Review: Hydrogeology of weathered crystalline/hard-rock aquifers—guidelines for the operational survey and management of their groundwater resources

Hard rocks or crystalline rocks (i.e., plutonic and metamorphic rocks) constitute the basement of all continents, and are particularly exposed at the surface in the large shields of Africa, India, North and South America, Australia and Europe. They were, and are still in some cases, exposed to deep weathering processes. The storativity and hydraulic conductivity of hard rocks, and thus their groundwater resources, are controlled by these weathering processes, which created weathering profiles. Hard-rock aquifers then develop mainly within the first 100 m below ground surface, within these weathering profiles. Where partially or noneroded, these weathering profiles comprise: (1) a capacitive but generally low-permeability unconsolidated layer (the saprolite), located immediately above (2) the permeable stratiform fractured layer (SFL). The development of the SFL’s fracture network is the consequence of the stress induced by the swelling of some minerals, notably biotite. To a much lesser extent, further weathering, and thus hydraulic conductivity, also develops deeper below the SFL, at the periphery of or within preexisting geological discontinuities (joints, dykes, veins, lithological contacts, etc.). The demonstration and recognition of this conceptual model have enabled understanding of the functioning of such aquifers. Moreover, this conceptual model has facilitated a comprehensive corpus of applied methodologies in hydrogeology and geology, which are described in this review paper such as water-well siting, mapping hydrogeological potentialities from local to country scale, quantitative management, hydrodynamical modeling, protection of hard-rock groundwater resources (even in thermal and mineral aquifers), computing the drainage discharge of tunnels, quarrying, etc.


Introduction
Crystalline rocks, or hard rocks (HR), are plutonic and metamorphic rocks (Michel and Fairbridge 1992); marbles, however, cannot be categorized as HR aquifers as they often constitute karstic aquifers. When unweathered and unfractured, hard rocks such as granites, have low primary porosity and low hydraulic conductivity, whereas metamorphic rocks have lost their original hydrodynamic characteristics through metamorphic processes (Achtziger-Zupancik et al. 2017;Ingebritsen and Gleeson 2017). Although limestones and volcanic rocks (e.g., Charlier et al. 2011 andLachassagne et al. 2014a for volcanic rocks) can be mechanically hard to drill, as their hydrogeological properties differ, they are not included as HR. Although plutonic granites and metamorphic metavolcanites, meta-sediments and meta-granites-forming schists, micaschists, quartzites, and gneisses show specific This article is part of the topical collection "Progress in fractured-rock hydrogeology" mineralogy, petrofabrics and textures, they exhibit similar hydrogeological properties and are considered as HR.
Hard rocks form continental basements (Ingebritsen and Gleeson 2017), outcrop over large areas within tectonically stable regions such as ancient cratons, and form over 20% of the earth's surface. Africa contributes about 35% of the world's HR; South and North America about 15% each; India and Australia also have large "HR regions" (Lachassagne et al. 2014b).
The groundwater resources of hard-rock aquifers (HRA) are small in terms of sustainable discharge per productive borehole: from 100 l/h to a few tens of m 3 /h (e.g. Courtois et al. 2010;Maurice et al. 2018) as compared to those of porous, karstic and (recent) volcanic aquifers. "Dry" boreholes are also common in HRA (e.g. Courtois et al. 2010). HRA are, however, able to supply water to scattered populations and small-to-medium-size villages, or the suburbs of larger towns. HRA groundwater is often of better quality than surface water and contributes to the well-being of the population and to economic development, especially in arid and semiarid areas where surface-water resources are seasonal and limited to sites adjacent to large rivers. Large parts of Africa, South America, Asia and Australia are reliant on such groundwater resources. In India, HRA water resources greatly contributed to the "Green Revolution" that initially allowed food self-sufficiency in that country (Foster 2012). This has now been partly lost due to aquifer overexploitation and groundwater quality degradation (Maréchal and Ahmed 2003;Dewandel et al. 2007bDewandel et al. , 2010Perrin et al. 2012).
Reviews and research works, particularly by Lachassagne et al. (2014bLachassagne et al. ( , 2011 and Dewandel et al. (2006), but also by Worthington et al. (2016), demonstrate that weathering is the main process driving the development of a significant groundwater resource in HR, thus driving the creation of the fractures and then of aquifers in hard rocks.
In fact, if one deals with an historical perspective, this role of the weathering had been described for decades, but mostly for the unconsolidated part of the weathering profile-see for instance Davies et al. (2014) who summarize works from as far back as the 1950s, or Acworth (1987), or Wright (1992). Moreover, as described in the following, several other hypotheses such as tectonics, unloading, etc., were also invoked by hydrogeologists, independently or conjunctly with weathering. These other hypotheses were mainly (and are still in some cases) influencing the methods developed to survey, manage and protect HRA groundwater resources, although not without any drawbacks-see for instance the recent papers from Alle et al. (2017) and Soro et al. (2017). These other hypotheses also constitute (conceptual) limitations that prevent the achievement of results in applied hydrogeology, for instance for the operational modelling of HRA groundwater resources at watershed scale. Lachassagne et al. (2011) presented a comprehensive review of available papers at the time of its writing, which demonstrates, argument by argument, (1) why the permeable fractures of hard rocks cannot be due to processes other than weathering, and (2) that all the other hypotheses (tectonics and unloading for the most commonly proposed, but several others are listed) either are not grounded on sound demonstrations, or are not demonstrated at all, and are then just hypotheses. This review was completed in 2014 (Lachassagne et al. 2014b).
This paper aims to describe the role of weathering in HRA hydrodynamic properties (hydraulic conductivity, storage) and to present a corpus of efficient operational applications that have been deduced from these genetic concepts. This paper builds upon previous review papers (particularly Lachassagne et al. 2011 and with up-to-date research/references that complete and reinforce these results. Most of the references cited in these two previous papers are not cited in this present paper; hence, the interested reader is recommended to consult them as well. Origin, structure and hydrodynamic properties of hard-rock aquifers Nonweathered HR have low matrix porosity and low hydraulic conductivity (below 10 −8 m/s); thus, these rock masses are well-suited for nuclear waste storage (see for instance Figueiredo et al. 2016) at depths greater than several hundreds of metres.
This section describes how HRA include two layers ( Fig. 1) forming a composite aquifer (Maréchal et al. 2004;Dewandel et al. 2006Dewandel et al. , 2011Lachassagne et al. 2011Lachassagne et al. , 2014b. Their hydraulic conductivity relies on a secondary fracture permeability due to weathering processes (stress induced by the swelling of some minerals, notably biotite). A borehole providing a significant yield of several m 3 /h (see section ' H y d r o g e o l o g i c a l f u n c t i o n i n g o f H R a q u i f e r s : hydrochemistry'), crosscuts a decametric (30 to >100 m) layer of low-permeability unweathered rock alternating with a few millimetres to decimetre-thick permeable fractures ("water strikes") created by weathering. Such fractures provide water to the pumped borehole. In granites, subhorizontal fractures appear to be more permeable than the other ones, and notably more permeable than the subvertical ones (Kxy ≈ 10 Kz; Maréchal et al. 2004).
This transmissive layer, the "stratiform fractured (or fissured) layer" (SFL) forms the aquifer. The unconsolidated superficial zone of the aquifer (the saprolite) is not pervious enough to deliver a significant discharge to common (small diameter) boreholes. However, it is characterised by a significant porosity, enabling leakage towards the underlying fractured layer. This upper layer thus ensures most of the capacitive properties of the whole composite aquifer (see also section 'Hydrogeological functioning of HR aquifers: hydrochemistry').
Several hypotheses that are not associated with weathering are still regularly invoked by hydrogeologists to explain the origin of the HRA fractured layer. Lachassagne et al. (2011) and 2014b describe in detail that several authors, in various regions of the world (Africa, India, Australia, Europe, etc.), have acknowledged for several decades the role of weathering in the creation of the unconsolidated part (the saprolite) of the HRA: see for instance Davies et al. (2014), who describe case studies in Africa (Tanzania, Zimbabwe, Nigeria, South-Africa), Madagascar, Asia (Hong-Kong (China), India), and Brazil, and also in Europe, in the UK. However, it is only in the last 15 years that the creation of the fractured layer was demonstrated to be also linked to the same weathering processes Lachassagne et al. 2011). This was demonstrated from observations performed by these authors all over the world: France (Dewandel et al. 2017a;Lachassagne et al. 2001cLachassagne et al. , 2014bMaréchal et al. 2014;Wyns et al. 2004), South Korea (Cho et al. 2002(Cho et al. , 2003, Burkina Faso (Courtois et al. 2010), and India (Dewandel et al. , 2010(Dewandel et al. , 2017bMaréchal et al. 2004Maréchal et al. , 2006, and also from several other observations performed in various regions of France (Brittany, Massif Central, Vosges, Corsica, Pyrénées), other African countries, French Guiana, Madagascar, etc., mostly reported within BRGM reports. The role of weathering in creation of the fractured layer is also directly or indirectly confirmed by the data sets of most of the authors working on HRA. See for instance Davies et al. (2014) and the countries cited previously, as well as Chilton and Smith-Carington (1984), Chilton and Foster (1995), Howard (1999, 2000), Chirindja et al. (2017), Dickson et al. (2018), Muchingami et al. (2019) and Vassolo et al. (2019) Krasny (1996) for Central Europe; Baiocchi et al. (2014Baiocchi et al. ( , 2016 for the Mediterranean Europe; Chou et al. (2014) and Chuang et al. (2016) for Taiwan; David et al. (2014) for Australia; Riber et al. (2017), Jahns (1943, Hart (2016), Sanford (2017), Meyzonnat et al. (2018) and Setlur et al. (2019) for North America, etc. Even if several authors have not attributed the fractures to weathering processes, some others keep an ambiguity on this genetic relationship, while others do not challenge or discuss this issue at all, at least for the uppermost crust (0-2 km; Achtziger-Zupancic et al. 2017;Ingebritsen and Gleeson 2017).
The main outcome is that the stratiform fractured layer is considered to belong to the weathering profile as well as the above-lying unconsolidated layer (the saprolite; Fig. 1). Worthington et al. (2016) confirm this conceptual model, which is today more and more accepted within the hydrogeological community and used by several authors to explain their observations (see for instance Allé et al. 2017;Baiocchi et al. 2014Baiocchi et al. , 2016Guihéneuf et al. 2017;Soro et al. 2017;Vouillamoz et al. 2014Vouillamoz et al. , 2015Wubda et al. 2017, etc.).

Previous concepts
Tectonics, unloading processes, and emplacement and cooling of plutonic rocks have been the most common genetic hypotheses formulated to explain the development of fractures within the fractured layer, but without any scientific demonstration of their validity. Lachassagne et al. 2011 and demonstrated that these hypotheses, briefly summarized in the following, are not relevant to explain the development of HRA. Jahns (1943) demonstrated that joints in plutonic rocks are unrelated to the emplacement and cooling of these rocks. More annecdotally, Hasenmueller et al. (2017) show that plant roots have no significant impact on HR weathering. . Note that all technical terms in this figure are described and explained in section 'Geological structure and hydrodynamic properties of weathering profiles' Lachassagne et al. (2011), synthesising previous works in a detailed review paper (see all the references in Lachassagne et al. 2011), demonstrated that: 1. The fractured layer cannot be explained by natural unloading processes. With the exception of some specific conditions such as high cliffs, deep sedimentary basins and artificial unloading (a rapid (hours) decrease of the vertical stress component as in deep wells and mine galleries), there is no existing natural phenomenon able to induce strong compression parallel to the surface. This is the only strain state that may produce subhorizontal fractures such as those commonly observed in granites (described later in this paper). Such an unloading process does not exist in the near subsurface; moreover, such a process would be observed for types of rocks other than HR. 2. Gravity-induced fractures can only be invoked in specific contexts such as high relief such as along the flanks of Alpine valleys, but they do not produce horizontal jointing such as observed in granites. 3. The fractured layer and its associated hydrodynamical properties cannot be explained by tectonic fracturing. Tectonically active areas are mainly at plate boundaries with, nevertheless, to a lesser extent, intraplate tectonics or other processes such as glacial unloading, accommodation in sedimentary basins, etc. Plate boundaries only occupy a few percent of the earth's emerged surface and not thousands to hundred of thousands of km 2 as observed in HR aquifers. The concentration of strain by definition makes them anomalous in regard to the average properties of the crust; therefore, the tectonic fracturing theory faces several inconsistencies. In tectonically stable areas such as most HR areas in the world it requires: a. A tectonic process to create the fracture. As explained previously, it is clear that occurrence of such fractures is rare both in time and space. Lachassagne et al. (2011) explain that, with the example of Brittany, France, where the contrast is high between, on one hand, the~20,000 wells evenly distributed in this region that are delivering HR groundwater, and, on the other hand, the few active tectonic fractures known in this region. Only a few of these wells are located in the vicinity of these fractures. The vast majority of the wells are far away from these active tectonic fractures and statistical analyses do not show any correlation with them. b. That the resulting fracture is permeable. It is obvious that a tectonic fracture is a complex structure that is far from being systematically permeable-for instance, Caine et al. (1996) comprehensively describe what a tectonic fracture looks like and show that it comprises several impervious parts. Lachassagne et al. (2011) also summarize several references demonstrating that tectonic fractures act as impervious bodies. c. That the resulting tectonic fracture reaches the subsurface, i.e. the depth of a HR borehole, whereas most recent/active tectonic fractures do not reach the last kilometres below ground surface. d. That a permeable fracture is not rapidly sealed at the geological timescale, when in fact rejuvenation is overbalanced by fast sealing (in a few years).
Outcropping tectonic fractures are often old and thus sealed. e. The ubiquity of such fractures. Tectonic fractures are spatially unevenly distributed and thus cannot account for the tens of millions of evenly distributed productive water wells in HR areas of the world (see the example of Brittany already presented).
In addition, further arguments can be put forward: the absence of thermal springs in most of such tectonically stable areas, whereas the main active faults in tectonically active regions quite systematically exhibit several of these thermal springs; and, that most tectonic fractures have a strong dip and should thus hardly ever be tapped by standard boreholes as opposed to the subhorizontal jointing of granitoids and the variously dipping fractures of the other HR fractured layer which are systematically cut by vertical boreholes. Another argument is that lineaments are used to locate "tectonic fractures" but, most of the time, without any demonstration that there is any tectonic fracture at all underlying the identified lineaments and, in some cases, the demonstration that there is no tectonic fracture at all (Soro et al. 2017).
Tröger (U. Tröger, Technische Universität Berlin, personal communication, 2020) raised attention to the fact that Twidale and collaborators "demonstrated" the tectonic origin of the sheet fractures, constituting the fractured layer mostly in granites. As part of this review, several of their papers were studied, and Vidal Romani and Twidale (1999) and Twidale and Bourne (2009) were found to be the most relevant and recent publications on this topic. First, Vidal Romani and Twidale (1999) convincingly show that sheet fractures are due to compressive stress, more precisely horizontal compressive stress. This agrees with the conclusions of, notably, Lachassagne et al. (2011). Second, they show that sheet fractures cannot derive from unloading processes. As exposed previously, this is also one of the conclusions of Lachassagne et al. (2011), independently from the work of Vidal Romani and Twidale (1999), but with several similar proofs and arguments. Third, the remaining issue is to identify the origin of this compressive stress.
Vidal Romani and Twidale (1999) argue that the stress is of tectonic origin, whereas the authors of this review argue (notably in Lachassagne et al. 2011 and) that this compressive stress is due to the weathering process at the time it occurred. However, briefly, it is observed that Vidal Romani and Twidale (1999), and also Twidale and Bourne (2009), do not demonstrate convincingly that the compressive stress that creates(ed) the sheet fractures are of tectonic origin. They propose this as a hypothesis (and use often the conditional argument), that, to the authors' opinion, is far from being formally demonstrated. Their argument is mostly to correlate the fact that "many parts of the continents are in substantial compression" (Vidal Romani and Twidale 1999) with the observation of these sheet fractures; however, correlation does not demonstrate causality. They do not consider the other hypothesis of weathering that is much better demonstrated, in the authors' opinion, and can be generalized to all hard rocks, and not only to the sheeting fractures of granites.
Description of the weathering processes that create the stratiform fractured layer (SFL) The development of the SFL's fracture network is the consequence of the stress induced by the swelling of some minerals. This process was evidenced by Wyns et al. (2003Wyns et al. ( , 2004 and is described in detail by Lachassagne et al. (2014b), who proposed a detailed description and sketches of the deformation ellipsoid, resulting stress and permeability ellipsoid, and the geometry of the induced fracture network as a result of the swelling minerals' orientation configurations: randomly oriented swelling minerals, vertically oriented, horizontally oriented, folded rock. This process is summarized in the following and completed with some additional findings and also references that complete the description of Lachassagne et al. (2014b).
The development of the SFL is thus the consequence of the stress induced by the swelling of some minerals, particularly biotite, olivine and pyroxene, being progressively hydrated and turning into hydro-biotite, then vermiculite, then mixed clay layers, etc. Chemical weathering reactions cause biotite swelling, then volume changes, increasing stress and finally fracturing the rock. In turn, these fractures locally enhance the transport and exportation of chemicals through the rock, accelerating there the weathering. This process of "reaction-induced cracking" is also described by Royne et al. (2008) and was numerically modelled by Rudge et al. (2010) for the carbonation and serpentinisation of peridotite. In all cases (HR, peridotite), the result of the weathering process is then a SFL parallel to the paleotopography contemporaneous with the weathering process (Fig. 1). The numerical model developed by Rudge et al. (2010) was applied by Vasseur and Lachassagne (2019) to granites in order to evaluate the thermal impact of weathering; this last paper also provides additional references on reaction-induced cracking.
The weathering of biotite is an early process. Nevertheless, Anovitz et al. (2021) suggest that it may be preceded by earlier stresses due to crystallization pressures caused by the growing of iron phases that follow pyrite oxidation. The increase in volume of biotite can reach 30% during weathering and that of the total rock 50% (example for a granite). In granular rocks like granites with a quasirandom orientation and location of swelling minerals, the potential expansion tensor is isotropic, but expansion is more difficult in the horizontal plane as the rock extends far away in this direction. Consequently, the horizontal stress component increases during the weathering. In the vertical axis, however, the stress increases until the lithostatic component is offset, then allows vertical expansion, while horizontal stress continues to increase. Consequently, the resulting stress tensor is characterised by a minor vertical component (σ 3 ), and two major ones (σ 1 and σ 2 ) that are horizontal (see detailed figures in Lachassagne et al. 2014b). When the stress deviator reaches the elastic limit of the rock, tension cracks appear and the SFL is forming.
Then, for granitic rocks, in accordance with rocks mechanics (Pollard and Aydin 1988;Mandl 2005), the resulting fractures are perpendicular to the minor stress (subvertical) and consequently are subhorizontal, parallel to the gentle topography contemporaneous with the weathering, and lead to the formation of the subhorizontal jointing of granites (Fig. 2). In foliated and folded rocks, the variability of the orientation and location of the minerals able to swell, as well as the ones of the weaker surfaces of the rock (foliation, schistosity), induce fracturing without preferential orientation.
It is the early stage of biotite weathering that induces fracturing. The later stages produce clays which show a tendency to fill newly formed pores of the fractured and weathered minerals. Even if the early weathering of biotites leads to local clogging of biotites and surroundings at a millimeter scale, the density of biotites in the parent rocks is not enough to clog the aperture of all the fractures of the fractured layer at a metre to decametre scale.

Geological structure and hydrodynamic properties of weathering profiles
A typical weathering profile ( Fig. 1) comprises layers that have specific hydrodynamic properties. Where and when saturated with groundwater, these layers form a composite aquifer. With depth, the four main layers are Lachassage et al. 2014b;Vassolo et al. 2019): 1. An iron or bauxitic crust, that may be absent. Where preserved from erosion and recharged by heavy rainfall, the iron or bauxitic crust can give rise to small perched aquifers with locally some springs that show an epikarst-type functioning. These flows are of short duration as the system normally dries up during the dry season; hence, the concentration of iron oxides within this zone aided by evapotranspiration and water throughflow. This is observed in tropical humid regions, as in French Guiana, or during wet season in western Africa. 2. The saprolite, alterite or proparte regolith. This is commonly a clay-rich material derived from prolonged in-situ decomposition of bedrock, several tens of metres thick, where this layer has not been eroded. This is commonly a clay-rich material, as most minerals (aluminosilicates) from HR transform into various kinds of clays. However, some rocks such as quartz or microgranite veins, quartzites, etc., mostly composed of silica, do not produce a clay-rich saprolite. In nonplutonics rocks, the saprolite layer is commonly divided into two subunits ( Fig. 1): the alloterite and the isalterite. The alloterite, at the top of the saprolite, is a clayey horizon where, through volume reduction due to chemical weathering and leaching, the structure of the parent rock is lost. In plutonic rocks, the structure of the parent rock is still identifiable at the top of the weathering profile, which explains that alloterites cannot be identified. In the underlying isalterite, weathering only induces slight or no change in volume thus preserving the rock structure. This layer forms half to two thirds of the thickness of the saprolite layer.
In plutonic rocks such as granites and gabbros, the base of the saprolite has a laminar form. This "laminated layer" is constituted by the partly consolidated weathered parent rock with coarse-grained clastic texture and a millimeterscale dense horizontal fracture lamination crosscutting the largest minerals, but preserving the original structure. It was recently demonstrated that the geophysical signature of this laminated layer is shown by an increase in resistivity as compared to both the overlying saprolite and the underlying top of the fractured layer (Belle et al. 2016(Belle et al. , 2017, particularly in granites, but also in micaschists. The saprolite layer has low hydraulic conductivity. In granite-type rocks, from several studies gathered in Africa and India , the (geometric) mean is 2.10 −6 m/s for the entire saprolite, with values ranging between 10 −7 and 5.10 −6 m/s. The hydraulic conductivity is less in the basal laminated layer (transmissivity about 5.10 −8 -10 −7 m 2 /s −1 ; Boisson et al. 2015). The hydraulic conductivity of the alloterite is less than that of the isalterite as the first is more clayey. The hydraulic conductivity is much less in fine-grained rocks' saprolite such as shales, for instance, than in coarse-grained rocks. In these coarse granites, the base of the saprolite can be permeable enough to sustain the discharge of a well, even if it may be difficult to identify if this permeability originates from the base of the saprolite or the top of the SFL.
Where saturated the saprolite layer ensures the capacitive function of the whole composite aquifer. With its clayey-sandy composition in coarse granites, the saprolite layer has porosities of 5-30% (Compaore et al. 1997). In fine grained or low quartz content schists, the saprolite is clayey with low porosity and hydraulic conductivity. 3. The SFL (Figs. 1 and 2), which is characterized by dense fracturing in the first few metres and decreases in density with depth. As described previously, the fractures are mostly subhorizontal in granite-type rocks or in vertically foliated rocks. They are randomly dipping in metamorphic folded rocks (no preferential orientation). In some papers, the top of the SFL, and/or the bottom of the saprolite, where saprolite and boulders coexist, are also referred to as "saprock" (see for instance Wright 1992;Riber et al. 2017). The thickness of the SFL is approximately two times that of the saprolite in places where the saprolite was not eroded ( Fig. 1) and can reach over 100 m thick (Dewandel et al. 2017a;Wyns 2020a, b). The fracture density decreases with depth towards the base of the weathering profile (the base of the SFL). Then, the observed decrease of the hydraulic conductivity of HRA with depth is not a consequence of a lower permeability of the fractures or of their "closure with depth due to lithostatic constraints", but is rather due to their disappearance in depth (Maréchal et al. 2004;Dewandel et al. 2006).
Geological logging (or coring) identifies several tens of such fractures in the SFL. Due to healing of some of these fractures, weathering, channeling within fracture plane, etc., only a maximum of four to five fractures are permeable enough to be detected as "water strikes" during borehole drilling and logging, the others being "dry" (see section 'Hydrogeological functioning of HR aquifers: hydrochemistry'). These (max) four to five fractures then ensure the discharge of the well. Sometimes, no fracture at all is permeable enough to ensure a significant discharge. Hence there is a large variability of borehole discharge even for close boreholes; drilling a "dry" , c upper part of the SFL, Lescastel quarry, Brittany, France, d the SFL in biotite-bearing metamorphic volcanic tuffs, Vensat, SW of Gannat, Northern Massif central, France (the height of the photograph is about 2 m), e subhorizontal productive fractures in a biotite granite in a 165-mm-diameter unscreened borehole, Choutuppal village, Andhra Pradesh (now Telangana), India. The picture is taken from top to down (e), and the yellow rope indicates scale (the well and its exact location are described in Boisson et al. 2015). Photographer R. Wyns (a-d) borehole close to a productive one, and vice-versa, is not rare (Maréchal et al. 2004;Dewandel et al. 2006). Then, the SFL of a weathering profile statistically exhibits relatively "homogeneous" properties (thickness, hydrodynamic parameters, etc.), when one considers several wells drilled in this SFL. However, considering the data from each individual well, the standard deviation of the hydrodynamic parameters (notably hydraulic conductivity) is very high, and notably much higher than in most porous aquifers.
Ancient weathering profiles may have been completely sealed as the result of diagenesis due to their burial by marine sediments Tan et al. 2017)for example, in the Vosges Massif (eastern France), the porosity of the SFL of the pre-Triassic weathering profile has been sealed by hydrothermal minerals (ankerite, dolomite, calcite, barite, hematite) during subsidence in the Rhine graben (Wyns 2012); nevertheless, it seems that this sealing can be localized only along faults in old hydrothermal loops (Wyns 2020b). This diagenesis is thought to contribute to the SFL hydraulic-conductivity decrease observed below Cenozoic sediments in another case study in France (Dewandel et al. 2017a).
The properties and hydrodynamic parameters of a granitic SFL in India are described in detail by Maréchal et al. (2004Maréchal et al. ( , 2002. The length of each subhorizontal fracture within the SFL is a few tens of metres in diameter (10-30 m). Such a horizontal permeable fracture set is on average 10 times more permeable, and more numerous, than the subvertical joint set. The hydraulic conductivity of these permeable fractures does not show much variability (Maréchal et al. 2004;Dewandel et al. 2006), which supports weathering as a unique origin. The transmissivity of the most conductive part of this 35-m-thick granite SFL reaches 1.10 −3 m 2 /s (Boisson et al. 2015).
The storage coefficient (S) in the fracture zone of the same granite aquifer is about 10 −6 (Boisson et al. 2015). The drainage porosity of this 35-m-thick SFL is 0.5-2% with a partition between blocs and clogged fractures (90%), and permeable fractures (10%) (Maréchal et al. 2004). Dewandel et al. (2010Dewandel et al. ( , 2017b show that the aquifer-scale vertical distribution of the drainage porosity (or specific yield) decreases with depth from 1.5% to 0.2% over a range of depth of 30-40 m below the base of the saprolite. Dewandel et al. (2012) demonstrate, in a granite aquifer in India (see also Dewandel et al. 2017c for similar results on a peridotite aquifer), the relatively good homogeneity of the SFL hydraulic conductivity at the scale of aquifer subareas. Hydraulic conductivity values range from 10 −7 to 10 −4 m/s in the whole aquifer, but are distributed over zones up to 1 km 2 in area with similar hydraulic conductivity. Dewandel et al. (2012) show, at watershed scale, that the effective porosity of the whole aquifer (here a thin saprolite and the SFL) ranges from 0.8 to 2.7% (mean of 1.5%, with an uncertainty below 0.5%). They regionalised and mapped these hydrodynamic parameters over 53 km 2 of aquifer showing that the SFL has homogeneous hydraulic conductivity at few hundred metres to a kilometer scale. These results are useful for modelling groundwater flux and solute transport (sections 'Quantitative management and modeling of HRA groundwater resources at the watershed scale: managed aquifer recharge' and 'Protecting and restoring HRA groundwater resources, groundwater quality modelling, and contaminant transfer'). Dewandel et al. (2017b) report the order of magnitude of the effective porosity at watershed scale, namely 0.3-2%, in two watersheds (700 and 1,000 km 2 ) in India. Moreover, they show that the vertical distribution of effective porosity can be very different from one place to another at the scale of the same aquifer, and that the fractured layer is not always characterized by a rapid decrease of the effective porosity with depth. Locally, lateral variations in effective porosity can be larger than vertical ones.
The SFL assumes the transmissive function of the composite aquifer. It is drawn from most of the boreholes drilled in hard-rock areas. In some regions, due to erosion, the covering saprolite layer may have been partially or totally eroded; it may also be unsaturated due to low piezometric levels. In these cases, the SFL assumes not only the transmissive function of the aquifer, but also its capacitive one, which is for instance the case in Brittany (France), where, due to the erosion of the saprolite, 80-90% of the stored groundwater is found in the fractured layer (Wyns et al. 2004). 4. The fresh basement This is only permeable where "deep fractures" are present (see section '"Deep vertical" discontinuities in the unweathered basement'). Even if these fractures are as permeable as the fractures of the SFL, their density with depth and laterally is lower (Cho et al. 2003).
Then, for catchment-scale water resources studies, the fresh basement is considered as impermeable with a negligible storativity (Maréchal et al. 2004) Main parameters governing HRA development (weathering profile development) and their structure The development of a weathering profile, and thus of HRA, requires several criteria: -The HR must be located above sea level and not covered by a thickness of sediments. -Development of a weathering profile tens of metres thick must have enough time for effective hydrogeological development; this takes millions to tens of millions of years (Myr) to create (Wyns 2002;Wyns et al. 2003). If there is no erosion (see the following), the regolith thickness increases as the square root of time (Braun et al. 2016). -As weathering requires water, hot or cold deserts are not favorable for weathering, nor areas of permafrost (Acworth 1987;Wyns et al. 2003). However, due to continental drift, climatic changes at the geological time scale, and duration of a weathering profile functioning, many cases of weathering profiles are observed in present desert countries (e.g. Saudi Arabia, Fig. 2b). -There must be stable geodynamic conditions. The erosion rate must be lower than the weathering rate Braun et al. 2016), allowing the weathering profile to grow and to thicken leading to peneplaination. Due to erosion, tectonically active processes do not favor weathering profile conservation. In fact, weathering profiles of lateritic type are characterized by the leached removal of dissolved compounds from the rock. Profiles submitted to a gentle regional uplift (such as long-term lithospheric deformation) favor the development of thick weathering profiles, with a SFL over 200 m thick (Wyns 2020a;Cho et al. 2002;Monod et al. 2016). -Temperature, linked to the climate, is a secondary factor that only affects the kinetics of the weathering process, with a factor of 1.7 between a tropical (28°C) and a temperate climate (15°C; Oliva et al. 2003). Moreover, the characteristic period of the climate stability, 10 4 to 10 5 years, cannot explain the time required to develop the weathering profiles which is much longer (10 7 -10 8 years; Wyns et al. 2002). The characteristic period is thus mostly of geodynamic origin (Wyns 2020a). In addition, the present climatic belts are not older than 10 Myr, due to the late Miocene to Quaternary formation of polar ice caps: palynology and stable isotope data show that lateritic weathering was possible up to 60°of latitude. Most of the iron or bauxitic surfaces in modern tropical countries are then also related to ancient weathering profiles Freyssinet 1999 and2002).
Differential weathering results in a local topographic scarp separating less weatherable rocks from lower more weatherable ones; this "scarp" is also present at depth at the interface between the various layers of the weathering profile. Leucogranites, quartz and microgranite veins and pegmatites, generally weather less than the surrounding biotite granite (Hill 1996;Lachassagne et al. 2001c;Dewandel et al. 2006Dewandel et al. , 2011. Within metamorphic lithologies, harder quartzite layers and green rock intrusions stand proud of surrounding schistose rocks in the weathering profile. Within calc-silicate gneisses, harder granite veins weather less and stand proud of the surrounding strata . To summarise, mineralogy and texture are the main controls of differential weathering, biotite-rich and coarse rocks being the more weatherable; however, at watershed scale, a case-by-case geological study is needed to show the effects of differential weathering. As the geodynamical context changes (i.e., when passing from no tectonic activity (peneplaination) to active tectonism) the weathering profile is partially or totally eroded. Conditions again favorable for the development of a new weathering profile favour the development of a polyphased weathering profile Lachassagne et al. 2014c); thus, most HR areas of the world show relicts of several generations of weathering profiles (Jahns 1943;Chilton and Smith-Carington 1984;Freyssinet 1999, 2002;Thiry et al. 2005;Dewandel et al. 2006).
In Europe, periods of weathering profile formation have been recognised from the Carboniferous, Permian, Triassic and Liassic. The Early Cretaceous and the Early and Middle Eocene, having lasted 45 and 25 Myr, respectively, are the most recent period during which weathering processes lasted long enough to develop 70-100-m-thick profiles that included a 50-70-m-thick fractured layer underlying a 20-30 m layer of saprolite Bauer et al. 2016). The late Miocene to Present period only produced thin profiles that have little hydrogeological significance (Lachassagne et al. 2014c).
Complex weathering profiles result from multiphase periods of weathering and erosion Howard 1999, 2000;Dewandel et al. 2006), and, in some cases, resedimentation as "pediment formations" (Chardon et al. 2018). For instance, recent weathering periods produced a 5-8 m thick (below the current topographic surface) weathering profile in South Korea (Cho et al. 2002;Fig. 3) and a 2-3-m thick weathering profile in France (Lachassagne et al. 2014c). This recent weathering clearly developed after the main earlier weathering phase(s): Late Cretaceous to early Eocene period in France, with hard rock weathering profiles about 80 m deep, including a thick hydraulically effective SFL. In fact, this recent weathering crosscuts all the horizons of the ancient Late Cretaceous to early Eocene weathering profile. This recent phase follows the present topographic surface. Since these recent weathering phases have not developed a thick fractured layer, they are of low hydrogeological interest, but must be considered as they can mislead weathering profile mapping (section 'Assessing HRA groundwater reserves from watershed to regional scales').
In the Cacao village area, French Guiana ( Fig. 4; Courtois et al. 2003), a polyphased weathering, uplift and erosion history formed the complex weathering profile, with four stepped palaeo-erosion surfaces including the present-day fluviatile topography. This case study comprises juxtaposed amphibolite metamorphic rocks and a granodiorite; hence, differential weathering is also observed between these rocks-the amphibolite metamorphic rocks weather less and stand "in relief" compared to the granodiorite.
Consequently, and as a conclusion, weathering profiles are not the result of present-day climate but of ancient ones. That is why Norway and Scotland (UK), as well as France, Central Europe, the Korean Peninsula, North America, etc., exhibit such weathering profiles. Globally, individual regions exhibit characteristic weathering profiles indicative of the paleo-(and present day in some cases) tectonic context (and erosion processes) they have experienced. The main determinants of such weathering profiles are not the climate, but geodynamical factors such as morphology, erosion rate, duration of tectonic stability, and the presence of water. Present-day arid or semiarid regions as well as permafrost zones may exhibit relicts of ancient weathering profiles, like in the Arabian and Scandinavian shields or in Greenland (Wyns 2020a).
In some areas of significant relief, ancient weathering profiles may be partly or totally eroded, but they may have been preserved by a sedimentary rock cover and latterly revealed by erosion. In the latter, an ancient weathering profile, like pre-Triassic or Early Cretaceous profiles of the Hercynian French Massif Central and Armorican Massif, can be reactivated. The late Tertiary and Quaternary weathering profiles formed on fresh rocks are less developed with a small thickness 5-30 m (Lachassagne et al. 2014c).  Courtois et al. (2003) Ferric oxide caps mainly occur in present-day tropical countries with a long dry season. However, relict weathering profiles are locally present in France (Thiry et al. 2005). The absence of iron duricrust at the top of lateritic profiles is due either to erosion of the profile surface before burial by sediments, or to rehydratation of the hematitic duricrust to iron hydroxides (goethite and limonite) within a latosol. This absence is not due to the present-day temperate climate.
Mapping the layers (saprolite, SFL) constituting the hard rock aquifers This conceptual model shows how HRA can be considered as "continuous aquifers" being formed of superimposed layers, even though hydraulic conductivity heterogeneities exist between each layer and in the SFL. Therefore, it is possible to map, or model in three dimensions, the elevation and thickness of HRA constituent layers (Lachassagne et al. 2001c;Wyns et al. 2004;Dewandel et al. 2006). The saprolite layer has mostly capacitive hydrodynamic properties, while the SFL has transmissive properties and some capacitive properties.
This mapping methodology is based on topographic maps, digital elevation models (DEM), field (outcrops, dug wells) geological observations, borehole data, field geophysical surveys (see Allé et al. (2017) (Chandra et al. 2016), and hydrodynamical data (Fig. 5). It is a two-fold methodology: -First, within a study area, characterise the local weathering history (eventually multiphase), and also the weathering profiles for the main rock types, to explain any differential weathering (Fig. 4). Identifying also "paleosurfaces" (see section 'Main parameters governing HRA development (weathering profile development) and their structure'). -Second, map the elevation of the interfaces separating the saprolite base at the SFL top and the SFL base (top of the unweathered basement). These interfaces are generally parallel to the paleosurfaces.
The methodology varies according to data availability and the presence of outcrops. The resulting maps (Fig. 6) show the intersection of these horizons with topography similarly in a geological map (Fig. 6a), isohypses of these interfaces (Fig. 6b), and map of layer thickness variation (Fig. 6c). Such a mapping methodology can be implemented at various scales, from local (Figs. 4 and 6) to country scale (e.g. Courtois et al. 2010), or even continental scale (Wilford et al. 2016).
The geometry of the saprolite base is determined using borehole data, geophysical data and geological field surveys. Where the saprolite basal surface is cut by the present-day ground surface, as by valleys, it is possible, in the field, to determine from outcrops the position of the saprolite/ weathered-fractured layer interface. Using observations/ computations of the SFL thickness from outcrop and analysis of borehole data (Courtois et al. 2010), the SFL base elevation is calculated by substracting SFL thickness from the elevation of the base of the saprolite.
The elevations of (1) the saprolite/SFL interface and (2) the SFL/fresh rock interface can be subtracted from ground surface from DEM data, to calculate the residual thickness of the saprolite and SFL. In regions where the Fig. 5 Methodology and data used for mapping HRA layers. See also the geological legend in Fig. 1. Note that the vertical scale is exaggerated, but that the relative thicknesses of the saprolite and SFL (1/3-2/3) are correct paleosurface(s) are partly preserved from erosion, slope analysis from DEM data can be used to reconstruct paleosurfaces confirmed by field observations (outcrops, drillings). This mapping methodology can be applied at local (a few hectares to square kilometres) to regional (a whole country) scales. Observation of weathering at HR outcrops requires differentiation of limited "recent weathering" (e.g. Fig. 3) from thicker weathering profiles; recent weathering is usually only a few metres deep (see for instance the nice picture of such a recent saprolite developed on an older eroded weathering profile, in Fig. A1a of Setlur et al. 2019). Limited "recent weathering" indicates that the associated SFL will not be hydrogeologically significant (section 'Main parameters governing HRA development (weathering profile development) and their structure'). Recent weathering can transform a near-surface weathering profile (SFL, fresh rock) into a thin few metres of saprolite where the previous layers cannot be yet distinguished. Thus, observation on shallow outcrops must be completed with deeper ones, for instance in quarries or in deep trenches, as well as with borehole or geophysical data, to distinguish shallow recent weathering (not significant hydrogeologically) from the older (significant) weathering profile; thus, characterizing the local weathering history is all important.
Observations on outcrops also enable the identification of the SFL (see for instance Fig. 2) and, more precisely the location of the observed outcrop within the SFL (top, medium, bottom) according to the density of the fractures. Electrical geophysical methods, e.g. electrical resistivity tomography (ERT) are particularly well suited for identifying and mapping the various layers constituting weathering profiles. Definitely, as weathering profiles are three-dimensional (3D) objects, at least two-dimensional (2D) methods such as ERT, must be employed. Alle et al. (2017) nicely show the important artifacts due to the implementation of one-dimensional (1D) methods; they show that a near-surface clayey anomaly (a patch) within the saprolite can be interpreted similarly as a deep subvertical conductive structure. Moreover, they also show that deep subvertical conductive structures such as those searched for by some hydrogeologists, can not at all be detected from the surface with 1D profiling or vertical electrical soundings. Belle et al. (2016Belle et al. ( , 2017 identify the specific electrical signature of the laminated layer in granites, and even in micaschists. It appears on most profiles as a more resistant layer than the overlying saprolite and the underlying SFL, which must not be mistaken for unweathered boulders (corestones) remaining within the bottom of the saprolite. The identification and location of this laminated layer is needed to measure the remaining thickness of the saprolite, thus its groundwater storage capacity, and also to locate the underlying SFL, which is of interest for water-well siting.
From a methodological standpoint, Belle et al. (2017) also show the influence of the chosen ERT inversion method. A given ERT inversion method can highlight or, on the opposite, partially or totally hide some geological structures. Some inversion methods provide evidence of subvertical structures (e.g., standard vertical), whereas some other methods provide evidence of subhorizontal structures (e.g., standard horizontal) such as can be found in the various layers of the weathering profile. In the Belle et al. (2017) case study, the resistive laminated layer is nevertheless found by all inversion modes, even where the layer thickness and extent are small.
"Deep vertical" discontinuities in the unweathered basement Authors such as Chilton and Smith-Carington (1984); Acworth (1987); Sander (1997); Owen et al. (2007); Dewandel et al. (2011) and Place et al. (2015), describe preexisting heterogeneities within the HR-such as quartz, pegmatite and aplite veins; dolerite dykes, ancient faults and joints; and contacts between lithological units (Neves and Morales 2007;Le Borgne et al. 2004Durand et al. 2006;Roques et al. 2016), that act as sites for preferential weathering. At such sites, local deepenings of the weathering profile are reported to reach several hundreds of metres below ground surface (Roques et al. 2014b). The weathering process is at the origin of fracture development and enhanced local hydraulic conductivity in the HR surrounding these heterogeneities. The weathering process is also at the origin of the fracturation within meter-to decameter-wide discontinuities such as quartz veins and pegmatite veins ; Figs. 1 and 7). This deep weathering can locally enhance hydraulic continuity between the subsurface (HR or alluvial) aquifers and these deep structures (Roques et al. 2014b) and special attention has to be paid for evaluating their hydrodynamic properties, their relationships with the surroundings weathered layers, and consequently their sustainable yield .
Near these discontinuities, the HR weathering-profile saprolite and SFL are characterised by a deepening of weathered layers in the HR close to the contact. There the stress tensor is deviated and σ 3 becomes suborthogonal to the heterogeneity; consequently, the fractured layer is "verticalised" as it develops parallel to the vein or the former fault/joint/contact (Fig. 7). There, the fractured layer and saprolite reach greater depthsby up to several hundred metres Roques et al. 2014b;Arias et al. 2016;Belle et al. 2017)-than within the classic subhorizontal previously described profile. However, these layers are not wide on the sides of the discontinuity. Such geological structures are of primary hydrogeological interest in areas where the SFL has been eroded or is unsaturated above the piezometric level. In these areas, they form the only targets for water-supply boreholes.
Dolerite dykes also form weathering products, however with properties that define them as having low hydrodynamic potential. Such discontinuities, not enhancing the thickness or hydraulic conductivity of the weathered saprolite or SFL in granite, nor its hydraulic conductivity, make poor hydrogeological targets Perrin et al. 2011a). They can even act as impermeable boundaries to groundwater flow, and, as such, have an impact on the piezometry. Some of the low-permeability structures are large, e.g. up to 500 m in a granite case study near the border of an Oligocene graben in the Massif Central in France (Belle et al. 2016, Dewandel et al. 2017a). There, the saprolite depth reaches more than 200 m (the depth of the deepest borehole, that did not reach the bottom of the saprolite). The two boreholes sited in these structures are of low productivity: only 0.5 m 3 /h for each of these 150-m deep boreholes totally screened in this saprolite-type formation. Similar deep weathered structures, to 250 m depth, are described along the contact between the metamorphic rock and an underlying granite (Arias et al. 2016). Such large structures need to be represented in the weathering-profile geological-mapping process as they must be avoided for siting water boreholes but included in groundwater reserve evaluation and groundwater modelling.

Hydrogeological functioning of HR aquifers: hydrochemistry
The HRA are characterized by a rather low hydraulic conductivity, showing then high hydraulic gradients (often over 1%), notably in the parts of the world where the recharge is significant. Consequently, piezometric heads are there subparallel to the topographic surface (Figs. 8 and Fig. 7 Conceptual hydrodynamic model of a vertical discontinuity in hard rocks . The parameters for the SFL are from Dewandel et al. (2006), cited in ; both papers provide ranges and variances. Km means matrix hydraulic conductivity. With permission from Elsevier  . Under natural conditions, piezometric contours beneath watersheds are often similar to the surface topography. Groundwater watersheds are very similar to surfacewater watersheds (Fig. 9).
Except in areas of low recharge and over-exploited areas, piezometric levels are shallow (Fig. 9a). Streams drain aquifers in wet temperate and tropical humid regions where streams flow during the wet season, and topographic depressions include "spring zones" (and/or wetland areas) with groundwater issuing from HRA. Due to the low hydraulic conductivity of these composite aquifers, rivers' specific discharges are low in HR areas, below 2 L/s/km 2 (Baiocchi et al. Fig. 9 Hydrogeological functioning of hard-rock aquifers a in natural flow conditions and b in pumping-induced conditions (borewell pumped in the SFL). The geological legend is the same as those of Figs. 1 and 5. The two main layers (saprolite and SFL) are clearly delimited, as well as the unsaturated zone, and the hydraulic heads in each layer are described (water table, piezometers). Legend: 1. Piezometric level (water table) in the saprolite (a-b) and also without any pumping or before pumping (b); 2. piezometric level (water table, b) in the saprolite after the well was put in operation and pumped for a while. Quadratic head losses in the borewell are not shown; 3. Hydrogeological water divide; 4. Groundwater flow lines (a, the thickness of the flow lines is proportional to the groundwater flow); 5. Borewell/piezometer with, from bottom to top: screened interval, tube filled with groundwater, and piezometric level (triangle). The tube is grouted; 6. Location of the main water strikes in the borewell/ piezometers (b) 2016; Dewandel et al. 2004). Consequently, most springs have also a low discharge. In areas where the piezometric level reaches the topographic surface, springs and rivers drain the aquifer, and groundwater/surface-water interactions occur. This is the aquifer's "discharge area". This "discharge area" cannot be opposed to an upstream "recharge area" of the aquifer. In fact, as the aquifer is unconfined, the whole aquifer watershed contributes to its recharge (Fig. 9a).
Both direct (diffuse) and indirect (localized) recharge, as defined by Lerner (cited by Marechal et al. 2006), are documented in hard-rock aquifers. Maréchal et al. (2006) quantify both direct and indirect recharge at the aquifer scale. In regions with a relatively high recharge amount-like in the Sudanian subsahelian zone in Africa (Wubda et al. 2017, Babaye et al. 2019 or in Ireland (Cai and Ofterdinger 2016)-direct recharge occurs either through the saprolite (or the SFL if it outcrops) and/or reworked superficial formations (Cai and Ofterdinger 2016). In regions with a low recharge amount (like in the Sahelian zone in Africa), indirect recharge prevails. The recharge occurs more in the lower parts of the watershed, in the flats or "bas-fonds", for two main reasons: 1. The highest parts of some watersheds can have thicker remains of low-hydraulic-conductivity clayey alloterites (or even duricrust) that inhibit infiltration and allow runoff that accumulates in the flats as seasonal ponds. 2. In arid environments, aquifer recharge is low. Effective recharge to aquifers occurs in valley bottoms where runoff accumulates into temporary ponds during the major rainfall events (Leblanc et al. 2008). Elsewhere in the watershed, direct recharge is very low to nil. Wubda et al. (2017), in Burkina Faso and Benin, show that the infiltration does not take place at the center of the pond to the aquifer, but occurs laterally in its banks.
Under "natural" conditions, most groundwater flow results from recharge in the upper aquifer layer (saprolite). Then the water flows to ephemeral or perennial streams or spring, through the saprolite (Fig. 9a). Due to compartmentation of the SFL, its "regional" hydraulic conductivity is poor; only a small proportion of the recharge reaches the deepest fractures (Alazard et al. 2016). Therefore, aquifer groundwater age increases with depth (Armandine Les Where the saprolite is thin (due to its erosion), as on the granites of central India (Alazard et al. 2016), the aquifer recharge process comprises several stages with (1) mostly vertical diffuse piston flow in the saprolite as well as preferential recharge fluxes through preserved fractures in the saprolite and, below, (2) preferential deep horizontal flow in the SFL. Direct vertical percolation is limited to the recharge period (monsoon); then newly recharged water mixes with groundwater in the aquifer both vertically and horizontally. Different "water bodies" of varied hydrochemistry can then be distinguished within the aquifer, evolving according to the hydrological conditions and recharge structures. In wet regions where the saprolite has been removed, for example due to glacial erosion, the stratiform fractured layer is often very shallow; significant discharge of interflow can originate from this layer.
Pumping groundwater from the SFL radically modifies flow patterns in the aquifer, as groundwater from the upper horizon, the saprolite, leaks vertically downward towards the pumped SFL and flow directions within the SFL can be drastically modified (Fig. 9b). The groundwater chemistry and ages are modified (rejuvenation; see for instance Roques et al. 2018), as the natural vertical water stratification described previously is altered. Incidently, mixing of subsurface groundwaters with deeper ones due to pumping enhances denitrification processes, where there are any (Roques et al. 2018).
Within a 50-km 2 granite watershed in Andrah Pradesh (now Telangana), India, Dewandel et al. (2012) evaluated the connectivity of the SFL fractures and aquifer compartmentalization in areas of about 650 × 650 m. However, the number of permeable fractures and their connectivity in the SFL decrease with depth (most of the fractures in granites are subhorizontal, see section 'Geological structure and hydrodynamic properties of weathering profiles'); then Guihéneuf et al. (2014) show that, depending on the elevation of the piezometric level, the aquifer can shift from a "watershed flow system" to an independent local flow system. Under high piezometric-level conditions, the interface including the bottom of the saprolite and the first flowing fractured zone in the upper part of the granite controls groundwater flow at "watershed-scale" (or at the scale of smaller areas: see Dewandel et al. 2012). Under low piezometric-level conditions, the aquifer is characterised by lateral compartmentalization due to decrease in the number of permeable fractures with depth and their connectivity. Then, individual compartments each exhibit a diameter of about 100 m only. This phenomenon, not presented in Fig. 9a,b, has important implications for watershed hydrology, groundwater chemistry, aquifer vulnerability and pollutant transfer.

Operational hydrogeological applications based on this conceptual model
Several practical hydrogeological applications (and for applied geosciences) are derived from these concepts and the mapping methodology. This new genetic conceptual model of HRA leads to a revision of the methodologies used for characterizing the structure and functioning of such aquifers.
Another asset is the improvement of the efficiency of some of these methodologies, particularly geophysical methods. Two-dimensional electrical profiling now allows a rather precise imaging of the complete weathering profile (Belle et al. 2016(Belle et al. , 2017Soro et al. 2017). Magnetic resonance soundings allow estimation of HRA hydrodynamic parameters (specific yield and transmissivity; Vouillamoz et al. 2014;Wyns et al. 2004). The following strategy for siting a water well relies on these concepts.

Strategy for water-well siting in hard-rock aquifers
The strategy for siting a water borehole is improved and depends on the following cases: 1. Generally, where the SFL is not eroded, it will efficiently be tapped by vertical boreholes targeting the subhorizontal (in granite) or randomly dipping (in most metamorphic rocks) fractures of the SFL (see Fig. 1). 2. In some cases, where the SFL has been removed by erosion, or is unsaturated, it is more efficient to target a vertical fractured layer located along the walls of a subvertical discontinuity (Fig. 7), but avoiding its inner saprolite core, and large weathered corridors (described in section '"Deep vertical" discontinuities in the unweathered basement'; Arias et al. 2016;Belle et al. 2016Belle et al. , 2017. In the first case (saturated SFL), the survey focuses on the SFL preserved from erosion, not clogged by weathering or diagenesis, with saturation (below the piezometric level). Vertical boreholes are used to reach the subhorizontal fractures of the granite SFL or the unevenly dipping fractures of the metamorphic rock SFL. The siting of hand-dug wells uses similar principles, looking for the coarse lower part of the saprolite and the underlying top of the SFL, which are rather easy to dig (Bakundukize et al. 2016).
In the second case (no or unsaturated SFL), techniques for well siting such as lineament analysis, geophysics and radon surveys (Lachassagne et al. 2001b;Reddy et al. 2006), can be used. As weathering is also the main driver of aquifer development in this context, linking open discontinuities to past or recent tectonic events and to recent or paleo stress-field data is not of interest, as indirectly shown by Holland and Witthüser (2011). This is because weathering develops along all heterogeneities whatever their orientation. Moreover, in most cases, regional plate-scale stresses do not influence fractures in the first hundreds of metres below the surface; topographic factors predominate (see references cited by Lachassagne et al. 2011). Still in this second case (no or unsaturated SFL), inclined boreholes are more efficient than vertical ones at increasing the probability of crosscutting a vertical structure, to intersect it. Such a well-located inclined borehole will avoid cross cutting only the low-hydraulic-conductivity saprolite core of the structure and will cross cut its two permeable surroundings.
In both cases, as explained in the preceding (for the SFL in section 'Geological structure and hydrodynamic properties of weathering profiles'), the success of water-well siting also relies on the financial and technical feasibility of drilling several wells (at least 2 or 3) to account for the possible drilling of a first dry well (and eventually a second) that would (case 1) only have crosscut impervious fractures of the SFL or (case 2) only crosscut impervious fractures along the subvertical discontinuity.

Mapping hydrogeological potentialities and waterwell siting in hard-rock aquifers
Water-borehole siting techniques in hard-rock aquifers, for population, agricultural and industrial water supply, have been developed based on the concepts and research outcomes summarised in the preceding. These rely on physical considerations about weathering and its hydrodynamic consequences on the SFL hydraulic conductivity (section 'Most favorable criteria and drivers for good aquifer properties') and on statistical and field observations (section 'Mapping hydrogeological potentialities and water-well siting at various scales: regional to country scale approach'). The SFL hydraulic conductivity is one of the main drivers of a borehole's productivity. Siting productive boreholes, with sufficient discharge threshold, is a challenge in HRA where many unsuccessful dry boreholes are drilled. The long-term exploitable discharge of a borehole or the sustainable abstraction from an aquifer relies on other parameters such as the available water resource. This issue is addressed further in this paper (section 'Quantitative management and modeling of HRA groundwater resources at the watershed scale: managed aquifer recharge').

Most favorable criteria and drivers for good aquifer properties
The mineralogy, texture and structure of HR influences the weathering and consequently the hydrodynamic properties of the HRA. This has been described in detail in Lachassagne et al. 2014b, and is summarised in the following: 1. Mineralogy. Biotite is the most common mineral that is responsible for creating fractures in hard rocks during weathering. It is found in many plutonic rocks (granitoids) and metamorphic rocks (micaschists, paragneisses and orthogneisses). The abundance of biotite fosters the fracturing. White micas also swell during weathering, but later than biotite; the weathering of muscovite and K-felspar begins when the rock has already been transformed into sandy saprolite. Therefore, the weathering of muscovite plays no part in the fracturing of hard rocks.
In some white mica-bearing rocks such as sericitoschists, the weathering of sericite can contribute to fracturing. Feldspars do not swell during weathering even if the transformation of plagioclase into clay minerals begins as early as the biotite transformation. In mafic and ultramafic rocks such as peridotites, weathering conduces swelling of olivine and pyroxene and consequently facilitates intense fracturing of the rock (Dewandel et al. 2005(Dewandel et al. , 2017cRudge et al. 2010). Secondarily, pseudo-karstification may develop in such hard rocks below the saprolite cover mainly made up of limonite (Genna et al. 2005). 2. Texture. During weathering, the pressure caused by swelling minerals depends upon mineral size. Therefore, hardrock fracturing is easier in coarse-grained rocks than in microgranular ones; hence, microgranites containing biotite are less fractured than granites. 3. Structure. During weathering, the swelling of biotite minerals occurs perpendicular to the cleavage. The layer spacing is changed from 10 Ǻ to 14 Ǻ as biotite is altered to hydrobiotite then to vermiculite. This is caused by the exchange of ions (K + ) initially between layers by larger, hydrated cations (Mg 2+ , Na + , Mn 2+ , Ca 2+ ) and water molecules. Swelling pressure exerted perpendicular to cleavage and the resulting stress ellipsoid shape depend on the orientation of the minerals. Thus, the rock permeability tensor is dependent on the swelling mineral(s) orientation; see Lachassagne et al. 2014b for a detailed description and figures presenting the various cases summarized here: a. When swelling minerals are almost randomly oriented, as in plutonic rocks, the theoretical deformation ellipsoid is isotropic. Vertical stress increases until the lithostatic charge is offset, whereas horizontal stresses continue to increase. Horizontal tension joints are created when the difference between minor and major stress components reaches the elastic limit. The resulting reservoir properties (permeability, porosity) are good, and the main permeability component is horizontal. b. When the swelling minerals are vertically oriented (e.g. gneiss with vertical folitation), the deformation ellipsoid is anisotropic, with the main axis orthogonal to mineral orientation (i.e., horizontal); the maximum stress component has horizontal orientation, and the resulting tension joints are horizontal and generally largely open. The resulting reservoir properties (permeability, porosity) are good, and the major permeability component is horizontal. c. When the swelling minerals are horizontally oriented, the swelling potential is vertical, and the horizontal stress is slight. In this case, there are no or few open tension joints, but the rock acquires a horizontal laminated structure. The resulting reservoir properties are poorer. d. In metasedimentary folded rocks (such as schists, micaschists, gneisses) the orientation of swelling minerals varies laterally, and the bedding creates lithological preferential brittleness planes. Breaking joints are created in random directions. 4. Synthesis. For plutonic rocks, granular, biotite-bearing granites, granodiorites and diorites (as with unmetamorphosed gabbros) exhibit good hydrodynamic properties within their SFL. For metamorphic rocks, orthogneisses, paragneisses and micaschists can exhibit good hydrodynamic properties if their foliation dip is near-vertical, but poorer when foliation dip is near-horizontal. Consequently, biotite-rich rocks (such as biotite-granites, biotite-gneisses or micaschists) show a well-developed SFL, with a relatively high hydraulic conductivity and thick saprolite, whereas biotite-poor rocks such as leucogranites, are poorly weathered/fractured. This results in a granitic landscape where leucogranite intrusions in biotite-granite form relatively unweathered hills, tens of metres high, standing above the weathered and eroded biotitegranite cover. A similar landscape results from differential weathering of metamorphic rocks where unweathered quartzite ridges steeply emerge from flat weathered schists or micaschists. In granite-type rocks, a higher-hydraulicconductivity SFL develops in coarse-grained rocks such as porphyric or pegmatitic granite, whereas fine-grained granites, like aplitic granites, develop a lower-hydraulicconductivity SFL. In foliated rocks, the fracturing intensity is maximum where the foliation is subvertical, whereas it decreases and reaches a minimum where the foliation is subhorizontal. Peridotites, which are considered as HR by several authors, can form good aquifers; however, they show a dual aquifer behavior, pseudo-karstic and fractured (Dewandel et al. 2005(Dewandel et al. , 2017cJoin et al. 2005;Jeanpert et al. 2019).
Mapping hydrogeological potentialities and water-well siting at various scales: regional to country scale approach In many hard-rock aquifer studies, the blow discharge of a borehole measured during and on completion of drilling has been regarded as a robust indicator of the local hydraulic conductivity/transmissivity of the aquifer, and thus of the SFL (or at least of the part of the SFL cross cut by the borewell). The blow discharge is, most of the time, obtained at the end (or during) drilling with a down-the-hole hammer. It consists of measuring the discharge of the well after a few minutes of air injection (to stabilize the discharge); the hammer is slightly lifted so that it is no longer in the position of drilling. The blow discharge is more rarely obtained with a specific device such as an air-lift system. The blow discharge is also regarded as a good indicator of the future exploitable discharge of the well, particularly for village water supply where (low) discharges of >0.5 m 3 /h are required (Lenck 1977), and where the aquifer-scale water resource is not (yet) an issue. The depth to the base of the saprolite can also be measured with reasonable accuracy during drilling, by a geologist or hydrogeologist, or by the driller. It is determined accurately as there is always a strong hardness difference between the (unconsolidated) saprolite and the underlying hard SFL. The total drilled depth of a borehole is also accurately measured and recorded; however, sometimes, there can be some recurrences of this interface, notably when the top of the SFL comprises boulders (corestones). These data (discharge, depth to the base of the saprolite, total depth of the well) are often reported and stored in national databases (Courtois et al. 2010). Statistics computed from these data, particularly those gathered in Africa during large village hydraulics programs, are used to infer, at regional scale, the hydrodynamic properties of the various lithologies underlying a study area (see for instance Lenck 1977). These statistical methodologies have been refined using the presented conceptual model (Courtois et al. 2010). In fact, the hydraulic conductivity of a borehole, or the blowing discharge, strongly correlates with the depth of the well below the base of the saprolite. The "linear discharge" (Courtois et al. 2010) is defined as the blowing discharge/the depth of the well below the base of the saprolite (Fig. 10). Then, (1) the boreholes that only cross-cut the saprolite have a low to nil discharge, and (2) the shallowest boreholes that reach the SFL cross-cut the more densely fractured part of the SFL and provide a higher linear discharge than the deeper ones that also cross-cut less-or unfractured levels.
The statistical processing of the linear discharge parameter, for each lithology, allows computation of two further important parameters (Courtois et al. 2010): 1. The thickness of the productive part of the SFL-in Burkina Faso for instance this thickness can be 35-40 m depending on the lithology (Fig. 10).

Hydrodynamical parameters such as statistical parameters
about the potential discharge of the boreholes to be drilled in a given area: mean, median, standard deviation, percentage of "dry" boreholes below a discharge threshold, etc.
Such results have direct practical uses. In association with the thickness of the saprolite (Fig. 6b), the results provide an estimation of optimal borehole depth, i.e. the depth beyond which it is not necessary to drill, as the small gain in discharge does not justify the increased drilling cost. Hence, the results are of interest for planning (duration) and for assessing the financial cost of drilling campaigns (cumulative length amount and depth of the wells to be drilled, including numbers of "dry" boreholes) including the preliminary geophysical surveys that can also be precisely sized. They provide an estimated minimum depth for water wells necessary to cross the saprolite and reach the SFL, and as such can be used to evaluate the drilling constraints, i.e. the technical characteristics of the drilling rig and borewell equipment. Other applications can be inferred from the results such as mapping the mean discharge of boreholes at regional/country scale for different depths and correlating these results with economic parameters such as drilling costs, and other fields beyond hydrogeology such as town and country planning, soils engineering, the search for areas suitable for quarrying, etc.

Mapping hydrogeological potentialities and water-well siting at various scales: local scale
Such results at large scale are also of interest for local-scale hydrogeological projects. However, available data from wells may be insufficient to compute robust statistics. In that case, high-resolution mapping (with principles such as shown in Figs. 4 and 5) is needed for water-well siting. By also using information on land-use, land-ownership, conditions of access, etc., favorable zones for geophysical surveys and then exploratory drillings can be selected (Fig. 11).
At a local scale, geophysical methods, especially 2D electrical geophysical methods, rarely delineate junctions between lithologies or enable identification of lithologies (Belle et al. 2016(Belle et al. , 2017. Nevertheless, they are very efficient at characterising the structure of the weathering profile (they can identify the saprolite, the SFL, their thickness and geometry, etc.).
The typical response in granites is, from top to bottom (Belle et al. 2016(Belle et al. , 2017Soro et al. 2017; Fig. 12): 1. High resistivity, up to several hundreds to 1,000 ohm-m; occurrence of (unsaturated) saprolite and/or, where preserved, iron or bauxitic duricrust. 2. Lower resistivity, down to a few tens of ohm-m or lower; this indicates clayey saprolite and/or the saturated clayey saprolite.
3. Steadily increasing resistivity within the sandy saprolite (isalterite) and then within the SFL. The fractured layer resistivity is often below 600 ohm-m, and locally can be less than 100 ohm-m. Where present and thick enough, the laminated-layer resistivity is identified by a local increase in the resistivity (up to 600 ohm-m), and then a decrease (to that of the SFL; Belle et al. 2016Belle et al. , 2017, within this increasing trend from the low resistivity saprolite to the SFL and its substratum (Fig. 12). 4. The highest resistivity (up to more than 10,000 ohm-m) is reached within the unweathered/unfractured rock.
As described in section 'Geological structure and hydrodynamic properties of weathering profiles', Dewandel et al. (2012Dewandel et al. ( , 2017bDewandel et al. ( , 2017c developed a method to map the spatial variations of hydrodynamic properties (hydraulic conductivity, specific yield) at the aquifer scale. Mapping can also use electrical geophysical data (Chandra et al. 2008). Vouillamoz et al. (2014) used magnetic resonance soundings (MRS) to estimate the specific yield and the transmissivity of HRA in Bénin, as otherwise determined from pumping tests (same scale). The MRS water content is about twice the specific yield. They then propose two linear relationships for calculating the specific yield from MRS water content, with an uncertainty of 10%, and from the pore-size parameter, with an uncertainty of 20%. The latter identifies nondrainable MRS water content and, thus, low groundwater reserve. Aquifer transmissivity is estimated using MRS with an accuracy of 70%. High-resolution mapping (with principles such as shown in Figs. 4 and 5) is used for the delineation of groundwater protection zones, and also for other applications such as the location of landfill sites on the upper saprolite where rock weathering forms a clayey layer, and the location of rock quarries with a minimum cover cap, preferably where the weathering profile is eroded or at the boundary between successive weathering profiles ( Fig. 11; section 'Other applications in applied geosciences').

Delineating groundwater watersheds and piezometric mapping in HRA
Piezometric heads in HRA well mimic the surface topography (see also section 'Hydrogeological functioning of HR aquifers: hydrochemistry'). Consequently, delineating groundwater watersheds from topographic data (groundwater watershed = surface-water watershed) is possible.
Using topographic data, and location and elevation of perennial stream water levels, with piezometric measurements, allows computation of piezometric maps (Wyns et al. 2004;Yao et al. 2015), even with limited data. Methodologies for the optimisation of piezometric networks in HRA have been proposed by Zaidi et al. (2007).
Quantitative management and modeling of HRA groundwater resources at the watershed scale: managed aquifer recharge Several methods have been developed for water resources management of HRA using, as input or calibration data (Dewandel et al. 2010(Dewandel et al. , 2017b: (1) the concept of stratiform layers, (2) the relative homogeneity of HRA hydrodynamic properties at a scale of a few 100 m, beyond the local scale of a single well (see for instance Dewandel et al. 2014Dewandel et al. , 2018Lachassagne et al. 2016) that can be dry , (3) the decrease of the specific yield with depth, as well as (4) the spatial variations of the specific yield. Maréchal et al. (2006) show that the combined use of water-table-fluctuation and groundwater-budget methods is appropriate at the watershed scale and relevant for this purpose; for instance, Chinnasamy et al. (2018) also Fig. 11 Various practical applications of a high-resolution mapping of the weathering cover at the local scale (see Fig. 4 for the legend and section 'Geological structure and hydrodynamic properties of weathering profiles'; from Courtois et al. 2003). The five orange ellipses represent potential targets, of various interest, for water-well siting use this combined method. Similar methods have been applied to overexploited HRA in India Dewandel et al. 2007bDewandel et al. , 2010, even through use of soil and water assessment tools (SWAT models) that are applicable in such a hydrogeological context where groundwater vertical fluxes prevail over regional lateral flow at km-scale (Ferrant et al. 2014;Perrin et al. 2012). Such models can be used for water resources mapping and simulation of scenarios such as land use changes, changes in water uses and climate change. Remote sensing data are also crucial to provide cheap and efficient input or calibration data for these models (see for instance Ferrant et al. 2017 for the context of Indian HRA). Lachassagne et al. (2016) forecast daily piezometric variations in a weathered-fractured metamorphic aquifer pumped by a well, taking account of its discharge and pumped aquifer characteristics, including recharge. They developed a simple lumped reservoir model (1) based on the hydrodynamic parameters, and shape and size of the aquifer diagnosed from the early months of time series discharge/piezometric level (D/PL) data from that well, and (2) calibrated from the D/PL time series data. The computed aquifer recharge time series appears realistic.
The demonstration that HRA are constituted by rather homogeneous layers (saprolite and SFL) at least at the scale of a few hundreds of meters allows their deterministic mathematical modelling with "classical models" similar to those used for sedimentary aquifers modelling (see for instance Lachassagne et al. 2001a;Leray et al. 2013;Baiocchi et al. 2014;Durand et al. 2017;Dickson et al. 2018). This was performed for water resources evaluation and to propose management scenarios.
The evaluation of managed aquifer recharge (MAR) from percolation tanks is assessed by Perrin et al. (2012), Boisson et al. (2014), Alazard et al. (2016), andNicolas et al. (2019). Recharge is delayed in the aquifer due to slow groundwater flow; the unsaturated saprolite acts as a buffer to aquifer recharge. Due to the compartmentalization of the aquifer, and despite a large number of pumping boreholes surrounding the tank, the radius of influence of the percolation tank is small, and only a few boreholes benefit from the tank artificial recharge. The characteristics (geometry, hydraulic conductivity) of the SFL fractures drive the lateral propagation of the aquifer recharge (Nicolas et al. 2019). MAR appears to be beneficial for groundwater quality, particularly with respect to reducing fluoride contamination (Pauwels et al. 2015;Pettenati et al. 2014). Assessing HRA groundwater reserves from watershed to regional scales These new concepts allow assessment of HRA groundwater reserves from the watershed to the regional scale through combined use of the 3D geometry of the weathering profile and the vertical distribution of effective porosity determined by magnetic resonance soundings (MRS) (Wyns et al. , 2004Baltassat et al. 2005;Vouillamoz et al. 2014Vouillamoz et al. , 2015 or its evaluation using 3D aquifer parameter regionalization methods (Dewandel et al. , 2017bMizan et al. 2019).
Magnetic resonance soundings allow determination of the mean effective porosity of the three parts of the aquifer: saprolite, upper fractured layer, and lower fractured layer. For a given lithology, several MRS are used to determine the mean porosity of each layer. The saturated thickness of each layer, multiplied by the porosity, gives the water content for each cell expressed in water height (Figs. 14 and 15). The total groundwater reserve is given by the addition of the water height of the three layers of the aquifer. Vouillamoz et al. (2015) estimate groundwater storage in Benin metamorphic rocks to be between 200 and 1,200 mm. At a 270-km 2 study area located in Brittany, France, underlain by Hercynian plutonic and metamorphic rocks (Wyns et al. 2004), values range between 100 and 1,500 mm-as proposed by Vouillamoz et al. (2015); the results from Wyns et al. (2004) were divided by 2 to account for calibration by pumping tests. The most recent and advanced methods for estimating groundwater reserves probably come from the use of groundwater-budget and water-table-fluctuation techniques, combined with satellite remote sensing methods (Dewandel et al. , 2017bFerrant et al. 2017; see also section 'Geological structure and hydrodynamic properties of weathering profiles'). Such methods were combined with the knowledge of the weathering profile geometry and geostatistical approaches for regionalizing. These methods allow evaluation of the 3D distribution of the effective porosity at the aquifer scale (Dewandel et al. , 2017b and the recharge, along with its spatial distribution (Mizan et al. 2019). Some recent studies (e.g., Chinnasamy et al. 2018) do not consider the vertical structure of the HRA and find less precise values of specific yield, which may yield inaccurate forecasts. Figure 14 shows, for two large watersheds (more than 700 km 2 ), how the effective porosity varies on average with depth into the weathering profile. Then, with data on water-table depth, it allows computation of the groundwater reserve at the watershed scale, and also establishes scarcity and vulnerability maps with respect to pumping, for assessing the areas most vulnerable to groundwater overexploitation (Fig. 15).
Protecting and restoring HRA groundwater resources, groundwater quality modelling, and contaminant transfer As discussed in the preceding (section 'Hydrogeological functioning of HR aquifers: hydrochemistry', dealing with HRA piezometry), the delineation of groundwater protection zones for water wells relies on knowledge of the geometry and hydrodynamic properties of the exploited HR aquifers (particularly the presence and extension of the clayey saprolite that can prevent local (pollutants) infiltration), and thereby ensure groundwater protection (section 'Hydrogeological functioning of HR aquifers: hydrochemistry' dealing with recharge of HRA). Knowledge of geometry and hydrodynamics also yields more precise piezometric maps that can be used to identify flow systems and subwatersheds (section 'Delineating groundwater watersheds and piezometric mapping in HRA').
Evaluation of HRA groundwater reserves (section 'Assessing HRA groundwater reserves from watershed to regional scales'), taking into account the recharge of the aquifer, allows for calculation of the average residence time of groundwater in the aquifer, and the duration of residence of non-point-source pollutants such as diffuse nitrate contamination. For the sites studied in Benin, Vouillamoz et al. (2015) compute a "buffer capacity" of 6 ± 4 years.
The methodologies developed for the quantitative management of HRA groundwater resources (section 'Quantitative management and modeling of HRA groundwater resources at the watershed scale: managed aquifer recharge') can be applied to the management of groundwater quality. Perrin et al. (2011b) designed a lumped reservoir model to simulate long-term piezometric and solute concentrations in a 53-km 2 watershed in India. They then reproduced chloride concentrations and show that, in that case study in a semiarid context and overexploitation for irrigation, solute recycling will induce long-term aquifer salinisation. In fact, evapotranspiration impacts the quality of the return flow from irrigation, through a higher concentration of salts in recharged water and soils, and a higher enrichment in stable isotopes of the water molecules that are successfully used to quantify such processes (Négrel et al. 2011). The same evapotranpiration processes induce fluoride concentration in aquifers prone to this geochemical influence (Pettenati et al. 2013;Pauwels et al. 2015). Paddy fields show large irrigation return flows (Dewandel et al. 2007a), but a higher rate of evaporation at the surface, due to their inundation, than other crops (Négrel et al. 2011). Perrin et al. (2012) simulated the impact of various scenarios on the groundwater quality (e.g., higher pumping The methodologies developed for the regionalisation and mapping of HRA hydrodynamic parameters such as specific yield and hydraulic conductivity (Dewandel et al. , 2017b, are of interest in terms of contaminant transport evaluation and modelling in such a hydrogeological context. Taylor et al. (2010) provide some estimates of saprolite dispersivity, and Guihéneuf et al. (2017) characterize the solute transport properties of the fractured layer of an Indian granite aquifer; they notably confirm that non-Fickian dispersion dominates.
Computing the drainage discharge and the hydrogeological and hydrological impacts of tunnels drilled in HR The drainage discharge and the surface impacts of shallow tunnels in HR were studied in the 4,000-m-long Violay Tunnel, France (Lachassagne et al. 2014c; Fig. 16). Hydrogeological impacts include piezometric drawdown in wells and drying up of springs; hydrological impacts include the decrease of stream discharge. These impacts have been forecasted applying the HRA concepts presented in this review paper. The depth of the tunnel was 0-300 m below the  (Fig. 16b,d) corresponds to the amount of net groundwater abstraction (in m 3 /year). From Dewandel et al. (2017b). With permission from Wiley ground surface, and consequently within and below the weathering profile.
The following computations were performed (Lachassagne et al. 2014c): 1. Characterisation of the location of the tunnel within or below the various layers constituting the weathering profile, which was polyphased and locally deepened by various deep veins and ancient faults. 2. Measurement of the steady-state groundwater discharge in existing tunnels in the same area and associated lithologies (and their location within the weathering profiles). 3. The analytical solutions for tunnel inflow.
The actual discharge of the now drilled tunnels, within the predicted range, validates the efficacy of the methodology. The impact of the recent shallow weathering profile is evidenced near the entrances of the tunnel (Fig. 16), as already shown by several other authors, even if they did not all relate this to a weathering profile. In a more subtle way, discharge variations related to the depth of the tunnel below the base of the saprolite are also shown (see the right section of Fig. 16).

Thermal and mineral water
Most of the thermal and mineral water occurrences, including carbo-gaseous ones, do emerge from outcropping hard rocks. In fact, a sedimentary cover may hinder upwards vertical flows (water, gas) from deep origin. Additionally, most thermal and mineral springs are a mix of such deep waters (and gas in some cases) and groundwater from near-surface aquifers. Then, interaction within the superficial HRA between these two types of waters is common; in the superficial HRA, water from the overlying saprolite mixes with deep thermo-mineral water .
The edges of grabens (Dewandel et al. 2017a) often show well-preserved weathering profiles, as the dominant process during the creation of the basin was sedimentation at these locations, not erosion. These profiles can form thick capacitive aquifers, fed by upwelling deep mineral water rising from a few fractures (notably the graben normal faults), and by downflowing low-mineralisation waters from the HRA (Fig. 17). The SFL of the HRA acts as the main aquifer and it is easily reached by production boreholes.
Generalisation of these observations reveals strong opposite characteristics between (1) low-inertia crystalline thermomineral (carbo-gaseous) hydrosystems without a subsurface reservoir (as the weathering profile was eroded) and (2) high-inertia ones as already just described. The former generally provide highly mineralized water, but at a very low discharge (Ledoux et al. 2016), whereas the latter provide more diluted water, but with a much higher discharge (Dewandel et al. 2017a).
The chemical reactions involved in the weathering process of rock fracturing are exothermic. This exothermic weathering process has been invoked as a possible hypothesis to explain geothermal anomalies (water warmer than expected) observed in some HRA, notably in the western part of France (Wyns et al. 2015), in quite shallow wells (about 100 m deep). However, Vasseur and Lachassagne (2019) demonstrated that Fig. 16 Violay (Loire Department, Massif Central, France) tunnel computed discharge Q 0 (blue crosses -L/s/50 m of tunnel) according to the lithology, the weathering grade and the depth of the tunnel (black line) below the topographic surface (brown profile). The highest-discharge areas (purple circles) are located at both entries of the tunnel, and in the Gantet River Valley, where the tunnel crosscuts the SFL of the weathering phases 2 and 3. The other parts of the tunnel are either within or below the phase 2 SFL. In the first kilometre of the tunnel, the discharge change is due to a lithological change. This computation provides a 4.8 L/s discharge for the whole tunnel. The estimated surfacewater loss in the Gantet watershed is about 2.7 L/s (in red). From Lachassagne et al. (2014c) this exothermic process of weathering has no significant geothermal impact at the aquifer scale, and at the weathering timescale. In fact, heat diffusion and advection quickly evacuate the small amount of generated heat, and so the temperature cannot increase. Thus, the process of HR weathering cannot explain these geothermal anomalies. More classical explanations such as the local upwards flow from deep fractures, should then be invoked.

Other applications in applied geosciences
Prior to improvement and development for hydrogeological application, weathering profile concepts have indicated geodynamic evolution (Wyns 1991) within geological mapping (Wyns et al. 1998), and aided radioactive-waste-storage safety analysis (Wyns 1991(Wyns , 1997. These concepts can help locate and identify faults in HR, as layers and interfaces in the weathering profile, to reveal the location and throw of faults (Cho et al. 2002;Lachassagne et al. 2001c).
Applications have been developed in the investigation of soil and rock mechanics to forecast mechanical properties for: the subsoil burial of electrical and telephonic cable networks at 1/25,000 to 1/100,000 scale (Wyns et al. 1999; quarrying in HR (Rocher et al. 2003;Fig. 11); evaluation of the thickness of the cover cap; the determination of optimal geo-mechanic quality of the materials, etc. Future applications will include preliminary surveys for highways and railways. The HRA conceptual model has been applied to hydrocarbon reservoirs in crystalline rocks (Wyns 2015), and it is acknowledged that some hydrocarbon reservoirs are found in HRA (see for instance Riber et al. 2017 andTan et al. 2017).

Conclusion and perspectives
The hydraulic conductivity and storativity of HR are mostly inherited from weathering processes. Consequently, HRA develop mostly within the first 100 m below ground surface. Where not partially or totally eroded, they comprise: (1) a capacitive but rather low-permeability unconsolidated weathered layer, the saprolite, located immediately above (2) the permeable stratiform fractured layer. To a lesser extent, weathering and hydraulic conductivity develop deeper in the unweathered rock, at the periphery of or within preexisting geological discontinuities (lithological contacts, joints, dykes, veins, etc.). The demonstration and recognition of this conceptual model allowed development of a comprehensive corpus of applied methodologies in hydrogeology and geology, which are described in this review paper: water well siting, mapping hydrogeological potentialities from the local to the country scale, quantitative management, modeling and protection of HRA groundwater resources, even in thermal and mineral aquifers, computing the discharge of tunnels, quarrying of HR, etc.
The properties of such HRA have been described and understood in granites, and several applications have been developed. Metamorphic rocks, being more complex and heterogeneous, will require more detailed characterisation in the future using applications developed for plutonic rocks.
One of the other challenges with respect to both plutonic and metamorphic rocks is to develop geophysical method(s) able to identify if (or where) the fractures of the SFL are permeable, from the surface, i.e. locally down to >50 m deep. This would be of great interest for water borehole siting. In fact, 3D high-resolution permeable-fracture imaging has already been achieved (see for instance Robinson et al. 2016), but only from drillings, and not from the surface.

Acknowledgements
This paper synthesizes more than 20 years of research projects on hard-rock aquifers, involving several scientists and collaborators; many of these projects were undertaken in France and India. Beyond the references cited in this paper, the list of all these contributors would be too long; however, we hope that they will all recognize themselves and will accept these acknowledgements.
The leader of this issue's topical collection on hard-rock aquifers, Uwe Tröger, and an anonymous reviewer, are also warmly thanked for their careful review, which helped improve the manuscript before its formal submission to Hydrogeology Journal. We thank Prof. Maurizio Barbieri and two more anonymous reviewers who also formally reviewed this paper.
We also want to thank the organizers of the International Conference on Groundwater in Fractured Rocks (GwFR'2017) organized by the International Association of Hydrogeologists (IAH) 'Network on Fractured Rock Hydrogeology' and the Portuguese and Spanish Chapters of IAH in Chaves, Portugal, in June 2017, and notably Raquel Sousa who encouraged us to write this review paper after our Conference's oral presentations (notably the following presentation: Lachassagne et al. 2017). As this paper was submitted in November 2017, it was later completed with additional recent references in August 2019, during its revision; recommendations from the reviewer were received in June 2019 and the revision completed in September 2019, before the last formal submission to Hydrogeology Journal in August 2020.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.