Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Deep”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 703 records · Page 39
All sourcesNASA NTRSDOE OSTI
DOE OSTI2025

Increasing the Stability of Deep Blue Phosphor‐Sensitized OLEDs Using the Polariton‐Enhanced Purcell Effect

Abstract The stability of efficient, deep blue organic light‐emitting diodes (OLEDs) remains a major challenge in the field of organic electronics. Poor device stability originates from the high probability for destructive non‐radiative triplet exciton annihilation events. Phosphor‐sensitized fluorescence (PSF) is proposed to achieve an efficient, deep blue color by energy transfer from phosphors to fluorophores. Recently, the polariton‐enhanced Purcell (PEP) effect is introduced to decrease the triplet radiative lifetime and density, resulting in an increase in blue phosphorescent OLED lifetime. Here, the PEP effect is introduced to enhance the stability of the PSF‐OLEDs. It is shown that the PEP effect increases all radiative decay rates of the phosphors and fluorophores, leading to a reduction in the triplet annihilation events. Using a Pt‐complex phosphor sensitizer and a so‐called multi‐resonance fluorescent emitter in a PEP cavity, a 3.1‐fold lifetime increase is observed at a current density ofJ= 10 mA cm −2 , reduced EQE roll‐off and deep blue color with Commission Internationale de l'Eclairage coordinates of (0.13, 0.09). The PEP effect maximally extends PSF‐OLED lifetimes in devices with the highest triplet‐to‐singlet energy transfer rates. Moreover, this work suggests that the benefits of PEPs apply to all triplet‐based OLEDs.

Chemistry↗
DOE OSTI2024

Report on the AAPM grand challenge on deep generative modeling for learning medical image statistics

Abstract Background The findings of the 2023 AAPM Grand Challenge on Deep Generative Modeling for Learning Medical Image Statistics are reported in this Special Report. Purpose The goal of this challenge was to promote the development of deep generative models for medical imaging and to emphasize the need for their domain‐relevant assessments via the analysis of relevant image statistics. Methods As part of this Grand Challenge, a common training dataset and an evaluation procedure was developed for benchmarking deep generative models for medical image synthesis. To create the training dataset, an established 3D virtual breast phantom was adapted. The resulting dataset comprised about 108 000 images of size 512 512. For the evaluation of submissions to the Challenge, an ensemble of 10 000 DGM‐generated images from each submission was employed. The evaluation procedure consisted of two stages. In the first stage, a preliminary check for memorization and image quality (via the Fréchet Inception Distance [FID]) was performed. Submissions that passed the first stage were then evaluated for the reproducibility of image statistics corresponding to several feature families including texture, morphology, image moments, fractal statistics, and skeleton statistics. A summary measure in this feature space was employed to rank the submissions. Additional analyses of submissions was performed to assess DGM performance specific to individual feature families, the four classes in the training data, and also to identify various artifacts. Results Fifty‐eight submissions from 12 unique users were received for this Challenge. Out of these 12 submissions, 9 submissions passed the first stage of evaluation and were eligible for ranking. The top‐ranked submission employed a conditional latent diffusion model, whereas the joint runners‐up employed a generative adversarial network, followed by another network for image superresolution. In general, we observed that the overall ranking of the top 9 submissions according to our evaluation method (i) did not match the FID‐based ranking, and (ii) differed with respect to individual feature families. Another important finding from our additional analyses was that different DGMs demonstrated similar kinds of artifacts. Conclusions This Grand Challenge highlighted the need for domain‐specific evaluation to further DGM design as well as deployment. It also demonstrated that the specification of a DGM may differ depending on its intended use.

Radiology, Nuclear Medicine & Medical Imaging↗
DOE OSTI2015

Deep Subgap Feature in Amorphous Indium Gallium Zinc Oxide. Evidence Against Reduced Indium

Amorphous indium gallium zinc oxide (a-IGZO) is the archetypal transparent amorphous oxide semiconductor. In spite of the gains made with a-IGZO over amorphous silicon in the last decade, the presence of deep subgap states in a-IGZO active layers facilitate instabilities in thin film transistor properties under negative bias illumination stress. Several candidates could contribute to the formation of states within the band gap. We present evidence against In + lone pair active electrons as the origin of the deep subgap features. No In + species are observed, only In0 nano-crystallites under certain oxygen deficient growth conditions. Our results further support under coordinated oxygen as the source of the deep subgap states.

36 MATERIALS SCIENCE↗
DOE OSTI2019

A Hybrid Deep Learning Architecture for Classification of Microscopic Damage on NIF Laser Optics

Accurately classifying microscopic damage helps automate the repair and recycling of National Ignition Facility optics and informs the study of damage initiation and growth. This complex 12-class problem previously required human experts to distinguish and label the various damage morphologies. Finding image analysis methods to extract and calculate distinguishing features would be time consuming and challenging, so we sought to automate this task by using convolutional neural networks (CNNs) pretrained on the ImageNet database to take advantage of its automated feature discovery and extraction. We compared three model architectures on this dataset and found the one with highest overall accuracy, 99.17%, was a novel hybrid architecture, one in which we removed the final decision-making layer of the deep learner and replaced it with an ensemble of decision trees (EDT). This combines the power of feature extraction by CNNs with the decision-making strength of EDT. The accuracy of the hybrid architecture over the deep learning alone is shown to be significantly improved. Furthermore, we applied this novel hybrid architecture to an entirely different dataset, one containing images of repaired damage sites, and improved on the previously published findings, also with a demonstrably significant increase in accuracy over using the deep learner alone.

47 OTHER INSTRUMENTATION↗
DOE OSTI2019

Using Deep Learning for Automated Communication Pattern Characterization: Little Steps and Big Challenges

Characterization of a parallel application’s communication patterns can be useful for performance analysis, debugging, and system design. However, obtaining and interpreting a characterization can be difficult. AChax implements an approach that uses search and a library of known communication patterns to automatically characterize communication patterns. Our approach has some limitations that reduce its effectiveness for the patterns and pattern combinations used by some real-world applications. By viewing AChax’s pattern recognition problem as an image recognition problem, it may be possible to use deep learning to address these limitations. In this position paper, we present our current ideas regarding the benefits and challenges of integrating deep learning into AChax and our conclusion that a hybrid approach combining deep learning classification, regression, and the existing AChax approach may be the best long-term solution to the problem of parameterizing recognized communication patterns.

Roth, Philip↗
DOE OSTI2025

Heterogeneous Mixtures of Dictionary Functions to Approximate Subspace Invariance in Koopman Operators: Why Deep Koopman Operators Work

Abstract Koopman operators model nonlinear dynamics as a linear dynamic system acting on a nonlinear function as the state. This nonstandard state is often called a Koopman observable and is usually approximated numerically by a superposition of functions drawn from a dictionary . In a widely used algorithm, extended dynamic mode decomposition (EDMD), the dictionary functions are drawn from a fixed class of functions. Deep learning combined with EDMD has been used to learn novel dictionary functions in an algorithm called deep dynamic mode decomposition (deepDMD). The learned representation both (1) accurately models and (2) scales well with the dimension of the original nonlinear system. In this paper, we analyze the learned dictionaries from deepDMD and explore the theoretical basis for their strong performance. We explore State-Inclusive Logistic Lifting (SILL) dictionary functions to approximate Koopman observables. Error analysis of these dictionary functions show they satisfy a property of subspace approximation, which we define as uniform finite approximate closure. Typically, a Koopman dictionary’s nonlinear functions are homogeneous. In this paper, we discover that structured mixing of heterogeneous dictionary functions drawn from different classes of nonlinear functions achieve the same accuracy and dimensional scaling as the deep-learning-based deepDMD algorithm Yeung et al. ( In: 2019 American Control Conference (ACC), 2019). We specifically show this by building a heterogeneous dictionary comprised of SILL functions and conjunctive radial basis functions (RBFs). This mixed dictionary achieves similar accuracy and dimensional scaling to deepDMD with an order of magnitude reduction in parameters, while maintaining geometric interpretability. These results strengthen the viability of dictionary-based Koopman models to solving high-dimensional nonlinear learning problems.

Johnson, Charles A.↗
DOE OSTI2015

Evolution of the deep Atlantic water masses since the last glacial maximum based on a transient run of NCAR-CCSM3

During the last deglaciation (from approximately 21 to 11 thousand years ago), the high latitudes of the Atlantic Ocean underwent major changes. Besides the continuous warming, the polar and subpolar ocean surface received a large amount of meltwater from the retracting ice sheets. These changes in temperature and salinity affected deep waters, such as the Antarctic Bottom Water (AABW) and the North Atlantic Deep Water (NADW), which are formed in the Southern Ocean and in the northern North Atlantic, respectively. In this study, we present the evolution of the physical properties and distribution of the AABW and the NADW since the last glacial maximum using the results of a transient simulation with NCAR-CCSM3. In this particular model scenario with a schematic freshwater forcing, we find that modern NADW, with its characteristic salinity maximum at depth, was absent in the beginning of the deglaciation, while its intermediate version—Glacial North Atlantic Intermediate Water (GNAIW)—was being formed. GNAIW was a cold and relatively fresh water mass that dominated intermediate depths between 60 and 20°N. At this time, most of the deep and abyssal Atlantic basin was dominated by AABW. Within the onset of the Bølling-Allerød period, at nearly 15 thousand years ago (ka), GNAIW expanded southwards when the simulated Meridional Overturning Circulation overshoots. The transition between GNAIW and NADW ocurred after that, when AABW was fresh enough to allow NADW to sink deeper in the water column. When the NADW appears (~11 ka), AABW retracts and is constrained to lie near the bottom.

Meteorology & Atmospheric Sciences↗
DOE OSTI2019

Deoxidation Mechanism in Reduced Titanium Powder Prepared by Multistage Deep Reduction of TiO{sub 2}

There are technical bottlenecks in the preparation of metallic titanium powder by conventional thermal reduction of TiO{sub 2} with calcium, magnesium, or other metals, such as low efficiency and incomplete deoxidation. Based on the thermodynamic equilibrium characteristics of the step-by-step reduction of TiO{sub 2}, a comparative study method was used, and the thermodynamic and kinetic differences between thermal reduction reactions of TiO{sub 2} with magnesia and calcium were investigated in this paper. Considering the difference in the electronegativity characteristics between magnesium and calcium, a new idea was put forward that the primary reduction product is prepared through a primary reduction process of TiO{sub 2} in the form of magnesiothermic self-propagating high-temperature synthesis (SHS), and then the reduced titanium powder is obtained through the primary reduction product for a deep calcium-thermal reduction reaction. The reduction degree and mechanism of TiO{sub 2} during the primary magnesium thermal reaction and deoxidation mechanism of primary reduction products in different deep reduction modes were investigated in this paper. The results showed that it is more conducive to undergo complete deoxygenation such that the primary reduction product is first acid impregnated, followed by deep calciothermic reduction. High-purity reduced titanium powder with a final oxygen content of only 0.21 pct and purity greater than 99.0 pct was obtained.

36 MATERIALS SCIENCE↗
DOE OSTI2019

Planar Anisotropy, Tension–Compression Asymmetry, and Deep Drawing Behavior of Commercially Pure Titanium at Room Temperature

The planar anisotropy (PA) and tension–compression asymmetry (TCA) of the CP-Ti were investigated via uniaxial tension and compression tests at room temperature. The formability and the earing behavior of the CP-Ti sheet were studied via deep drawing experiment. The deep drawing simulations using the uniaxial tensile and the compressive curves as the hardening rules were compared with each other and with the experimental results. The CP-Ti sheet showed PA and TCA in yielding and strain hardening. The PA was characterized by the plastic strain ratio r{sub 0}, r{sub 45}, and r{sub 90} of 1.47, 2.06, and 2.05. The TCA showed PA, which showed tension–compression yield strength ratios of 1.12, 1.08, and 1.04 in 0°, 45°, and 90° in the rolling direction, and tensile and compressive hardening exponent ratios of 0.86, 0.8, and 0.62. The orientation distribution functions (ODFs) analysis demonstrated that the tensile and the compressive deformation textures were different and showed PA. The simulative results, including the simulated forming force and the earing profiles using the uniaxial tensile and compressive curves as the hardening rules, were not in agreement with each other. The results were not in good agreement with the experimental results, implying that the TCA had an important effect on the formability of the sheet. The TCA tended to reduce the thickness of the deep drawing parts, increase the earing ratio, and affect the drawing force.

36 MATERIALS SCIENCE↗
DOE OSTI2025

Karhunen–Loève deep learning method for surrogate modeling and approximate Bayesian parameter estimation

We evaluate the performance of the Karhunen-Loève Deep Neural Network (KL-DNN) framework for surrogate modeling and approximate Bayesian parameter estimation in partial differential equation models. In the surrogate model, the Karhunen-Loève (KL) expansions are used for the dimensionality reduction of the number of unknown parameters and variables, and a deep neural network is employed to relate the reduced space of parameters to that of the state variables. The KL-DNN surrogate model is used to formulate a maximum-a-posteriori-like least-squares problem, which is randomized to draw samples of the posterior distribution of the parameters. We test the proposed framework for a hypothetical unconfined aquifer via comparison with the forward MODFLOW and inverse PEST++ iterative ensemble smoother (IES) solutions as well as the state-of-the-art Fourier neural operator (FNO) and deep operator networks (DeepONets) operator learning surrogate models. Our results show that the KL-DNN surrogate model outperforms FNO and DeepONet for forward predictions. For solving inverse problems, the randomized algorithm provides the same or more accurate Bayesian predictions of the parameters than IES as evidenced by the higher log-predictive probability of both the estimated parameter field and the forecast hydraulic head. The posterior mean obtained from the randomized algorithm is closer to the reference parameter field than that obtained with FNO as the maximum a posteriori estimate.

Approximate Bayesian inference↗
DOE OSTI2013

Deep subsurface drip irrigation using coal-bed sodic water: Part I. Water and solute movement

Water co-produced with coal-bed methane (CBM) in the semi-arid Powder River Basin of Wyoming and Montana commonly has relatively low salinity and high sodium adsorption ratios that can degrade soil permeability where used for irrigation. Nevertheless, a desire to derive beneficial use from the water and a need to dispose of large volumes of it have motivated the design of a deep subsurface drip irrigation (SDI) system capable of utilizing that water. Drip tubing is buried 92 cm deep and irrigates at a relatively constant rate year-round, while evapotranspiration by the alfalfa and grass crops grown is seasonal. We use field data from two sites and computer simulations of unsaturated flow to understand water and solute movements in the SDI fields. Combined irrigation and precipitation exceed potential evapotranspiration by 300–480 mm annually. Initially, excess water contributes to increased storage in the unsaturated zone, and then drainage causes cyclical rises in the water table beneath the fields. Native chloride and nitrate below 200 cm depth are leached by the drainage. Some CBM water moves upward from the drip tubing, drawn by drier conditions above. Chloride from CBM water accumulates there as root uptake removes the water. Year over year accumulations indicated by computer simulations illustrate that infiltration of precipitation water from the surface only partially leaches such accumulations away. Field data show that 7% and 27% of added chloride has accumulated above the drip tubing in an alfalfa and grass field, respectively, following 6 years of irrigation. Maximum chloride concentrations in the alfalfa field are around 45 cm depth but reach the surface in parts of the grass field, illustrating differences driven by crop physiology. Deep SDI offers a means of utilizing marginal quality irrigation waters and managing the accumulation of their associated solutes in the crop rooting zone.

01 COAL, LIGNITE, AND PEAT↗
DOE OSTI2016

Daylight performance of a microstructured prismatic window film in deep open plan offices

Daylight redirecting systems with vertical windows have the potential to offset lighting energy use in deep perimeter zones. Here, a microstructured prismatic film designed for such use was characterized using goniophotometric measurements and ray tracing simulations. The synthetically-generated bidirectional scattering distribution function (BSDF) data were shown to have good agreement with limited measured data for normal incident angles (0–60°). Measured data indicated that the prismatic film was most efficient when vertical angles of incidence were between 18 and 35° and within ±45° of normal incidence to the plane of the window so maximum energy savings across the full depth of the zone occurred over the equinox to winter solstice period. Annual lighting energy use and visual comfort in a deep open plan office zone were evaluated using the Radiance three-phase method in several climates and for south and east-facing window orientations. Lighting energy savings were 39–43% for a 12 m (40 ft) deep south-facing perimeter zone compared to the same zone with no lighting controls. The prismatic film with and without a diffuser controlled glare for views parallel to the window but produced glare for seated viewpoints looking toward the window. At mature market costs, the system was projected to have a simple payback of 2–6 years. Technical challenges encountered throughout the evaluation led to improvements in measurement and modeling tools and stressed the importance of having accurate input data for product development.

29 ENERGY PLANNING, POLICY, AND ECONOMY↗
DOE OSTI2021

Physicochemical properties of the Portland cement-based mortar exposed to deep seafloor conditions at a depth of 1680 m

Achieving breakthroughs in marine technologies requires the development of infrastructures submerged in deep-sea environments, whose physicochemical effects on cement-based materials considerably differ from those of shallow seas. However, very few studies focused on the cement-based materials subjected to deep-sea conditions. This work investigates the changes in the compressive strength and phase composition of the cement mortar kept on the seafloor with a depth of 1680 m for 608 d. The mortar specimens salvaged from the seafloor exhibited severe visible damages, including softened mashy structures and significantly decreased compressive strengths. The obtained X-ray diffraction, scanning electron microscopy, and nuclear magnetic resonance data revealed that the dissolution of portlandite, decalcification of calcium (alumino) silicate hydrate, and formation of brucite, magnesium (alumino) silicate hydrate, a hydrotalcite-like phase, thaumasite, and ettringite likely contributed to the disintegration of mortar, which could be further accelerated by the low temperature of the deep-sea environment.

36 MATERIALS SCIENCE↗
DOE OSTI2024

Bayesian reduced-order deep learning surrogate model for dynamic systems described by partial differential equations

We propose a reduced-order deep-learning surrogate model for dynamic systems described by time-dependent partial differential equations. This method employs space–time Karhunen–Loève expansions (KLEs) of the state variables and space-dependent KLEs of space-varying parameters to identify the reduced (latent) dimensions. Subsequently, a deep neural network (DNN) is used to map the parameter latent space to the state variable latent space. An approximate Bayesian method is developed for uncertainty quantification (UQ) in the proposed KL-DNN surrogate model. The KL-DNN method is tested for the linear advection–diffusion and nonlinear diffusion equations, and the Bayesian approach for UQ is compared with the deep ensembling (DE) approach, commonly used for quantifying uncertainty in DNN models. It was found that the approximate Bayesian method provides a more informative distribution of the PDE solutions in terms of the coverage of the reference PDE solutions (the percentage of nodes where the reference solution is within the confidence interval predicted by the UQ methods) and log predictive probability. The DE method is found to underestimate uncertainty and introduce bias. For the nonlinear diffusion equation, we compare the KL-DNN method with the Fourier Neural Operator (FNO) method and find that KL-DNN is 10% more accurate and needs less training time than the FNO method.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

Investigation of enhanced CO 2 storage in deep saline aquifers by WAG and brine extraction in the Minnelusa sandstone, Wyoming

Geological CO 2 sequestration in deep saline aquifers has been extendedly investigated to reach the goal of carbon neutral as large amount of CO 2 can be reduced in a short time. Here, this study investigates the feasibility of water-alternating-gas (WAG) injection and brine extraction on enhancing CO 2 storage in a target deep saline aquifer, Minnelusa Sandstone in the Powder River Basin (PRB) of Wyoming. An integrated numerical model is developed to account for four scenarios: (1) CO 2 continuous injection (CI) through one injection well, (2) WAG injection through one injection well, (3) CO 2 CI through one injection well and brine extraction through one producer, (4) WAG injection through one injection well and brine extraction through one producer. Simulation results suggest that WAG injection and brine extraction corporately make it possible to enhance CO 2 injectivity while securing CO 2 storage safety. For instance, WAG injection considerably reduces structural trapping contribution while enhances dissolved and residual trapping contributions. As for brine extraction, it can decrease the maximum averaged reservoir pressure by 37% and 22% for CI and WAG injection, respectively. Besides, sensitive analyses of the operational parameters for the fourth scenarios are performed. Results reveal that, when the amount of total CO 2 injection has been predetermined, CO 2 injection time and rate within each period of WAG should be small whereas it is preferable to have large water injection time and rate. This is to secure desirable dissolved and residual trapping contributions (store CO 2 with safety). Sensitive degree results confirm that CO 2 injection rate, water injection time, and CO 2 injection time of WAG injection exhibit significant effects on CO 2 storage whereas the impacts of producer bottom-hole pressure (BHP), water injection rate, and CO 2 -water injection time are relatively weaker. This study not only sheds light on achieving double-win goal: enhance CO 2 injectivity and store CO 2 with safety, but also can be a critical reference for other CO 2 storage in deep saline aquifers.

58 GEOSCIENCES↗
DOE OSTI2023

Towards deep computer vision for in-line defect detection in polymer electrolyte membrane fuel cell materials

Polymer Electrolyte Membrane (PEM) fuel cells are a promising source of alternative energy. However, their production is limited by a lack of well-established methods for quality control of their constituent materials like the membrane-electrode assembly during roll-to-roll manufacturing. One potential solution is the implementation of deep learning methods to detect unwanted defects through their detection in scanned images. Here we explore the detection of defects like scratches, pinholes, and scuffs in a sample dataset of PEM optical images using two deep learning algorithms: Patch Distribution Modeling (PaDiM) for unsupervised anomaly detection and Faster-RCNN for supervised object detection. Both methods achieve scores on performance metrics (ROC-AUC and PRO-AUC for PaDiM and AP for Faster-RCNN) that are comparable to their scores on benchmark datasets. These methods also have the potential to detect a wider range of defects compared to IR thermography and previous optical inspection methods. Overall, deep learning shows promise at detecting relevant defects of interest and has the potential to achieve real-time defect detection.

30 DIRECT ENERGY CONVERSION↗
DOE OSTI2026

Arm and shoulder muscle segmentation in axial MRI with UNet deep learning model

Quantifying individual upper-limb muscle volumes from MRI provides key insight into muscle-specific strength, deficits, and adaptations. Manual delineation is the gold standard but time‑intensive, and the performance of current deep learning approaches, particularly for small or anatomically complex muscles, remains incompletely characterized. We evaluated a state‑of‑the‑art deep learning framework across the entire upper limb and analyzed factors governing segmentation performance, with attention to the forearm. Three previously published MRI datasets (1.5 T, 3D GRE T1‑weighted; total n = 39) spanning young, middle‑aged, and older adults were curated and quality‑checked, including expert manual segmentations for 31 muscles. Following multiclass mask reconstruction, we trained three 3D nnU‑Net multiclass models matched to the muscle subsets present across datasets, using five‑fold cross‑validation and a composite Dice Similarity Coefficient (DSC) + cross entropy loss. Segmentation accuracy was assessed with DSC. Performance varied across muscles (mean DSC = 0.806 ± 0.098), ranging from 0.920 (Deltoid) to 0.461 (Extensor pollicis brevis). In uncertainty‑weighted regressions, muscle volume was positively associated with DSC (R2 = 0.36, p < 0.001), whereas training segmentation count and muscle orientation showed negligible associations (R2 ≤ 0.06). A weighted mixed‑effects model identified volume as the strongest evaluated predictor, explaining 23.9% of variance in DSC; orientation and training count each contributed <1%, leaving 61.5% unexplained. These results indicate that deep learning–based segmentation can accurately quantify muscle volume for many upper‑limb muscles but remains constrained for small, low‑contrast forearm muscles.

Gillespie, Samuel↗
DOE OSTI2022

Large-scale paleo water-table rise in a deep desert aquifer recorded by dissolved noble gases

Illuminating past hydrological and climatological conditions in arid regions may provide insights into future trends in groundwater availability. The main objective of this study is to explore paleorecharge processes in the deep regional Nubian Sandstone Aquifer (NSA), which stretches below the hyperarid deserts of the Sinai Peninsula (Egypt) and the Negev (Israel), using dissolved noble gas data. Extremely high amounts of dissolved excess air (Ne concentrations up to 4 × solubility equilibrium) were observed in the ancient ( 81 Kr-depleted) groundwater that was recharged during past pluvial epochs in the basin. Here, the observed unique excess air signal and the clear spatiotemporal structure of the noble gas compositions in the aquifer are hypothesized to reflect two major characteristics of the groundwater system: (a) large-scale, long-term rises in the water table that facilitate the entrapment and dissolution of air bubbles under increased hydrostatic pressure, and (b) the geological settings in the southern recharge area in Sinai, including an almost horizontal position of the layers and intercalation of low-permeability rock formations within the aquifer, which inefficiently reject air bubbles from groundwater, thereby allowing for substantial entrapment of bubbles that dissolve as the water table rises. Enhanced dissolution of entrapped air was also observed in other paleo-groundwater-containing regional sandstone aquifers across North Africa, which, together with the recent findings from the NSA, suggest that largescale water table fluctuations have likely extensively occurred during past pluvial periods over these (and possibly other) arid regions. A preliminary assessment of noble gas recharge temperatures (NGTs) indicates an apparent decoupling of surface and water table temperatures in the case of deep aquifers with a thick (hundreds of m) unsaturated zone. This observed decoupling calls for a re-evaluation of previously obtained NGT records and a need for future work to consider the modification of NGTs due to soil air fractionation and geothermal heating of the underlying deep unsaturated zones.

54 ENVIRONMENTAL SCIENCES↗