A Rusty Record of Weathering and Groundwater Movement in the Hyperarid Central Andes

The Atacama Desert, on the western margin of the Central Andes, hosts some of the world's largest porphyry copper deposits (PCDs). Despite a hyperarid climate, many of these PCDs have undergone secondary “supergene” enrichment, whereby copper has been concentrated via groundwater‐driven leaching and reprecipitation, yielding supergene profiles containing valuable records of weathering and landscape evolution. We combine hematite (U‐Th‐Sm)/He geochronology and oxygen isotope analysis to compare the weathering histories of two Andean PCDs and test the relative importance of climate and tectonics in controlling both enrichment and water table movement. At Cerro Colorado, in the Precordillera, hematite precipitation records prolonged weathering from ∼31 to ∼2 Ma, tracking water table descent following aridity‐induced canyon incision from the late Miocene onward. By contrast, hematite at Spence, within the Central Depression, is mostly younger than ∼10.5 Ma, suggesting exhumation ended much later. A heavy oxygen isotopic signature for Spence hematite suggests that upwelling formation water has been an important source of groundwater, accounting for a high modern water table despite persistent hyperaridity, whereas isotopically light hematite at Cerro Colorado formed in the presence of meteoric water. Compared with published paleo‐environmental and sedimentological records, our data show that weathering can persist beneath appreciable post‐exhumation cover, under hyperarid conditions unconducive to enrichment. The susceptibility of each deposit to aridity‐induced water table descent, canyon incision and deep weathering has been controlled by recharge characteristics and morphotectonic setting. Erosional exhumation, rather than aridity‐induced water table decay, appears to be more important for the development of supergene enrichment.

. For weathering and enrichment to progress, tectonic or climatic factors must trigger relative water table descent for fresh rock to be exposed to oxidation and leaching (Ague & Brimhall, 1989;Alpers & Brimhall, 1988;Anderson, 1982;Brimhall et al., 1985). This could be caused by uplift of rock through the water table (Sillitoe, 2005) (hereon referred to as "exhumation-driven water table descent," where the water table is the reference datum), or climate-induced canyon incision or reduction in aquifer recharge (Cooper et al., 2016), lowering the water table (hereon referred to as "water table decay," where the surface is the reference datum). Understanding of the relative contribution of these factors to water table movement during the development of preserved weathering profiles is limited (García et al., 2011), and as mineral deposits become harder to find, greater knowledge of the controls on paleo-water tables in different settings will be essential for future exploration.
Previous studies have constrained the timing of enrichment of Andean PCDs via K-Ar and 40 Ar/ 39 Ar dating of supergene alunite group minerals; K-bearing sulfates formed as by-products of sulfide leaching (Alpers & Brimhall, 1988;Arancibia et al., 2006;Bouzari & Clark, 2002;Sillitoe & McKee, 1996). A major limitation on the usefulness of alunite is that it can form under oxidising conditions above the water table (Sillitoe & McKee, 1996) and under reducing conditions beneath it (Alpers & Brimhall, 1988), implying that samples separated by hundreds of meters within a weathering profile may have formed "anytime from several million years apart to synchronously" (Sillitoe, 2005). Therefore, a more redox-sensitive indicator mineral is required to understand the relationship between water tables, enrichment, and the development of PCD weathering profiles, and identify controls on water table movement and groundwater flow in a tectonically and climatically dynamic area.
Iron oxides, such as hematite (Fe 2 O 3 ), are common authigenic weathering minerals and can be dated using (U-Th-Sm)/He geochronology, providing a tool for constraining periods of weathering (Monteiro et al., 2014) and tracking water table movement (Cooper et al., 2016;Deng et al., 2017;Heim et al., 2006), as these minerals, and their metastable precursors, predominantly form under oxygenated conditions at or above the atmosphere-groundwater interface (Heim et al., 2006). The only application of hematite (U-Th-Sm)/He geochronology to an Andean PCD weathering profile, which tracked water table decay at Cerro Colorado in the Andean Precordillera (Figure 1) was conducted by Cooper et al. (2016). Here, geochronology data show that weathering had commenced by ∼31 Ma and persisted until at least ∼2 Ma. An apparent younging-with-depth trend observed after 16 Ma was attributed to aridity-induced incision of a nearby canyon (the Quebrada de Parca) which continues to control the local water table today. These geochronology data show that regional climate can control water table movement and the progression of weathering via canyon incision, and that hematite precipitation persisted long after supergene enrichment ended (at ∼14.6 Ma; Bouzari & Clark, 2002), suggesting that hematite is not a proxy for enrichment. However, the response of the water table in different locations to desiccation of the Atacama Desert, and the impacts of this response on the supergene development of different PCDs are yet to be determined.
This contribution re-examines the history of canyon incision and water table decay at Cerro Colorado (based on satellite imagery of truncated drainage networks and published ages of volcanic deposits contained within the gravel cover proximal to the deposit), and investigates the timing of weathering at Spence, an enriched PCD situated within the Central Depression, ∼300 km south. In contrast to Cerro Colorado, Spence has not been affected by canyon incision and the modern water table is shallow despite persistent hyperaridity (Cameron & Leybourne, 2005). We report oxygen isotopic measurements (δ 18 O) for Fe-oxides from the weathering profiles at Spence and Cerro Colorado, enabling recharge mechanisms and groundwater sources during weathering to be constrained and compared.

PCD Formation and Supergene Enrichment in the Central Andes
PCDs are typically located within arc-parallel belts at convergent plate boundaries, such as the western Andean margin of South America, where magmatism, PCD formation and orogenesis have resulted from subduction of the Farallon and Nazca plates beneath the South American plate from the mid to late Jurassic onward (Armijo et al., 2015;Mpodozis & Cornejo, 2012;Richards, 2013;Sillitoe, 2010). PCDs form when heat and water expulsion from stocks and dykes, emplaced above granitoid plutons at paleo-depths of several kilometers, drive hydrothermal alteration of large volumes of rock and generate disseminated and vein-hosted sulfide mineralization (e.g., pyrite [FeS 2 ] and chalcopyrite [CuFeS 2 ]; Sillitoe, 2010).
Erosional exhumation has brought many PCDs to the near-surface, where copper present in hypogene ("formed from below") orebodies may be concentrated by up to a factor of three by supergene processes (Sillitoe, 2005). In the supergene environment, Cu enrichment is produced via chemical weathering by descending meteoric waters. Oxidation of sulfides at or above the water table forms sulfuric acid, enabling downward leaching of Cu. Under reducing conditions, beneath the water table, Cu is reprecipitated, forming enriched secondary sulfides (e.g., chalcocite [Cu 2 S] and covellite [CuS]; Ague & Brimhall, 1989;Sillitoe, 2005). Where acidity or water availability are insufficient for leaching, in situ oxidation may produce an "oxide zone" containing secondary Cu minerals such as atacamite and brochantite (Sillitoe, 2005;Vasconcelos, 1999). Mature supergene weathering profiles thus comprise an upper, weathered zone, underlain by a sulfide enrichment blanket grading downward into the hypogene orebody. The boundary between the weathered zone and enrichment blanket represents the deepest paleo-water table position-the "ultimate redox front." Although there is evidence supergene enrichment is most effective during periods of landscape stability, characterized by pediplanation and low erosion rates Sanchez et al., 2018), sufficient lowering of the water table for deep weathering is aided by periods of tectonic uplift (Sillitoe, 2005). Between 26° and 27°S, the deeply leached and strongly enriched El Salvador deposit, which experienced multiple phases of tectonic uplift and erosion during the Oligocene and middle Miocene, contrasts with the Potrerillos deposit, which is less deeply exhumed and hosts a less mature supergene profile (Bissig & Riquelme, 2009). This shows the relative importance of climatic and tectonic factors and the extent to which they affect water tables are spatially variable.
Many Andean PCDs are buried beneath Miocene gravels which form regionally extensive paleo-surfaces (Evenstar et al., 2017;Hollingworth, 1964). Cover deposition may end enrichment by stopping precipitation reaching the water table, especially in areas with low infiltration rates and high evapotranspiration such as the Atacama (Davis et al., 2010). Furthermore, burial may cause the ultimate redox front to be "drowned" beneath an elevated water table, so that infiltrating water is moving through rocks already leached of copper (Brimhall et al., 1985;Enders et al., 2006;Sillitoe & McKee, 1996). Ages of paleo-surfaces and ash layers within gravel units allow PCD weathering ages to be viewed within a sedimentological framework of pre-versus post-cover deposition (Bouzari & Clark, 2002).
Hyperaridity in the Atacama Desert is sustained by its latitudinal location within the Inter-Tropical Convergence Zone, the cold Humboldt ocean current moving northward along the west coast of South America, its distance from the Atlantic Ocean, and the effect of the Andean rain shadow, which blocks moisture from the Amazon basin (Houston, 2006b;Houston & Hartley, 2003). Modern mean annual rainfall (MAR) at Cerro Colorado is 20 mm (Jordan et al., 2014) and Spence receives <10 mm (Sun et al., 2018). Estimates for the onset of hyperaridity range from Oligocene (Dunai et al., 2005) to Pliocene (Hartley & Rice, 2005), although it is generally considered to have begun during the middle to late Miocene. Evidence for strengthening aridity in the middle Miocene includes the apparent cessation of PCD supergene enrichment at ∼14 Ma (Alpers & Brimhall, 1988;Sillitoe & McKee, 1996), and increasingly heavy oxygen isotopic compositions of soil 10.1029/2021GC009759 5 of 24 carbonates by 15 Ma (Rech et al., 2019). However, multiple sedimentological and isotopic records indicate climate desiccation, from (semi-)arid to hyperarid conditions, in the late Miocene, between ∼12 and ∼10 Ma, due to a strengthening of the Andean rain shadow (Rech et al., 2019). Gypsum-cemented ("gypcrete") horizons, formed when MAR was <20 mm, are preserved beneath ∼9.5 Ma ignimbrites in the Precordillera and Central Depression (Hartley & May, 1998;Jordan et al., 2014;Rech et al., 2019). Similarly, non-reworked soluble salt horizons in the gravels covering Spence, preserved beneath a 9.47 ± 0.04 Ma ash layer (Sun et al., 2018), suggest precipitation has been insufficient for direct recharge since this time. Isotopic evidence for hyperaridity includes increasingly heavy δ 18 O and δ 13 C signatures in palustrine carbonates in the Calama Basin after 12 Ma (Rech et al., 2010) and higher δ 18 O in travertine deposits at Barrancos Blancas (24°S) after ∼11.5 Ma (Quade et al., 2017). Global datasets suggest that direct recharge is negligible when MAR is <200 mm (Houston, 2009;Scanlon et al., 2006). Therefore, it is unlikely that either Cerro Colorado or Spence have experienced direct recharge since the onset of hyperaridity, especially through volcanic and sedimentary cover. Instead, indirect recharge by precipitation higher in the Andes and deep formation water (saline porewaters and meteoric water with a long residence time) are likely contributors to groundwater at lower elevations in the Precordillera and Central Depression (Hoke et al., 2004;Houston, 2002;Magaritz et al., 1990).

Weathering and Water Tables at Cerro Colorado and Spence
Comparing Cerro Colorado and Spence is useful because, despite their similar hypogene mineralization ages (53.5-50 Ma at Cerro Colorado [Bouzari & Clark, 2002;Tsang et al., 2018] and ∼57 Ma at Spence [Rowland & Clark, 2001]), subsequent exhumation, weathering, and water table movement at each deposit have operated under contrasting tectonic, sedimentological, and geomorphological conditions. At Cerro Colorado, situated 2,600 m a.s.l. within the Andean Precordillera (Figures 1 and 2), (U-Th-Sm)/He hematite ages record persistent weathering from early Oligocene to Pleistocene (∼31-2 Ma; Cooper et al., 2016). By comparison, 40 Ar/ 39 Ar alunite ages suggest the onset of supergene enrichment was coeval at ∼35 Ma but ceased in the middle Miocene at ∼14.6 Ma (Bouzari & Clark, 2002). A minimum age for the end of exhumation is provided by the 19.25 ± 0.43 Ma Tambillo ignimbrite, which covers much of the deposit (Bouzari & Clark, 2002), ruling out regional, exhumation-driven water table descent as the cause of the post-middle Miocene water table decay observed by Cooper et al. (2016). Cover deposition at Cerro Colorado, represented by gravels of the El Diablo Formation, continued until ∼11 Ma (Blanco et al., 2012), when climate desiccation led to surface abandonment (Evenstar et al., 2017) and decreased river discharge led to channel steepening and incision of the Quebrada de Parca (Cooper et al., 2016).
Spence lies within the Central Depression, at 1,700 m a.s.l., along the Antofagasta-Calama Lineament (ACL), a SW-NE-striking crustal-scale fracture which facilitated magma emplacement during formation of the deposit   (Figures 1 and 2). Unpublished alunite ages suggest Spence was enriched between ∼44 and ∼21 Ma (Rowland & Clark, 2001), although it is unclear whether all dated samples were of supergene origin. Spence is covered by 50-100 m of Atacama Gravels (Cameron & Leybourne, 2005). A 9.47 ± 0.04 Ma ash layer, situated 37 m above the gravel-bedrock contact, provides a minimum age for the end of exhumation (Sun et al., 2018) (Figures 2c and 2d). Unlike at Cerro Colorado, which is cut by the Quebrada de Parca, the nearest river is the Rio Loa, 34 km to the north (Figure 1b). Although it has been suggested that Spence lies within the groundwater catchment of the Rio Loa (Jordan et al., 2015), the water table slopes toward the southwest (Cameron & Leybourne, 2005). Furthermore, the Rio Loa did not breach the Coastal Cordillera until the late Pliocene to late Pleistocene, shifting the river base level from >1,000 m a.s.l. in the Central Depression to sea level at the Pacific Ocean, and causing upstream channel incision (May et al., 2005;Ritter et al., 2018). Incision of the Rio Loa is therefore unlikely to have influenced the water table at Spence. In contrast to Cerro Colorado, the modern water table at Spence is elevated relative to the ultimate redox front, approximating the gravel-bedrock contact, despite persistent hyperaridity. Oxygen isotope analysis of local groundwater suggests this is caused by upwelling formation water along the ACL, which acts as a fluid pathway for deep groundwater recharge (Cameron & Leybourne, 2005).

Hematite as a Record of Water Table Movement and Weathering Fluid Composition
Fe-oxide formation during PCD weathering is thought to be caused by oxidation of sulfides at or above the water table (Ague & Brimhall, 1989), with minor sub-water table sulfide oxidation along high-permeability pathways (Lichtner & Biino, 1992). For pyrite, the summary reactions are approximated below (Dold, 2003): Sulfide oxidation (Equation 1) requires free oxygen, sourced either from the atmosphere or from oxygenated groundwater. In arid areas with low permeability and slow-moving groundwater, such as our study sites, this limits efficient weathering to the unsaturated zone, at or above the water table. Oxygen involved in subsequent Fe-oxide precipitation is primarily contributed by water (Equations 2 and 3), enabling the isotopic composition, and source, of groundwater during Fe-oxide precipitation to be determined. If hematite precipitation, via sulfide oxidation, genuinely persisted until more recently than supergene enrichment, liberated S and Cu (the latter from chalcopyrite) likely contributed to the oxide zone mineral assemblages observed at both deposits, which include atacamite (Cu-chloride), brochantite (Cu-sulfate) and gypsum (Bouzari & Clark, 2002;Cameron et al., 2007;Reich et al., 2008). Reich et al. (2009) reported Pleistocene 230 Th-234 U ages for intergrown gypsum and atacamite from several PCDs, including Spence, showing that S and Cu mobility continued until much more recently than proposed by previous models of weathering and enrichment.

Sample Collection, Characterization, and Preparation
At Spence, the hematitic weathered zone varies from 30-100 m thick. Minor hematite also occurs within the subjacent enriched zone, on fracture surfaces and within partially weathered veins which formed high-permeability pathways for descending, weakly oxidising fluids (Lichtner & Biino, 1992). Hematite-bearing veins and fracture surfaces were sampled from two drill cores in the north zone (SPD1848 and SPD3024, 300 m apart) and one in the south zone (SPD0402) (Figure 1), where the weathered zone is 32 and 48 m thick, respectively. Samples from SPD1848 span the full thickness of the weathered zone, whereas samples from SPD0402 span the lower half of the weathered zone due to the availability of suitable material. Hematite from SPD3024 was taken from the zone of sulfide enrichment, extending the depth range of sampling beneath the ultimate redox front. Hematite was identified in drill core by its dark red to black appearance, textural characteristics (boxwork or botryoidal habit), and red streak. Mineral identification was confirmed via SEM analysis and Raman spectroscopy of polished thin sections. Dated samples were selected from mm to cm scale veins or fracture fills free from contamination and mineral intergrowth.

Analytical Methodology
Samples were extracted from drill core using a micro-saw and crushed to ≤1 mm. Seventy-two fragments of hematite (mean fragment weight = 46 μg) were picked manually under a binocular microscope and their mineralogy confirmed by Raman spectroscopy. (U-Th-Sm)/He dating was undertaken at the Caltech Noble Gas Laboratory, following the single aliquot method (Farley, 2002;House et al., 2000). Individual hematite fragments were loaded into Pt tubes and degassed by incremental heating with a Nd-YAG laser. To ensure complete extraction of He, samples were heated to >1,000°C. To avoid parentless He and erroneously high dates due to partial U-loss during hematite-magnetite transformation, He extraction was conducted under high pO 2 (100 torr), buffering the transition to >1,200°C; above the temperature of complete He degassing (Hofmann et al., 2020). Isotope-dilution measurements of 4 He were made with an enriched 3 He spike, using a Pfeiffer Vacuum quadrupole mass spectrometer. U, Th, and Sm were measured on the same aliquots (dissolved in HCl for 12 h at 95°C, then dried and the precipitate re-dissolved in HNO 3 ) using an Agilent 8800 triple-quadrupole ICP-MS (Hofmann et al., 2017).
(U-Th-Sm)/He ages were calculated according to Farley (2002). Since aliquots were taken from structures much larger than the typical alpha-particle stopping distance in hematite of 13-16 μm (Ketcham et al., 2011), no correction for alpha-ejection or -implantation was applied.

Oxygen Isotope Analysis
The δ 18 O composition of weathering-derived hematite (and goethite) (δ 18 O H(G) ) is fixed during crystallisation (Yapp, 2001). Rapid isotopic exchange between water and ferrihydrite promotes equilibrium, and preserved δ 18 O H(G) signatures reflect the temperature and average fluid composition during mineral formation (Bao & Koch, 1999;Yapp, 1987). Transformation of ferrihydrite likely occurs over days to ∼100 years (much shorter timescales than the multi-Myr weathering periods which affect PCDs), depending on temperature and pH (Das et al., 2011). Fe-oxides remain closed to later oxygen exchange with groundwater at ambient temperatures (Bao & Koch, 1999). Thus, δ 18 O H(G) compositions have been used to constrain continental climate change (Yapp, 2001), identify the nature of fluids present during alteration of hypogene base-metal deposits (Cruise et al., 1999), and track latitudinal variation in the isotopic composition of meteoric water during weathering (Miller et al., 2017).

Oxygen Isotope Sample Collection, Characterization, and Preparation
Hematite samples, identified by appearance in hand sample and SEM observations on corresponding thin sections, were selected from three drill holes at Spence and four holes at Cerro Colorado. Samples were crushed/ micro-drilled to obtain several milligrams of chips and powder for analysis. Raman spectroscopy was used to confirm the mineralogy of samples JSC17-069 and FC1649. The composition of Fe-oxide extracted from fracture surfaces for two samples from drill hole SPD0551 (JSC17-068 and JSC17-072), and all of the Cerro Colorado samples, for which no thin sections were available, was determined by non-quantitative powder XRD following standard analytical procedures (Bish & Post, 2018). These samples yielded similar spectra, best interpreted as a mixture of hematite and goethite (referred to as hematite(goethite)). At equilibrium, low-temperature (25-120°C) hematite and goethite are isotopically indistinguishable (Bao & Koch, 1999;Yapp, 1990), allowing calculation of fluid values using the same mineral-water fractionation equation. The isotopic effect of quartz present in several Cerro Colorado samples was accounted for via aqua regia dissolution and isotopic measurements on quartz separates (see Supporting Information).

Oxygen Isotope (δ 18 O) Analysis Via Laser Fluorination
Oxygen isotope analysis of 21 Fe-oxide samples (15 from Spence and six from Cerro Colorado) was conducted at the SUERC Stable Isotope Facility via laser fluorination (Giuliani et al., 2005). For each measurement, 3-5 mg of Fe-oxide was heated with a CO 2 laser in the presence of a fluorine-based reagent (ClF 3 ). After passing through an in-line Hg-diffusion pump, liberated oxygen was converted to CO 2 using a heated rod of platinized graphite. Isotopic measurements were made using a BG Optima dual-inlet mass spectrometer. Isotopic values were calculated using the mass measurements of CO 2 isotopologues and are reported as δ 18 O‰ relative to Vienna Standard Mean Ocean Water (VSMOW). Calibration on three secondary standards (UWG2, GP147 (international garnet standards) and YP2 (internal quartz standard)) yielded a standard error of 0.07 and R 2 = 0.9999. Internal uncertainty on the isotopic measurements is <0.1‰ (1σ).

(U-Th-Sm)/He Geochronology
At our study sites, empirical evidence such as the presence of hematite in former hypogene sulfide veins (originally containing pyrite ± chalcopyrite and minor quartz) (Figure 3a), boxwork texture (semi-pseudomorphic replacement of cubic pyrite crystals by hematite) (Figure 3b) and microscale textural relationships between spherulitic Fe-oxides and embayed pyrite ( Figure 3c) show that hematite has formed through sulfide oxidation, supporting the use of hematite dating to track the paleo-weathering front.
Hematite occurs as amalgamated micro-spheres or polycrystalline aggregates, commonly exhibiting layering on the scale of microns to tens of microns ( Figure 3d). It is unclear whether layers are Liesegang bands, formed by geochemical self-organization during precipitation from a supersaturated solution in a single depositional event, or growth layers formed by discrete precipitation events. In the latter case, all ages (new and previously published discussed here) will be averages of the individual growth events contained within each dated fragment (Heim et al., 2006).
Hematite precipitation at Spence records continuous weathering from the middle Miocene to the Pleistocene, with ages tightly clustered in both the North and South Zones ( To account for the westward slope of the land surface and the water table (Section 5.1), we replot these data, normalised to a sloping water table (Figure 4g).

Iron Oxide Oxygen Isotope Analysis
Fe-oxide samples from Spence yielded δ 18 O values between +5.7‰ and +11.2‰, whereas samples from Cerro Colorado were found to be isotopically lighter (−3.14‰ to +6.76‰) ( Table 2). Published hematite(goethite)-water  fractionation factors at 25°C range from −8.96‰ (Zheng & Simon, 1991; based on thermodynamic calculations) to 6.04‰ (Yapp, 1990; based on synthesis experiments), which is important because the isotopic composition of the parental water is the predominant control on that of the precipitate (Miller et al., 2017;Sultan, 2015;Yapp, 1987). We apply the fractionation factor of Yapp (1990) as it is representative of samples from natural environments (Miller et al., 2017;Yapp, 2000), and yields calculated fluid compositions within previously published ranges for meteoric and groundwater documented near Cerro Colorado (Aravena et al., 1999;Fritz et al., 1981) and Spence (Cameron & Leybourne, 2005). We assume a temperature of 25°C when calculating parental fluid isotopic compositions based on groundwater temperatures of 20-29°C (BHP data) measured in drill holes around Spence and elsewhere in the Central Depression and Precordillera close to Cerro Colorado (Fritz et al., 1981; see Supporting Information).

Landscape Evolution, Canyon Incision, and Water Table Movement at Cerro Colorado
The north side of Cerro Colorado is cut by the Quebrada de Parca, one of several endorheic drainages linking the Andean Precordillera with regional base level in the Central Depression (Figures 1b, 1c, and 2d). Cooper et al.'s (2016) hematite (U-Th-Sm)/He data appear to support a prolonged period of water table stability between ∼31 and ∼16 Ma, with water table descent <16 Ma linked to aridity-induced incision of the Quebrada de Parca. However, we suggest that at least some of the age-elevation trend observed by Cooper et al. (2016) is an artifact of sample locations and the slope of the water table, and present a reinterpretation of these data constrained by published paleo-climate and sedimentological records, and geomorphological relationships observed in satellite imagery, which imply later incision at ∼11 Ma.
Outcropping both north and south of the Quebrada de Parca are Miocene El Diablo Formation (EDF) gravels (Blanco et al., 2012), deposited by a large distributive fluvial system (García et al., 2011;Jordan et al., 2010Jordan et al., , 2014. K-Ar ages of volcanic units within the uppermost EDF sediments, including samples located ∼5 km downslope of Cerro Colorado, suggest that deposition ended between ∼11.9 and ∼11.2 Ma (Blanco et al., 2012;Farías et al., 2005;García et al., 2004García et al., , 2011, forming a regional aggradational paleo-surface (Evenstar et al., 2017(Evenstar et al., , 2020. The Quebrada de Parca dissects a weakly developed drainage network on this surface, showing that canyon incision must post-date EDF deposition. This supports the view that the onset of modern hyperaridity, which created the conditions for both surface abandonment and canyon incision, occurred between ∼12 and 10 Ma, in agreement with the onset of canyon incision elsewhere in the Precordillera at ∼10 Ma (e.g., García et al., 2011;Hoke et al., 2007;Schlunegger et al., 2006).
Proximal to Cerro Colorado, the surface of the EDF slopes westward 2.7-3.7°. Between the Precordillera and the Central Depression, the Quebrada de Parca drops from ∼4,000 to ∼1,000 m a.s.l., and recent geophysical (TDEM) data from the lower reaches of the Precordillera near to Cerro Colorado show the water table approximates a subdued version of the surface topography (Viguier et al., 2018). Based on geomorphic analysis of the channel profile, Cooper et al. (2016) showed that canyon incision was driven by climatic factors and that postincision tilting of this section of the Precordillera has been negligible since the middle Miocene. Therefore, we assume the modern surface and water table slopes are representative of those during development of the preserved weathering profile at Cerro Colorado. To account for the lateral separation of sampled drill holes (1.5 km in the direction of slope), we replotted the (U-Th-Sm)/He data in terms of elevation relative to a water table with a westward slope of 2.7° (Figure 4g). By normalizing the data in this way, the previously apparent break in slope at ca. 16 Ma is no longer obvious. We suggest that following the end of exhumation-driven relative water table descent, possibly by ∼30 Ma (which marks the beginning of prolonged hematite precipitation), and certainly by 19.25 Ma (minimum age constrained by the Tambillo ignimbrite; Bouzari & Clark, 2002), the water table at Cerro Colorado remained relatively stable, until a reduction in MAR, from ∼130 to <20 mm, between 12 and 11 Ma (Jordan et al., 2014) ended deposition of EDF sediments and drove incision of the Quebrada de Parca and associated water table decay (Figure 5a). Our reassessment implies 300 m incision (the depth of the Quebrada de Parca proximal to Cerro Colorado) within ∼11 Myr; an overall incision rate of ∼27 m/Myr, in agreement with estimates for other canyons in the Precordillera (Evenstar et al., 2020). However, the overall impact of incision on the elevation of the water table is of lower magnitude (∼85 m) as the water table was already situated at a depth of ∼150-200 m when incision began (Figure 5a).

Landscape Evolution and Water Table Movement at Spence
Most hematite ages above the ultimate redox front at Spence are younger than ∼10.5 Ma (Figure 4). The cluster of hematite ages starting at this time across a range of depths likely reflects pervasive weathering following relative water table descent. We attribute two older ages from beneath the ultimate redox front in the North Zone (14.7 and 12.4 Ma; Figure 4a) to incipient oxidation along high-permeability pathways, such as faults and fractures, which allowed oxygenated water to penetrate beneath the paleo-water table (Lichtner & Biino, 1992). Similarly, two middle Miocene-aged samples from the South Zone ( Figure 4c) are interpreted as hematite that formed along high-permeability pathways prior to water table descent.
In the absence of canyon incision, we suggest that weathering front propagation at Spence was controlled by exhumation-driven relative water table descent. The position of the ultimate redox front was attained when exhumation ceased, and the estimated rates of relative water table descent (23.9 ± 19.7 m/Myr in the north and 17.6 ± 9.2 m/Myr in the south) approximate rates of exhumation prior to cover deposition. The apparent end of exhumation in the North Zone, constrained by a 9.50 ± 0.67 Ma (2σ) hematite at the ultimate redox front (Figure 4a), coincides with the age of a 9.47 ± 0.04 Ma ash layer within the overlying gravels (Figure 5b; Sun et al., 2018). Assuming steady-state erosion and water table descent, the switch from erosion to local cover deposition occurred in the late Miocene (Figure 5b), at least 10 Myr after Cerro Colorado and coeval with the onset of hyperaridity. Our observations at Spence bear some similarity to the thin cover of Arriero Gravels overlain by a 9.52 Ma tuff layer at the Mirador mine in the Centinela District, described by Riquelme et al. (2018), although the supergene enrichment at Spence (∼44-20 Ma; Rowland & Clark, 2001) occurred earlier than in the Centinela District (∼25.2-12.6 Ma; Riquelme et al., 2018).
At Spence, the modern water table depth (86 m in the North Zone and 39 m in the South Zone) is not significantly different to the depth of the ultimate redox front below the erosional paleo-surface/base of the gravel cover (32 m in the North Zone and 48 m in the South Zone). Although increasing aridity was probably important in the switch from erosion to deposition (first of sheet flood sediments, followed by drier debris flows: Riquelme et al., 2007;Sun et al., 2018), and then to surface abandonment, prolonged hyperaridity has had little effect on the position of the water table, which did not get progressively deeper from the late Miocene onward (Figures 4a, 4c, and 5b). Similarly, the depth of the modern water table in the Central Depression west of Cerro Colorado is ∼50 m in some areas (Viguier et al., 2018). Although it is unclear when the water table at Spence attained its present position near the gravel-bedrock contact, this likely occurred during the Pliocene, based on the youngest clustered ages observed in both the North and South Zones.

The Relative Importance of Climate and Tectonics on Water Table Descent and Supergene Enrichment
As climate desiccation in the Atacama occurred on a regional scale, subjecting many PCDs to comparable environmental conditions, variations in water table depth, supergene enrichment, and weathering zone thickness between different deposits (5-50 m at Spence, 50-200 m at Cerro Colorado, 10-500 m at La Escondida) cannot be explained by spatially variable precipitation. Instead, this variation suggests differences in exhumation histories (Bissig & Riquelme, 2009), local lithological and tectonic controls on aquifer architecture (Jordan et al., 2014), medium to long-range groundwater recharge characteristics (Houston, 2002;Magaritz et al., 1990;Scheihing et al., 2017) and canyon incision may be important factors.

Oxygen Isotopes, Groundwater Sources, and the Susceptibility of Water Tables to Post-Hyperaridity Descent
δ 18 O H(G) compositions have been shown to rapidly equilibrate with, and record, those of weathering fluids, according to the fractionation factor of Yapp (1990) (Miller et al., 2017;Yapp, 1990Yapp, , 2000. Thus, our calculated groundwater isotopic values demonstrate the relative importance of meteoric versus formation water during PCD weathering and offer clues as to the relative susceptibility of different locations to water table decay in response to increased aridity. Calculated fluid compositions for Cerro Colorado Fe-oxides (δ 18 O = (−3.14‰ to +6.76‰)) largely overlap the range of measured values for precipitation falling within the Quebrada de Parca catchment at or above the modern elevation of the deposit (Aravena et al. [1999] and Fritz et al. [1981] data; Figure 6). A meteoric signature is indicative of direct recharge (pre-hyperaridity) or short to medium-range (catchment-scale) indirect recharge (post-hyperaridity), recording groundwater flow through the shallow subsurface according to the Andean basin fill recharge model (Houston 2002; Figure 7). Prior to climate desiccation, groundwater at Cerro Colorado would have been replenished by more rainfall in the upper reaches of the Precordillera infiltrating down to the water Note. The "*" refers to calculated using the fractionation factor of Yapp (1990).  Yapp [1990]; 1,000lnα = 1.63 × (10 6 /T 2 ) − 12.3), Where 1,000lnα Refers to the Fractionation Factor and T is Temperature in Kelvin   (Jordan et al., 2014;Rech et al., 2019), but the deposit has since been susceptible to water table decay due to decreased meteoric recharge in the Quebrada de Parca catchment, and aridity-induced canyon incision.
At Spence, isotopically light meteoric water, derived from precipitation between 2,500 and 3,000 m a.s.l., has been documented in the eastern area of the deposit, upslope of the ACL, whereas isotopically heavy, saline water has been documented west of the ACL (Figure 6c; Cameron & Leybourne, 2005). Our calculated fluid isotopic values for Spence hematite(goethite) (δ 18 O = −0.37‰ to +5.13‰), from both sides of the ACL (Figure 6c) are much heavier than for Cerro Colorado (Table 2; Figure 6b). Although the elevation-dependent trend of precipitation compositions in Figure 6b could be extrapolated to intercept the range of calculated parent fluid values for Spence hematite, most hematite at Spence formed after climate desiccation and beneath cover, making local precipitation an unlikely water source during weathering. Isotopically distinct groundwaters at Spence (Cameron & Leybourne, 2005) show that meteoric water that arrives via indirect recharge retains its light isotopic signature under hyperarid conditions. Therefore, isotopically heavy groundwater cannot be explained by evaporation or reactive flow of meteoric water through the shallow subsurface. Isotopically heavy groundwater is likely deep formation water, upwelling along weaknesses associated with the ACL (Cameron & Leybourne, 2005). We suggest that the maintenance of a shallow water table at Spence is due to long-range groundwater recharge of basal aquifers in the Pampa del Tamarugal, over long timescales (10 4 -10 5 years; Jayne et al., 2016), by more consistent precipitation in the high Andes (>100 mm/year MAR above 4,000 m a.s.l.; Jordan et al., 2014). The heavy isotopic signature of hematite-forming water at Spence may be the result of prolonged reactive transport and mixing with hydrothermally circulating waters within the basement (Magaritz et al., 1990) prior to upwelling along the ACL in line with the Toth (1963) model of groundwater flow (Figure 7).
The destructive nature of both techniques precluded obtaining isotopic data for the same hematite fragments that were dated. However, in many cases both measurements were made on hematite from the same veins/fractures and we therefore assume that our isotopic results reflect the composition of groundwater during weathering from the late Miocene onward. The isotopic composition of hematite-forming water at Spence shows deep formation water has contributed to recharge for ≥10 Myr, suggesting that the groundwater regime proposed by Cameron & Leybourne (2005) has been long-lived. We propose that the water table at Spence has remained shallow, despite climate desiccation, because of deep recharge fed by consistently higher MAR in the high Andes.

Conclusions
We use hematite (U-Th-Sm)/He geochronology and oxygen isotope analysis to compare the timing of weathering and sources of groundwater at two Andean PCDs in different morphotectonic settings-Cerro Colorado, within the Precordillera, and Spence, within the Central Depression. By combining our data with field observations and published sedimentological and geochronological constraints, we draw the following conclusions:  3. Supergene alunite ages suggest enrichment ended by ∼14.6 Ma at Cerro Colorado (Bouzari & Clark, 2002) and ∼21 Ma at Spence (Rowland & Clark, 2001), whereas hematite precipitation at both deposits persisted into the Pleistocene. Post-hyperaridity hematite precipitation suggests weathering profiles continue to develop after the end of supergene enrichment-therefore hematite ages are not necessarily indicative of periods of enrichment.  Aravena et al. (1999) and Fritz et al. (1981). Locations of isotopic and sedimentological hyperaridity indicators are from Hartley and May (1998) Figure 6a). Arrows schematically show indirect recharge pathways. The Houston (2002) basin fill recharge model accounts for water table decay at Cerro Colorado in response to increased aridity and incision of the Quebrada de Parca. At Spence, the water table has remained shallow as the ACL has provided a pathway for upwelling formation water, originating as precipitation in the high Andes and moving through the basement according to the recharge and hydrothermal mixing model of Magaritz et al. (1990). 4. Fluid isotopic compositions show deep formation water was present during hematite precipitation at Spence. Long-range groundwater recharge and the low-relief setting of Spence maintained a relatively shallow water table despite hyperaridity. Conversely, Cerro Colorado has been more susceptible to water table decay linked to aridity-induced canyon incision in the Precordillera. These results are consistent with models of groundwater recharge in the Atacama, within different morphotectonic settings (Houston 2002;Magaritz et al., 1990).
Both Spence and Cerro Colorado are enriched, yet there has been no canyon incision local to Spence, and incision of the Quebrada de Parca at Cerro Colorado began ∼3.5 Myr after the apparent end of supergene enrichment. We suggest that exhumation-driven relative water table descent, rather than incision-driven water table decay, has been more important for the propagation of weathering fronts during supergene enrichment of both PCDs. Therefore, deeply exhumed areas (tectonic control) are more likely to host supergene enrichment than incised areas (climatic and/or tectonic control), unless incision occurred while conditions were conducive to copper leaching (prior to climate desiccation). The greater importance of exhumation, compared to canyon incision, is also demonstrated by the potential magnitude of each process. The cumulative thickness of rock which can be exposed above the water table through exhumation is on the order of kilometers (bringing PCDs from their depth of emplacement to the surface), whereas canyon incision can only exert a second-order control on the water table, over a smaller depth range. It appears that climate variability alone is insufficient to produce relative water table descent over the depth range recorded by weathering profiles in the Atacama.

Data Availability Statement
Data reported in this paper can be accessed at the BGS NGDC repository. Data reported in Cooper et al. (2016) can be accessed at the GSA repository: https://gsapubs.figshare.com/articles/journal_contribution/ Supplemental_material_Aridity-induced_Miocene_canyon_incision_in_the_Central_Andes/12533894. This work was funded by a NERC GW4+ UK DTP Studentship (NE/L002434/1) with CASE support from BHP. Additional funding for the (U-Th-Sm)/He work was provided by a Student Research Grant from SEG (SRG18-104; J. Shaw) and a Royal Society Grant (RG140683; F. Cooper). Isotopic analyses were supported by a NERC Isotope Geosciences Facilities Grant (IP-1752(IP- -1117. The authors are grateful to BHP for permitting mine visits, Ed Bunker for sample collection, and Guillerme Santos, Reinaldo Guzman, and Lia Ituarte for advice and logistical assistance. The authors thank Martin Smith for feedback on a draft manuscript and Mark Cuthbert for discussion on water tables. The authors thank Pete Reiners and Carlos Marquardt for their constructive reviews, and Peter van der Beek for his editorial handling.