the geochemical residence times of the Algerian CI waters using 14 C (t 1/2 ¼ 5730 a: Gonfiantini et al., 1974;Sonntag et al., 1978;Guendouz, 1985;Elliot, 1990;Guendouz et al., 1997;Edmunds et al., 2003) identified that, apart from aquifer margins, the sampled CI waters all have low radiocarbon activities (<5 per cent modern carbon, pmc). Towards the centre of the basin, waters are very close to the limit of age discrimination by the radiocarbon method (~25e30 ka). This along with past climatic signatures archived in the waters through their stable isotopes of water signatures (d 2 H, d 18 O) and dissolved noble gas contents (Elliot, 1990;Guendouz et al., 1997) appear to confirm their palaeowater status. Moreover, Guendouz and Michelot (2006) report 36 Cl (t 1/2 ¼ 3 Â 10 5 a) dates for CI waters suggesting minimum model groundwater ages of 0e134 ka and maximum ages 49e223 ka for relevant samples on the M'Zab ridge (Berriane, Metlili) and ages > 100 ka (Zelfana, El-Hadjira), although initial 36 Cl/Cl data are being revised (Petersen et al., 2014a, in press).
Recent literature however has queried the significance and presumption of the paleowater/fossil water status (and by implication a stagnant/null recharge or disconnected flow system) particularly for the North-Western Sahara Aquifer System. Al-Gamal (2011) invokes stratification and regional mixing of modern and palaeowaters generally in the system on the basis of moderately-depleted d 2 H, d 18 O signatures seen in recharge zones, and states particularly that tritium ( 3 H; t 1/2 ¼ 12.32 a) is widespread e although few data are presented and even the presence of 14 C signatures > 2 pmc would translate to significant geochemical residence times (Annexe 8, OSS, 2003). From piezometric modelling of the CI aquifer Ould Baba Sy (2005), suggests that a null recharge presumption for the Tademait and Tinrhert plateaux of southern Algeria is reasonable, however he queries null recharge in the Algerian Saharan Atlas and the Dahar Hills (Tunisia) and also the Algerian M'zab (at least for the CT aquifer). Gonçalv es et al.
(2013) deploy a regional water balance approach to assess natural recharge, although since all aquifers are distributed flow systems water recharging at any location then is flowing somewhere specific and for water resources management and sustainability of any aquifer equating safe aquifer yield to its natural recharge can be problematic (e.g. Elliot et al., 1998Elliot et al., , 2001Sophocleous, 1997).
Radioactive 3 H, 14 C and 36 Cl dating methods are all based on a 'decay clock' of atmospheric inputs at recharge (albeit with various correction mechanisms for mixing sources and dilution within an aquifer). Natural 238 U (t 1/2 ¼ 4.5 Â 10 9 a)-and 232 Th (t 1/ 2 ¼ 1.405 Â 10 10 a) decay is internal to the system, which can provide an 'accumulation clock' for their products e including other U-and Th-isotopes, Ra, Rn (all radioactive) and also (stable) dissolved 4 He (since the U-and Th-series decay mechanism is predominantly by a-emission). Whilst natural dissolved U distributions and 234 U/ 238 U activity ratios (A.R.s) in Continental Intercalaire (CI) waters have been reported (Edmunds et al., 2003;Chkir et al., 2009), the radiogenic and radioactive daughters 226 Ra and 222 Rn have not been reported previously. Moreover, 234 U/ 238 U disequilibria modelling for these deep basin waters in terms of groundwater dating and as a comparative check for consistency of groundwater ages by other methods in the Algerian aquifer has not been attempted to date. Preliminary results for CI waters in the Tunisian aquifer by Petersen et al. (2013) and Fr€ ohlich (2013) suggest residence times~500 ka based on an apparent decreasing trend of 234 U/ 238 U activity ratios ( 234 U t 1/2 ¼ 244.5 Â 10 3 a) with increasing U-contents (see also Bonotto, 2006). In the current study, the evolution of U, 234 U/ 238 U A.R., 226 Ra (including 228 Ra/ 226 Ra A.R.), and 222 Rn systematics particularly along a flow line in the Algerian CI are discussed in terms of 238 U-series systematics and groundwater dating (including also 4 He ages).

Study area
The Continental Intercalaire (CI) aquifer underlies continuously an area~600,000 km 2 in Algeria and Tunisia (Castany, 1982) and 100,000 km 2 overall (including Libya). In Algeria, the M'Zab Ridge running NeS (Fig. 1) structurally provides a watershed divide between (to the west) the Grand Erg Occidental and (to the east) the Grand Erg Oriental hydrogeological basins. The shallower, overlying Complexe Terminal (CT) aquifer covers~35,000 km 2 . In the Grand Erg Occidental the two aquifers are hydraulically connected (cf. Moulla et al., 2012), whereas in the Grand Erg Oriental the two aquifers are separated by semi-permeable/impermeable layers and confined, artesian conditions exist for the CI aquifer towards the centre of this basin.
The CI formation comprises permeable continental detrital deposits of sand-sandstone and argillaceous sands with intercalations of marine clays and arenaceous clays of Lower Cretaceous (Albian) age (Furon, 1963). Except at its borders and in the western and Djeffara sub-basins, the CI aquifer is confined over the major part by a series of Upper Cretaceous (predominantly Cenomanian) transgressive clays with evaporites. Underlying the whole of the central region from Hassi Messaoud to the great salt-lake Chotts in the N is the confining basal Upper Jurassic Malm. The CT aquifer groups under the same name several very heterogeneous formations: permeable beds of (Upper Cretaceous) Senonian limestones, with Turonian dolomites on the borders (Dahar, M'Zab), and (Tertiary) Mio-Pliocene sands (the CT proper). Guendouz (1985) includes within this CT unit the phreatic aquifer system of the Quaternary aeolian dunes.
The recharge area for the CI aquifer in the Algerian study area is in the Atlas Mountains~400 km to the NW (Gonçalv es et al., 2013; Fig. 1). The aquifer is hydraulically continuous from here to the Chotts of Tunisia where it discharges. Groundwater samples for radioelements and their isotopes have been taken ( Fig. 1) from 12 wells in the Eastern Great Erg (Grand Erg Oriental) basin of Algeria, samples A1eA10 follow a NWeSE radial flow direction identified originally by Guendouz (1985, his Fig. 3) from piezometric data for the CI aquifer and confirmed from the latest piezometric map (OSS, 2003, Planches 10e13). Two samples (A12, A13) were taken from the overlying, shallower CT aquifer, and a further CI sample from a new well (A14) to the North.
A dozen groundwater samples from Tunisia are also reported here: 7 in the CI proper, 2 in the CT, and 3 associated with both aquifers (CI/CT). Major flow directions in the extensive CI aquifer system ( Fig. 1) appear broadly to converge on the major Chotts west of Gab es in Tunisia: WeE from the Saharan Atlas to Chott Djerid and the Gulf of Gab es; SWeNE from the M'zab ridge region of southwestern Algeria and the Tademait Plateau and/or the Tinrhert Plateau (SeN) towards Chott Fedjej/Gulf of Gab es; SeN from local recharge in the Dahar uplands in southern Tunisia. The Tunisian CI samples presented here generally are located to the East of the horst structure identified around the region of Tozeur (T9; cf. Edmunds et al., 2003, their Fig. 13) and likely therefore the flow direction is oriented predominantly SeN from the Tinrhert plateau of southern Algeria or the Tunisian Dahar Hills as the possible recharge areas, following the piezometric contours (OSS, 2003). Sample numbering then is SeN for the Tunisian CI samples (T1eT7) proper and also other samples (T8eT10 are CI/CT; T11, T12 are CT).
Two borehole sites in Tunisia were also sampled at depth for aquifer solids (Chott Fedjej F1 in the CI; Negga N6 in the CT) giving possible representative Th/U/K rock data for these aquifers in the study region.

Material and methods
Borehole and field sampling details for the Algerian samples are given in Elliot (1990). Water samples for the determination of total dissolved U-content and 234 U/ 238 U activity ratio were collected in 60 L acid-washed polyethylene containers. On collection, samples were acidified in the field to pH < 2 and a tracer spike of 236 U (0.185 ± 0.002 Bq, representing a coverage factor k ¼ 2) added to quantify the chemical yield of U extraction, in addition to 900 mg of Fe 3þ carrier. In the laboratory, U was co-precipitated with Fe 3þ under alkaline conditions, then separated from Fe, Ca, Mg and other elements in the precipitate by solvent extraction and ion-exchange procedures (Andrews and Kay, 1983). The isotopes of uranium were co-precipitated on Fe(OH) 3 by increasing the pH to 7e8 through addition of concentrated NH 4 OH solution. The precipitate was recovered, dissolved in 9 mol/dm 3 HCl and Fe 3þ was extracted into an equal volume of methyl isobutyl ketone. The acid solution of uranium was further purified by anion exchange, first on a Cl À and then on a NO 3 À column of Biorad AG1-X8 100e200 mesh resin. U was finally eluted from the NO 3 À column with 0.1 mol/dm 3 HCl, evaporated to dryness, dissolved in 10-cm 3 2 mol/dm 3 (NH 4 ) 2 SO 4 solution, and transferred to a teflon electrolysis cell. Electrodeposition of U on a stainless steel planchet was complete after 3 h at a current density of 1 A/cm 2 . The a activities were determined with 100 mm depletion depth, 450 mm 2 area Passivated Implanted Planar Silicon (PIPS) detectors, whose typical backgrounds in the 238 U, 236 U and 234 U energy regions were (0.001 ± 0.0002), (0.0009 ± 0.0002) and (0.0028 ± 0.0003) cpm (counts per minute) respectively. The spectra for natural U and 236 U tracer extracted were recorded on a Canberra, 2048-channel, multi-channel analyser, where the concentration data were calculated from the counting rates of 238 U and 236 U peaks and the 234 U/ 238 U activity ratio data were calculated from the counting rates of 238 U and 234 U peaks. Samples for dissolved radiogenic 222 Rn determination were collected in gas-tight throughflow bottles with sealable inlet and outlet tubes. Samples were generally analysed within two weeks of collection. The 222 Rn was outgassed into a scintillation flask using a N 2 stream and its activity determined by a-scintillation counting.
The counting efficiency of each scintillation flask was determined using standard 226 Ra solutions that allowed estimation of an average detection limit of 98 mBq/L. Samples for Ra determinations were filtered (0.45 mm) into a 5 L acid-washed container, and acidified. Recovered samples for 226 Ra contents were transferred to glass de-emanation bottles and 222 Rn outgassed using N 2 . The bottles were sealed and left one month for ingrowth of 222 Rn to its equilibrium activity with 226 Ra, subsequently determined as for dissolved Rn samples. Samples for Ra activity ratios were recovered using the MnO 2 powder scavenging method, subsequently leached with 2 M HNO 3 and co-precipitated as Ba(Ra)SO 4 . Radium isotope activities were then determined by g-spectrometry with Ge(Li) detector (cf. Michel et al., 1981).
For a-spectrometric determinations on core samples representative 1 g splits from bulk samples crushed to <10 mm were totally dissolved by acid treatment (HF, HClO 4 ) following the addition of tracer spikes of 236 U (0.153 ± 0.003 Bq, representing a coverage factor k ¼ 2) and 229 Th (0.201 ± 0.004 Bq, representing a coverage factor k ¼ 2). Following separation by anion exchange, U and Th were electrodeposited onto stainless steel discs. For g-spectrometric determinations samples were crushed to <1 mm and 200 g weighed in to a tin, sealed and stored for 20 days to establish secular equilibration with 222 Rn, 214 Pb, 214 Bi. The sample then was counted on a 6 00 (diameter) Â 4 00 NaI (Tl) crystal using a multichannel analyser.
Geochemical speciation modelling including mineral saturation indices (SI) was performed using the original WATEQ4F code (Ball et al., 1987) incorporating the uranium thermodynamic database of Ball et al. (1981).

Results and discussion
For the Algerian CI aquifer samples, borehole screened intervals for samples are 200e500 m (M'Zab Ridge: A1eA3) with water sampling temperatures (ST) in the range 30.7e31.5 C. Away from the M'Zab Ridge sampling depths are 800e1600 m with ST up to 57.7 C, demonstrating the depth (and geothermal potential) of these waters. For CT aquifer samples, sampling depths are 145 m (A12), but 1000 m at A13 with sampling temperatures 24.4 and 47.1 C, respectively. Dissolved O 2 (DO) values were 3e8 ppm on and close to the M'Zab Ridge (A1eA5) in the CI aquifer and showing oxidising conditions (Eh~þ79 to þ287 mV), and 4.5 ppm in the CT aquifer (A12) even though the recharge area to the NW (Atlas Mountains) is distant. All other CI samples were anoxic (DO below detection) and reducing (Eh À39 to À177 mV). There is therefore a downgradient redox zonation identified along the flow direction ( Fig. 2). High DO (and NO 3 ) for the M'Zab waters confirms this region as a possible recharge direction, but these contents do not necessarily imply recent recharge and a modern water component; a 3 H level of 0.2 ± 1.0 TU (N.B. the TU represents one molecule of 3 H 1 HO in 10 18 molecules of (stable) 1 H 2 O, with 1TU activity approximately equivalent to 0.118 Bq/kg e Stonestrom et al., 2013; this sample equivalent therefore to 0.02 ± 0.12 Bq/l) at Ghardaia (Guendouz, 1985) confirmed that water components at this site on the M'Zab Ridge have residence times > 30 a. Persistent DO has been observed also in deep waters up to 10 ka old, and up to 80 km from their point of recharge (Edmunds et al., 1982;Winograd and Robertson, 1982) possibly reflecting low levels of reducing agents such as organic carbon in the aquifer.
Uranium, Radium and Radon contents for both Algerian and Tunisian groundwater samples are given in Table 1, and whole-rock U/Th/K contents and U/Th activity ratios (A.R.s) in samples from core material for the (Tunisian) CI and CT aquifers in Table 2.

Uranium and 234 U/ 238 U activity ratios
Bulk sample 234 U/ 238 U and also 230 Th/ 238 U A.R.s for both the CI and CT cores samples (Table 2) are (within errors)~1.0 confirming closed (rock) system conditions. Under closed system conditions, such as in rocks, 238 U-series radionuclides should come to secular radioequilibrium due to radioactive decay after 1.25 Ma such that their 234 U/ 238 U A.R. would be~1.0 within the bulk of the rock matrix (Andrews, 1991). Surficial (open system), lacustrine/palaeolake carbonate deposits show significant deviation from unity (Fontes et al., 1992;Causse et al., 2003); however, Causse et al. (2003) have shown clustering of 234 U/ 238 U values discriminating Chott Djerid and Chott Fedjej deposits in Tunisia which might reflect differing groundwater discharge components from the CI or CT aquifers.
The downgradient evolution of dissolved U and 234 U/ 238 U in the Algerian CI groundwaters is shown in Fig. 2. Oxidising waters show dissolved U-contents of a few ppb (mg/kg). There is also an apparent decrease in 234 U/ 238 U A.R. with distance and increasing dissolved U-content to the identified redox barrier. After the redox barrier, the reducing waters (excepting El-Hadjira, A6) exhibit Ucontents < 0.5 ppb, but an A.R. increase then in the reducing waters along the flow direction. Clearly the dominant factor controlling U in solution is the redox character of the waters.
All Algerian samples showed pH within the range 6.79e7.63, with values averaging 7.2 for the oxidising waters but trending in the reducing waters towards the lowest sampled pH value occurring at Gassi Touil (cf. Elliot, 1990). Uranium speciation modelling (Elliot, 1990)    level of only 8 ppb dissolved silica may be needed for mineral stability between USiO 4 and UO 2 (c), but a threshold of 60 ppm silica has been suggested on the basis of the common occurrence of coexisting quartz and uraninite mineral phases. Unfortunately, geochemical data are not available on the U-mineralogy of the CI aquifer formation to confirm the presence of either UO 2 (c) or USiO 4 (c) as the controlling mineral phases for U(IV)-deposition in the reducing waters.
In terms of 234 U/ 238 U evolution, on a standard plot ( Fig. 3; Cowart and Osmond, 1980, their Fig. 4) oxidising waters (samples A1eA5) plot in Zone III (Forming accumulation) with a trend towards decreasing 234 U/ 238 U A.R. with increasing U-content downgradient. Sample A6, identified from redox measurements as a reducing water, plots (Zone III) at its boundary suggesting active deposition. The reducing waters generally and characteristically plot in Zone IV (Normal Reduced), with increasing A.R. (samples A7eA10) for decreasing U-content. These trends are confirmed in a plot of 234 U/ 238 U versus 1/U (Fig. 4). Tunisian CI waters are also shown on Fig. 3, with samples T4, T6, T7 and also T2 plotting in Zone IV (Normal Reduced) and the first three samples suggesting a trend similarly to increasing A.R. for decreasing U-content as for the Algerian reducing waters (although their A.R. values do in fact lie within error of each). On a 234 U/ 238 U versus 234 U content plot (Fig. 5) the Algerian reducing water samples (A7eA10) follow an exponential trend of increasing A.R. with decreasing 234 U-content characteristic of deposition of U with a-recoil of 234 U from the surface of the solids into solution (or rather due to the a-recoil of its short-lived precursor 234 Th; cf. Andrews, 1991, his Fig. 15.4). The oxidising waters trend (cf. Fig. 2) suggests initially enhanced 234 U/ 238 U in the M'Zab waters then reflecting progressive U dissolution of the solids matrix back towards a bulk rock measured value of unity ( Table 2). The initially enhanced values of 234 U/ 238 U here typically reflect the preferential dissolution of 234 U from the solids matrix due to radiation damage of the solids matrix during the decay process and/or a-recoil of the sort-lived 234 Th parent (Kigoshi, 1971;Andrews and Wood, 1972). In the reducing zone, as  Figs. 2 and 4). Further downgradient of any active reducing zone (zone of precipitation) the injected 234 U sourced from the U precipitate becomes unsupported and decays.
For the Tunisian CI waters, only one 3 H measurement relevant to the sampled sites (Ksar Ghilane, 4 ± 3 TU; equivalent to around 0.5 ± 0.0.4 Bq/l, Stonestrom et al., 2013) is reported and cited radiocarbon activities again are generally low suggesting geochemical residence times again >20 ka (Annexe 8, OSS, 2003). Petersen et al. (2013) have suggested that in the Tunisian CI aquifer for U-contents <1 ppb the apparent trend is towards a decreasing 234 U/ 238 U A.R. following decreasing U contents and invoke an exponential decay relation (see Eq. (1b), Table 3) to explain this general trend and age date these waters. However, for the given samples here, Fig. 3

Radium and radon
Dissolved radium activities lie in the range 16e102 mBq/l for 226 Ra, higher than for bottled mineral waters but lower than the range seen for the Chott El Hodna groundwaters; a similar range is calculated for dissolved 228 Ra (11e134 mBq/l) in the deep Algerian CI waters. The two Algerian CT waters show similar activity levels to these CI waters. The Tunisian CI waters show enhanced Ra activities (71e567 mBq/l measured for 226 Ra; 106e595 mBq/l estimated for 228 Ra). The two Tunisian CT waters show just~20 mBq/l. In the oxidising Algerian CI waters their mean 226 Ra activity is similar to their mean 238 U activity, indicating chemical dissolution and weathering and in line with the trend to decreasing 238 U/ 234 U A.R. (Fig. 2). In the reducing waters, high measured dissolved 226 Ra contents occur with low U contents (similarly for the association of dissolved 228 Ra and U), simply reflecting the redox control on the solubility of the latter (Figs. 2 and 6). Nevertheless, the dissolved 228 Ra/ 226 Ra A.R.s within error generally cluster around the Th/U activity ratio of the aquifer core material (Fig. 7). Radium speciates predominantly as Ra 2þ in groundwaters (Langmuir and Riese, 1985), and as a divalent alkaline-earth element the residence time of 226 Ra in solution is likely also controlled by congruent/ incongruent solution of Ba and Ca as well as its half-life (t 1/ 2 ¼ 1620 a). The downgradient trend in the Algerian CI waters is to increasing dissolved Ra following also enhanced contents in both Ca and Ba (Elliot, 1990). Ca-and Ba-contents generally are higher also in the Tunisian CI waters than the Algerian CI waters (Edmunds et al., 2003).
Dissolved 222 Rn activities lie in the range 3e32 Bq/l for the Algerian CI waters, higher than Algerian mineral waters but of a similar range to Chott El Hodna samples. The Tunisian CI waters show a similar range (6e14 Bq/l), with the single CT sample giving an activity at the lower end of this range. The 222 Rn activities in the Algerian and Tunisian CI waters generally are~2e3 orders of  on the y-axis axis with decreasing 234 U content for the samples A7eA10 (and including A14) and also T4, T6, T7 and which follow deposition of U with a-recoil solution of 234 U (cf. Andrews, 1991, his Fig. 15.4, trend 1b). Error bars are 2s. magnitude greater than precursor dissolved 226 Ra activities (Figs. 6 and 7). For 222 Rn, its solution in intergranular pore fluids is predominantly controlled by a-recoil from the rock surface and diffusional processes (Andrews and Wood, 1972). With a short halflife (t 1/2 ¼ 3.825 d) equilibration is achieved in a groundwater residence time of just 25 d. The dissolved 222 Rn content, [Rn], in radioequilibrium with uranium in the solids matrix can be simply modelled as: where, 12.3 is the conversion factor for the specific degradation rate for 238 U, L Rn is the fractional release efficiency for radon from the rock matrix, r is the rock density (g/cm 3 ), [U] r the U contents of the rock surface (mg/g) and f is the fractional rock matrix porosity. Petersen et al. (in press) have highlighted that porosity measurements for the Continental Intercalaire are scarce, but suggest values between 22 and 26% in Algeria and 18% and 21% in Tunisia. For reasonable values then for a sandstone and for a coverage factor k ¼ 1 (Elliot, 1990: Table 2); f ¼ 0.2 ± 0.02) then using Eq.
(2) the calculated [Rn] ¼ 15 ± 3 (1s) Bq/l, generally in good agreement with observed activities in many of the CI waters (Fig. 7). Higher contents could reflect higher U contents or lower porosity. However, particularly the apparent "spike" in 222 Rn value at sample A7 (Fig. 6) considered alongside the downgradient profile for U (Fig. 2) might identify the position of a stable (Type 1) reducing barrier (Vogel et al., 1999, their Fig. 10); a stable barrier being one that has been established a long time compared to the half-life of 230 Th (t 1/2 ¼ 75.2 ka; the precusor to 226 Ra). Potentially calculation of the recoil supply rate of 222 Rn to groundwater also can be used to assess the recoil supply of other nuclides to the aquifer, which for short-lived radionuclides then can be used to assess sorption characteristics (sorption rate constants and retardation factors) for the aquifer (Krishnaswami et al., 1982). The 222 Rn activities in the Algerian CI waters generally are~2e3 orders of magnitude greater than precursor dissolved 226 Ra activities and similarly (up to two orders of magnitude) in the Tunisian CI waters, confirming that 226 Ra generally is not controlled simply by a-recoil, but likely by exchange at the rock surface.

Groundwater dating
For recoil-dominated environments 228 Ra/ 226 Ra A.R.s potentially can be up to the order of twice that of the parent radionuclides in 'old' groundwaters, such that higher 228 Ra/ 226 Ra A.R.s might be an indicator of groundwater maturation (Davidson and Dickson, 1986). The closeness of measured 228 Ra/ 226 Ra A.R.s to the rock Th/U production ratio suggests that the a-recoil coefficients for the parent nuclides are about equal, and that Th adsorption from weathering is minimal (Porcelli, 2008). 222 Rn/ 226 Ra A.R.s presuming a rock surface in which 230 Th, 226 Ra and 222 Rn are in radioequilibrium suggest groundwater residence times of only 2e20 a (Andrews, 1983;Elliot, 1990), but this estimate likely reflects other controls on 226 Ra than simply an a-recoil mechanism, as discussed previously.
Once a groundwater has become so reducing in character that chemical leaching of 234 U ceases then the 234 U/ 238 U A.R. may evolve in time as a balance between 234 U leaching (a-recoil) and its decay (Eq. (1a), Table 3). Table 3 shows estimated A.R. evolution with time (t) for the Algerian CI samples A7eA10 based on the trend seen in Figs. 3 and 5 and using Ouargla (A7) as starting 234 U/ 238 U value influent to this zone (given this is the initial low U sample encountered downgradient). Given infinite time, the dissolved 234 U/ 238 U of waters could evolve to a maximum value~3.4 under these conditions. An increase of A.R. from 1.76 to 3.17 potentially could take 500e750 ka (the generated A.R.s for these residence times bracketing the sample value seen at Rhourde El Baguel (A9); Table 3). For the flow distance Hassi Messaoud e Rhourde El Baguel/Gassi Touil the groundwater residence time~65 ka. These (geochemical) age estimates are relative residence times starting from the location sample A7 to that of sample A10 on top of the absolute age of the water at location sample A7. Nevertheless, such residence times clearly suggest a palaeoage status for the waters downgradient of Ouargla (A7).
For the Tunisian waters, using Menchia (T4) as the AR i ¼ 1.03 and mean U content for the Normal Reducing waters (Figs. 3 and 5), would suggest a potential maximum A.R. generated~5.3 (Eq. (1a), Table 3), and a residence time~20 ka then from the observed A.R. of 1.24 at Chott Fedjej (T7).
Further support for the characterisation of the water samples as being old, palaeowaters is provided both by dissolved 4 He-contents and recharge temperature (RT) estimates based on other dissolved noble gases (Ne, Ar, Kr, Xe; Andrews and Lee, 1979;Elliot, 1990). Along the given flow direction in the Algerian CI, 4 He-contents cumulatively increase (Fig. 8). Even without a detailed model for He-release (cf. Castro et al., 2000) this downgradient 4 He trend Table 3 Calculated 234 U/ 238 U activity ratios (AR t ) for 234 U ingrowth following 234 Th recoil into pore fluids in the reducing zone of the Algerian CI sandstone aquifer after the model of Andrews et al. (1982). Where the enhanced 234 U contents become unsupported then the right-hand term disappears (effectively, at the rock surface [U] r / 0) and a "decaying regime" occurs such that age of waters progressively downgradient can be dated as where A 0 is the A.R. upflow of the redox front (time t ¼ 0 for the "decaying regime") and A the downgradient A.R. sample (Osmond and Cowart, 2000). Thus the excess 234 U injected into the water then decays. This is thought to happen where injected 234 U is substantially reduced in deeper zones (i.e. effectively a decrease in leach rates because of progressive depletion of 234 U in mineral surfaces downgradient) with U deposition and consequent (unsupported) excess 234 U decay.
supports the idea of progressively more mature waters along this flow direction. A groundwater "age" can be calculated based simply on the production terms for the radioactive decay of U, Th and their radioactive a-emitting daughters in the rock matrix:  (3)) assumes all the 4 He generated is dissolved in the porosity, and as such these are conservative, minimum estimates of ages. An alternative formulation using 4 He/ 222 Rn ratios (Torgersen, 1980) would suggest residence times up to an order of magnitude greater for the youngest waters than those calculated simply by Eq. (3) (Elliot, 1990). Elliot (1990) has also modelled ages based on crustal diffusive loss model (Andrews, 1985) which suggests that the generating function potentially could be just 1e2% of the cumulative 4 He produced over the age of the Continental Intercalaire formation. Thus, although generally giving enhanced age estimates in absolute terms, these relative 4 He ages nevertheless clearly suggest  increasing groundwater ages downgradient for the Algerian CI waters; a general linear trend of increasing 4 He with Cl À is also evidenced for these waters (Elliot, 1990). Concomitantly, recharge temperatures (RTs) calculated from dissolved noble gas contents (Fig. 8) are !19 C for the M'Zab Ridge (samples A1eA5) but appear to decrease significantly in the central basin (~15 C at (A10) Gassi Touil). The three Tunisian CI waters sampled for dissolved noble gases (T1, T3, T6) suggest 4 He ages from around 660 ka up to 1.3 Ma, and show RTs around 18e19 C. Recent 3 He/ 4 He measurements in the Tunisian study area attest that high He contents here are crustal in origin but that the enhanced contents may be associated also with local complex geology (e.g. faulting) and tectonics (Fourr e et al., 2011).
The estimated noble gas RTs generally are somewhat cooler than current WMO Climate Normals (CLINO) data with climatological standards for 1961e1990 (the latest global standard normals period) for dry bulb annual average temperatures being 21.8 C (Biskra, Algeria), 21.7 (Tamanrasset, southern Algeria) and 19.5 C (Gab es, Tunisia) (WMO, 2014). Thus RTs generally suggest cooler waters than present climatic conditions, and alongside the given age estimates affirm the palaeowater status of both Algerian samples (especially towards the centre of the Grand Erg Oriental basin) and some CI Tunisian waters.

Conclusions
Dissolved Rn contents in the deep Continental Intercalaire (CI) aquifer of Algeria and Tunisia show activities 3e32 Bq/l, slightly higher than activities shown in Algerian bottled mineral waters but similar to activities seen in other deep wells from sandstone aquifers in SE Algeria and also shallow groundwaters from around the Chott El Hodna to the North of the current study area. The Algerian and Tunisian CI activities for 222 Rn appear to be generally in secular radioequilibrium with a rock U-content of~1 ppm (as measured in (Tunisian) CI core sample) and for a porosity~20%%; however the enhanced value at Ouargla may identify the presence of a stable-type redox barrier in this locale. Radium contents ( 226 Ra, 228 Ra) are 10e135 mBq/l in the Algerian CI waters, and up to 600 mBq/l locally associated with Chott Fedjej in Tunisia. These activities are higher than for mineral bottled waters and reported groundwaters in Algeria. Activity ratios for 226 Ra/ 228 Ra generally cluster around the Th/U activity ratio for the CI core material as the predominant source. Activities are orders of magnitude less than for dissolved 222 Rn, showing that Ra contents are not recoildominated as they are for 222 Rn. The highest Ra activities appear positively correlated with Ca and Ba contents.
U-contents and 234 U/ 238 U activity ratios following a flow line NWeSE from the M'Zab Ridge in the Algerian CI identify a redox zone downgradient where A.R.s are enhanced as dissolved U-contents decrease to <1 ppb, and showing a-recoil of 234 U from the rock surface enhanced by U precipitate as the dominant source. Modelling of the residence times to generate the enhanced A.R.s, starting from a value of from 1.76 (at the start of the recoildominated zone) to maximum 3.17 downgradient, suggest resi-dence times up to~600 ka; the distance between Hassi Messaoud and Rhourde El Baguel (some 100 km) apparently taking~65 ka. In Tunisia, a similar trend U-contents and 234 U/ 238 U activity ratios is suggested going SeN and implicating a recharge direction from the Tinrhert Plateau of SE Algeria or the Tunisian Dahar Hills towards the Chott Fedjej discharge zone, as indicated also by the aquifer piezometry. This trend indicates also a redox barrier occurring in this SeN direction. Age modelling of the 234 U/ 238 U evolution suggests geochemical residence times~20 ka between Menchia and Chott El Fedjej in Tunisia.
The paleoage status of these Algerian and Tunisian waters is supported by simple modelling of 4 He accumulation (and 4 He/ 222 Rn) in the waters which suggest water ages of the order tens of thousands up to 1 Ma old. The palaeowater status of these waters is supported also by recharge temperatures measured for the CI waters, which are generally cooler than current Climate Normal values for annual temperature in the region and in the case of Gassi Touil (RT~15 C) indicate recharge under significantly cooler (past) climate conditions. The 234 U/ 234 U activity ratios seen in waters from both the Algerian and Tunisian aquifers therefore support previous 14 C and 36 Cl age estimates that identify the Continental Intercalaire waters from the M'Zab ridge towards the centre of the Grand Erg Oriental sub-basin in Algeria as being palaeowaters, and similarly for waters in the locality of Chott Fedjej to the East of the major horst structure that occurs around Chott Djerid in Tunisia.
Finally, old water ages (residence times) are suggested also for the few Complexe Terminal (CT) waters presented from Algeria and Tunisia, since these samples again have enhanced (radiogenic) 4 He contents.
For the regions studied therefore these waters should be regarded as "fossil" waters and treated effectively as a nonrenewable resource. Nevertheless, any such geochemical residence time estimates may require reconciliation with hydraulic and emerging geophysical approaches, especially on the margins as recognised by Gonçalv es et al. (2013). For example, for identified palaeowaters in the centre of the regional, fissured London Basin Chalk aquifer (UK) Elliot (1999) invokes matrix exchange effects on solute transport to reconcile apparent geochemical residence times with the much younger hydraulic transit time estimates.