Search for: All records

Abstract The assimilation of radar reflectivity requires an accurate and efficient forward operator that links the model state variables to radar observations. In this study, newly developed parameterized forward operators (PFO) for radar reflectivity with a new continuous melting model are implemented to assimilate observed radar data. To assess the impact of the novel parameterized reflectivity forward operators on convective storm analysis and forecasting, two distinct sets of cycled assimilation and forecast experiments are conducted. One set of experiments (ExpRFO) uses a conventional Rayleigh‐scattering‐approximation‐based forward operator (RFO) with hydrometeor classification, while the other uses the PFO (ExpPFO_New) for radar reflectivity with a new continuous melting model. Eight high‐impact severe convective weather events from the Hazardous Weather Testbed (HWT) 2019 Spring Experiments are selected for this study. The analysis and forecast results are first examined in detail for a classic tornadic supercell case on 24 May 2019, with the potential benefits provided by the PFO then evaluated for all eight cases. It is demonstrated that ExpPFO_New provides more robust results in terms of improving the short‐term severe weather forecasts. Compared to ExpRFO, ExpPFO_New better reproduces all observed supercells in the analysis field, yields a more continuous and reasonable reflectivity distribution near the melting layer, and improves the strength of the cold pool compared to observations. Overall, ExpPFO_New, initialized from the more accurate analysis fields, produces better forecasts of reflectivity and hourly precipitation with smaller biases, especially at heavy precipitation thresholds. 

The burgeoning field of brain health research increasingly leverages artificial intelligence (AI) to analyze and interpret neuroimaging data. Medical foundation models have shown promise of superior performance with better sample efficiency. This work introduces a novel approach towards creating 3-dimensional (3D) medical foundation models for multimodal neuroimage segmentation through self-supervised training. Our approach involves a novel two-stage pretraining approach using vision transformers. The first stage encodes anatomical structures in generally healthy brains from the large-scale unlabeled neuroimage dataset of multimodal brain magnetic resonance imaging (MRI) images from 41,400 participants. This stage of pertaining focuses on identifying key features such as shapes and sizes of different brain structures. The second pretraining stage identifies disease-specific attributes, such as geometric shapes of tumors and lesions and spatial placements within the brain. This dual-phase methodology significantly reduces the extensive data requirements usually necessary for AI model training in neuroimage segmentation with the flexibility to adapt to various imaging modalities. We rigorously evaluate our model, BrainSegFounder, using the Brain Tumor Segmentation (BraTS) challenge and Anatomical Tracings of Lesions After Stroke v2.0 (ATLAS v2.0) datasets. BrainSegFounder demonstrates a significant performance gain, surpassing the achievements of the previous winning solutions using fully supervised learning. Our findings underscore the impact of scaling up both the model complexity and the volume of unlabeled training data derived from generally healthy brains. Both of these factors enhance the accuracy and predictive capabilities of the model in neuroimage segmentation tasks. Our pretrained models and code are at https://github.com/lab-smile/BrainSegFounder. 

While network attacks play a critical role in many advanced persistent threat (APT) campaigns, an arms race exists between the network defenders and the adversary: to make APT campaigns stealthy, the adversary is strongly motivated to evade the detection system. However, new studies have shown that neural network is likely a game-changer in the arms race: neural network could be applied to achieve accurate, signature-free, and low-false-alarm-rate detection. In this work, we investigate whether the adversary could fight back during the next phase of the arms race. In particular, noticing that none of the existing adversarial example generation methods could generate malicious packets (and sessions) that can simultaneously compromise the target machine and evade the neural network detection model, we propose a novel attack method to achieve this goal. We have designed and implemented the new attack. We have also used Address Resolution Protocol (ARP) Poisoning and Domain Name System (DNS) Cache Poisoning as the case study to demonstrate the effectiveness of the proposed attack. 

Due to the scarcity of C–F bond-forming enzymatic reactions in nature and the contrasting prevalence of organofluorine moieties in bioactive compounds, developing biocatalytic fluorination reactions represents a pre-eminent challenge in enzymology, biocatalysis and synthetic biology. Additionally, catalytic enantioselective C(sp3)–H fluorination remains a challenging problem facing synthetic chemists. Although many non-haem iron halogenases have been discovered to promote C(sp3)–H halogenation reactions, efforts to convert these iron halogenases to fluorinases have remained unsuccessful. Here we report the development of an enantioselective C(sp3)–H fluorination reaction, catalysed by a plant-derived non-haem enzyme 1-aminocyclopropane-1-carboxylic acid oxidase (ACCO), which is repurposed for radical rebound fluorination. Directed evolution afforded a C(sp3)–H fluorinating enzyme ACCOCHF displaying 200-fold higher activity, substantially improved chemoselectivity and excellent enantioselectivity, converting a range of substrates into enantioenriched organofluorine products. Notably, almost all the beneficial mutations were found to be distal to the iron centre, underscoring the importance of substrate tunnel engineering in non-haem iron biocatalysis. Computational studies reveal that the radical rebound step with the Fe(III)–F intermediate has a low activation barrier of 3.4 kcal mol−1 and is kinetically facile. 

Abstract To improve short‐term severe weather forecasts through assimilation of polarimetric radar data (PRD), the use of accurate and efficient forward operators for polarimetric radar variables is required. In this study, a new melting model is proposed to estimate the mixing ratio and number concentration of melting hydrometeor species and incorporated in a set of parameterized polarimetric radar forward operators. The new melting model depends only on the mixing ratio and number concentration of rain and ice species and is characterized by its independence from ambient temperature and its simplicity and ease of linearization. To assess the impact of this newly proposed melting model on the simulated polarimetric radar variables, a real mesoscale convective system is simulated using three double‐moment microphysics schemes. Compared with the output of the original implementation of the parameterized forward operators (PFO_Old) that rely on an “old” melting model which only estimates the mixing ratio of the melting species, the updated implementation with the new melting model (PFO_New) that estimates both the mixing ratio and number concentration of melting species eliminates the very large mass/volume‐weighted mean diameter (D_m) at the bottom of the melting layer and produces more reasonable melting layer signatures for all three double‐moment microphysics schemes that more closely match the corresponding radar observations. This suggests that the new melting model has more reasonable implicit estimates of mixing ratios and number concentrations of melting hydrometeor species than the “old” melting model.