Tara P. A. Tahseen, Nikolaos Nikolaou, Luís F. Simões
et al.
Complex physical simulations often require trade-offs between model fidelity and computational feasibility. We introduce Adaptive Online Emulation (AOE), which dynamically learns neural network surrogates during simulation execution to accelerate expensive components. Unlike existing methods requiring extensive offline training, AOE uses Online Sequential Extreme Learning Machines (OS-ELMs) to continuously adapt emulators along the actual simulation trajectory. We employ a numerically stable variant of the OS-ELM using cumulative sufficient statistics to avoid matrix inversion instabilities. AOE integrates with time-stepping frameworks through a three-phase strategy balancing data collection, updates, and surrogate usage, while requiring orders of magnitude less training data than conventional surrogate approaches. Demonstrated on a 1D atmospheric model of exoplanet GJ1214b, AOE achieves 11.1 times speedup (91% time reduction) across 200,000 timesteps while maintaining accuracy, potentially making previously intractable high-fidelity time-stepping simulations computationally feasible.
Supernova remnants (SNRs) are the aftermath of massive stellar explosions or of a white dwarf in a binary system, representing critical phases in the life cycle of stars and playing an important role in galactic evolution. Physical properties of SNRs such as their shock velocity, density and age are important elements for constraining models for their evolution and understanding the physical processes responsible for their morphological appearance and emission processes. Our study provides, for the first time, a comprehensive statistical analysis of the physical parameters in 64 Galactic SNRs both as a population as well as regions within individual objects. These 64 objects represent the subset of the 310 known Galactic SNRs for which there are published optical data, from which we compiled their physical parameters through an exhaustive literature survey. Through a systematic statistical analysis accounting for uncertainties and/or upper and lower limits in these parameters we obtain distributions of the electron density and shock velocity in the studied SNRs and regions within them. This information is combined with constraints on their age and type. Analysis of electron density and shock velocity distributions for the entire sample of SNRs shows that they are consistent with a log-normal distribution and a skewed log-normal distribution, respectively. Within individual remnants, our study reveals that electron density and shock velocity show larger scatter in younger objects, reflecting the varying conditions of the ambient medium immediately surrounding the explosion epicenter and their impact on SNR evolution. Comparison of the dependence of the shock velocity and density on the supernova age with expectations from theoretical models shows good agreement.
Jasmine R. Kobayashi, Daniela Martin, Valmir P Moraes Filho
et al.
Labeling or classifying time series is a persistent challenge in the physical sciences, where expert annotations are scarce, costly, and often inconsistent. Yet robust labeling is essential to enable machine learning models for understanding, prediction, and forecasting. We present the \textit{Clustering and Indexation Pipeline with Human Evaluation for Recognition} (CIPHER), a framework designed to accelerate large-scale labeling of complex time series in physics. CIPHER integrates \textit{indexable Symbolic Aggregate approXimation} (iSAX) for interpretable compression and indexing, density-based clustering (HDBSCAN) to group recurring phenomena, and a human-in-the-loop step for efficient expert validation. Representative samples are labeled by domain scientists, and these annotations are propagated across clusters to yield systematic, scalable classifications. We evaluate CIPHER on the task of classifying solar wind phenomena in OMNI data, a central challenge in space weather research, showing that the framework recovers meaningful phenomena such as coronal mass ejections and stream interaction regions. Beyond this case study, CIPHER highlights a general strategy for combining symbolic representations, unsupervised learning, and expert knowledge to address label scarcity in time series across the physical sciences. The code and configuration files used in this study are publicly available to support reproducibility.
Matilde Valente, Tiago C. Dias, Vasco Guerra
et al.
Data-driven machine learning models often require extensive datasets, which can be costly or inaccessible, and their predictions may fail to comply with established physical laws. Current approaches for incorporating physical priors mitigate these issues by penalizing deviations from known physical laws, as in physics-informed neural networks, or by designing architectures that automatically satisfy specific invariants. However, penalization approaches do not guarantee compliance with physical constraints for unseen inputs, and invariant-based methods lack flexibility and generality. We propose a novel physics-consistent machine learning method that directly enforces compliance with physical principles by projecting model outputs onto the manifold defined by these laws. This procedure ensures that predictions inherently adhere to the chosen physical constraints, improving reliability and interpretability. Our method is demonstrated on two systems: a spring-mass system and a low-temperature reactive plasma. Compared to purely data-driven models, our approach significantly reduces errors in physical law compliance, enhances predictive accuracy of physical quantities, and outperforms alternatives when working with simpler models or limited datasets. The proposed projection-based technique is versatile and can function independently or in conjunction with existing physics-informed neural networks, offering a powerful, general, and scalable solution for developing fast and reliable surrogate models of complex physical systems, particularly in resource-constrained scenarios.
Tunnel mapping systems are essential for tunnel inspection, integrating sensors like LiDAR, cameras, and odometers to enhance data accuracy. However, calibration is challenging due to mechanical constraints and repetitive sensor installations, especially for LiDAR-Camera alignment. Existing methods struggle in tunnels with poor lighting and low texture, and they fail to address irregular vibrations from the flashing light system, causing instability. We propose a robust online calibration technique for LiDAR-Camera extrinsic parameters. By establishing a reversible mapping through surface parameterization, our approach ensures accurate cross-modality alignment. Additionally, we use depth constraints to stabilize adjacent camera stations, which are typically short-edge connections and prone to instability in photogrammetric bundle adjustment. This effectively mitigates irregular vibration effects. Validation in real-world tunnels confirms persistent vibration issues despite mechanical reinforcement. Our algorithm achieves precise point cloud and image alignment, reducing back-projection errors by over 50% and significantly improving data fusion accuracy in challenging conditions.
Cinthya Soledad Manjarrez-Rangel, Silvana Raquel Halac, Luciana Del Valle Mengo
et al.
Eutrophication has intensified in lacustrine systems across the American continent, which has been primarily driven by human activities such as intensive agriculture, wastewater discharge, and land-use change. This phenomenon adversely affects water quality, biodiversity, and ecosystem functioning. However, studies addressing the historical evolution of trophic states in lakes and reservoirs remain limited—particularly in tropical and subtropical regions. In this context, sedimentary records serve as invaluable archives for reconstructing the environmental history of water bodies. Paleolimnological approaches enable the development of robust chronologies to further analyze physical, geochemical, and biological proxies to infer long-term changes in primary productivity and trophic status. This review synthesizes the main methodologies used in paleolimnological research focused on trophic state reconstruction with particular attention to the utility of proxies such as fossil pigments, diatoms, chironomids, and elemental geochemistry. It further underscores the need to broaden spatial research coverage, fostering interdisciplinary integration and the use of emerging tools such as sedimentary DNA among others. High-resolution temporal records are critical for disentangling natural variability from anthropogenically induced changes, providing essential evidence to inform science-based lake management and restoration strategies under anthropogenic and climate pressures.
Study region: Horqin Sandy Land (HQSL), a typical agro-pastoral transition zone in northern China. Study focus: This study assessed groundwater storage anomaly (GWSA) in HQSL under climate change and human activity, using Gravity Recovery and Climate Experiment (GRACE) satellite data, Global Land Data Assimilation System (GLDAS) models and in-situ well observations. Herein, Mann–Kendall test, Empirical Orthogonal Functions (EOFs) and Partial Least Squares Structural Equation Modeling (PLS-SEM) model were applied to analyze the seasonal and long-term trend changes of GWSA and quantify the process of GWSA in HQSL. New hydrological insights: HQSL was divided into sandy plain of intensive water use (Zone I) and mountain of water source (Zone II). GWSA from GRACE fitted well with in-situ data (r² = 0.65, p < 0.01). GWSA declined at −0.15 ± 0.12 mm/yr from 1985 to 2002 derived by in-situ observation wells and at −7.79 ± 0.87 mm/yr from 2002 to 2020 evaluated by GRACE. The decline was more pronounced in Zone I (−5.04 ± 0.23 mm/yr) than in Zone II (−3.84 ± 0.18 mm/yr) from 2002 to 2020. Monthly variations peaked in June (−30.78 mm), mitigated by precipitation in August (−8.10 mm) from 2002 to 2020. Spatially, GWSA loss intensified after 2013, particularly in northern mountains. Climate factors consistently influenced GWSA while growing human activity impacts intensified after 2010. These findings provide valuable insights for locals to mitigate climate change impacts through optimization of human activity, as water-saving land use strategies.
Doris Esenarro, Jimena Ccalla, Guisela Yabar
et al.
The objective of this research is to conduct a spatial analysis of the functional Andean worldview of the Ankasmarka Archaeological Site, located in Calca, Peru. The preservation of cultural heritage in Latin America faces significant challenges that threaten the integrity of key sites such as Ankasmarka. Despite its historical relevance, this site lacks available open access information and data, collected in accessible reports, which hinders the attraction of attention and funding necessary for its conservation. Furthermore, urbanization and uncontrolled tourism negatively impact both cultural traditions and the connection of local communities with their past. The methodology employed is based on a systematic review of primary information, supplemented by excavation reports and official sources. Specialized software such as AutoCAD Architecture and Revit were used to carry out the topographic and architectural survey of the site, enabling the precise and rigorous interpretation of the data. This article focuses on the spatial and functional description of the site, with the aim of paving the way for future research in specific areas such as formal and structural analysis, as well as social and political dynamics. The results reveal a complex organizational structure at Ankasmarka, with enclosures designated for various functions, particularly storage and agricultural activities. The site is divided into three sectors: Sector A, which includes housing, storage areas, and tombs; and Sector B and C, with the highest concentration of housing and agricultural zones with storage areas, respectively. The findings underscore the interrelationship between agriculture, funerary practices, and architecture, highlighting the importance of Ankasmarka in the lives of its ancient inhabitants and the need for continued future research.
Nearshoring is a strategic approach that allows companies to establish operations close to their consumer markets, offering a key opportunity for Mexico given its geographic location, skilled workforce, and trade agreements with North America. This strategy helps businesses reduce costs, improve supply chain resilience, and enhance operational efficiency by relocating production or services to closer locations. Infrastructure plays a crucial role in nearshoring, as efficient connectivity between roads, railways, ports, airports, and intermodal terminals is essential for logistics and transportation. This research evaluates multimodal connectivity in Mexico using geographic information systems (GIS) to identify the states best positioned for nearshoring operations. The methodology consists of two stages: the first includes a comprehensive literature review and data collection, while the second focuses on spatial data analysis by applying the concepts of location, distribution, accessibility, and spatial association. The results determine multimodal connectivity levels across Mexican states and identify those with the highest potential, advantages, or challenges for nearshoring development and investment.
Matthew Jörke, Shardul Sapkota, Lyndsea Warkenthien
et al.
Mobile health applications show promise for scalable physical activity promotion but are often insufficiently personalized. In contrast, health coaching offers highly personalized support but can be prohibitively expensive and inaccessible. This study draws inspiration from health coaching to explore how large language models (LLMs) might address personalization challenges in mobile health. We conduct formative interviews with 12 health professionals and 10 potential coaching recipients to develop design principles for an LLM-based health coach. We then built GPTCoach, a chatbot that implements the onboarding conversation from an evidence-based coaching program, uses conversational strategies from motivational interviewing, and incorporates wearable data to create personalized physical activity plans. In a lab study with 16 participants using three months of historical data, we find promising evidence that GPTCoach gathers rich qualitative information to offer personalized support, with users feeling comfortable sharing concerns. We conclude with implications for future research on LLM-based physical activity support.
Manuel Tonneau, Diyi Liu, Samuel Fraiberger
et al.
Perceptions of hate can vary greatly across cultural contexts. Hate speech (HS) datasets, however, have traditionally been developed by language. This hides potential cultural biases, as one language may be spoken in different countries home to different cultures. In this work, we evaluate cultural bias in HS datasets by leveraging two interrelated cultural proxies: language and geography. We conduct a systematic survey of HS datasets in eight languages and confirm past findings on their English-language bias, but also show that this bias has been steadily decreasing in the past few years. For three geographically-widespread languages -- English, Arabic and Spanish -- we then leverage geographical metadata from tweets to approximate geo-cultural contexts by pairing language and country information. We find that HS datasets for these languages exhibit a strong geo-cultural bias, largely overrepresenting a handful of countries (e.g., US and UK for English) relative to their prominence in both the broader social media population and the general population speaking these languages. Based on these findings, we formulate recommendations for the creation of future HS datasets.
Formulation of the problem. This is the second part of a trilogy dedicated to the analysis of climate indicators in central Ukraine over the entire period of instrumental observations, which analyzes air temperature data from the weather stations of Uman, Kropyvnytskyi, and Poltava. This work addresses issues related to the 13th Sustainable Development Goal, which is to combat climate change and strengthen resilience and adaptation to climate-related hazards and disasters in all countries.
The purpose of this study was to analyze data from weather stations in central Ukraine that have the longest period of observation and to find patterns in the dynamics of temperature indicators over the past 140-200 years.
Data and methods. To characterize the climate of central Ukraine, we analyzed the average monthly and average annual temperatures of Uman, Kropyvnytskyi, and Poltava, which have the longest continuous or almost continuous period of observation. Based on these data, we have constructed graphs of changes in the average annual and average monthly temperatures for the winter and spring seasons. To analyze the dynamics of temperature indicators, we constructed linear and 11-year moving trends.
Results. At all weather stations, there is a trend towards an increase in both average annual air temperatures and temperatures for certain months. In particular, in Uman, the average annual temperature over the entire observation period (138 years) has increased from +6.80C to +8.60C, i.e. by 1.8 degrees. In Kropyvnytskyi, average annual temperatures over 149 years increased from +7.40C to +8.90C, i.e. by 1.5 degrees. In Poltava, the average annual temperature over 199 years has increased from +5.90C to +8.70C, i.e. by 2.8 degrees (since 1886 from +6.40C to +8.70C, i.e. by 2.3 degrees). At all weather stations, the most significant increase in average annual temperatures occurred between 1989 and 2023. Temperatures in the autumn months increased the least. Over the entire observation period, average monthly temperatures in September/October/November increased from 0.3/0.1/0.10C in Uman, 0.6/0.1/1.80C in Kropyvnytskyi to 1.5/1.2/1.90C (since 1886 – 0.9/0.9/1.70C) in Poltava. All three meteorological stations have common periods of temperature increases and decreases, in particular, a decrease in average monthly summer temperatures occurred from 1947-1969 to 1985-1995; from 1986-1996 to 2023, an increase in air temperature. Air temperatures in the summer months have increased quite significantly. Over the entire period of observation, the average monthly temperature in June/July/August increased from 0.9/0.3/0.70C in Kropyvnytskyi, 1.9/1.3/1.60C (since 1886 – 1.3/1.2/1.40C) in Poltava to 2.0/1.1/1.10C in Uman. The greatest increase in average monthly autumn temperatures occurred from 1999-2001 to 2023. Analyzing the graphs of 11-year moving averages, one can see the presence of periods of increase and decrease in average monthly temperatures lasting about 33 years or doubled periods lasting about 66 years.
Scientific novelty. For the first time, the data of meteorological stations in central Ukraine for the entire period of observation (138 years – Uman, 149 years – Kropyvnytskyi, 199 years – Poltava) were analyzed and regularities in the dynamics of temperature indicators were determined.
The practical significance lies in the possibility of using the results of the study to predict future climate change.
David Bertermann, Mario Rammler, Mark Wernsdorfer
et al.
For large infrastructure projects, such as high-voltage underground cables or for evaluating the very shallow geothermal potential (vSGP) of small-scale horizontal geothermal systems, large-scale geothermal collector systems (LSCs), and fifth generation low temperature district heating and cooling networks (5GDHC), the thermal conductivity (λ) of the subsurface is a decisive soil parameter in terms of dimensioning and design. In the planning phase, when direct measurements of the thermal conductivity are not yet available or possible, λ must therefore often be estimated. Various empirical literature models can be used for this purpose, based on the knowledge of bulk density, moisture content, and grain size distribution. In this study, selected models were validated using 59 series of thermal conductivity measurements performed on soil samples taken from different sites in Germany. By considering different soil texture and moisture categories, a practicable guideline in the form of a decision tree, employed by empirical models to calculate the thermal conductivity of unconsolidated soils, was developed. The Hu et al. (2001) model showed the smallest deviations from the measured values for clayey and silty soils, with an RMSE value of 0.20 W/(m∙K). The Markert et al. (2017) model was determined to be the best-fitting model for sandy soils, with an RMSE value of 0.29 W/(m∙K).
Tobias Golling, Takuya Nobe, Dimitrios Proios
et al.
In the hunt for new and unobserved phenomena in particle physics, attention has turned in recent years to using advanced machine learning techniques for model independent searches. In this paper we highlight the main challenge of applying anomaly detection to jet physics, where preserving an unbiased estimator of the jet mass remains a critical piece of any model independent search. Using Variational Autoencoders and multiple industry-standard anomaly detection metrics, we demonstrate the unavoidable nature of this problem.
The simulation of vegetation dynamics is essential for guiding regional ecological remediation and environmental management. Recent progress in deep learning methods has provided possible solutions to vegetation simulations. The gated recurrent unit (GRU) is one of the latest deep learning algorithms that can effectively process dynamic data. However, static and dynamic data, which typically coexist in the datasets of vegetation dynamic changes, are typically processed indistinguishably. To efficiently extract spatiotemporal patterns and improve our ability to simulate potential vegetation changes, we introduced GRU into vegetation simulation and further amended the original structure of GRU according to the characteristics of the simulation dataset. The new model, the vegetation dynamics model (VDM), can independently process static and dynamic data using a more appropriate algorithm, thereby improving the simulation accuracy. Moreover, we presented a model test applied in the Luntai Desert-Oasis Ecotone in Northwest China and compared the performance of the VDM with baseline models. The results showed that the VDM produced a 7.51% higher coefficient of determination (R2) value, 7.51% higher adjusted R2 value, 16.67% lower mean squared error, and 10.78% lower mean absolute error than those of the GRU, which is the best baseline model. The proposed VDM is the first GRU-based simulation model of vegetation dynamics that has the potential to detect the time-order characteristics of dynamic factors by comprehensively considering the static information that affects vegetation changes. Moreover, the flexibility of the VDM, in combination with the wide availability of data from different data sources, aids the broader application of the VDM.
Land-use and land-cover analyses based on satellite image classification are used in most, if not all, sub-disciplines of physical geography. Data availability and increasingly simple image classification techniques – nowadays, even implemented in simple geographic information systems – increase the use of such analyses. To assess the quality of such land-use analyses, accuracy metrics are applied. The results are considered to have sufficient quality, exceeding thresholds published in the literature. A typical practice in many studies is to confuse accuracy in remote sensing with quality, as required by physical geography. However, notions such as quality are subject to normative considerations and performative practices, which differ between scientific domains. Recent calls for critical physical geography have stressed that scientific results cannot be understood separately from the values and practices underlying them. This article critically discusses the specific understanding of quality in remote sensing, outlining norms and practices shaping it and their relation to physical geography. It points out that, as a seeming paradox, results considered more accurate in remote sensing terms can be less informative – or meaningful – in geographical terms. Finally, a roadmap of how to apply remote sensing land-use analyses more constructively in physical geography is proposed.