GRO.publications
Not a member yet
137257 research outputs found
Sort by
Enhancing Debris Flow Warning via Machine Learning Feature Reduction and Model Selection
No full textAbstract The advent of machine learning has significantly improved the accuracy of identifying mass movements through the seismic waves they generate, making it possible to implement real‐time early warning systems for debris flows. However, we lack a profound understanding of the effective seismic features and the limitations of different machine learning models. In this work, we investigate eighty seismic features and three machine learning models for single‐station‐based binary debris flow classification and multi‐station‐based warning tasks. These seismic features, derived from physical and statistical knowledge of impact sources, are grouped into five sets: Benford's law, waveform, spectra, spectrogram, and network. The machine learning models belong to two families: two ensemble models, Random Forest and eXtreme Gradient Boosting (XGBoost); one recurrent neural network model, Long Short‐Term Memory (LSTM). We analyzed feature importance from the ensemble models and found that the number and even the types of seismic features are not critical for training an effective binary classifier for debris flow. When using models designed to capture patterns in sequential data rather than focusing on information only in one given window, using the LSTM does not significantly improve the performance of binary debris flow classification task over Random Forest and XGBoost. For the multi‐station‐based debris flow warning task, the LSTM model predicts debris flow probability more consistently and provides longer warning times. Our proposed framework simplifies machine learning‐driven debris flow classification and lays the foundation for affordable seismic signal‐driven early warning using a sparse seismic network.Plain Language Summary Machine learning can enhance the accuracy of detecting debris flow, landslides, and rockfall through seismic signals and add more time to warn people. However, it remains to be seen which machine learning models and seismic features are the most effective, especially to identify debris flows. In this study, we utilized two machine learning models, the Random Forest model and the eXtreme Gradient Boosting model (XGBoost), to evaluate the importance of seismic features in distinguishing debris flows from other events of interest. Our study found that using over seventy seismic features to train a machine learning model to classify debris flows includes unnecessary input features. The number and even the types of seismic features are not critical for training an effective binary classifier for debris flow. We also experimented with a different algorithm, Long Short‐Term Memory (LSTM), but it did not improve classification performance compared to Random Forest and XGBoost. The LSTM was more effective in detecting the onset of debris flows and producing fewer false warnings for alarms. Our approach simplifies the use of seismic signals to build early warning systems and may be applied to other types of mass movement events as well.Key Points We compare the performance of three machine learning models and eighty seismic features in five sets to enhance debris flow early warning The number and even the types of seismic features are not critical for training an effective binary classifier for debris flow Long Short‐Term Memory performs like ensemble models in classification task but it excels in predicting debris flow probability for warningState Key Laboratory of Geohazard Prevention and Geoenvironment Protection https://doi.org/10.13039/501100011171National Natural Science Foundation of China https://doi.org/10.13039/50110000180
Microbiota as Potential Functional Traits Facilitating Springtail Activity in Winter
No full textABSTRACT Understanding the role of microbiota in supporting animal survival and activity under extreme environmental conditions provides valuable insights into adaptation and resilience mechanisms in ecosystems. While vertebrate microbiota have received considerable attention, those associated with arthropods, particularly species capable of surviving sub‐zero temperatures, remain poorly understood. Springtails (Collembola), key contributors to litter decomposition and soil ecosystem functioning, require specialized adaptations to endure harsh winter conditions. Using the α‐ and β‐niche trait concept and phylogenetic comparative approaches, we investigated the microbiota of 10 coexisting springtail species with different overwintering strategies. Our results revealed that certain bacterial genera, including Marmoricola , Mycobacterium , Rhodococcus , and Vibrionimonas , exhibited phylogenetic signal, suggesting evolutionary constraints on their potential roles in hosts. Winter‐active springtail species harbored higher bacterial diversity and distinct microbial community compositions compared to inactive species, with enrichment in bacteria such as Wolbachia , Morganellaceae , and Micrococcaceae . Additionally, winter‐active species exhibited higher energy metabolism and lower lipid metabolism, alongside more frequent positive interactions within bacterial networks. These findings suggest that microbiota may play a functional role in supporting the metabolic demands of winter‐active springtails, potentially contributing to their adaptation to cold environments. Overall, our study highlights the role of microbiota in shaping ecological success and adaptation of arthropods to extreme conditions, providing new perspectives for soil animal research by integrating microbial functional traits with the evolutionary context of microbe‐host interactions
Valley-controlled photoswitching of metal–insulator nanotextures
No full textAbstract Spatial heterogeneity and phase competition are hallmarks of strongly correlated materials, influencing phenomena such as colossal magnetoresistance and high-temperature superconductivity. Active control over phase textures further promises tunable functionality at the nanoscale. Although light-induced switching of a correlated insulator to a metallic state is well established, optical excitation generally lacks the specificity to select subwavelength domains and determine final textures. Here we drive the domain-specific quench of a textured Peierls insulator using valley-selective photodoping. Polarized excitation exploits the anisotropy of quasi-one-dimensional states at the charge-density-wave gap to initiate an insulator–metal transition with minimal electronic heating. We find that averting dissipation facilitates domain-specific carrier confinement, control over nanotextured phases and reduction in thermal relaxation from the metastable metallic state. This valley-selective photoexcitation approach will enable the activation of electronic phase separation beyond thermodynamic limitations, facilitating optically controlled hidden states, engineered heterostructures and polarization-sensitive percolation networks
Double-peaked Ca II traces a relativistic broad-line region disk in NGC 4593
No full textContext. Double-peaked emission lines are observed in a small percentage of active galactic nuclei (AGN). These lines allow the determination of fundamental properties of the line-emitting region, known as the broad-line region (BLR). Aims. We investigated the structure and kinematics of the BLR in the nearby Seyfert galaxy NGC 4593 through an analysis of the near-infrared (NIR) line blend of Ca II λ 8498, λ 8542, λ 8662, and O I λ 8446 observed in a 2019 VLT/MUSE spectrum. Methods. We performed a detailed decomposition of the NIR Ca II triplet and O I λ 8446 blend, extracting clean profiles of Ca II λ 8498, λ 8542, λ 8662, and O I λ 8446. We then fitted Ca II λ 8662 with a relativistic elliptical line-emitting accretion disk model. Results. The extracted line profiles are double-peaked with a full width at half maximum (FWHM) of ∼3700 km s −1 and exhibit a redward asymmetry with a red-to-blue peak ratio of 4:3. The Ca II triplet lines have an intensity ratio of 1:1:1 and show no evidence of a central narrow or intermediate-width component. The profiles of Ca II and O I are remarkably similar, suggesting a common region of origin. Given the 1:1:1 ratio of the Ca II triplet, this region is likely a high-density emission zone, and the Ca II λ 8662 profile is well described by a mildly eccentric, low-inclination relativistic line-emitting disk with minimal internal turbulence. The profile represents one of the clearest kinematic signatures of a relativistic disk observed in BLR emission lines to date. Conclusions. The double-peaked profiles of the NIR Ca II triplet and O I λ 8446 in NGC 4593 represent the first detection of double-peaked Ca II and O I λ 8446 in a nontransient AGN spectrum. The minimal intrinsic turbulence (the lowest value reported for an AGN emission line to date) and the absence of narrow or intermediate-width components in Ca II λ 8662 make it a powerful diagnostic tool of BLR structure and kinematics. Further investigations of the profiles of Ca II and O I in other AGN are recommended to better constrain BLR properties and the nature of the underlying accretion flow
Microtubule dynamics are defined by conformations and stability of clustered protofilaments
No full textMicrotubules are dynamic cytoskeletal polymers that add and lose tubulin dimers at their ends. Microtubule growth, shortening, and transitions between them are linked to GTP hydrolysis. Recent evidence suggests that flexible tubulin protofilaments at microtubule ends adopt a variety of shapes, complicating structural analysis using conventional techniques. Therefore, the link between GTP hydrolysis, protofilament structure and microtubule polymerization state is poorly understood. Here, we investigate the conformational dynamics of microtubule ends using coarse-grained modeling supported by atomistic simulations and cryoelectron tomography. We show that individual bent protofilaments organize in clusters, transient precursors to the straight microtubule lattice, with GTP-bound ends showing elevated and more persistent cluster formation. Differences in the mechanical properties of GTP- and GDP-protofilaments result in differences in intracluster tension, determining both clustering propensity and protofilament length. We propose that conformational selection at microtubule ends favors long-lived clusters of short GTP-protofilaments that are more prone to forming a straight microtubule lattice and accommodating new tubulin dimers. Conversely, microtubule ends trapped in states with unevenly long and stiff GDP-protofilaments are more prone to shortening. We conclude that protofilament clustering is the key phenomenon that links the hydrolysis state of single tubulins to the polymerization state of the entire microtubule.Max-Planck-Gesellschaft 501100004189Deutsche Forschungsgemeinschaft 501100001659Queen Mary University of London 100009148University of Dundee 10000889
Multiple-Attribute Lorenz Functions and Gini Indices: A Measure Transportation Approach
No full textCzech Science Foundation http://dx.doi.org/10.13039/501100001824Deustche Forschungsgemeinschaf
Evaluation of amino acid digestibility of black soldier fly larvae reared on different substrates in caecectomised laying hens
No full textOffice of the Ministry of Higher Education, Science, Research and Innovation 10.13039/501100002385Thailand Science Research and Innovation through the Kasetsart University Reinventing University Program 2021the National Research Council of Thailand 50110000470
Urinary peptide signature distinguishes autosomal recessive polycystic kidney disease from other causes of chronic kidney disease
No full textABSTRACT Background The diagnosis of autosomal recessive polycystic kidney disease (ARPKD) can be hampered by its pronounced phenotypic variability and ARPKD-mimicking phenocopies. Here, for the first time we specifically studied the urinary peptidome of patients with ARPKD with the aim of distinguishing ARPKD from other causes of chronic kidney disease (CKD). Methods Fifty-eight urine samples from patients with ARPKD, 662 urine samples from paediatric patients with CKD with various other CKD aetiologies and 45 samples from healthy children were included. The urinary peptidome was analysed by capillary electrophoresis/mass spectrometry. Results A 77-peptide signature specific for ARPKD was identified. Application of this signature in a matched random validation set of 19 samples of patients with ARPKD, 23 samples from patients with other CKD and 21 samples from healthy individuals led to a sensitivity of 84.2% [95% confidence interval (CI) 60.4–96.6], a specificity of 100% (95% CI 92.0–100%) and an area under the receiver operating characteristics curve (AUC) of 0.994 (95% CI 0.93–1.00). The 77-peptide signature displayed a specificity of 76.1% (95% CI 72.4–79.5) and an AUC of 0.88 (95% CI 0.85–0.90) in 591 samples from non-matched children with various CKD aetiologies. The signature was primarily (83%) composed of collagen fragments indicating structural damage. Of the remaining peptides, five originated from proteins known to bind to calcium potentially linking the current work to defaults in calcium signalling in polycystic disease. Conclusions We determined a urinary peptide signature that identifies paediatric patients with ARPKD with high precision among a population of children with CKD. Knowledge of the identity of the underlying peptides offers a novel starting point for discussion of possible pathophysiological processes involved in ARPKD
Combination of searches for singly produced vectorlike top quarks in p p collisions at s = 13 TeV with the ATLAS detector
No full textA combination of searches for the single production of vectorlike top quarks ( T ) is presented. These analyses are based on proton-proton collisions at s = 13 TeV recorded in 2015–2018 with the ATLAS detector at the Large Hadron Collider, corresponding to an integrated luminosity of 139 fb − 1 . The T decay modes considered in this combination are into a top quark and either a Standard Model Higgs boson or a Z boson ( T → H t and T → Z t ). The individual searches used in the combination are differentiated by the number of leptons ( e , μ ) in the final state. The observed data are found to be in good agreement with the Standard Model background prediction. Interpretations are provided for a range of masses and couplings of the vectorlike top quark for benchmark models and generalized representations in terms of 95% confidence level limits. For a benchmark signal prediction of a vectorlike top quark SU(2) singlet with electroweak coupling, κ , of 0.5, masses below 2.1 TeV are excluded, resulting in the most restrictive limits to date. © 2025 CERN, for the ATLAS Collaboration 2025 CER
Australian continental‐scale heavy mineral patterns track climate, weathering and erosion
No full textABSTRACT The heavy mineral cargo of sediment is influenced by a complex and variable overprint of climate and environmental processes acting on weathered crystalline basement sources. Here, a continental‐scale heavy mineral compositional dataset of 1315 surface sediment samples, analysed via automated mineralogy, from across Australia is utilized to differentiate between primary source signals and secondary, weathering‐related and transport‐related sediment modification. The results show that heavy mineral patterns not only retain provenance information from the underlying crystalline basement but also have imprinted apparent relationships to measurements of climatic and environmental conditions. Sediment properties change from source to sink and indicate that sediment storage duration considerably influences mineral composition, progressively overprinting primary source information over hundreds of thousands of years. These findings offer a quantitative means to assess and spatially evaluate palaeo‐weathering intensity within ancient landscapes, estimate information loss, and enhance interpretation of the detrital record.Minerals Research Institute of Western Australia https://doi.org/10.13039/50110000910