Paleoenvironmental Evolution of a Forearc in Response to Forcings by Drainage, Climate, Volcanism, and Tectonics: The Quillagua Depocenter, Chile

The Quillagua depocenter in the hyperarid Atacama Desert of the southern Tropics in Chile (Figure 1) shifted over the last 9 million years among conditions ranging between internal drainage that created a lake which was at times fully evaporated and other times fresh water, to the erosion of a 260 m deep canyon. In this paper, we dissect the complex roles of drainage basin evolution, deformation, volcanic activity, and climate change during the accumulation of 200 m of upper Miocene to Pliocene fluvial, shallow lacustrine, and evaporite deposits, as well as during their postdepositional diagenetic alternation and partial erosional destruction. This paper’s new analysis is based largely on new regional mapping, tephra dating, and facies descriptions by the Chilean geological survey [1].

The Quillagua depocenter is a distal part of the latest Oligocene to present Pampa del Tamarugal (PdT) forearc basin (Figures 1 and 2) [2–4]. The 400 km long Loa River drains through the Quillagua depocenter. Today, the entire study area is hyperarid, yet Miocene–Quaternary climate and hydrological inputs varied.

A series of studies evaluated three basins arrayed along the modern path of the Loa River: the Quillagua depocenter, the minor Cerro Batea basin 150 km south of Quillagua in the PdT, and the Calama basin (Figure 3). The stratigraphic and environmental evolution of the Quillagua depocenter was established by Jensen [2], Sáez et al. [3], Bao et al. [5], and Pueyo et al. [6]. A series of studies beginning with Naranjo and Paskoff [7] led Jensen [2] and May [8] to infer that deposition near Cerro Batea in the southernmost PdT was connected to the Calama basin during the Early and Middle Miocene, yet those two basins were apparently disconnected during the early Late Miocene (Figure 3(d)). Later, during the Late Miocene, a surface-water connection of the upper Loa River catchment through the Cerro Batea area and on to the Quillagua basin developed, roughly 6 Ma (preceding Figure 3(c)). May [8] and May et al. [9] documented that the Late Miocene Quillagua Formation strata are approximately age equivalent to the Opache carbonate-rich lacustrine deposits of the Calama Basin (Figure 3(c)). Hartley and Chong [10] and Sáez et al. [11] evaluated the relationship of those environments to paleoclimate history. Chong et al. [12] and Vásquez et al. [1] analyzed a connected basin within the Coastal Cordillera, the Salar Grande basin (Figure 3(a)). This halite-dominated depositional system today is separated from the surface water of the Quillagua depocenter by a low topographic threshold, yet Chong et al. [12] interpreted that the Quillagua depocenter leaked groundwater to the Salar Grande basin along faults (Figure 3(b)). The environmental evolution of contemporaneous features and landforms marginal to the depocenter was established by studies of the distribution and ages of alluvial facies of the eastern and central sectors of the PdT basin [13, 14] and by Ritter et al.’s [15] studies of relict fluvial landforms on the western flank of the depocenter, constrained by Terrestrial Cosmogenic Nuclide (TCN) exposure ages.

Jensen [2], Sáez et al. [3], May et al. [9], and Chong et al. [12] all identified the incision of the Loa Canyon into the Quillagua basin as the trigger for termination of lacustrine conditions in most parts of the Quillagua depocenter and Calama basin (Figure 3, time between (a) and (b)). May [8] and May et al. [9, 16] document a maximum age of first incision of the Loa Canyon system in the zone between the Calama and Quillagua basins to be younger than approximately 3 Ma.

The 3000 km² Quillagua lacustrine depocenter lies in the northern Chile forearc between 21.0–21.8°S and 69.3-69.9°W, at 700 to 1000 meters above sea level (masl). The western basin flank is a deeply eroded coastal mountain range (Coastal Cordillera, 1000-2000 masl), and the eastern flank is the foothills (Precordillera, to 4500 masl) of the Andes (Figure 1). A north-trending series of isolated hills that expose Paleozoic and Mesozoic units bisects the basin. Quillagua depocenter strata occur in the subsurface along the central axis of the PdT valley and crop out widely along the margin of the Coastal Cordillera. Paleoaltimetry based on the distribution of fog during the Pliocene to Holocene across the Coastal Cordillera and PdT suggests that the elevation of the Quillagua depocenter has changed little since 5 Ma, with localized subsidence of >50 m [17].

Two water sources flow into the modern hyperarid Quillagua depocenter. The first is surface water that enters infrequently through seven major canyons that drain the Precordillera (e.g., Figures 1 and 2); the northern three drainages extend short distances into the volcanic arc. Today, the Precordillera-sourced streams only carry surface water and siliciclastic detritus into the Quillagua depocenter following infrequent summer rain events [18–20]. The second is the 400 km long perennial Loa River, whose headwaters are in the less arid high volcanic peaks of the Andes (to 6000 masl). The Loa crosses the southwestern margin of the Quillagua depocenter in a 50 to 260 m deep incised canyon. No streams drain into the western or northern margins of the basin.

Today, Quillagua depocenter water is exported through the Loa Canyon and the Amarga Canyon, a tributary to the Loa sourced by groundwater derived from the Precordillera. From the western margin of the Quillagua depocenter, the Loa descends >550 m over 30 km distance to the Pacific shore. Because the Loa River flows entirely within a deep canyon, it is not now nor has it been recently a part of the Quillagua depocenter lacustrine history.

The mean annual rainfall throughout the Quillagua depocenter and adjacent Coastal Cordillera is <1 or 2 mm [19, 21]. Extreme longevity of geomorphic surfaces indicates that hyperaridity dominated the last 20 million years in the Coastal Cordillera [22–26]. In the Quillagua depocenter itself and in the lower part of the Precordillera, regional and persistent hyperarid condition throughout the last 12 Ma was interrupted by short-lived wetter intervals [13, 27–29]. Within the Quillagua watershed, late Pleistocene intervals of enhanced rainfall on the highlands of the Precordillera and headwaters of the Loa River persisted for intervals of a few thousand years [30–32]. Although the lowlands remained hyperarid [26, 32], the enhanced surface water input produced riparian wetlands or ephemeral playa lakes [33]. In the lowest elevation sector of the Quillagua depocenter, groundwater levels are approximately 5 to 100 m below the ground surface [34]; water is exposed in a set of sink holes.

During the Late Miocene and Pliocene, the Altiplano-Puna volcanic complex ejected extraordinary volumes of silicic volcanic materials from calderas located only 100 to 300 km east and southeast of the Quillagua depocenter (Table 1) [35–37]. Eruptions that were contemporaneous with and proximal to areas now draining into the Quillagua depocenter include numerous examples with volumes exceeding 1000 km³ (Figures 3(b)–3(d)). Similarly, voluminous, explosive eruptions from slightly more distant calderas likely also contributed airfall materials (Table 1).

Carving the Loa Canyon through the basin-fill column and into the basement is the largest example of water-driven erosion since widespread deposition of the Soledad Formation ceased near the end of the Pliocene (~2.6 Ma) [1]. Deflation and sedimentary transfer by wind removed or added a few meters height (e.g., 1-3 m) from the surface in much of the central basin [38]. Surrounding Cerro Soledad, in the middle of the Quillagua depocenter, occur topographic benches cut into the Quillagua and Soledad Formations as well as into basement lithologies and thus must be Pliocene or younger. These were first interpreted by Brüggen [39] to be wavecut lake shorelines, and Ritter et al. [15] used Ne, Be, and Al cosmogenic dating of these topographic benches to propose that a series of lakes filled the Quillagua depocenter between ~2.5 Ma and 0.27 Ma. Jordan et al. [38] offered the alternative interpretation that the benches express differing resistance to erosion among successive lithologies within the Quillagua and Soledad Formations and that their ages record the denudation of the Quillagua depocenter.

Groundwater also impacted the Quillagua depocenter, expressed by diagenesis and dissolution [6]. Surface features indicative of widespread dissolution include circular closed basins and elongate valleys with no surface water exit, yet a systematic study of the karst is lacking [1, 6]. Units dominated by gypsum and anhydrite dissolved to generate the sinkholes and caverns, and overlying units were distorted into complex fold forms that are evidence of collapse.

Despite a near-total absence of vegetation, exposure of the basin filling strata in the central sector of the basin is poor due to cover by a 0.5-2.0 m thick CaSO₄-dominated soil (e.g., [13]) or to a halite crust on the order of one meter thick [38].

Despite its location in the active Andean plate margin system, few Late Miocene or younger faults or folds occur within the PdT and the Quillagua depocenter [1, 23, 40]. Although the vertical displacement on even the largest nearby faults is less than a few hundred meters, these fault scarps are major controls on the organization of the very slowly eroding landscape of the Coastal Cordillera [23, 26].

Structures that are significant for depocenter evolution include a set of east-northeast trending faults along the northern margin of the Quillagua depocenter and a set of north-trending faults and folds in the basin interior, which deform strata at least as young as 2.6 Ma (Figure 2) [1]. The structures within the Quillagua depocenter are not well documented. Examples include the Cerro Soledad and Cerro Mogote ridges (Figure 2), where north-trending bands of pre-Cenozoic basement are juxtaposed by faults against folded basin-fill strata (Figure S11A, B) [1]. The asymmetry of north-trending Lomas de Sal hill and abrupt changes of elevations and facies of age-equivalent strata across its western margin motivate the hypothesis that this 150 m high feature is a fault-controlled anticline [17, 41].

The foundations of this study are field geological mapping, outcrop-scale sedimentary facies analysis, geochronology, satellite image and topographic analysis of landforms, and construction geological cross-sections using geophysical and geological data. Geological mapping and facies analysis was conducted systematically over the region from 21 to 21° 30$′$ South latitude and 69° 30$′$–70° West longitude at a scale of 1 : 100,000 [1] and on Cerro Soledad at a scale of 1 : 10,000 [38]. Targeted field observations extended the area of mapping and facies analysis east to 69° 25$′$ W and south to 21° 42$′$ S (Figure 2). Field facies analysis utilized the criteria of primary sediment composition, grain size, sorting, lamination geometry, bed contact relations, and vertical variations of those properties, in addition to bioturbation abundance and trace fossil style. Karst features are identified on satellite images (Google Earth© image catalog) using criteria of circular shape, shadows, and other steep-walled pits and valleys without surface drainage outlets. The discrimination between dissolution features and wind-deflation hollows is a challenge. The elevation data provided by Google Earth©, derived from the 90 m SRTM satellite, was adequate for most parts of the analysis. For analysis of river terraces, the 30 m ALOS satellite topography was used, yet even this resolution fails to reveal key details of the terrace sets that are visible in the field.

Data for three petroleum exploration boreholes, all exceeding 600 m depth and two penetrating the basement, include cuttings, cores, and electric logs (Figure 2). Data for the Soledad 1 and Hilaricos 1 wells were presented by Gallardo [42, 43], and the lithologic succession with depth was reported by Nester [41] and published by Labbé et al. [44]. The Hilaricos 2 well samples and logs were provided by a private operator for use in Vásquez et al.’s [1] study. Physical samples for all these boreholes were viewed by the authors and volcanic tephra were sampled and dated (Table 2). Seismic reflection profiles z1f-005 and 99-12 were used in construction of cross-sections (Figure 4). Our interpretations of the Cenozoic strata revise those of Nester [41] and Labbé et al. [44]. For 99-12, Labbé et al.’s [44] depth conversion was used. For z1f-005, depths were based on comparing the interpreted top-of-basement reflector to the Soledad well materials. West of the western limit of seismic lines, the interpretations presented are constrained by the depth of basin fill reported in the Hilaricos boreholes, López et al.’s [34] reports of basement depth from water wells, and are guided by outcrop relations in canyon walls.

We compile radiometric constraints spanning two decades of publication, among which the tephra dates first reported by Vásquez et al. [1], and those first reported here were completed in the Isotope Geology laboratory of Servicio Nacional de Geología y Minería (SERNAGEOMIN), in Santiago, Chile. Most of these new results were dated using the ⁴⁰Ar/³⁹Ar method; three samples were analyzed by U-Pb zircon LA-ICP-MS and one by K-Ar. The SERNAGEOMIN preparation and analysis steps are detailed in Supplemental File 1. U-Pb zircon ages were calculated using “Zircon Age Extractor” of the program Isoplot, followed by treating the results as a weighted average.

New data for the characteristics and distributions of facies are integrated with published information for the strata and presented in Table 3. Based on utilization of informal lithologic units in regional mapping at a scale of 1 : 100,000 [1], which illuminated their regional extent and lateral variations, and with the insight provided by new tephra dates, this paper departs somewhat from prior publications in its assignment of particular outcrops to lithostratigraphic formations or informal units. Two lithologic units that were treated informally for decades are elevated to formation rank (Table 4; supplemental file 2). An interval known previously as the Hilaricos unit [2], Hilaricos anhydrite [3], and Hilaricos strata [1] is formalized as the Hilaricos Formation, whose most diagnostic lithology is anhydrite. A stratotype and a reference section are established in the exposures on the western wall of the Loa Canyon, 15 km distant from one another (supplemental file 2). Strata treated previously as either the lower part of the Quillagua Formation [1] or as the redbeds of the Aduana de Quillagua and an unnamed overlying interval of gray ashes [45] constitute the newly defined Ancachi Formation, which is characterized by laminated siliciclastic alluvium that is rich in tephra and reworked volcanic ash. A stratotype is identified near the Customs Station (aduana de Quillagua) along the Panamerican highway, and the unit is named for a nearby archeological site. By division of the new Ancachi Formation from the lower Quillagua Formation, the Quillagua Formation is of more uniform lithologic properties than before amendment, with medium to thick interbeds of diatomite and marl plus occasional thin layers of chert and sepiolite among fluvial deposits as the diagnostic lithologies (Table 3). Strata comprising the Lomas de Sal uplifted structural block of the eastern Quillagua depocenter region (Figure 2) that were assigned previously to the Hilaricos informal unit and/or the Soledad Formation are treated here as the “unnamed unit of Lomas de Sal” (Table 3) because their lithologies do not fit the prior formation assignments.

Lacustrine conditions are documented within the depocenter during each of the principal intervals of sediment accumulation. Yet, only in the Quillagua Formation are the lacustrine facies indicative of a persistent and relatively fresh-water lake, with abundant micro- and macrofauna [3, 5, 46]. During deposition of the Ancachi Formation, the largely fine-grained and well laminated siliciclastics suggest that alluvial fans bordering the Precordillera graded westward into a sandflat and mudflat that was, intermittently, the eastern margin of a body of standing water [45]. During all depositional intervals, either the major part of the depocenter (Hilaricos and Soledad Formations) or localized zones in the eastern area functioned as groundwater-fed salars, indicative of a sufficiently arid climate to drive evaporite precipitation where the landscape intersected the groundwater table.

Erosion since the end of deposition of the Soledad Formation occurred broadly in the central part of the basin and narrowly in what is now the canyon of the Loa River. We estimate that roughly 15-20 km³ of upper Miocene to Pliocene strata were removed by erosion since ~2.6 Ma, constituting a 15-30% loss of the stratigraphic record.

More than 80 discrete features attributed to dissolution were mapped. In cases in which multiple small pits are close together, these have been integrated into a single-mapped polygon, resulting in a total of 54 polygons to mark dissolution features.

The availability of dated tephra interbedded in strata of the Quillagua depocenter and within unconsolidated sediments west of the western margin of the depocenter expanded greatly with the work of Vásquez et al. [1]. Here, their results, a revision of the age calculation for one of the Vásquez et al. [1] U-Pb dates (supplemental file 3), and five additional tephra dates (supplemental files 4, 5, 6) are integrated with numerous previously published dates (Table 2) to document the space-time relationships among sediment accumulation, erosion, and stasis (Figure 5).

For the Hilaricos Formation, Vásquez et al. [1] dated one tephra near the base, with 28 single zircons yielding U-Pb ages ranging between $8.2±0.3Ma$ and $6.8±0.2Ma$⁠, plus one outlier (supplemental file 3). Those data yield a weighted average age estimate of $7.51±0.08Ma$ (Table 2). Sáez et al. [11] reported a tephra dated $8.76±0.05Ma$ (biotite ⁴⁰Ar/³⁹Ar) within the Hilaricos unit in another location. Together, these sparse data and those in the overlying unit suggest that the Hilaricos Formation accumulated between approximately 9 and 7.2 Ma.

Three previously published dates and one first reported here correspond to the Ancachi Formation (Table 2). The two ⁴⁰Ar/³⁹Ar dates and a U-Pb zircon date imply deposition between 6.4 and 5.5 Ma, in conflict with the single K-Ar date. Reassignment to the Ancachi Formation of part of section that was the basis for Sáez et al.’s [11] magnetic polarity stratigraphy leads to inclusion of four normal and four reversed polarity zones. According to the 2004 magnetic polarity time scale of Lourens et al. [47], the polarity zones imply that deposition of the Ancachi Formation began prior to 7 Ma and ended around 5.8 Ma. Collectively, the data imply Ancachi accumulation approximately 7.1–5.5 Ma.

The six tephras assigned to the Quillagua Formation indicate accumulation between 5.7 and 5.0 Ma (Table 2). After reassignment of some strata to the underlying formation, the magnetic polarity of the Quillagua Formation is dominantly reversed, with only one high-confidence normal zone in each of two sections [11]. Both the tephra ages and the magnetic stratigraphy suggest that the Quillagua Formation accumulated in no more than 1 million years, between 5.7 Ma and greater than 4.7 Ma.

Jensen and Siglic [45] dated three slightly reworked tephra layers near the top of an interval of anhydrite strata in the lower part of the unnamed unit of Lomas de Sal Lomas (Figure 2; Table 2). The results span 5.56-6.0 Ma (Table 2).

If all the ⁴⁰Ar/³⁹Ar and U-Pb tephra ages of the Quillagua Formation and Ancachi Formation are accurate, there is a slight overlap, suggesting lateral facies interdigitation. This interpretation is consistent with the gradational formation boundary in southeastern outcrops (Figure 5). Correlative strata sampled in deep boreholes occur also in the southeastern and northeastern sectors of the Quillagua depocenter where tephra (Figure 2) are dated as 5.7-5.8 Ma (Table 2). Along the eastern flank of the PdT basin, lowland piedmont deposits of ages similar to the Ancachi and the Quillagua formations are well documented [14, 48, 49]. The lower anhydrite-rich strata of the unnamed unit of Lomas de Sal also correlate to the interval of overlap of the Quillagua and Ancachi formations.

The depositional age range of the Soledad Formation emerged from the numerous tephra dates by Vásquez et al. [1] and sparse earlier data [50] (Table 2). Fifteen dates for samples that are unambiguously in Soledad Formation range from 2.61 to 4.3 Ma. Four other samples give discordant ages spanning 6.7 to 5.5 Ma and were interpreted by Vásquez et al. [1] to be reworked volcanic materials. All of these Soledad Formation tephras predate by at least 2 m.y. the age for what Sáez et al. [11] reported to be Soledad Formation tephras. Consequently, we interpret the materials sampled by Sáez et al. [11] at Lomas de Sal to occur in terrace deposits and not within the Soledad Formation. The field relations for a final tephra sample that Vásquez et al. [1] assigned to the Quillagua Formation, which yielded an age of 2.75 Ma, permit its placement within the Soledad Formation (Table 2). Collectively, ⁴⁰Ar/³⁹Ar and U-Pb data imply that the Soledad Formation accumulated from approximately 4.1 Ma to 2.6 Ma.

Six or more distinct terraces with tread vertical separations of ≤10 m have been identified at various points along the Loa Canyon. The oldest well-dated terrace is the sixth above modern river level in a region upstream of the junction with the Amarga River where the canyon is 160 m deep (Figure 6). In it, a tephra interbedded in terrace fill sediments yielded an age of $0.255±0.017Ma$ (Table 2) [1]. At a position downstream of the Amarga confluence where the Loa Canyon is 250 m deep, terraces 14 m above the modern channel are as young as 4800 years (Table 2).

Borehole and geophysical data reveal that the Late Miocene-Pliocene fluvio-lacustrine strata of the Quillagua depocenter overlie the zone in which the Precordillera-sourced Oligo-Miocene PdT alluvial fan deposits thin westward and pinch out (Figure 4). The lake systems occupied the distal fringe of the forearc basin, and their deposits lapped onto the Coastal Cordillera basement. Although the upper Miocene to Pliocene lacustrine strata are best known from outcrops in the southern part of the basin, they are comparatively thin in the south (only 150 m at the Hilaricos boreholes) and thicken to the north (about 500 m at Soledad borehole) (Figure 4).

Exposed in the Loa Canyon walls, in the canyons and mountains east of the Quillagua depocenter, and drilled in three deep boreholes within the Quillagua depocenter, are upper Oligocene to middle Miocene siliciclastic conglomerates and sandstones deposited under nonmarine environmental conditions [1, 2, 14, 49]. Vásquez et al. [1] identified in the Loa Canyon a very mature calcium sulfate paleosol developed on upper Oligocene to lower Miocene alluvial deposits of the Quebrada Amarga informal unit and overlain by upper Miocene strata. This paleosol is correlated with gypsum-dominated paleosols of late middle Miocene age that occur widely in the Central Depression and Calama Basin [13, 51, 52]. The regional soils indicate that there was a noteworthy absence of sediment accumulation between 11 and 9 Ma due to hyperarid conditions [51, 52]. We treat that environmental state as the starting condition for analysis.

The 10 m thick Hilaricos unit is characterized by anhydrite pseudomorphed after gypsum, with up to 35% siliciclastics inclusive of a basal conglomerate and interbedded mudstone and sandstone (Tables 3 and 4; supplemental file 2). North of 21° 38$′$ S along the Loa Canyon, it lies unconformably over Oligocene-Lower Miocene conglomerate on whose upper surface occurs a calcium sulfate paleosol [1]. The Hilaricos upper contact is an erosional unconformity which causes marked thickness changes over horizontal distances on the scale of 100 meters along the walls of the Loa Canyon (supplemental file 7). The unit thins northward, although difficulty differentiating calcium sulfate-dominated Pleistocene terrace deposits from the Hilaricos Formation creates uncertainty whether the Hilaricos Formation pinches out near 21° 27$′$ S or persists as far north as the Amarga Canyon. In the Soledad borehole (Figure 2, “S”), approximately 30 m of gypsum older than $5.76±0.07Ma$ may correspond to the Hilaricos Formation. In the Hilaricos 1 and 2 boreholes (Figure 2, “H”), greater than 60 m of interbedded gypsum and siliciclastic mudstone are inferred to be Hilaricos Formation. These tentative subsurface identifications place the eastern limit east of 69° 23$′$ W.

The basal conglomerate accumulated in an alluvial plain, where the distance from the source area was insufficient to fully round the clasts. Knowledge of the depositional environment of the Hilaricos Formation in most locations lacks detail, because of extensive diagenetic overprinting by anhydrite (Supplemental file 8, Figure S8A, B). Pueyo et al. [6] interpreted prediagenetic relicts to reveal that a groundwater-fed salar interfingered with a low-relief mudflat (Figure 7(f)). Locally, pseudomorphs of upperward-directed gypsum crystals are indicative of subaqueous crystal growth (Supplemental file 8, Figure S8E). More widespread are pseudomorphs of many-centimeter length randomly oriented gypsum crystals, indicative of interstitial crystal growth in a muddy sediment. Elsewhere, pseudomorphs of decimeter-scale blocky gypsum crystals suggest that secondary gypsum cemented cavernous porosity (Supplemental file 8, Figure S8A) and brecciated crystals accompanying swirling mudstones (Supplemental file 8, Figure S8B) suggest collapse features. The Hilaricos Formation lacks direct evidence that either its groundwater or surface water was supplied from a paleo-Loa River.

The Ancachi Formation (Table 3) occurs south of 21° 25$′$ S overlying an erosional unconformity above the Hilaricos Formation or locally the Quebrada Amarga beds and thins from the stratotype in the southeastern sector of the depocenter toward the north and west. A lower 50 m thick interval of reddish brown mudstone and sandstone and an upper 20 m thick gray interval of volcanic ash and ash-rich mudstone and sandstone strata comprise this formation (Table 3) [45] (Supplemental file 2, Figure S2-3, 4). Thin to thick beds with common horizontal lamination and ripple lamination are characteristic (Supplemental File 2, Figures S2-5, 6). Minor interbeds of conglomerate typically coarsen and thicken eastward, indicating interdigitation with distal alluvial fan deposits from the Precordillera. Locally, some conglomerate occurs near the western limit of the unit. The upper member is widespread in the southeasternmost sector of the depocenter. To the north and west, this member may either transition to tuffaceous pinkish brown fine grained siliciclastics or transition to Quillagua Formation facies or pinch out. There is a comparable thickness of sandy mudstone with ash beds in the Hilaricos-1 borehole (Figure 2, H), for which a lack of sulfate minerals leads to correlation to the Ancachi Formation. The redbeds are discontinuous in exposures along the Loa Canyon [6], due to westward depositional thinning and to erosion prior to the Quillagua Formation (Supplemental File 7). The northernmost known occurrence is near 21° 25$′$ S, 69° 35$′$ W.

The lower redbeds represent sandflats and mudflats along the toes of alluvial fans (Figure 7(e)). Sepiolite lamina (Table 3) is consistent with very shallow lacustrine conditions of high Mg and Si activity [2, 53], which suggests that a fluctuating water table at the toe of the depositional system sometimes submerged the mudflat. The upper volcanic ash-rich sandstone interval reflects the margin between the shallow early Quillagua lake and an alluvial zone supplied by the local Precordillera catchments. These deposits provide no direct evidence of sediment supply from a paleo-Loa River. The lack of identified similar strata in the northern half of the basin likely reflects the much greater width of the northern Quillagua depocenter (e.g., approximately 90 km, compared to 40 km at the latitude of Quillagua village). In the southern area, sediment from the Precordillera-sourced alluvial fans aggraded at a comparatively higher rate and thereby initially kept up with the rising baselevel of the lake system, delaying eastward shoreline transgression.

The Quillagua Formation as identified by Sáez et al. [3] and Jensen and Siglic [45] is gray to nearly white and, while volumetrically dominated by sandstone, its diatomite beds and marls are diagnostic (Table 3; Supplemental Figure S9A, B). This facies crops out from near 22° 23$′$S to near 21° 13$′$ S (north of Cerro Soledad), a distance of approximately 130 km (Figure 7(d)). Lateral and vertical facies variations have been well described [3, 5, 46]. The major facies includes well-sorted, thin bedded sandstones and conglomerate, siliciclastic mudstones with ostracods, marls, and diatomites.

Strata assigned to the Quillagua Formation by Vásquez et al. [1] thicken near the northern limit of the depocenter and change facies to almost entirely siliciclastic (Figure 2, Pampa Maya; Supplemental file, Figure S9C,D). Although lacking the characteristic diatomite, the stratigraphic position of these strata below the dated Soledad Formation suggests correlation to the traditional Quillagua Formation. Features such as convoluted laminations as well as ostracod and algal limestones imply deposition in water-rich conditions (Table 3). With paleocurrents directed to the south-southwest and south-southeast, these strata are interpreted to have accumulated along the northern shoreline of Quillagua lake. Although the Quillagua Formation is inferred to be comparatively thick below the modern Salar de Llamara (Figure 4) and age-equivalent halite occurs in Salar Grande [1], there are no data for the position of the northwesternmost Quillagua lacustrine facies.

The Quillagua Formation indicates a renewed supply of sediment to all parts of the basin, and a marked change in the hydrological budget. Sáez et al. [3, 11] interpreted the Quillagua Formation to represent a combination of fine-grained siliciclastic strata fed into the eastern lake margin from alluvial fans, fluvial channel deposits along a north-oriented paleo-Loa River which entered the south end of the basin, and open-lake low-energy conditions in the central basin (Figure 7(d)). The fauna indicates open-lake conditions [3, 5, 46]. In the basin center (e.g., Cerro Soledad), knobs of basement rocks were islands within the lake. The Quillagua-age strata in the northernmost Quillagua depocenter represent a sand-flat and mud-flat environment with local wetlands, across which the lake shoreline rarely transgressed.

Strata correlative in part to the Quillagua Formation and in part to the Soledad Formation but of distinctive facies occur in Lomas de Sal, the easternmost exposures in the depocenter. Previously, a lower part was assigned to the Hilaricos informal unit and an upper halite-dominated part to the Soledad Formation [3, 6, 11, 45, 54]. Yet, their lithologies, lateral relationship with the Quillagua Formation, and tephra ages are inconsistent with these assignments.

The strata exposed in Tambillo Canyon (Figure 2) begin with 2 m of thin to thick beds of anhydrite, overlain by 2 m of thin to medium interbeds of anhydrite and tephra (Table 3), topped by a sharp, irregular contact (Supplemental file Figure S9F). A thick unit above that unconformity is characterized by a large percentage of halite. Its base is a meter-thick conglomerate that fills steep-walled channels and connected paleo-caverns, composed of basement rock fragments and cobbles of anhydrite breccia. For outcrops in Tambillo Canyon, the primary lithology of the thick upper unit is well sorted and laminated siliciclastic sandstone (Table 3; Supplemental Figure S9E), within which the halite is cement [45]. Although published reports emphasize a nearly pure halite, that description seems to apply specifically to outcrops in a quarry along the fault at the northern boundary of Lomas de Sal [6] where secondary halite crystals grew in fracture zones and steeply inclined laminated sands may be fracture fills. Anhydrite-rich and siliciclastic strata a few meters thick overlie the halite body, with facies consistent with assignment to the Soledad Formation.

Tephra dates and mapping relations suggest correlation of the lower anhydrite with the Quillagua Formation. Similarly, a 100 m thick interval of mudstone with cement and interbeds of gypsum and anhydrite in the Soledad borehole near the northeastern limit of the depocenter (Figure 2, “S”) contain a tephra (Table 2, PS-531-546d) that indicates age equivalence with the Quillagua Formation. These occurrences lead to the interpretation that evaporative mudflats lay to the east of fresher water Quillagua lake (Figure 7(d)). The position of the sandstone beds at Lomas de Sal overlying an erosional unconformity and a thin alluvial conglomerate is similar to local examples of the contact between Quillagua and Soledad formations elsewhere, leading to tentative correlation of the sandstone beds to the Soledad Formation. The provenance of clasts in the basal conglomerate is the adjacent Precordillera. The sandstone beds are interpreted as a localized dune field in which a rising water table gave rise to early halite cementation.

The evaporitic Soledad Formation overlies the Quillagua Formation across what is widely a paraconformity with local erosional relief and locally is an angular unconformity within the interior of and along the northern limit of the depocenter. The Soledad Formation occurs only north of 21° 40$′$ S, and its characteristic facies extends continuously into the southern part of Salar Grande (Figure 7(c)).

The Soledad Formation is characterized by repeated beds of calcium sulfate of various textures intercalated with siliciclastic siltstone and sandstone, often containing reworked volcanic ashes, with scarce conglomerate (Table 3). The diagenetic transformation of most of the calcium sulfate to anhydrite obliterated most textural clues of environmental processes (Supplemental file, Figure 10A, E). The relative abundance of siliciclastic and evaporite types varies across the depocenter: along the northeastern rim (Figure 2, Pampa Maya), siltstone, and fine sandstone constitute about half of the beds (Supplemental file, Figure 10B); in the western and central areas, calcium sulfate dominates.

Sáez et al. [3] interpreted the Soledad Formation to have accumulated in ephemeral playas and groundwater-fed evaporative salars. Vasquez et al. [1] show that, near the northern boundary of the basin, fine-grained siliciclastics accumulated initially in shallow, ephemeral lakes, and then transitioned upward to distal alluvial fan conditions, with intermittent groundwater-fed evaporite deposition and surface-water reworking of gypsum. Even near Cerro Soledad, salar conditions in which gypsum crystals grew interstitially in mud (Supplemental File, Figure 10C) were intermittently replaced by distal alluvial fan conditions. In the western exposures occur crystal textures and bedding suggestive of cyclic variations between times when evaporites crystallized in saline pools and times of desiccation and erosion of those pools [1, 6].

Water-lain deposits younger than the Soledad Formation in the southern and western sectors of the depocenter are distributed in localized bands (Figure 7(b)). Rounded-clast gravels are the surface deposits at many locations south of an 18 km long north-facing scarp at about 780 masl, which spans from near the confluence of the Loa and Amarga rivers to near 21° 24$′$ S, 61° 30$′$ W. The unconsolidated gravels are typically 1-2 m thick yet widespread within zones 0.5–5 km in width, overlying the Quillagua Formation east of Quillagua village as well as north and west of the Hilaricos hills (Figure 2; Figure S3B; Figure 7(b)). These gravels occur where Soledad Formation is not identified. The pebbles and cobbles are of mixed clast compositions, subangular to rounded, and occur in medium thick beds. Adjacent to the eastern wall of the Loa Canyon (21.453° S, 69.685° W), similar gravels reach approximately 10 m thick and are inset within an elongate broad paleo-valley. Below this gravel, an anomalously thin Quillagua Formation (~5 m) displays highly variable bedding dips and breccia. Here, the Quillagua Formation overlies an extensively recrystallized top of the Hilaricos Formation, with large gypsum crystals that fill macroscale cavities. Water-lain deposits constituting the distal toes of alluvial systems fed from the Precordillera are widespread in the eastern sector of the depocenter (Figure 7(a)).

Meter-scale beds of sand, silt, and clay, commonly with root traces and locally with thin limestone interbeds or fossil gastropods, are superficial sediments in and around Salar de Llamara [1, 33]. Interbedded tephra and ¹⁴C dating indicate episodic deposition spanning at least 340,000 years (Table 2) [1] to 12,000 years BP [33]. These deposits include examples interpreted to represent playa conditions, palustrine conditions, and low-energy unconfined fluvial conditions that produced sand sheets [1, 33]. Similar facies in likely older deposits overlies the Quillagua Formation southwest of Salar de Llamara and north of the Amarga river valley, filling a 2–3 km wide north-trending belt that is inset 5-10 m lower than the surrounding landscape. The absence of the Soledad Formation suggests that the area was first eroded by fluvial action (Figure 7(b), denoted by ephemeral stream) and then transformed into a depositional area.

The walls of much of the Amarga Canyon expose tens of meters of subhorizontal, medium to thick bedded anhydrite and fine-grained siliciclastic sediment, with sporadic medium to thick diatomite interbeds. These strata overlie the Quillagua Formation but, due to their localization adjacent to the canyon, they are considered terraces rather than part of the Soledad Formation. In the central and eastern sectors of the canyon, large crystals of gypsum and anhydrite pseudomorphs of gypsum are distinctive, ranging from millimeter to centimeter-scale blades locally horizontally oriented and elsewhere vertical. Multicentimeter crystal roses and centimeter-scale gypsum crystals that terminate downward are suggestive of crystal growth in void space. The decimeter-thick diatomites are well laminated and friable. Siliciclastic interbeds include anhydrite-cemented very fine to medium sandstone. Within the westernmost 3–4 km length of the Amarga Canyon, a localized gypsum-dominated unit occurs as terraces terminating at the junction with the Loa Canyon. These strata, which originally filled the incised canyon to a depth of approximately 20 m, are composed of horizontal and continuous beds of gypsum, diatomite, and laminated siltstone to very fine sandstone. A significant gypsum facies is 10–30 cm thick beds of vertical selenite crystals. Those horizontal beds overlie canyon-margin beds that are steeply inclined toward the canyon axis, which include a basal pebble to cobble conglomerate of basement and angular anhydrite clasts, interbeds of sandstone, and sedimentary breccias whose boulders are largely gypsum.

Both weakly indurated CaSO₄-rich silts [1] and mature CaSO₄-rich soils blanket the upper depositional surfaces of the upper Miocene and Pliocene formations as well as covering bedrock exposed in numerous hills of the western part of the basin. Immature CaSO₄-rich soils occur above the gravels of even the late Pleistocene alluvial fan deposits along the eastern margin of the depocenter [13].

We interpret the Quaternary sediments located across most of the Quillagua depocenter plain to have been created in a combination of fluvial, alluvial fan, playa, and eolian environments, or they are specifically associated with soil development. Of most paleogeographic importance, the rounded-clast conglomerates reflect transport and deposition in a fluvial system with persistent flow of surface water. The common association of these channel gravels with an omission of the Soledad Formation indicates the initial erosion or dissolution of Soledad evaporites by the stream, followed by deposition. The anomalously thick gravels at the Loa Canyon wall reflect a major paleo-channel (Figure 7(b)), below which Soledad Formation evaporites were fully removed and Hilaricos Formation evaporites partially dissolved, leading to collapse and brecciation of the Quillagua Formation strata. We hypothesize that these fluvial gravels reveal positions of the paleo-Loa River prior to the beginning of incision of the modern canyons (Figure 7(a), designated “Loa starting position”).

The gypsum-dominated strata in the lowermost Amarga Canyon represent deposition first by gravitational transport of detritus sourced from the canyon walls and sediment transport in channelized flow and later deposition from a body of water in a small lake. These strata indicate a local baselevel rise due to a temporary blockage of the mouth of the Amarga Canyon.

The largest change in the landscape since the end of widespread accumulation of sedimentary rocks is the incision of the Loa River and its tributary, the Amarga River, into the western part of the Quillagua depocenter. The mapping reported by Vásquez et al. [1] and the chronological and geomorphological studies reported by Ritter et al. [15] reveal that there must have been several intermediate steps between the end of Soledad Formation deposition approximately 2.6 Ma and canyon incision.

An extensive drainage network has not formed yet to convey surface water from most of the former depositional basin into the Loa Canyon (Figure 2), despite the large topographic difference between the bottom of the Loa Canyon at the confluence of the Amarga Canyon (590 masl) and the central parts of the Quillagua depocenter (e.g., approximately 760 masl in the Salar de Llamara center and 880 masl east of the Hilaricos hill). Prior to canyon incision, across roughly one-third of the area where Soledad Formation evaporites originally accumulated (Figure 7(c)), erosion removed the 2-25 m thickness of Soledad Formation and exposed the Quillagua Formation. Erosional mechanisms included channel incision and channel widening by surface water erosion and wind-deflation [13, 20]. Ritter et al. [15] found that exhumation of Cerro Soledad spanned $1.27±0.47$ to $0.274±0.075Ma$⁠. Given that two of Ritter et al.’s [15] ¹⁰Be exposure dates were for exhumation to the top of the Quillagua Formation (geological unit designations in Jordan et al. [38]), we interpret that these denudation ages represent the broad landscape lowering that removed the Soledad Formation. Removal of the Soledad Formation predated initiation of incision of the modern Loa Canyon in some locations (e.g., paleo-Loa paths, Figure 7(b)) and likely persisted until headward erosion of the Loa Canyon and Amarga Canyon significantly reorganized the drainage pattern.

The time at which incision of the Loa Canyon progressed from points west of the basin through the depocenter is constrained by a single data point from Vásquez et al. [1]. They established that the highest of the prominent terraces south of the confluence with the Amarga, inset 110 m below the rim of the canyon, had been eroded and then partially infilled with sediment by $0.255±0.017Ma$ (Figure 6). Another 5 erosional phases that terminated with terrace development occurred more recently, carving 50 m to the modern channel floor. Were the local rate of incision since 255 ka (0.18 m/ka) to be applied to the full 160 m depth of the canyon (Figure 6), then incision at this location began approximately 0.870 Ma. Given the 12 km Loa Canyon length from this dated location downstream to the western limit of the Quillagua depocenter, it is likely that capture of the Loa catchment by a stream that reached the Pacific shoreline was complete prior to 1 Ma. That capture led to an upstream-propagating wave of canyon incision that entered the west-central extreme of the depocenter around 1 Ma. The erosion that created topographic benches around Cerro Soledad ongoing at 0.274 Ma [15] likely was the last vestige of broad exhumation of the interior of the Quillagua depocenter as erosional activity shifted to be more focused along fixed river channel positions, as expressed by the 0.255 Ma terrace of the Loa Canyon.

Significant subsurface dissolution of the Hilaricos and Soledad formations occurred due to groundwater interaction with evaporite minerals and may also have occurred at the surface. In most of the regions of hilly terrain, beds in the Soledad and Quillagua Formations are commonly either brecciated or inclined at moderate angles that vary over distances of tens to hundreds of meters (e.g., the northwestern margin of Lomas de Sal, within the Hilaricos hills (Figure S11E), on the flanks of Cerro Soledad [38], and near Morro Gutierrez; Figure 2). We interpret that this disturbance of the strata was generated by either collapse into underlying voids or slumps into lateral voids created by dissolution. Sinkholes are more common in the plains (Figure 7(a)) (e.g., southeast of Quillagua village (Figure S11C) and within Salar de Llamara (Figure S11D)). Tentatively, we hypothesize that areas of extensive collapse of the Soledad Formation indicate paths of focused flow of surface and/or groundwater. Given the occurrence of Quaternary fluvial gravels directly overlying the Quillagua Formation adjacent to where the Soledad Formation collapsed, we speculate that elongate areas of karst and elongate areas of gravel may jointly indicate long-lasting positions of the paleo-Loa River (Figure 7(b)), prior to incision of the modern canyons.

The time of dissolution is constrained only as post-Soledad Formation. Prior to capture of the Quillagua depocenter by a river that reached the Pacific Ocean, stagnant groundwater in a hyperarid climate likely drove diagenetic recrystallization and cementation. Later, at the time incision of the Loa and Amarga canyons began, the ground water table throughout the Quillagua depocenter would have begun to decline [12]. A corresponding increase in the water table gradient probably increased the groundwater flow rate, decreased the length of time of contact of groundwater with the evaporite-bearing strata, and accelerated dissolution. Nevertheless, incision to depths approaching the modern Loa Canyon, >90 m below the base of the Quillagua depocenter evaporite units, lowered the groundwater of the southern half of the basin below depths of interaction with the easily dissolvable horizons (e.g., today the only reported spring within the Loa Canyon south of the confluence with the Amarga River drains an aquifer in the Oligo-Miocene conglomerate, underlying the Hilaricos Formation [55]). South of the Amarga River, the groundwater decline began prior to 0.255 Ma. North of the Amarga River and near the eastern margin, dissolution likely continues.

The Late Miocene–Pliocene Quillagua depocenter began in the distal zone of the PdT forearc basin, following a regional hiatus spanning 12 to 9 Ma formed when extreme hyperaridity deterred erosion and sediment transport. Thereafter, small variations in water and siliciclastic sediment fed into the Quillagua depocenter reflected either climate excursions [11, 13] or modifications in the catchment organization or area [3, 9]. The distribution of water within the basin responded to subtle tectonic deformation.

The basic tectonic elements of the PdT forearc basin have remained stable since the Oligocene: topographic asymmetry with higher mountains to the east than to the west, and maximum accommodation space along a north-south axial zone. Before the Late Miocene, the PdT filled with detritus from tectonically uplifting Andean source areas that interfingered with much smaller volumes shed from the tectonically quiescent Coastal Cordillera [1, 4, 44, 50, 56]. Whereas deformation since 10 Ma has been of small magnitude, it contributed to major paleogeographic changes in the Quillagua depocenter.

From 9 Ma to a time more recent than 4.7 Ma, drainage out of the basin exported surface water and/or groundwater to Salar Grande, and the corresponding drainage across the Quillagua depocenter flowed north of what are today Cerro Soledad and the Amarga River (Figures 7(c) and 7(d)). New dates in the Quillagua (Table 2) and Salar Grande basins [1] confirm Chong et al.’s [12] interpretation that halite of the Estratos del Salar Grande [1] accumulated in the Salar Grande basin at least partially contemporaneously with sulfate-dominated Soledad Formation deposition in the Quillagua depocenter. Those data also imply that the terminal Soledad Formation sulfates of Salar Grande are younger than 3.5 Ma, revealing that the northwestern drainage from the Quillagua depocenter persisted during much of the Pliocene. Thereafter, the topographic axis shifted sufficiently that surface water began to flow southwestward (Figure 7(b)).

The structural features adjacent to and within the Quillagua depocenter forced the progressive change of drainage. Uplifted blocks controlled by ENE-trending faults that define the northern margin of the depocenter became active during the early Pliocene [23]. During the hiatus between the Quillagua and Soledad formations, 4.7–4.1 Ma, the north-flank deformation is expressed in the Pampa Maya (Figure 3) by an angular unconformity (Figure S9D). We interpret that this progressive deformation of the northern basin flank uplifted the former drainage passageway at Montón de Gloria (today 830 masl) (Figure 2), so that the drainage axis shifted toward the main area of Salar de Llamara (today approximately 755 m elevation). Newly established hydraulic gradients then favored flow out of the west-central margin of the basin.

We hypothesize that there was an intervening step prior to incision of the modern Loa Canyon, when a pathway for groundwater escape from the Quillagua depocenter developed through porous conglomerates of the Oligocene–lower Miocene Quebrada Amarga beds and the Eocene–Oligocene Cañón del Loa beds [1]. The Cañón del Loa conglomerates filled an east-trending low region within the Coastal Cordillera. That topographic low was near but not originally connected to the western margin of the Quillagua depocenter [1]. The hypothesized groundwater flow may have initiated headward erosion via stream sapping, promoting the upstream progression of canyon incision that eventually captured the main Loa River. During the 1 to 1.5 million years between the end of the Soledad Formation and the arrival of deep canyon incision at the western edge of the Quillagua depocenter, sparsely distributed fluvial gravels within the southern depocenter suggest that the path of the Loa River remained in flux (Figures 7(a) and 7(b)).

Within the interior of the Quillagua depocenter, structural relief was created by near-surface folds and faults due to dissolution of tens of meters thickness of evaporite facies as well as small magnitude displacement across north-trending tectonic folds and faults. Under hyperarid conditions, structural relief by even a few tens of meters would have created surface relief, which would have impacted surface water drainage patterns. Water from the Andean foothills would have either deflected around a rising block (e.g., Cerro Soledad) or persisted as antecedent streams across an uplift (e.g., Tambillo Canyon across Lomas de Sal) (Figure 2). Also, prior to development of surface relief, the initial offset across the Lomas de Sal structure may have slowed groundwater transfer westward, leading to a rising water table east of the blind fault. This local groundwater rise may have localized the unusual salar facies during Quillagua lake time and thereafter the early cementation of a dune field.

A strong role for variable catchment area is expressed in the first major hydrological transition during the Late Miocene, from the Hilaricos salar to the Quillagua lake. The only sediment and water source known prior to 5.7 Ma is the immediately adjacent Andean Precordillera foothills [1]. In the time interval from 5.7 to 4.7 Ma (Figure 7(d)), the depocenter commonly held a standing water lake and produced the Quillagua Formation. The diverse fauna of diatoms, gastropods, and sponges are suggestive of slightly brackish water chemistry, very high concentration of dissolved silica, and a high pH [3, 5, 46]. Pisera and Sáez [46] note that malformation of sponge and diatom individuals is indicative of pollution of the water body by heavy metals. These chemical attributes, the lateral continuity of the Quillagua Formation 70 km south of the depocenter along the modern valley of the Loa River into the minor basin near Cerro Batea (Figures 1 and 3), and the relative change at the initiation of the Quillagua Formation to a positive water balance all imply the addition of a new southern source of water to the Quillagua depocenter. Although the Opache Formation facies indicate no integrated Loa surface drainage out of the Calama basin [9, 57], groundwater discharge from the westernmost Calama basin is implicated (Figure 3(c)), whose source was the volcanically active Andes.

The climate within the basin remained dominantly hyperarid, such that the 0.5–1.5 million year span between successive lake expansions entailed little erosion (Figure 5). Relationships of water-driven erosional and depositional landforms to gypsic soils serve as a proxy for alternations between hyperarid soil development and four short intervals of arid conditions that would have contributed more siliciclastic detritus and water to the alluvial fans [13, 17]. The first such slightly wetter interval is well dated, centered around 5.3 Ma [13, 58], marking Precordillera contributions to the early stage of Quillagua lake. Two later times of increased sediment input likely predated incision of the Loa in its modern canyon, yet they lack direct age constraints. Consequently, this climate proxy evidence cannot be compared to Quillagua lake system variability. The most recent interval of significantly increased precipitation occurred in the latest Pleistocene [31–33].

The Quillagua depocenter’s record of two stages of markedly different environments after it became connected to the Loa catchment (e.g., the fresh water Quillagua lake, the Soledad salar) contrasts with the lacustrine history in the upstream Calama Basin. In Calama, the Opache eastern lake and western wetlands developed at a similar time as the Quillagua lake. At least the eastern Opache zone persisted during the first half of Soledad Formation accumulation (Figures 3(b)and 3(c)) [16, 57, 59]. De Wet et al. [57] emphasize that groundwater was the primary input to the Calama’s Opache lake and wetlands, initially under semiarid conditions. A lack of well-resolved chronology in the Calama Basin prevents further analysis of the differences.

The evolving surface water input patterns and amounts had consequences for the Quillagua depocenter groundwater system, expressed in water table depth, hydraulic gradient, and composition. Whereas data for the groundwater itself are lacking, the history of groundwater changes is expressed in depositional and diagenetic changes revealed by the strata and the dissolution landforms. Figure 8 displays the major historical changes to the groundwater system.

The salt content of the groundwater was determined by climate, catchment geology, and eruptive history of the calderas. The salt source for the dominant evaporite minerals of the Quillagua and Salar Grande depocenters (i.e., anhydrite, gypsum, and halite) was examined by Pueyo et al. [6], who concluded that sulfur and bromine in the groundwater were sourced from the contemporaneous volcanic activity and hydrothermal systems. The number of dated tephra suggests two peak intervals for volcanic ash introduction into the Quillagua depocenter, 5.3-5.9 Ma (15 dated samples) and 2.8-4.1 Ma (12 dated samples) (Table 2), corresponding to eruptions of the Chuhuilla, Gaucha, and Carcote Ignimbrites and to the eruptions of the Pastos Grandes II, Tara, Atana, and Puripicar Ignimbrites, respectively (Table 1). Water from the Loa catchment would have included sources with direct contact to hydrothermal systems at the margins of calderas, a component not available within the Precordillera catchments (Figure 3). Water supplied from the Precordillera catchments likely was compositionally distinct from the paleo-Loa water composition. An indirect salt source was by reworking of sulfate-dominated soils of aeolian origin which were widespread throughout the hyperarid Atacama Desert during the Late Miocene and Pliocene [13, 17, 60]. For example, considering that sulfate paleosols preserved in the lowland sectors of the Precordillera catchments immediately east of the Quillagua depocenter are typically 0.5–1 m thick [13, 17], remobilization of similar air-distributed salts originally deposited across upland sectors of the Precordillera catchments into surface and groundwater could plausibly provide a significant fraction of the Soledad Formation salt volume.

Four categories of groundwater state persisted for lengthy time intervals (Figure 8). For one category, the top of the groundwater was almost at the surface across much of the hyperarid Quillagua depocenter area, which created a salar at the surface and highly saline groundwater, expressed by deposition of evaporite sediments (during the periods 9–7.2 Ma [Hilaricos] and 4.1–2.6 Ma [Soledad]) (Figures 8(a) and 8(e)). The second category encompasses repeated lengthy intervals with no sediment accumulation under hyperarid conditions. The lack of sediment implies an even more negative water balance, which forced lowering the water table to several meters below the ground surface and increasing groundwater salinity, corresponding to 7.2–7.1 Ma, 4.7–4.1 Ma, and 2.6 to less than 1 Ma (Figures 8(b), 8(d), and 8(f)). A third category featured a regionally much higher water table, whose corresponding lake was only moderately brackish (5.7–4.7 Ma [Quillagua]) (Figure 7(d)), which leads to the inference that the shallow groundwater was also of relatively low salinity. We infer that there would have been strong stratification of the groundwater, with relict dense brines beneath near-surface low salinity groundwater. The fourth groundwater state developed during the last million years and exists today, in which the water table is generally >20 m deep, controlled by the connection to the Loa Canyon (Figure 8(g)). A high gradient of water table depth across the basin would have increased the rate of groundwater lateral flow. The lowering of the water table would have progressively lowered the elevations of the interface of very saline relict brine with low salinity recent inflow from the Precordillera, triggering dissolution expressed by karst, and a new wave of diagenetic alteration. Today’s water table is at shallow depth only around Cerro Soledad, where water from ephemeral streams sourced in the Andean foothills (i.e., Mani stream, Figure 3(a)) is funneled, and this groundwater forms springs that supply the Amarga River. We infer that there exist strong lateral gradients in groundwater salinity. Throughout this groundwater history, the highly reactive evaporite strata of the Hilaricos and Soledad formations were repeatedly subjected to changing chemical conditions.

Since the Late Miocene, each new balance in the Quillagua depocenter between surface water drainage, climate, and tectonically driven accommodation was short lived, leading to rearrangements of the environmental system eight times in less than 10 million years. Consistent with studies elsewhere of the roles of drainage basin integration in mountain belts, e.g., [61], important environmental changes reflect the histories of changing balances among locations of water influx into the Quillagua depocenter and of changing points of water drainage out of the western side of the basin. Identified causes of shifts of water pathways were tectonic. One example is the relatively small magnitude structurally controlled changes of within-basin topography that uplifted the northern depocenter margin and caused a southward shift in the axis of drainage within the basin. The major extrabasinal change in Loa River drainage pattern also likely responded to subtle tectonic shifts of forearc topography, but specification of the location(s) and nature of that tectonic activity is beyond the scope of this study.

In addition to the environmental changes driven by the shifting Loa River path, the surface environmental conditions in the Quillagua depocenter were unstable. Even modest variations of the climate of the Central Depression and Precordillera were important because they shifted parts of the Precordillera catchment across the threshold for erosion and runoff. Each time the climate became slightly less hyperarid, water delivered siliciclastic materials and sulphates to localized salars, where groundwater was near the surface. Variations among degrees of aridity likely contributed to the demise of the Hilaricos salar, the onset of Ancachi alluvial fans, the expansion of the Quillagua lake, the desiccation of Quillagua lake, and changes in water balance that created and then ended salar deposition during the time of the Soledad Formation.

The groundwater chemistry would have changed markedly when the balance of inflow waters shifted from domination by Precordillera catchments, which lacked active volcanic centers, to domination by the much larger Loa catchment. The Loa River headwaters in the Andean volcanic arc would tap chemically more reactive source water, charged with ions from hydrothermally active volcanic centers and magma degassing. Whichever water chemistry dominated the influx during a given time, climate changes led to variations in the ionic concentration of the groundwater within the Quillagua depocenter aquifers, from relatively low salinity conditions during times of wetter climate in the catchment (e.g., early Quillagua lake) to highly concentrated brines during times of more widespread hyperaridity. The changing salinity and chemistry drove extensive diagenetic modification of primary sediments and of the early cements. An improved understanding of the groundwater evolution, diagenesis, and karst history could be derived from detailed studies of cement paragenesis, trace element chemistry, and possibly fluid inclusions.

Beginning prior to 1 Ma, up-river translation of incision of the Loa River from the Pacific Coast toward its current position within the Calama Basin changed dramatically the water balance within the Quillagua depocenter. Once the knickpoint entered the Quillagua depocenter (Figure 7(a)), the Loa efficiently drained both the Quillagua depocenter surface water and its groundwater.

Geochronological data included in Table 2 that are not previously published can be found in Supplemental files 3-6. Details for most previously published samples may be obtained by emailing a request to author Fernándo Sepúlveda. Map relationships are previously published (see reference list).

No conflicts of interest exist.

The authors are grateful to José Luis Díaz, Gonzalo Nuñez, Hugo Neira, Mario Martínez, Eduardo Martinez, Héctor Toro, Luís Guerra, Eugenio Orellana, and the late Antonio Díaz and Israel Acuña, for logistical services during field work. Lisandro Rojas and Jorge Arriagada of ENAP provided subsurface samples and data. H. Echaurren of Minera K+S and A. Gordon Miller of Golden Dragon Resources provided subsurface information and core samples. Andrew Tomlinson provided geochronological and technical advice, and Richard Allmendinger provided technical support. The efforts by authors Vásquez, Sepúlveda, Blanco, and Quezada, field expenses for those authors and in part for Jensen, as well as costs of radiometric dating were funded by the Servicio Nacional de Geología y Minería of Santiago, Chile. Field expenses for Jordan, and in part for Jensen, were supported by the United States National Science Foundation (grant numbers EAR-0609621 and EAR-1049978). Costs of publication are supported by the Cornell University College of Engineering.