The Atacama Desert is widely regarded as one of the most extreme deserts on the planet, with its hyperarid core today receiving less than two millimeters of precipitation per year. This remarkable dryness stems from a combination of factors: the desert’s position under the subsiding branch of the Hadley circulation, the rain-shadow effect of the Andes, and the cooling influence of the Humboldt Current, whose cold upwelled waters suppress moisture transport from the Pacific. Geological evidence indicates that arid conditions may date back as far as the Late Cretaceous or Oligocene, intensifying through the Miocene. Yet the picture is not one of unbroken desiccation. Sedimentary records from surrounding regions reveal that the overall hyperaridity was repeatedly interrupted by wetter phases, and shorter archives such as paleo-wetland deposits and rodent middens document vegetation shifts over the past 50,000 years linked to variations in the South American Summer Monsoon.

The problem, the researchers explain, is that most existing records do not actually reflect local rainfall. Many of the sediment sequences studied so far received their moisture through groundwater or surface flows originating in the wetter Precordillera and Altiplano, blurring the local climate signal. Archives from the Coastal Cordillera, which are hydrologically disconnected from these external water sources, are extremely rare. This is precisely what makes the clay pans so valuable. These endorheic basins formed when tectonic activity along fault scarps blocked ancient river channels, creating terminal basins that today are fed only episodically by runoff from small local catchments. Any sediment that accumulates in them records precipitation that fell within the desert itself.

The two study sites differ in their settings but share a common origin. The PAG clay pan, at 940 meters altitude, sits in the driest part of the desert with mean precipitation below one millimeter per year. It was formed when movement along the reverse Adamito Fault, active since at least the Miocene, dammed a meandering paleo-drainage with a fault scarp offset of up to 130 meters. The Paranal clay pan, at 2,231 meters on the Pampa Remiendos, receives around five millimeters of rain annually, mostly during the austral winter. It formed when a trace of the Quebrada Grande Fault System, part of the Atacama Fault System, blocked a channel that once drained the Sierra Vicuña Mackenna toward the Pacific. Cosmogenic nuclide exposure dating of quartz pebbles from the abandoned outflow channels indicates that both drainages were finally abandoned during the Middle to Late Pliocene, providing minimum ages for the basins’ formation.

Before committing to expensive deep drilling, the team conducted pilot coring with a handheld percussion system. A 6.2-meter core from PAG proved remarkably informative: luminescence and paleomagnetic dating suggest it spans at least the last 215,000 years, with coarse-grained deposits indicating stronger fluvial input during Marine Isotope Stages 7 to 5 and finer sediments recording a progressive drying from MIS 4 onward. A shorter 1.88-meter core from Paranal, covering roughly the last 32,000 years, revealed that significant runoff into the basin today occurs only during infrequent ENSO-related heavy rains exceeding 20 millimeters, which strike on average every six years. Intriguingly, coarser layers intercalated at intervals of roughly 1,000 to 1,400 years point to millennial-scale flood events far exceeding anything in the historical record.

To peer deeper, the researchers deployed transient electromagnetic (TEM) soundings and active seismic refraction surveys across both basins. At PAG, 56 TEM soundings and a 576-meter seismic profile revealed a consistent three-layer structure: a resistive bedrock basement, overlain by a highly conductive basal sediment unit reaching about 70 meters in thickness, capped by roughly 30 meters of more resistive sediments. The maximum sediment thickness was estimated at 103 plus or minus 10 meters, with the deepest point coinciding with the head of the abandoned paleo-channel. At Paranal, where 116 soundings were recorded, the same three-layer architecture emerged, but with a dramatic twist: the conductive layer follows a channel-like depression incised up to 80 meters into the granitic basement, interpreted as the buried course of the ancient river that once flowed through the site. Total sediment thickness there reaches 160 plus or minus 10 meters in the basin center.

The drilling campaigns themselves pushed the limits of technology in an environment with no water to spare. The team used a truck-mounted sonic drilling rig, which advances without cooling liquids, making it ideal for remote deserts. At PAG in 2017, four boreholes were drilled to a maximum depth of 51 meters before technical issues halted operations, leaving the deepest 50 meters of the predicted sequence unrecovered. At Paranal in 2022, sonic drilling reached 52.3 meters before a massive cobble layer forced a switch to rotary diamond drilling, which ultimately penetrated to 174 meters, about three meters into weathered bedrock. Core recoveries were excellent, often exceeding 100 percent due to sediment expansion during sonic coring. Downhole logging of natural gamma radiation, magnetic susceptibility, and electrical conductivity provided continuous physical records to complement the visual core descriptions.

The cores tell strikingly different stories in their lower sections. At PAG, the basal unit consists of fine-grained clay and silt deposited in a persistent playa lake, interbedded with calcium sulfate layers formed during episodes of enhanced evaporation. The absence of erosional surfaces suggests the lake never dried completely, implying that precipitation once balanced or exceeded evaporation over extended periods, conditions far wetter than today. At Paranal, by contrast, the lower 110 meters comprise coarse, poorly sorted conglomerates with clasts up to 1.3 meters across, deposited by a powerful river system before the tectonic blocking, followed by well-sorted fluvial sands. Even the weathered granodioritic bedrock at the bottom of the Paranal hole bears chemical alteration signatures pointing to a much wetter past. In both cores, the upper sections shift to alluvial-fan deposition, including proximal mudflows and debris flows, sheetflood deposits, distal sediment flows, and pedogenic calcium sulfate paleosols, a facies assemblage typical of the torrential-rain-dominated regime seen in the Atacama today.

One especially evocative find was a roughly 15-centimeter volcanic ash layer in the Paranal core at 14 meters depth, confirmed by microscopic volcanic glass shards and likely sourced from a distant eruption in the Altiplano–Puna volcanic complex or the southern Central Volcanic Zone of the Andes. Once its glass geochemistry is analyzed, this tephra could serve as a precise chronological marker. The team also documented a fossilized desiccation crack at PAG, filled with loose sand and gravel, marking a paleo-surface and a significant hiatus when the ancient lake bed was exposed to subaerial conditions, a process the researchers even replicated in a 14-day irrigation experiment on the modern clay pan surface.

Taken together, the two records sketch an evolution from wetter conditions with more persistent rainfall toward the episodic, torrential precipitation patterns of the present, with increasing paleosol formation and alluvial-fan starvation signaling progressive aridification. The differences between the two sites, only 250 kilometers apart, may reflect different precipitation histories, different basin sensitivities, or gaps in their chronologies. Establishing robust age models is now the greatest challenge, since radiocarbon and luminescence methods reach their limits within the first core meters. The team plans to apply tephrochronology, magnetostratigraphy, U-series dating of evaporites, and cosmogenic nuclide burial dating, techniques capable in principle of reaching back to the Middle Miocene. If successful, these humble desert basins will yield an unprecedented account of how the driest heart of the Atacama responded to global tectonic, oceanic, and atmospheric changes over millions of years.

Subject of Research: Deep drilling of Neogene sediment archives in Atacama Desert clay pans to reconstruct long-term precipitation history

Article Title: Unearthing the climate history of the Atacama Desert in northern Chile – deep drilling in two clay pans of the Coastal Cordillera

Article References: Wennrich, V., Diederich-Leicher, J., Blanco-Arrué, B. N., Büttner, C., Buske, S., Campos Sepulveda, E., Dunai, T., Feller, J., Galego, E., Hasberg, A., Leicher, N., López, D. A., Maldonado, J., Medialdea, A., Ninnemann, L., Perryman, R., Ríos-Contesse, J. C., Ritter, B., Scheidt, S., … Melles, M. (2025). Unearthing the climate history of the Atacama Desert in northern Chile – deep drilling in two clay pans of the Coastal Cordillera. Scientific Drilling, 34(1/2), 1-20. https://doi.org/10.5194/sd-34-1-2025

Image Credits: AI Generated

DOI: 10.5194/sd-34-1-2025

Keywords: Atacama Desert, paleoclimate, scientific drilling, clay pans, Coastal Cordillera, hyperaridity, Neogene, sediment cores, geophysics, transient electromagnetics, Chile, precipitation history

The research, led by Yan Lévesque of the Institut National de La Recherche Scientifique and Université du Québec à Chicoutimi, together with Romain Chesnaux, Julien Walter and Lamine Boumaiza, tackles one of the most persistent problems in hydrogeology: calibrating a numerical model when direct measurements of groundwater levels are sparse. Groundwater models are the workhorses of water resource management, used to predict how aquifers will respond to pumping, drought and climate change. Yet a model is only as good as the data used to constrain it, and in many regions of the world, observation wells are few and far between.

To build their model, the team assembled an unusually rich geological picture of the moraine system. They drew on 94 boreholes drilled through the glacial deposits, five stratigraphic cross-sections that traced the layering of sediments across the landscape, and three complementary geophysical techniques: 20 transient electromagnetic surveys, 6 electrical resistivity tomography surveys, and 6 ground-penetrating radar surveys. Each method probes the subsurface in a different way. Transient electromagnetic surveys induce currents in the ground and measure how they decay, revealing contrasts in electrical conductivity that distinguish saturated from unsaturated materials. Electrical resistivity tomography injects current through electrodes and maps resistivity variations in two-dimensional slices, while ground-penetrating radar uses high-frequency radio waves to image shallow stratigraphy with centimeter-scale resolution.

From this combined dataset, the researchers constructed an integrated three-dimensional geological framework capturing the complex, heterogeneous and anisotropic character of the aquifer system. Glacial deposits such as moraines are notoriously difficult to model because they were deposited by ice in chaotic, laterally discontinuous layers. A single borehole may reveal coarse sand at one location while a neighboring hole, only a few hundred meters away, encounters dense till. Geophysics helps bridge those gaps by interpolating between the hard constraints of the boreholes with continuous geophysical profiles.

With the geological architecture in place, the team turned to numerical simulation. They used FEFLOW, a widely applied finite element software package for modeling groundwater flow, mass and heat transport in porous media, to run flow simulations and define initial estimates of hydraulic parameters such as hydraulic conductivity, which describes how easily water moves through the subsurface. These initial values were then refined through calibration, the process of adjusting model parameters until simulated groundwater levels match observed ones as closely as possible.

The crucial innovation lay in what the researchers used as calibration targets. Instead of relying solely on water levels measured in wells, they employed groundwater levels derived from the geophysical data themselves. Because the electrical resistivity of sediments changes sharply at the water table, where dry unsaturated material gives way to saturated material, geophysical surveys can be inverted to estimate the depth of the saturated zone. The team used the FePEST module, an interface between FEFLOW and the PEST parameter estimation software, to automatically adjust hydraulic parameters until the model reproduced these geophysics-derived water levels.

The real test came afterward. The researchers evaluated the calibrated model against an entirely independent dataset: piezometric measurements from 26 observation wells distributed across the study area. These wells had not been used in the calibration, so they provided an unbiased measure of the model’s predictive skill. The comparison between calibrated and observed groundwater levels yielded a root mean square error of 3.69 meters, a relatively low value for a regional-scale model of heterogeneous glacial terrain, indicating good agreement with field observations.

That result carries weight beyond one moraine in Quebec. In many parts of the world, including remote northern communities, agricultural regions with sparse monitoring networks, and developing countries investing in groundwater resources for the first time, the cost of drilling observation wells limits how well aquifers can be understood. Geophysical surveys are faster and often cheaper per unit of spatial coverage than drilling, and this study demonstrates that their output can serve not merely as a qualitative check on a model but as quantitative calibration data that actively refines hydraulic parameters.

The approach also reframes the traditional division of labor between geophysics and hydrogeology. Conventionally, geophysical data have been used to build geological models, delineating aquifer boundaries and layer thicknesses, while calibration has depended on hydraulic measurements. The new work shows that geophysically derived groundwater levels can play both roles, supporting the geological framework and simultaneously constraining the flow model. This dual use of a single dataset maximizes the value of every field campaign and reduces the uncertainty that accumulates when models are calibrated against too few points.

The findings arrive at a moment when pressure on groundwater is intensifying worldwide. Aquifers feed rivers, sustain ecosystems and supply roughly half of the world’s drinking water, yet many are being depleted faster than they recharge. Reliable regional models are essential for managing this invisible resource, and the Quebec study offers a practical template: combine boreholes, multiple geophysical methods and automated calibration to extract the maximum information from limited field data. As the authors note, integrating geophysical datasets with conventional hydrogeological observations improves parameter estimation and enhances the reliability of regional groundwater flow models, a step toward water management decisions grounded in a clearer picture of what lies beneath.

Subject of Research: Calibration of a 3D regional groundwater flow model using groundwater levels derived from geophysical surveys in Quebec

Article Title: Calibration of a 3D regional groundwater model using geophysics-derived groundwater levels

Article References: Lévesque, Y., Chesnaux, R., Walter, J., & Boumaiza, L. (2026). Calibration of a 3D regional groundwater model using geophysics-derived groundwater levels. Hydrogeology Journal. https://doi.org/10.1007/s10040-026-03144-w

Image Credits: AI Generated

DOI: 10.1007/s10040-026-03144-w

Keywords: groundwater modeling, geophysics, hydrogeology, aquifer calibration, transient electromagnetics, electrical resistivity tomography, ground-penetrating radar, FEFLOW, Saint-Narcisse Moraine, Quebec, hydraulic conductivity, water table