The study, led by Sanaa N. Al-Haj Ali of the Department of Orthodontics and Pediatric Dentistry and Ra’fat I. Farah of the Department of Prosthetic Dental Sciences, moves beyond the familiar accuracy benchmarks that dominate AI evaluation in medical education. Accuracy alone, the authors argue, cannot characterize how AI models engage with assessment tasks. Workload — the cost of achieving a particular level of performance, emerging from the interaction between task requirements and the operator’s skills and perceptions — offers a complementary lens. In humans, working memory is limited in capacity and duration, and when task demands exceed this capacity, performance is compromised. An operator may maintain accurate performance at an unacceptably high mental cost, which is why performance and subjective workload must be measured independently. This dissociation makes workload a valuable evaluative dimension for AI systems, where models and students may achieve similar accuracy while experiencing vastly different processing demands.

To capture that dimension, the researchers turned to the National Aeronautics and Space Administration Task Load Index (NASA-TLX), the most widely used multidimensional subjective workload scale in human factors research. The original instrument comprises six subscales — Mental Demand, Physical Demand, Temporal Demand, Effort, Performance, and Frustration — each rated on a 21-point scale. For this study, the team selected three subscales tailored to the OSPDE context: mental demand as the primary indicator of cognitive processing, temporal demand given the timed, bell-ringer station format of the examinations, and frustration to probe whether AI models show any sensitivity to task-related affective load. Physical demand was excluded as inapplicable to both the examinations, which involve no psychomotor component, and to AI; performance was excluded to avoid overlap with measured accuracy; and effort was excluded because, in a task with no physical component, it reduces largely to mental effort. Subscales were analyzed individually rather than as a composite, consistent with the diagnostic approach recommended for identifying discrete workload sources.

The experimental setup was meticulous. Data were sourced from OSPDEs administered to twenty fourth-year dental students during a preclinical orthodontics and pediatric dentistry course in 2025. The question pool comprised seventy items across four formats: ten plain multiple-choice questions without visual support, and twenty short-answer questions, twenty multiple-choice questions, and twenty True/False questions, all supported by visuals such as clinical photographs, radiographs, illustrations, and tables. By clinical discipline, the pool included thirty-three pediatric dentistry items (47 percent) and thirty-seven orthodontics items (53 percent). Correct answers were validated against three standard reference texts, and all visuals passed a three-component quality checklist scored by a single faculty member whose intra-rater reliability exceeded Cohen’s kappa of 0.8.

Ten multimodal AI models were evaluated from three providers: four from Anthropic (Claude Sonnet 3.5, Claude Sonnet 4.5, Claude Haiku 4.5, and Claude Opus 4.6), three from OpenAI (GPT-5.2 Pro, GPT-5.1, and GPT-4o), and three from Google Gemini (Gemini 3 Pro, Gemini 2.5 Pro, and Gemini 2 Flash). All models were accessed via a paid subscription to POE between January 25 and March 5, 2026. A single investigator, blinded to study objectives and student performance data, inputted each question into the models via a new chat session to prevent context contamination, transcribing questions verbatim with visuals uploaded as JPEGs at 300 dpi using a zero-shot prompting approach. The team employed a two-step elicitation protocol: after providing its answer, the complete question–response pair was resubmitted to the model with a prompt requesting ratings on the modified NASA-TLX, each subscale rated on a seven-point scale from one (very low) to seven (very high). To assess intra-model response stability, each question was tested in three independent sessions, with final responses determined by majority rule and NASA-TLX ratings by median value across trials. Students completed the same modified NASA-TLX in print immediately following the examination, with a 100 percent response rate.

The accuracy results set the stage for the workload findings. Overall accuracy differed significantly across responders (chi-square = 95.309, p < .001): students achieved the highest score at 77.9 percent, followed by GPT-5.2 Pro at 74.3 percent and Gemini 3 Pro at 72.9 percent, while Claude Haiku 4.5 scored lowest at 44.3 percent. Only GPT-5.2 Pro, Gemini 3 Pro, Gemini 2.5 Pro, and Claude Opus 4.6 did not differ significantly from students. Performance gaps were largest on visual short-answer questions (chi-square = 44.676, p < .001), where several Claude models scored at or below 25 percent, and smallest on plain multiple-choice questions, where the difference was not statistically significant. Intra-model reliability, measured by intraclass correlation coefficients, ranged from 0.677 to 0.955 across all models and dimensions (all p < .001), with mental demand showing the highest consistency and frustration the most variability — a gradient the authors attribute to frustration capturing an emotional reaction that is inherently more state-dependent than ratings referencing concrete task properties.

The workload comparisons produced the study’s central contribution. Seven of ten models reported significantly higher mental demand than students, six reported significantly higher temporal demand, and nine reported significantly higher frustration. The greatest human–AI divergence occurred on visual short-answer questions, where nine of ten models reported significantly higher mental demand and all ten reported significantly higher temporal demand. Yet frustration told a different story on this format: only three models (Gemini 2.5 Pro, Gemini 3 Pro, and GPT-5.1) reported significantly higher frustration than students, compared with seven to nine models on other formats, despite visual short-answer questions producing the lowest AI accuracy. On plain multiple-choice questions, models performed comparably to students on both accuracy and mental demand, indicating that human–AI convergence is greatest when visual processing is absent and recognition-based options constrain response demands.

This dissociation between demand ratings and frustration on the hardest format is not paradoxical, the authors argue, once the cue sources underlying each dimension are distinguished. Mental and temporal demand are stimulus-facing judgments: their cues derive from observable task properties such as inferential steps, visual integration requirements, and time constraints. Frustration, by contrast, is a self-evaluative judgment that requires comparing output quality against some performance standard. In multiple-choice and True/False formats, fixed alternatives provide an implicit benchmark against which the model can gauge correctness; short-answer questions strip away this external scaffold entirely, leaving the model with impoverished evaluative cues. Koriat’s cue-utilization framework supports this interpretation: metacognitive judgments lose diagnostic sensitivity when available cues are restricted. Additionally, the higher processing demands on visual short-answer items may consume resources that would otherwise support self-evaluative reflection, analogous to Sweller’s finding that heavy cognitive load diminishes capacity for secondary processes. Hence, the models had ample cues to detect that the task was difficult, but insufficient cues to evaluate how poorly they performed.

The authors also caution against anthropomorphic readings of the frustration data. The NASA-TLX frustration item asks how insecure, discouraged, irritated, or stressed the respondent felt — wording that presupposes an affective state current models do not possess. A model’s response is therefore generated text about frustration rather than a report of an experienced emotion, and the researchers treat the affective subscale strictly as a behavioral output rather than as evidence of internal experience. A parallel caveat applies to temporal demand: language models generate output without an internal clock or representation of elapsed time, so the instruction that two minutes were available was processed as semantic content rather than experienced as time pressure. Elevated temporal demand ratings are therefore more plausibly an inference from task characteristics — item length, information density, reasoning steps, and visual parsing load — than a report of temporal stress. This mirrors the broader pattern: models tracked content-derived indices of difficulty while showing attenuated responses on the affective dimension, which carries no comparable task signal.

Correlational analyses between workload and accuracy revealed a predominantly inverse relationship among both students and AI models, though among students significant correlations were weak and confined to frustration (rs = −0.08, p < .01). Among AI models, the pattern was architecture-dependent: Claude Sonnet 4.5 and Claude Opus 4.6 showed significant negative correlations across all three dimensions overall, while GPT-4o, Claude Sonnet 3.5, and the Gemini models showed none. Claude Haiku 4.5 showed significance for temporal demand (rs = −0.35) and frustration (rs = −0.40), and GPT-5.2 Pro for temporal demand (rs = −0.30) and frustration (rs = −0.33). At the format level, significant negative correlations were most concentrated on visual multiple-choice and visual short-answer items and least on visual True/False items and plain multiple-choice questions, suggesting that the diagnostic sensitivity of AI workload ratings scales with format complexity. This variability has practical significance: models with significant negative load–accuracy associations and acceptable accuracy — notably Claude Opus 4.6 and GPT-5.2 Pro — could serve as automated item difficulty estimators, flagging disproportionately demanding questions before student administration during examination development.

The findings carry implications for educators designing assessments intended to differentiate human from AI performance. Because human–AI convergence was greatest on plain multiple-choice questions and diverged most on visually integrated open-ended formats, educators seeking to distinguish human from AI capability should favor visually integrated open-ended question types. The study also establishes workload profiling as a complementary dimension of AI evaluation that extends beyond accuracy, offering insight not only into whether AI can answer correctly but into how it represents the difficulty of doing so — and where that representation breaks down.

Several limitations warrant acknowledgment. AI answers were determined by majority rule across three sessions whereas students had a single attempt, an asymmetry that, for accuracy, can only maintain or improve an AI’s score relative to a single attempt — meaning the finding that students outperformed most models is conservative. The student sample comprised twenty participants from a single institution, representing the entire eligible cohort but limiting generalizability, and the seventy-item question pool was smaller than licensing examinations. Potential data contamination cannot be fully excluded, as the training corpora of commercial models are proprietary, though the authors note that the study’s central finding — a systematic cross-item pattern in which workload tracked question format and visual complexity with a consistent demand–affect dissociation — would not be expected from memorization of individual images. Whether AI workload ratings reflect genuine metacognitive self-monitoring or statistically patterned output remains open; the reliability and correlational data favor a structured interpretation but cannot establish introspective capacity. Future work, the authors suggest, should compare both parties under identical single-attempt conditions and explore whether format-specific time allocations reduce the human–AI gap on open-ended items.

Subject of Research: Science News

Article Title: NASA TLX workload profiles of multimodal artificial intelligence models and dental students during objective structured practical dental examinations

Article References: Al-Haj Ali, S. N., & Farah, R. I. (2026). NASA TLX workload profiles of multimodal artificial intelligence models and dental students during objective structured practical dental examinations. Discover Education, 5(1), Article 835. https://doi.org/10.1007/s44217-026-02116-2

Image Credits: AI Generated

DOI: 10.1007/s44217-026-02116-2

Keywords: artificial intelligence, NASA-TLX, workload, dental education, multimodal AI, large language models, metacognition, cognitive load, objective structured practical dental examinations, AI evaluation, NASA, profiles

The software, built by Floyd Nichols and Grant Major of Virginia Tech’s Department of Geosciences and Ryan Felton of NASA Ames, is a web-based graphical user interface that lets educators, students, and researchers train and interpret machine learning models without writing a single line of code. It is hosted on Hugging Face, the popular open-source machine learning platform, and runs on two virtual CPUs with 16 gigabytes of RAM. The team describes it as an entry point for novices and a rapid prototyping tool for experts, a way to standardize the machine learning workflow across a community that has historically approached it in ad hoc ways.

The need for such a tool is not abstract. Within the last five years, machine learning has crept into nearly every corner of astrobiology: analyzing laboratory data on biosignatures, processing remote sensing and telescope observations, running simulations of planetary environments, and even enabling autonomous decision-making on space missions. NASA’s own strategic documents have repeatedly called for broader adoption. The agency’s Science Mission Directorate Strategy for Data Management and Computing recommended incentivizing and educating the community on AI and machine learning approaches, while the Origins, Worlds, and Life decadal survey explicitly endorsed machine learning for supporting autonomy on remote planetary missions. The NASA ROSES program has signaled the same need through dedicated solicitations for Solar System Science and Foundational Artificial Intelligence for the Moon and Mars.

Yet the barrier to entry remains steep. Machine learning algorithms, even those available as off-the-shelf functions in Python or R, demand a working knowledge of both the underlying mathematics and the programming language itself. Many astrobiologists are domain experts in geology, chemistry, or biology but have neither the time nor the institutional support to become competent programmers. This is precisely the gap AI² is designed to fill, offering a guided pipeline that walks users through the three canonical stages of a machine learning project: data preprocessing, exploratory data analysis with unsupervised learning, and supervised learning for prediction.

The preprocessing stage handles the unglamorous but critical work that determines whether any downstream model will be trustworthy. Users upload a CSV file, and the application immediately computes Pearson correlation coefficients to reveal pairwise linear relationships between variables and variance inflation factors to flag multicollinearity, a statistical pathology in which highly correlated predictors can silently degrade a model’s ability to generalize. The tool then offers options for handling missing values through removal or imputation using mean, median, or most-frequent strategies, encoding categorical variables into numerical form, and scaling features via min-max normalization or z-score standardization. These steps matter because many algorithms implicitly assume features are centered around zero with comparable variance; a single feature with orders of magnitude more spread can dominate the model and prevent it from learning anything useful from the others.

Once the data is clean, the unsupervised learning module offers dimensionality reduction and clustering techniques chosen specifically because they have appeared in prior astrobiology research. For dimensionality reduction, users can apply Principal Component Analysis, which captures maximum variance between features through linear transformation; Multidimensional Scaling, which instead preserves pairwise distances between data points; or t-Distributed Stochastic Neighbor Embedding, a nonlinear manifold learning method that preserves local structure and is particularly useful for visualizing high-dimensional data in two dimensions. For clustering, the options include K-means, which partitions data into k groups of equal variance around centroids; Gaussian Mixture Models, which identify clusters based on probabilistic Gaussian densities; and DBSCAN, which can detect clusters of arbitrary shape, size, and density without requiring the user to predefine how many groups exist.

The supervised learning module covers both classification and regression tasks. Classification options include Logistic Regression, k-Nearest Neighbors, Support Vector Classifier, and Random Forest, while regression offerings include Random Forest Regressor, Histogram Gradient Boosting, Support Vector Regressor, and Ridge Regression. After training, the interface produces standard evaluation metrics: accuracy, precision, recall, and F1 score for classification, and mean squared error, root mean squared error, mean absolute error, and mean absolute percentage error for regression. Cross-validation is incorporated to assess generalizability, and the tool generates interactive visualizations including confusion matrices for classification tasks and predicted-versus-actual crossplots for regression. Crucially, it also produces feature importance plots based on Shapley Additive Explanations, a technique that attributes each prediction to individual input features, keeping the entire workflow within the realm of explainable rather than black-box AI.

What distinguishes AI² from similar tools is its built-in statistical guardrails. The developers recognized that even experienced researchers sometimes select models inappropriately because they skip the diagnostic tests that reveal whether a dataset’s assumptions match a given algorithm’s requirements. The application therefore runs Henze-Zirkler’s test for multivariate normality, the Breusch-Pagan test for heteroscedasticity, and isolation forest analysis to quantify outlier rates, all under the hood. If a user selects a model that is statistically inappropriate for their data, the tool issues a warning and suggests a more robust alternative. For example, ridge regression, often chosen for its speed and resistance to multicollinearity, breaks down when data contains high outlier rates, heteroscedasticity, or non-normal distributions; in such cases the interface recommends tree-based models like random forest or histogram gradient boosting instead.

The team benchmarked the web-hosted version across datasets ranging from 100 to 10,000 observations with 10 features, sizes chosen to represent typical astrobiological datasets from biological measurements to stellar spectral catalogs. Most model scenarios completed in seconds to tens of seconds. The outlier was t-SNE with 10,000 observations, which averaged 194 seconds, a reflection of that algorithm’s inherent computational intensity rather than any flaw in the implementation.

The developers are candid about the limitations. The algorithm menu is deliberately restricted to methods most relevant to astrobiology, and neural networks are excluded for now because of their black-box nature, though simple artificial neural networks are planned for future versions. The team also acknowledges that large language models like ChatGPT and Claude can help novices write machine learning code, but argues that this path is fraught: LLMs require well-crafted prompts that themselves demand machine learning understanding, and studies have shown that novices using LLMs frequently make inappropriate model selections, misuse data, and skip systematic hyperparameter tuning, errors that then get baked into the training data of future models. Compared with existing tools like Google’s Teachable Machine, which handles only image and audio recognition, or Orange Data Mining, which requires moderate machine learning knowledge and lacks statistical assumption testing, AI² is purpose-built for tabular astrobiology data with guardrails baked in. Future plans include video tutorials and integration with the AI-Astrobiology portal at NASA Ames, a community resource that hosts seminars, literature reviews, and tutorial guides. The broader hope is that by lowering the technical barrier, the tool will help a generation of astrobiologists treat machine learning not as an arcane specialization but as a standard part of the scientific toolkit, one that complements rather than replaces human domain expertise in the ongoing search for life in the universe.

Subject of Research: An open-source graphical user interface for applying explainable machine learning to astrobiology datasets

Article Title: Artificial intelligence for astrobiology investigations (AI2): an open-source graphical user interface

Article References: Nichols, F., Major, G., & Felton, R. (2026). Artificial intelligence for astrobiology investigations (AI2): an open-source graphical user interface. Frontiers in Astronomy and Space Sciences, 13, Article 1933473. https://doi.org/10.3389/fspas.2026.1933473

Image Credits: AI Generated

DOI: 10.3389/fspas.2026.1933473

Keywords: astrobiology, machine learning, open-source software, graphical user interface, NASA, biosignatures, data preprocessing, unsupervised learning, supervised learning, explainable AI, Hugging Face, scientific education

What makes MISR extraordinary among Earth-observing instruments is its geometry. Flying on Terra, the flagship spacecraft of NASA’s Earth Observing System, the push-broom imager has been collecting radiance data since late February 2000, viewing the planet at nine angles ranging from straight down to 70.5 degrees fore and aft along the flight path, in four spectral bands centered at 446, 558, 672, and 866 nanometers. Because aerosol particles scatter and absorb light differently depending on the angle of observation, this multi-angle approach can constrain particle size, shape, and light-absorbing properties in ways that single-view imagers simply cannot. The instrument covers the globe roughly once a week, and even though Terra began constellation exit procedures in October 2022, MISR leaves behind a legacy of more than 24 years of multi-angle, multi-spectral observations of the entire planet.

The new MRPA is the culmination of a long-running parallel effort. NASA’s operational MISR Standard Algorithm, currently at version 23, was designed from the start to run automatically on the entire dataset, producing aerosol retrievals at 4.4 kilometer resolution. Alongside it, the research team built the MISR Research Algorithm, a more experimental code designed to push the sensitivity of the retrieval further, exploring innovations such as corrections for radiometric calibration errors like flat-field anomalies and ghosting in high-contrast scenes, and improved performance over shallow, turbid, and biologically productive waters. The catch was that the research algorithm demanded considerable hands-on user involvement and could only be run case by case, on limited subsets of data. The MRPA changes that equation by streamlining and automating the research approach so that any user can generate research-quality results across most of the MISR data record, spanning February 2000 through October 2022.

At the heart of the algorithm lies a clever dual strategy for dealing with the hardest problem in satellite aerosol retrieval: separating the signal of particles in the air from the light reflected by the ground below. The Retrieved-Surface Algorithm solves for surface reflectance self-consistently along with the aerosol properties, using a shape-similarity assumption in which the angular shape of surface reflectance is treated as independent of wavelength. In the updated MRPA, the shape-similarity coefficient is now calculated exclusively from the near-infrared band, because near-infrared light penetrates the atmosphere more reliably than the shorter blue and green wavelengths, giving better constraint on the surface. The Prescribed Surface Algorithm, by contrast, takes surface brightness from an external dataset, the Multi-Angle Implementation of Atmospheric Correction product derived from NASA’s MODIS instrument, and then solves for aerosol quantities directly using a bounded-variable least squares approach. This prescribed-surface route excels precisely where the retrieved-surface method struggles, namely in high-aerosol-optical-depth cases where the thick haze itself makes it hard to pin down what the surface is doing.

The team blends the two approaches into a Combined Surface Algorithm, weighting the contributions of each according to the Normalized Difference Vegetation Index of the underlying terrain. Bright, barren landscapes such as deserts are where small errors in prescribed surface reflectance can balloon into large aerosol errors, so the new weighting scheme, tuned against an extensive validation dataset, leans more heavily on the retrieved-surface results in low-vegetation scenes. The payoff is striking: in scenes where the prescribed-surface NDVI falls below 0.1, the new combined algorithm shows a root-mean-square error 45 percent lower than the previous version, with a 43 percent lower bias. The MRPA also introduces new cloud masking interpolated to the MISR swath from the MAIAC cloud mask over land, and over deep ocean it compares 12-micrometer brightness temperatures from MODIS-Terra against MERRA-2 reanalysis sea surface temperatures, flagging pixels that run 7.5 kelvin warmer as cloudy. Column ozone for absorption corrections now comes from MERRA-2 as well, and ocean surface roughness is prescribed from Cross-Calibrated Multiplatform wind speeds.

Validation against ground truth is where the new product truly shines. The researchers compared the retrievals against the AErosol RObotic NETwork, or AERONET, a global federation of calibrated sun photometers whose direct-sun aerosol optical depth measurements carry an uncertainty of about 0.01. Using more than 75,000 over-land coincidences between February 2000 and October 2022, the combined algorithm achieved an AOD root-mean-square error of 0.088, compared with 0.154 for the MISR V23 standard product, roughly half the error, along with a correlation coefficient of 0.938 versus 0.81. The bias of 0.008 is somewhat higher than the standard product’s −0.002, one of the compromises of automation, but the overall statistical improvement is substantial. Over water, where more than 29,000 coincidences were available, the algorithm performed even better, with an RMSE of 0.069 and a correlation of 0.923, and the expected error envelope works out to roughly 17 percent of the retrieved AOD plus 0.01 over land.

Particle-type information, the real prize of multi-angle imaging, is harder to validate quantitatively because the ground-based inversion retrievals of size, sphericity, and single-scattering albedo carry their own assumptions and limitations, particularly at low aerosol load. Still, the Ångström exponent, a robust measure of particle size derived from how aerosol optical depth varies with wavelength, shows two clear clusters in the validation data, around 0.25 and 1.5, corresponding to coarse-mode dust and fine-mode smoke or pollution respectively, and the agreement with AERONET improves as aerosol loading increases, with RMSE dropping from 0.43 to 0.25 at higher AOD thresholds. When the team examined regions and seasons dominated by a single aerosol source, the retrievals behaved exactly as physics demands. Smoke-dominated scenes, once the mid-visible AOD exceeded roughly 0.2 over water and 0.7 over land, showed a preponderance of small-to-medium, spherical, light-absorbing particles with single-scattering albedo between about 0.9 and 0.975. Dust-dominated regions near the Sahara in late spring and early summer yielded distinctly larger, weakly absorbing, non-spherical particles, with the coarse-mode signal especially clear over the high-contrast backdrop of dark ocean water.

The research algorithm behind the MRPA has already proven its worth in a remarkable range of scientific applications, from characterizing the properties and evolution of wildfire smoke plumes across North America, Alaska, and Siberia, to tracking volcanic eruption plumes including the catastrophic 2022 Hunga Tonga-Hunga Ha’apai event, to air quality assessments and studies of how dust particles age and change during trans-Atlantic transport. With the algorithm now automated, publicly released, and archived for general use, researchers anywhere can run it on any portion of the MISR archive that interests them, opening the door to systematic studies of how particle properties vary and evolve across the entire 22-year record rather than only in hand-picked case studies. The code’s availability marks a shift from a specialist tool requiring expert knowledge to a community resource.

The broader significance of this work lies in what it preserves and what it promises. As single-view instruments like MODIS have built their own long aerosol records, MISR’s multi-angle perspective offers a complementary and uniquely information-rich dataset, one that can distinguish, for example, a plume of absorbing smoke from a layer of scattering desert dust even when both produce similar optical depths. The compromises required to automate the research algorithm do come at a cost, with somewhat diminished particle-type sensitivity compared to the fully manual version, and the MRPA is currently configured to process data only through October 2022 to maintain continuity with its calibration corrections, though new Level 1 radiance products accounting for Terra’s orbit drift are planned. Even so, with about half the retrieval error of the operational standard product, four times finer pixel resolution, and demonstrated ability to separate smoke from dust by size, shape, and absorption, the MRPA hands atmospheric scientists a powerful new lens on the invisible particles that shape our climate and our air, and it ensures that the full scientific value of one of NASA’s most inventive instruments will be realized for years to come.

Subject of Research: Automated multi-angle satellite retrieval of global aerosol optical depth and particle properties from MISR

Article Title: The MISR Research Product Algorithm – producing global, self-consistent, pixel-level aerosol retrievals for the Multi-angle Imaging SpectroRadiometer

Article References: Anstett, M., Limbacher, J. A., & Kahn, R. A. (2026). The MISR Research Product Algorithm – producing global, self-consistent, pixel-level aerosol retrievals for the Multi-angle Imaging SpectroRadiometer. Atmospheric Measurement Techniques, 19(20), 6503-6516. https://doi.org/10.5194/amt-19-6503-2026

Image Credits: AI Generated

DOI: 10.5194/amt-19-6503-2026

Keywords: MISR, aerosol retrieval, Terra satellite, NASA, AERONET, aerosol optical depth, remote sensing, wildfire smoke, mineral dust, air quality, atmospheric measurement, multi-angle imaging

The research, led by Patrick K. Pischulti of the University of Colorado Boulder together with Min Young Hwang and Christopher McComb of Carnegie Mellon University and Katya Arquilla of CU Boulder, introduces what the team calls an Interaction Digital Twin, or IDT. Digital twins are virtual replicas of physical systems that mirror their structure, behavior, and context, updated in real time through bidirectional data flows. The concept, originally developed in manufacturing, has been adopted across healthcare, smart cities, and aerospace to answer

The distinction between system autonomy and crew autonomy sits at the heart of why the researchers felt an entirely new modeling layer was needed. Digital twins have already proven their worth in aerospace contexts by allowing engineers to monitor the health of spacecraft components, diagnose faults before they become critical, predict future states of hardware, and optimize operations without waiting for ground-based analysis. But these applications treat the human operator as, at best, a peripheral element or, at worst, an afterthought. The authors surveyed the existing literature and identified three recurring patterns that fall short of what deep space missions demand: efforts that twin a single human without modeling the environment that person inhabits, efforts that twin crowds or groups without capturing individual-level interactions, and efforts that twin human-to-human exchanges alongside environmental models but only in non-aerospace settings where participants are not collaborating toward a unified objective. None of these approaches, the researchers argue, adequately addresses the synergies NASA has identified between a human and a system, between a crew and a system, and among the members of a human team.

The consequences of this gap become clearer when one considers what mission control actually does during a typical mission. Beyond monitoring telemetry, flight controllers build and continuously evolve nominal operating procedures based on their observations of how crew members interact with each other and with their equipment, and how those interactions produce outcomes. This adaptive, relationship-aware form of oversight is something no purely hardware-focused digital twin can replicate. When a crew is 22 light-minutes from home, the loss of these time-critical cognitive and coordination functions cannot be compensated by faster hardware models alone. The framework proposed in the paper therefore treats the interaction itself as the thing being twinned, elevating it to what the authors describe as a first-class, inference-capable element within the architecture rather than a byproduct of separately modeling people and machines.

The distributed nature of the architecture deserves particular attention. Rather than concentrating all computational intelligence in a single monolithic model, the framework distributes twin responsibilities across multiple nodes with bidirectional information flow between them. Three complementary types of twins operate in concert: conventional Digital Twins representing habitat systems, Human Digital Twins representing individual crew members, and Interaction Digital Twins capturing the emergent phenomena that arise when people and systems engage with one another. These emergent phenomena include team cohesion, trust calibration, coordination patterns, and adaptive autonomy—the subtle, dynamic qualities that determine whether a crew functions smoothly or fractures under stress. By modeling these phenomena explicitly, the framework can in principle support inference about team state, not just system state, allowing onboard software to recognize when workload is unevenly distributed, when trust in an automated system has drifted too high or too low, or when coordination between crew members is degrading.

A practical consequence of this design is that aspects of command and control migrate onboard the vehicle. The authors describe this as giving the crew mission-control-like capabilities during periods of communication delay. In their vision, a Vehicle Systems Manager—an entity that understands and synthesizes the results of the twinned interactions—would augment nominal operations even when Earth is effectively out of reach for real-time consultation. This does not mean eliminating ground control, but rather ensuring that the crew retains the analytical depth normally provided by flight controllers when the delay makes that support impractical. The concept aligns with NASA’s long-standing recognition that even small communication delays in the range of seconds, such as those experienced on the International Space Station, can measurably harm performance, increase workload, and erode situational awareness.

The intellectual lineage of the digital twin concept helps explain why this extension is significant. Grieves, who originated the concept in the manufacturing sector, defined a digital twin as a set of virtual information constructs that fully describes a potential or actual physical manufactured product. Since that introduction, the idea has spread into advanced manufacturing, human health and medicine, smart cities, and aerospace. The medical community offers a particularly instructive analogy: clinicians use digital twins to perform interventions aimed at answering what-if questions, such as predicting how a patient would respond to a treatment without actually administering the drug. The Interaction Digital Twin applies the same logic to team dynamics—allowing mission planners and onboard systems to ask what-if questions about crew coordination, workload allocation, and trust before committing to a course of action during a mission.

To ground the framework in real mission needs, the authors began with a systematic review of NASA’s own published requirements. The 2009 Design Reference Architecture documents for human exploration of Mars, produced by NASA headquarters, remain among the most recent and comprehensive descriptions of the capabilities and operations needed for a human deep space mission. Beyond the traditional engineering concerns of propulsion and in-situ power generation, those documents emphasized the importance of information support provided by the habitat system—monitoring and diagnosing the habitat’s environmental health and delivering actionable insights to the crew. NASA noted that such capabilities are essential for reducing individual crew member stress, decreasing cognitive workload, and mitigating the downstream effects these factors have on team dynamics and overall mission performance. This causal chain, from information support to stress reduction to workload management to team stability, is precisely the chain the Interaction Digital Twin is designed to strengthen.

Building on the Mars objective established in the Design Reference Architecture, NASA developed the Moon to Mars mission plan to define concrete steps for achieving travel beyond low Earth orbit. The NASA Marshall Space Center expanded the agency’s reference architecture for this plan by elaborating on Ground Rules and Assumptions in conjunction with functional allocations for habitat design. While that document focuses primarily on the autonomy of an uncrewed habitat, it also details expectations of increased support from the habitat system for humans during the crewed portion of the flight, including aiding crew awareness. The framework’s authors treated these documents as a requirements specification, deriving from them the criteria that a digital twin architecture must satisfy to genuinely support crew autonomy rather than merely automate spacecraft functions.

The application scenario chosen to demonstrate the framework is an Artemis Phase II mission, which places the architecture in the context of NASA’s near-term lunar ambitions rather than a distant Mars voyage. This choice is strategically meaningful. Artemis Phase II operations will already involve communication conditions and operational complexity that stress current mission control paradigms, making the scenario a realistic proving ground for interaction-level twinning. The demonstration shows how the framework extends system-level modeling to support cognitive workload management, information sharing, and human-autonomy teaming—the practice of pairing human operators with automated agents in ways that leverage the strengths of both. Human-autonomy teaming has been an active area of research in aerospace and other safety-critical domains, but embedding it within a digital twin architecture that also captures team cohesion and trust calibration represents a notable synthesis.

The research team itself reflects the interdisciplinary nature of the challenge. Pischulti and Arquilla are based in the Ann and H. J. Smead Aerospace Engineering Sciences department at the University of Colorado Boulder, while Hwang and McComb bring expertise from Carnegie Mellon University’s Department of Mechanical Engineering. The collaboration between aerospace engineering and mechanical engineering groups mirrors the framework’s own philosophy: that human teaming constructs and technical system models must be developed together rather than in isolation. The paper was published as an original research article in the Intelligent Aerospace Systems section of Frontiers in Aerospace Engineering on March 27, 2026, appearing in Volume 5 of the journal.

Looking ahead, the implications of treating interactions as twinnable, inferable entities extend beyond spaceflight. Any domain where human teams operate alongside complex systems under constraints on communication or oversight—deep-sea operations, remote disaster response, autonomous shipping—faces analogous challenges in sustaining coordination without real-time expert support. The framework’s insistence that emergent phenomena like trust and cohesion can be modeled, monitored, and reasoned about onboard suggests a future in which the soft dynamics of teamwork become as quantifiable and actionable as the hard telemetry of a spacecraft’s power system. For the Moon to Mars campaign specifically, the work offers a concrete architectural answer to a question NASA has posed repeatedly in its exploration roadmaps: how to preserve the full richness of mission control’s judgment when the crew must carry that judgment alone.

Subject of Research: A distributed digital twin architecture that models team interactions and dynamics to support crew autonomy in deep space missions

Article Title: “Interaction Twin in the middle”: a distributed digital twin architecture to model team interactions and dynamics for deep space missions

Article References: Pischulti, P. K., Hwang, M. Y., McComb, C., & Arquilla, K. (2026). “Interaction Twin in the middle”: a distributed digital twin architecture to model team interactions and dynamics for deep space missions. Frontiers in Aerospace Engineering, 5, Article 1736392. https://doi.org/10.3389/fpace.2026.1736392

Image Credits: AI Generated

DOI: 10.3389/fpace.2026.1736392

Keywords: digital twin, crew autonomy, deep space exploration, human-system interaction, team dynamics, NASA, Artemis, human-autonomy teaming, distributed architecture, cognitive workload, behavioral health, Mars missions

The study, published in the journal Biogeosciences by Joaquín Chaves, Crystal Thomas, and Antonio Mannino, examined nearly 20,000 individual water samples analyzed by high-performance liquid chromatography, or HPLC, at NASA’s centralized pigment facility since its establishment in 2011. The facility, which has processed more than 30,000 samples in total from every major ocean basin, exists because NASA carries a mandate to distribute in situ data of the highest quality for calibrating and validating satellite ocean color missions. When a satellite sensor measures the greenness of the sea, that greenness comes chiefly from chlorophyll a and a suite of accessory pigments inside phytoplankton cells. If the ground-based measurements of those pigments are sloppy, the entire validation chain wobbles.

The technical heart of the analysis is the coefficient of variation, or CV%, calculated among replicate samples, predominantly duplicate filters collected from the same water mass and processed independently. Across the full suite of 26 routinely reported pigments, mean precision ranged from a remarkably tight 3.2% for divinyl chlorophyll a, a signature pigment of oceanic cyanobacteria, to 17.1% for chlorophyllide a, a degradation product that serves mainly as a red flag for sample-handling problems rather than a biogeochemical quantity in its own right. Crucially, the study tested the results against benchmarks forged during the SeaWiFS HPLC Analysis Round-Robin Experiments, known as SeaHARRE, which set a precision target of 5% for total chlorophyll a and 8% for the twelve primary pigments most relevant to ocean color validation. Total chlorophyll a came in at 4.3%, comfortably within spec, and ten of the twelve primary pigments also passed. Only diatoxanthin, at 8.6%, and peridinin, at 9.2%, slightly exceeded the bar, likely because both tend to occur at low concentrations where quantification uncertainty grows.

What makes the study genuinely surprising is what did not drive the variability. The team hypothesized that pigment concentration, or the mass of pigment actually injected into the instrument, would be the dominant factor, since analytical methods are expected to degrade near their detection limits. Instead, multivariate regression models incorporating concentration, filtered volume, inferred phytoplankton size structure, and coastal versus oceanic origin explained no more than 3% of the precision variability across the full dataset. In other words, once pigments sit comfortably above their detection thresholds, the validated NASA protocols deliver essentially concentration-independent precision, a testament to the quality assurance plan built on years of intercalibration exercises. The instrument itself, an Agilent system with a C8 column held at 60 degrees Celsius and dual-wavelength detection at 450 and 665 nanometers, achieves injection repeatability averaging just 0.6%, thanks in part to a clever vitamin E internal standard that corrects for variations in extraction volume without interfering with pigment signals.

The picture changed, however, when the researchers stripped out invariant replicates, those duplicate pairs that produced identical concentration values and thus a CV of exactly zero. In this censored dataset, regression models suddenly explained up to 44% of the variability for divinyl chlorophyll b, and concentration emerged as the dominant variable for 22 of 25 pigments. Precision deteriorated toward detection limits for secondary and tertiary pigments, the less abundant compounds that often serve as taxonomic fingerprints for specific phytoplankton groups. The authors caution that this shift partly reflects a statistical artifact: pigments with narrow, low concentration ranges, such as divinyl chlorophyll b, where invariant replicates made up 70% of the sets, are simply more likely to yield identical values by chance, so removing them concentrates the analysis on the rare cases where variability was detectable at all.

One of the most intriguing findings concerns the difference between analytical precision and natural sample heterogeneity. For taxon-specific carotenoids like peridinin, the marker for dinoflagellates, precision worsened at low concentrations when plotted against concentration but showed a much weaker relationship when plotted against the total pigment mass injected. The researchers interpret this asymmetry as evidence of stochastic cell capture: when rare organisms are sparsely scattered through the water, two replicate filters may trap genuinely different numbers of cells, producing variability that reflects the patchiness of the real ocean rather than any flaw in the instrument. Filtering larger volumes of dilute water improves precision precisely because it captures more cells, pointing to sampling statistics, not laboratory sensitivity, as the bottleneck for rare pigments.

Filtration volume itself emerged as a practical lever. For most primary pigments, filtering more than 1000 milliliters of seawater pushed precision below 10%, though the benefit plateaued well below that threshold for many compounds, and a subset including alloxanthin, diatoxanthin, peridinin, and all the tertiary pigments actually showed precision degradation at very large volumes in the censored analysis. The authors suggest a combination of detection limit constraints, where even large volumes cannot lift trace pigments far above the limit of quantification, and possible physical stresses during extended filtration, such as cell lysis under prolonged vacuum. Their recommendations align with NASA’s Ocean Optics Protocols: roughly 0.5 to 1 liter for nutrient-rich waters, and 1 to 4 liters for the oligotrophic open ocean where accessory pigments must be coaxed above detection limits.

Geography told a subtler story. In the full dataset, oceanic samples more than 200 kilometers from shore showed significantly better precision than coastal samples for nearly every pigment, and total chlorophyll a retained that oceanic advantage even in the censored analysis. Yet the team argues this does not mean coastal waters are inherently harder to measure. The SeaHARRE-4 and SeaHARRE-5 intercomparisons, which used exclusively coastal samples from Danish fjords and rivers in New England and Tasmania, found that quality-assured laboratories achieved precision essentially indistinguishable from open-ocean exercises, with total chlorophyll a differences of only about 1.4% to 2.4% among validated methods. The coastal penalty in the NASA dataset more likely reflects differences in field sampling procedures, filtration volumes, or adherence to collection protocols across research groups, factors that remain largely invisible because the necessary metadata were not systematically recorded.

The study’s conclusions carry real weight for the future of ocean color science, particularly as new satellite missions demand ever more stringent calibration and validation. The clearest path to improvement, the authors argue, lies not in the laboratory but on the ship deck: better field replication, careful vacuum pressure monitoring during filtration, immediate preservation in liquid nitrogen, and an unbroken cold chain from collection to analysis. Only about 30% of samples in the dataset were collected as replicates, and 97% of those replicate sets were mere duplicates, which statistically limits how well precision can be characterized. NASA recommends that at least 5% of samples be replicated, but the team urges investigators to exceed that minimum substantially, and to collect triplicate filters where logistics allow. As satellites continue to chart the pulse of the global ocean, this massive audit confirms that the ground truth beneath them is solid, and shows exactly where the next gains in confidence will come from.

Subject of Research: Precision assessment of HPLC phytoplankton pigment analysis for global ocean color satellite validation

Article Title: Precision of phytoplankton pigment analysis by high-performance liquid chromatography: an assessment of the global ocean color validation dataset analyzed by NASA

Article References: Chaves, J. E., Thomas, C. S., & Mannino, A. (2026). Precision of phytoplankton pigment analysis by high-performance liquid chromatography: an assessment of the global ocean color validation dataset analyzed by NASA. Biogeosciences, 23(19), 7043-7065. https://doi.org/10.5194/bg-23-7043-2026

Image Credits: AI Generated

DOI: 10.5194/bg-23-7043-2026

Keywords: phytoplankton, HPLC, chlorophyll a, ocean color, satellite validation, NASA, pigment analysis, SeaHARRE, precision, biogeosciences, coastal ocean, quality assurance

The award places the University of Arizona at the center of an unusual corner of NASA’s astrophysics portfolio. PUFFINS is the second mission the university has led under the Astrophysics Pioneers program, an initiative launched in 2020 to deliver high-value science on small, cost-capped platforms. The university also leads Aspera, a first-generation Pioneers mission scheduled to launch early next year, and contributed significantly to Pandora, which launched in January of this year. According to university leadership, Arizona has emerged as the only university leading multiple missions in the Pioneers program, while also running mission operations for Pandora. The new mission is a joint effort with the McDonald Observatory at The University of Texas at Austin and an international network of early-career and senior scientists, with corporate partners including Tucson-based Ascending Node Technologies and SFL Missions Inc. of Toronto.

At the heart of the satellite sits a specialized telescope just 10 inches across, roughly the size of a large dinner plate, designed to capture ultraviolet light from distant stars. Despite its modest aperture, the instrument’s value lies not in collecting vast quantities of light but in dissecting it with extraordinary care. PUFFINS will measure the polarization properties of starlight across a range of ultraviolet wavelengths using a technique called spectropolarimetry, an approach that has not been deployed in space in nearly thirty years. Polarization describes the orientation of light waves as they travel, and it carries information that ordinary brightness measurements cannot reveal. When unpolarized starlight passes through a cloud of interstellar dust, the dust imprints a preferred orientation on the light, and the precise way that imprint varies with wavelength encodes the size, composition, and alignment of the grains themselves.

The physics behind the technique is elegant. Interstellar dust grains made of silicates, when spun up by collisions and other processes, tend to align with the galaxy’s magnetic field, much like tiny compass needles. Once aligned, they act collectively as a polarizing filter for starlight passing through them. By measuring that polarization in the ultraviolet, PUFFINS can probe the smallest grain sizes, those smaller than about 50 nanometers, roughly the size of the common cold virus. Such grains are effectively invisible to optical and infrared telescopes, which is precisely why the ultraviolet window matters. The small grains are believed to play an outsized role in the chemistry and energy balance of the interstellar medium, yet their properties remain poorly constrained by existing observations.

Ramya Anche, an assistant research professor at Steward Observatory who serves as the mission’s principal investigator, described the stakes in plain terms. PUFFINS, she said, will change how we understand the dust between stars, and understanding these tiny grains is essential to building a complete picture of how interstellar dust forms, evolves over time, and ends up incorporated into new stars and planets. For Anche, the mission is also a personal milestone. She said PUFFINS represents the culmination of more than a decade of instrument-building work in polarimetry, and that it will be the first instrument she helped design that will actually be built and launched into space. She led the winning proposal to NASA while still a postdoctoral researcher at Steward Observatory, a detail that Steward Observatory director Buell Jannuzi highlighted as evidence of the institution’s unusual willingness to let junior scientists lead major programs.

B-G Andersson, the mission’s science principal investigator and assistant director for research at McDonald Observatory, has spent more than two decades studying how interstellar dust grains align with magnetic fields. He framed the mission’s scientific ambition around one of the oldest puzzles in the field. Interstellar polarization was discovered in 1949, he noted, and astronomers are finally beginning to understand its origin in dust grains. PUFFINS, he said, will work on fully solving what for more than fifty years was the second-longest-standing mystery of interstellar medium physics. The connection between grain alignment and magnetic fields means the mission’s data will also inform studies of the galaxy’s magnetic structure, a topic that touches everything from star formation to the propagation of cosmic rays.

The reach of interstellar dust extends far beyond the specialists who study it directly. Dust makes up only about one percent of the visible mass of the Milky Way, yet its effects ripple through nearly every corner of astrophysics. It dims and reddens the light of distant supernovae, complicating measurements of cosmic expansion. It shields molecules in cold clouds, enabling the chemistry that precedes star birth. It injects heavy elements, forged inside dying stars, back into the galactic ecosystem, where they later condense into the next generation of stars and planets. Because dust properties vary with environment, uncertainties about dust translate directly into uncertainties about nearly every astronomical measurement made from within the Milky Way.

PUFFINS is designed to attack that variability head-on. By observing stars at different distances and in different environments, including nearby stellar nurseries, reflection nebulae where dust scatters starlight back toward Earth, and even the Magellanic Clouds, our closest extra-galactic neighbors, the mission will build a detailed map of how interstellar dust varies across the sky. Comparing polarization signatures across such diverse settings allows scientists to separate the effects of grain composition from those of grain size distribution and magnetic environment. The target list of about 70 stars is chosen so that each sightline probes a distinct combination of dust conditions, turning the spacecraft into a sampling network for the galaxy’s dusty byways.

The mission is currently in its early design phase, with a launch target of 2030. Its planners see it as more than a standalone experiment: PUFFINS is intended to serve as a critical bridge to NASA’s Habitable Worlds Observatory, the agency’s next flagship space telescope currently in development for the 2040s. That flagship may carry an ultraviolet spectropolarimeter in one of its instrument ports, and PUFFINS data can help define what science such an instrument could accomplish. Demonstrating spectropolarimetry in orbit on a small platform, decades after the last such instrument flew, would de-risk the technology and sharpen the scientific case for its inclusion on a far larger observatory.

Behind the hardware, the team has been deliberate about designing PUFFINS as a training ground as much as a science mission. Graduate students and undergraduate researchers are embedded in the instrument development effort from the start, and the University of Arizona roster spans faculty, postdoctoral researchers, and students across Steward Observatory. Kyle Van Gorkom, an early-career faculty member at Steward Observatory who serves as deputy principal investigator and leads the spacecraft interface work, said the team is excited to start from conceptualizing a space telescope and see it through from build to launch to operating in orbit. While the payload is being designed and assembled, the science team will acquire spectropolarimetry data of the mission’s target stars in visible and infrared light using University of Arizona telescopes and other facilities around the world, groundwork that will enhance the ultraviolet data PUFFINS collects and enable a more complete analysis. Erika Hamden, a professor of astrophysics at Steward Observatory who leads the Arizona Space Institute, said the institute was set up to enable exactly these types of missions and that the team has a great science idea. If all goes to plan, by the early 2030s a dinner-plate-sized telescope will be quietly reading the light of distant stars, and with it the story of the dust that fills the space between them.

Subject of Research: Ultraviolet spectropolarimetry of starlight to characterize interstellar dust grains

Article Title: Stories from star dust: U of A-led mission will study light from distant stars

Article References: Stories from star dust: U of A-led mission will study light from distant stars. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: PUFFINS, NASA, interstellar dust, polarization, ultraviolet astronomy, spectropolarimetry, University of Arizona, Steward Observatory, Astrophysics Pioneers, small satellites, magnetic fields, star formation

Roman, as astronomers call the observatory, is designed from the ground up for speed and breadth. Rudnick described it as a thousand times faster than the Hubble Space Telescope at taking data over large portions of the sky. That advantage comes from Roman’s combination of a sensitive wide-field camera and a spectrographic mode that can capture the light of many objects at once. Where Hubble painstakingly stares at narrow patches of sky, Roman can sweep vast regions in a fraction of the time, making it the ideal instrument for a survey that aims to chart the distant universe on a scale never attempted before.

The GRACE team, assembled with colleagues at NASA’s Goddard Space Flight Center and a large group of international researchers, proposed to observe for 420 hours with Roman, covering an area of sky ten times the size of the full moon. Within that enormous footprint, the survey will take spectra of every galaxy by exposing for 40 hours at each location. Spectroscopy splits the light from each galaxy into its individual colors, and from those split spectra astronomers can extract a remarkable range of physical information: how fast the galaxy is receding, what chemical elements it contains, how vigorously it is forming new stars, and how its stars are arranged into a coherent shape.

The payoff will be a detailed three-dimensional map of the universe covering hundreds of thousands of galaxies. For each of them, GRACE team members plan to measure precise properties including distances, the mix of elements within them, the rate at which they form new stars, and their detailed morphologies. Rudnick emphasized that this combination of ultrasensitive spectroscopy over such a large area of sky is unprecedented and is expected to revolutionize knowledge of the distant universe. Because light from remote galaxies takes billions of years to reach Earth, the map will function as a time machine, showing astronomers how galaxies looked when the universe was far younger than it is today.

For Sangeeta Malhotra, a supervisory research astrophysicist at Goddard and co-principal investigator of GRACE, the mission is the culmination of a dream more than two decades in the making. She recalled that the team started imagining a project of this kind in 2005, as soon as they saw what Hubble could do for spectra of the earliest galaxies. From that point forward, she said, the group did not merely dream but worked with colleagues to build the scientific case for the wide-field space telescope that was ultimately realized as Roman. Now that the mission has launched, the team is looking forward to seeing the first data from GRACE arrive.

The survey’s observations are aimed at two pivotal epochs in cosmic history that astronomers refer to as cosmic dawn and cosmic noon. Rudnick leads the portion of the program focused on cosmic noon, a name he acknowledged can be misleading. The universe is 13.8 billion years old, and cosmic noon actually spans the interval from roughly 2 to 6 billion years after the Big Bang, nowhere near the midpoint of cosmic time. The label refers instead to what happened during that era: it was the period when galaxies across the universe were forming stars most rapidly. In that sense, cosmic noon marks the peak epoch of galaxy formation, the moment when the stellar output of the cosmos reached its zenith.

Among his multiple science goals with the GRACE data, Rudnick will investigate protoclusters, the earliest large assemblies of galaxies in the universe. In the present-day cosmos, galaxies are distributed across a hierarchy of structures. Galaxy clusters contain hundreds to thousands of galaxies bound together by their own gravity, moving around within a common gravitational well. Galaxy groups hold tens to hundreds of members. Beyond the groups and clusters, galaxies line filaments, long threads of matter that connect these systems into the vast network cosmologists call the cosmic web. The clusters and groups sit at the nodes, the intersections of that web, while still other galaxies drift in complete isolation, far from any neighbors.

A central question for Rudnick is how galaxies are shaped by the environments in which they live. Protoclusters contain all the galaxies that will eventually end up in a present-day galaxy cluster, but at these distant epochs they are spread out over very large areas of the sky. Observing a nearby cluster is comparatively easy because its member galaxies sit close together, but catching a cluster in the act of assembling requires the ability to study galaxies in exquisite detail across a wide field, which is precisely what Roman provides. Rudnick also hopes to learn how these structures connect with one another and how filaments tie the densest regions of the universe together. By his assessment, GRACE will be by far the biggest systematic survey to assess the cosmic web in the distant universe.

The environmental question extends to the striking differences between galaxies in crowded and empty regions today. In sparse, almost rural environments, galaxies tend to be spirals: rich in gas, actively forming stars, and rotating in an orderly fashion like a Frisbee. In dense regions such as clusters and groups, the picture changes dramatically. There, galaxies are no longer forming new stars, have exhausted much of their gas, and appear as conglomerations of stars moving in random orbits, like bees in a swarm. Astronomers call these elliptical galaxies because they look like fuzzy elliptical blobs on the sky, and their reddish color betrays a population dominated by old stars. What remains unknown is why galaxies in these different settings look so different, and by probing the earliest times at which galaxies could have been influenced by their surroundings, GRACE is positioned to make fundamental advances on that problem.

The practical timeline is already taking shape. Rudnick expects the first data from the Nancy Grace Roman Space Telescope to arrive a few months after commissioning, with the first GRACE-specific observations coming within the first year of operations. The impact will reach well beyond the professional community. NASA’s support will fund a postdoctoral researcher at Kansas, undergraduates at the university will work directly with data from the cutting-edge observatory, and Rudnick plans to incorporate the survey into a long-running outreach program at Lawrence High School. He noted that the program could define his scientific direction for the next five to ten years, and he highlighted what he considers one of the most meaningful aspects of GRACE: bringing data from NASA’s next flagship observatory to Kansas, giving researchers in the state first access to game-changing research on how the universe built its galaxies.

Subject of Research: Galaxy evolution and the cosmic web during cosmic noon, studied with the Nancy Grace Roman Space Telescope's GRACE survey

Article Title: Researcher will examine galactic evolution with new ‘Roman’ space telescope

Article References: Researcher will examine galactic evolution with new ‘Roman’ space telescope. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: Nancy Grace Roman Space Telescope, GRACE survey, galaxy evolution, cosmic noon, cosmic dawn, protoclusters, cosmic web, spectroscopy, NASA, Goddard Space Flight Center, University of Kansas, 3D galaxy map

The timing of the study is significant. NASA is currently planning multiple missions with durations exceeding six months in space, spanning commercial activity in low Earth orbit, the Artemis IV mission that will carry the first crewed flight to the lunar south pole in 2028, and the development of sustainable surface capabilities on the moon in preparation for eventual human journeys to Mars. Each of these endeavors will demand more human participation and more cumulative time spent in extraterrestrial environments, and each will draw on crews selected precisely because of their exceptional physical fitness and optimal health. The paradox highlighted by the new research is that the very individuals best equipped to endure the rigors of spaceflight may nonetheless carry forward hidden skeletal vulnerabilities that standard post-flight assessments do not fully capture.

The underlying physiology is well established, even if its long-term consequences have been harder to quantify. Muscle and bone loss in the microgravity environment of space is a common physiologic phenomenon among astronauts, and it is understood as an adaptive response to disuse. In the absence of gravitational loading, the skeleton receives far less mechanical stimulation than it does on Earth, and the body responds by reducing bone formation while resorption continues, gradually eroding bone mass and altering the internal architecture of load-bearing structures such as the hip. What makes this process insidious, as lead investigator Jean D. Sibonga, PhD, of NASA’s Johnson Space Center has noted, is that humans do not feel the effects of bone loss. An astronaut can return from space, feel stable, and even believe themselves fully recovered, then resume physically demanding activities despite underlying skeletal changes caused by prolonged disuse.

That disconnect between how astronauts feel and what their bones can actually withstand is central to why the new findings matter. A better understanding of bone health after spaceflight, the researchers argue, could help astronauts, patients, and clinicians assess long-term health concerns, determine appropriate care, and reduce the risk of injury. The hip is a particularly consequential site in this regard. Hip fractures are among the most serious fragility fractures in any population, associated with substantial morbidity, long recovery periods, and elevated mortality risk, especially when they occur at ages earlier than typical. A fracture rate that rises after long-duration flight, and that emerges at younger ages than would be expected in the general terrestrial population, therefore represents a warning signal that extends well beyond the immediate post-mission period.

One of the persistent obstacles in spaceflight medicine has been the sheer difficulty of gathering statistically robust data. Matthew T. Drake, MD, PhD, of the Endocrinology and Metabolic Bone Disease Service at the Hospital for Special Surgery in New York, and co-author of an accompanying editorial in the same journal, observed that while data to date on bone loss associated with spaceflight are consistent, a major limitation is accuracy, because astronaut cohort sample sizes are small and spaceflights are rare. With only a few hundred people having ever flown in space, and only a subset of those undertaking missions longer than 90 days, conventional epidemiological comparisons often lack the statistical power to detect meaningful differences in fracture outcomes. Small cohorts and rare events make it easy for real effects to hide within the noise.

To overcome this statistical limitation, the investigators took a different analytical approach. They surveyed a cohort of US-based astronauts for all fractures reported during annual clinical exams, assembling a longitudinal record of skeletal outcomes across careers that spanned missions of varying lengths. They then analyzed these data using Bayesian probabilistic modeling, a framework that allows researchers to draw more stable inferences from small samples by incorporating prior knowledge and quantifying uncertainty explicitly. The results were striking in their specificity. While the study did not find a higher incidence rate for fractures in general, the incidence rate for hip fractures in astronauts was increased after a long-duration spaceflight of more than 90 days. The effect was not a diffuse elevation of injury risk across the skeleton, but a concentrated signal at one of the body’s most mechanically critical sites.

Two features of that signal stand out. First, the increased hip fracture rate occurred at a younger age than would be expected in a non-astronaut terrestrial population, meaning astronauts were experiencing these injuries earlier in life than standard fracture epidemiology would predict. Second, the rate was greater than the rate observed before a long-duration mission, or with no spaceflight exposure at all, which points to the spaceflight itself, rather than age or occupational hazards alone, as the differentiating factor. Moshe Gertzulin, MD, also of the Endocrinology and Metabolic Bone Disease Service at the Hospital for Special Surgery and co-author of the accompanying editorial, summarized the implication: taken together, the finding of hip fracture occurrence at an age earlier than expected after long-duration spaceflight is clinical evidence that spaceflight-induced bone loss can lead to real long-term fracture consequences, which should be addressed accordingly.

The study’s authors argue that these conclusions carry direct implications for how space agencies monitor skeletal health. Dr. Sibonga emphasized the need to expand NASA’s current bone surveillance program beyond dual-energy X-ray absorptiometry, or DXA, measurements of bone mineral density, the long-standing standard for assessing skeletal status. DXA provides a two-dimensional summary of density but offers limited insight into the three-dimensional structure of bone, which is increasingly recognized as a critical determinant of fracture resistance. NASA has recently begun ordering quantitative computed tomography scans before and after spaceflight, a modality that can provide more detailed information about different areas of the bone and about whether astronauts have truly fully recovered. Without that finer-grained insight, astronauts could risk placing too much strain on bones that have not yet regained their preflight condition, precisely the scenario in which a fragility fracture is most likely to occur.

Beyond surveillance, the investigators advocate for prophylactic countermeasures designed to preserve the baseline, pre-flight skeletal health of the astronaut, rather than attempting to reverse damage after the fact. A fuller characterization of spaceflight-induced changes to bone mass, density, structure, and microarchitecture will determine which interventions, whether pharmaceutical agents, exercise regimens, or dietary adjustments, are most beneficial for long-term skeletal health. Current in-flight countermeasures centered on resistive exercise have slowed but not fully prevented bone deterioration on long missions, and the new fracture data suggest that residual deficits matter over timescales of years, not weeks. Identifying the right combination of countermeasures before crews embark on months-long lunar surface stays and multi-year Mars transits could mean the difference between a mission that ends with healthy crews and one that seeds decades of orthopedic problems.

The broader significance of the study extends beyond the astronaut corps. The disuse-related bone loss that astronauts experience is, in essence, an accelerated model of the immobilization and unloading that affects patients on Earth, from those confined to bed rest to individuals with spinal cord injuries or age-related osteoporosis. Insights gained from studying spaceflight-induced skeletal changes, and from the advanced imaging and modeling techniques being deployed to study them, may inform how clinicians on Earth assess and manage fragility in their own patients. For now, the message from Mayo Clinic Proceedings is unambiguous: the skeleton keeps a record of every day spent in microgravity, and for missions longer than 90 days, that record includes a heightened risk of hip fracture that surfaces earlier in life than anyone would want. As humanity prepares to spend longer stretches of time on the moon and beyond, safeguarding the bones that carry astronauts home has become a scientific challenge worthy of the ambition of the missions themselves.

Subject of Research: Long-duration spaceflight-induced bone loss and hip fracture risk in astronauts

Article Title: Study links 90+-day spaceflights to higher rates of hip fractures in astronauts

Article References: Study links 90+-day spaceflights to higher rates of hip fractures in astronauts. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: astronauts, hip fracture, bone loss, microgravity, spaceflight, NASA, Mayo Clinic Proceedings, bone mineral density, quantitative computed tomography, osteoporosis, Artemis, space medicine

The study, titled “Multiple simulated spaceflight stressors impact cardiac fibrosis, calcium dynamics, immune function, cytokines and gene variants in rat, Rattus norvegicus,” examined how several simulated spaceflight stressors influenced the rodent heart across multiple biological domains. Rather than focusing on a single endpoint, the investigators assessed structural changes in cardiac tissue, calcium handling within heart cells, immune function, circulating cytokine profiles and gene variants. This breadth matters because the cardiovascular system is integrated: a shift in immune signaling or calcium dynamics can ripple outward into tissue remodeling, electrical instability and impaired pumping function over time.

Galactic cosmic radiation is one of the defining hazards of missions beyond the protective envelope of Earth’s magnetosphere. Unlike the charged particles trapped in the Van Allen belts or the sporadic bursts of solar particle events, galactic cosmic rays consist largely of high-energy protons and heavier atomic nuclei traveling at near-relativistic speeds. When these particles, sometimes called HZE ions, pass through biological tissue, they deposit dense tracks of ionization that can damage DNA, proteins and cell membranes in ways that differ from the sparse ionizations produced by X-rays or gamma rays. Because Earth’s magnetic field and atmosphere shield people on the ground and largely protect crews in low Earth orbit, this radiation environment is essentially impossible to study directly in humans without leaving the planet.

To approximate that environment on the ground, researchers use particle accelerators to deliver carefully calibrated doses of heavy ions and protons to animal models. In the rat study, exposure to this simulated galactic cosmic radiation was associated with the accumulation of collagen in and around the coronary vessel wall. Collagen is the structural protein that forms scar tissue, and its excessive deposition, known as fibrosis, stiffens tissue and interferes with normal function. Perivascular fibrosis, specifically, thickens the supporting structures around blood vessels, which can restrict the ability of coronary vessels to dilate and deliver blood on demand. In the heart, such changes are associated with reduced compliance and an elevated risk of diastolic dysfunction in clinical medicine.

What makes the result particularly noteworthy is its timing. The fibrotic changes emerged during long-term follow-up, meaning the hearts of the exposed animals did not show the full extent of remodeling immediately after the radiation exposure. This delayed pattern is consistent with a broader theme in radiation biology: some forms of injury unfold slowly as cells respond to initial damage, inflammatory signals persist, and tissue repair processes go awry. For mission planners and flight surgeons, delayed effects are the most difficult kind to manage, because they may not announce themselves during a mission or even in the years immediately following it.

The study’s multi-domain design also reflects how modern space biology frames health risk. Calcium dynamics within cardiac cells are central to every heartbeat: calcium ions trigger the contraction of the heart muscle and must be rapidly cleared to allow relaxation. Disruption of this cycling can impair both the strength and rhythm of the heartbeat. Immune function and cytokines, the signaling molecules that coordinate inflammatory responses, connect to cardiac health because chronic low-grade inflammation is a recognized driver of fibrosis and vascular disease. By measuring these systems alongside structural tissue changes and genetic variation among the animals, the researchers could build a more complete picture of how simulated spaceflight stressors propagate through biology.

The inclusion of gene variants in the analysis acknowledges an important reality of radiation risk: individual susceptibility is not uniform. Variation in genes governing DNA repair, antioxidant defenses, inflammatory signaling and calcium handling could plausibly influence how strongly a given organism responds to the same radiation exposure. In a population of rats, as in a population of astronauts, some individuals may be intrinsically more vulnerable to fibrotic remodeling or immune dysregulation than others. Understanding that variability is a prerequisite for personalized risk assessment, which agencies such as NASA have identified as a priority for protecting crews on long-duration missions.

The research was supported by funding from the National Aeronautics and Space Administration under award 80NSSC19K0498 and by the Foundation for Heart Science under award 081735, with the funders having no role in study design, data collection, analysis or manuscript preparation. The authors declared no competing interests. The work was conducted by researchers based in the United States and published as a peer-reviewed article in PLOS One under the DOI 10.1371/journal.pone.0357485. An accompanying illustration by Dave Kiehl, released under a CC-BY 4.0 license, depicts the potential cardiovascular effects of galactic cosmic radiation during a Mars mission, including the perivascular fibrosis observed in the study.

The relevance to human spaceflight is direct. A crewed mission to Mars would expose astronauts to months of galactic cosmic radiation with no option for early return, and current mitigation strategies, primarily spacecraft shielding and mission duration limits, offer only partial protection against high-energy heavy ions. Cardiac fibrosis is a slow, cumulative process in terrestrial medicine, linked over decades to hypertension, aging and cancer therapies involving radiation to the chest. If deep-space radiation accelerates or initiates similar remodeling, the cardiovascular consequences might only surface years after a mission concludes, complicating both health monitoring and long-term care of former astronauts.

At the same time, the researchers and the field at large caution against overinterpreting a single animal study. Rats are not humans, simulated radiation exposures involve dose rates and particle spectra that only approximate the true space environment, and the doses used in laboratory studies are often delivered over compressed timeframes compared with the chronic, low-dose exposure of an actual mission. Translating these findings into human risk estimates requires replication across species, dose ranges and exposure schedules, as well as mechanistic work to identify the biological pathways linking heavy-ion exposure to collagen deposition in coronary vessels. Still, the study strengthens the case that the heart belongs on the list of organs that space agencies must actively monitor and protect, and it underscores a sobering possibility: the most consequential injuries of a journey to Mars may be the ones that reveal themselves only after the homecoming.

Subject of Research: Cardiac effects of simulated galactic cosmic radiation exposure in rats

Article Title: Simulated space radiation may damage rat hearts long after exposure

Article References: Simulated space radiation may damage rat hearts long after exposure. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: space radiation, galactic cosmic rays, cardiac fibrosis, rats, PLOS One, NASA, Mars missions, coronary vessels, collagen deposition, calcium dynamics, immune function, astronaut health

The satellite at the center of the discovery is the Surface Water and Ocean Topography mission, known as SWOT, a joint effort between NASA and international partners that has been in orbit since December 2022. SWOT was built to measure the water level of oceans, lakes, rivers and other large bodies of water at least once every three weeks. It does so using interferometric radar, a technique that compares multiple radar scans over time to detect subtle elevation changes on the water surface. Oceans and big lakes are relatively straightforward targets for such an instrument. Irrigation canals, however, are frequently narrower than 20 meters, placing them at the extreme low end of the satellite’s observational power. The mission’s architects assumed that water-level fluctuations inside such small channels would be invisible to the spacecraft.

Faisal Hossain, a professor of civil and environmental engineering at the University of Washington, was not convinced that the canals were beyond reach. In 2025, he and his collaborators published the Global Registry of Agricultural Irrigation Networks, or GRAIN, a global dataset that used open-source mapping data and machine learning to chart 3.8 million kilometers of canal networks worldwide. That registry gave researchers, for the first time, a comprehensive map of where the world’s agricultural canals actually are. What it could not show was what was happening inside them, which is precisely the kind of information farmers and irrigation authorities need to manage water deliveries.

Lead author Mridul Sharma, a graduate research assistant in civil and environmental engineering at the University of Washington, took the logical next step. He overlaid SWOT’s radar data on top of the GRAIN canal map to see whether meaningful elevation changes could be detected where the mapped canals ran. The result surprised even the team. Roughly 800,000 kilometers of canals in Asia showed water levels that the satellite could measure with moderate to high confidence at more than 85 percent of locations. “It was quite the accidental and pleasant discovery to see that SWOT is able to track flow direction and water levels in most of the canals skillfully,” Sharma said. “GRAIN mapped where canals are. SWOT revealed what was happening inside them. They complemented each other beautifully.”

As a proof of concept, the researchers concentrated on canals in 22 countries across Asia, a region where irrigation plays a critical role in feeding roughly 3 billion people. They examined the satellite’s data along 800,000 kilometers of canals within that region and assigned a confidence score to each kilometer. The scoring was based on two criteria: how closely the satellite’s view of the canal water matched the known contours of the surrounding land, and how strong the radar signal was relative to the noise in the data. Where the observed water surface tracked the terrain smoothly and the signal stood out clearly from background noise, confidence was high. Where the two diverged or the signal was weak, confidence dropped.

The team also validated the approach against ground truth. They compared the satellite’s data to real canal measurements in the United States and found that the confidence levels produced by their scoring system matched the system’s actual abilities. This cross-continental check matters because it suggests the confidence scores are not merely statistical artifacts but reliable indicators of where SWOT’s canal observations can be trusted. Of all the kilometers studied, the team designated 37.5 percent as highly observable, 46.9 percent as moderately observable and 15.6 percent as poorly observable. In other words, more than eight in every ten kilometers of canal in the study region could be monitored with at least moderate confidence by a satellite that was never designed to see them.

The patterns behind those numbers are instructive. The highest-confidence areas corresponded to well-organized, wider canals with smooth slopes and open surroundings, conditions that give the radar a clean view of the water surface and a stable signal to track. Dense vegetation around canals proved to be the satellite’s biggest obstacle, since foliage scatters and absorbs radar energy and obscures the narrow ribbon of water beneath. That limitation is likely to matter most in tropical and subtropical regions where canal banks are heavily vegetated. Encouragingly, the researchers expect performance to improve over time as Hossain and other scientists working with SWOT data build better algorithms and analysis methods tailored to small water bodies.

The study, published September 30 in Geophysical Research Letters, points toward a fundamentally new way of managing the conveyance systems that sustain modern agriculture. “What we are seeing is not simply a new satellite capability,” Hossain said. “It may represent a fundamentally new way of managing the water conveyance systems that sustain modern agriculture.” The next step for the research team is to turn the discovery into practical tools. Hossain and collaborators at the University of Washington are already building systems that use SWOT data to monitor and improve water delivery in South Asia and to aid canal management in the western United States, where aging irrigation infrastructure faces growing pressure from drought and competing demands.

The practical implications for farmers could be far-reaching. In time, growers around the world could use SWOT-powered tools to monitor which canals are carrying sufficient water, see where water levels drop unexpectedly, identify stretches where canal flow might be stagnant, and understand how water delivery evolves across an entire growing season. The researchers offer a concrete example: a farmer preparing to plant rice, one of the most water-intensive staple crops, might see through SWOT-based monitoring that there will not be enough water arriving to the field to support the crop, and switch instead to a less water-intensive option such as corn or wheat. Such decisions, made with timely and objective information rather than guesswork, could reduce crop losses and stretch scarce water supplies further.

Beyond individual farms, the new view of agriculture’s circulatory system could improve drought preparedness, strengthen food security and help researchers map humanity’s transformation of the global freshwater cycle. Irrigation is one of the largest human uses of freshwater on the planet, and canal networks are the arteries through which most of that water moves. Measuring them systematically from space adds a missing piece to the global water balance. “Some of the most important scientific discoveries happen when a tool designed for one purpose unexpectedly reveals something else,” Hossain said. “SWOT may be offering a similar surprise. A mission built to study oceans and rivers has begun revealing the hidden dynamics of the canals that sustain agricultural production around the globe. What was once invisible is becoming measurable.” Co-authors on the study include Sanchit Minocha and Shahzaib Khan, graduate research assistants at the University of Washington; Sarath Suresh, a postdoctoral researcher at Virginia Tech who completed the research as a UW doctoral student; and Tamlin Pavelsky of the University of North Carolina at Chapel Hill. The research was funded by NASA.

Subject of Research: Satellite-based monitoring of irrigation canal water levels using the NASA SWOT mission

Article Title: NASA satellite can monitor most of world’s irrigation canals, study finds

Article References: NASA satellite can monitor most of world’s irrigation canals, study finds. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: SWOT satellite, irrigation canals, water monitoring, NASA, University of Washington, remote sensing, interferometric radar, food security, drought preparedness, agriculture, freshwater, Geophysical Research Letters