Global Model Research Articles

Sensitivity of global hydrological models to potential evapotranspiration estimation methods in the Senegal River Basin (West Africa)

Study regionSenegal River Basin in West Africa Study focusThis paper aims to evaluate the sensitivity of global hydrological models to potential evapotranspiration (PET) methods in the Senegal River Basin. Potential evapotranspiration is estimated using 21 methods and its influence on the performance of three GR models (GR4J, GR5J and GR6J) is investigated in five catchments. The data used are mean rainfall, discharge and observed and calculated PET over the period 1984-1995. PET is calculated based on observed climate data and those from NASA POWER reanalysis data. The methodology consists in: (i) comparing the consistency of reanalysis data with respect to the observed PET, (ii) assessing the robustness of GR models and their sensitivity to different PET estimation methods. The evaluation criteria used to assess the performance of the hydrological model are KGE and PBIAS. New hydrological insights for the regionGood consistency is obtained between PET calculated with observed and reanalysis data. The GR4J and GR5J are more efficient to simulate flows in the Senegal River sub-basins. The aerodynamic PET methods perform well with the three hydrological models. However, in this context where data are scarce, temperature methods such as Droogers and Allen are a good choice for hydrological modeling. The results also show that the GR models have the ability to adapt to poorly estimated PET.

Dengue hemorrhagic fever prediction in coastal area using geographically weighted regression

Dengue hemorrhagic fever (DHF) is still a health problem globally, including in Indonesia. Geographical and climatic conditions in coastal areas are different from other areas, which may impact differences in environmental risk factors for dengue. This study aims to create a prediction model for the incidence of DHF in coastal areas. The research was conducted in Bantul Regency, Indonesia, involving data from 2015-2019. Dengue incidence data were collected from the health office. Climatic data were from climatology station. Data on altitude and shoreline distance were obtained by geographic information system (GIS) processing. Population density and wide settlement area are obtained from the Bureau of Statistics. The geographically weighted regression (GWR) analysis was carried out using GWR4. The results showed that GWR with a weighting of Fixed Bi-Square Kernel obtained an R2 value of 0.7768, better than the global model (R2 0.5254). It indicates that DHF (Y) in Bantul Regency is 77.68% determined by population density (X1), altitude (X2), settlement area (X3), shoreline distance (X4) and rainfall (X5) and the remaining 22.32% are influenced by other variables which are not investigated. Geographically, the predictor variables explain the DHF incidence with a strong category in the central region, and weak in coastal area.

Federated Gaussian Process: Convergence, Automatic Personalization and Multi-fidelity Modeling.

In this paper, we propose FGPR: a Federated Gaussian process ( GP) regression framework that uses an averaging strategy for model aggregation and stochastic gradient descent for local computations. Notably, the resulting global model excels in personalization as FGPR jointly learns a shared prior across all devices. The predictive posterior is then obtained by exploiting this shared prior and conditioning on local data, which encodes personalized features from a specific dataset. Theoretically, we show that FGPR converges to a critical point of the full log-marginal likelihood function, subject to statistical errors. This result offers standalone value as it brings federated learning theoretical results to correlated paradigms. Through extensive case studies on several regression tasks, we show that FGPR excels in a wide range of applications and is a promising approach for privacy-preserving multi-fidelity data modeling.

Assessment of global geomagnetic models from aeromagnetic data in Cameroon, Central Africa

ABSTRACT Global geomagnetic models are a compilation of existing magnetic data which were reprocessed and freely made available as 2 or 3 arc minute resolution grids at 0-km or 4–5-km altitude relative to the geoid. Thus, many researchers rely on those data due to their open access. This study aims at evaluating freely released global geomagnetic models with respect to aeromagnetic data which are the original data, usually subjected to third party or commercial restrictions in Cameroon. Hence, both global geomagnetic models and aeromagnetic data were sampled along 4 profiles for assessment. The extracted data were analysed using Pearson’s correlation coefficient depicted in tables. The Power Spectrum Density curves of both aeromagnetic and their respective global geomagnetic grids were computed and displayed as graphs. The results show that the so-claimed 4–5-km altitude relative to the geoid of global geomagnetic models is heterogeneous in Cameroon since the spectral content of those models vary from 5–20 km. Besides, considerable discrepancies were observed with Earth Magnetic Anomaly Grid version 3 (EMAG2v3) data, indicating that they are not reliable for scientific studies based in Cameroon. This study provides insights on global geomagnetic models and could aid in optimizing the use of those models in Cameroon.

FedUB: Federated Learning Algorithm Based on Update Bias

Federated learning, as a distributed machine learning framework, aims to protect data privacy while addressing the issue of data silos by collaboratively training models across multiple clients. However, a significant challenge to federated learning arises from the non-independent and identically distributed (non-iid) nature of data across different clients. non-iid data can lead to inconsistencies between the minimal loss experienced by individual clients and the global loss observed after the central server aggregates the local models, affecting the model’s convergence speed and generalization capability. To address this challenge, we propose a novel federated learning algorithm based on update bias (FedUB). Unlike traditional federated learning approaches such as FedAvg and FedProx, which independently update model parameters on each client before direct aggregation to form a global model, the FedUB algorithm incorporates an update bias in the loss function of local models—specifically, the difference between each round’s local model updates and the global model updates. This design aims to reduce discrepancies between local and global updates, thus aligning the parameters of locally updated models more closely with those of the globally aggregated model, thereby mitigating the fundamental conflict between local and global optima. Additionally, during the aggregation phase at the server side, we introduce a metric called the bias metric, which assesses the similarity between each client’s local model and the global model. This metric adaptively sets the weight of each client during aggregation after each training round to achieve a better global model. Extensive experiments conducted on multiple datasets have confirmed the effectiveness of the FedUB algorithm. The results indicate that FedUB generally outperforms methods such as FedDC, FedDyn, and Scaffold, especially in scenarios involving partial client participation and non-iid data distributions. It demonstrates superior performance and faster convergence in tasks such as image classification.

Using HawkEye Level-2 Satellite Data for Remote Sensing Tasks in the Presence of Dust Aerosol

This paper is the first to examine the operation of the HawkEye satellite in the presence of dust aerosol. The study region is the Black Sea. Dust transport dates were identified using visual inspection of satellite imagery, back-kinematic HYSPLIT trajectory analysis, CALIPSO aerosol stratification and typing maps, and the global forecasting model SILAM. In a comparative analysis of in-situ and satellite measurements of the remote sensing reflectance, an error in the atmospheric correction of HawkEye measurements was found both for a clean atmosphere and in the presence of an absorbing aerosol. It is shown that, on average, the dependence of the atmospheric correction error on wavelength has the form of a power function of the form from λ−3 to λ−9. The largest errors are in the short-wavelength region of the spectrum (412–443 nm) for the dust and dusty marine aerosol domination dates. A comparative analysis of satellite and in situ measurements of the optical characteristics of the atmosphere, namely the AOD and the Ångström parameter, was carried out. It is shown that the aerosol model used by HawkEye underestimates the Angström parameter and, most likely, large errors and outliers in satellite measurements are associated with this.

CMIP6 Models Rarely Simulate Antarctic Winter Sea‐Ice Anomalies as Large as Observed in 2023

AbstractIn 2023, Antarctic sea‐ice extent (SIE) reached record lows, with winter SIE falling to 2.5Mkm2 below the satellite era average. With this multi‐model study, we investigate the occurrence of anomalies of this magnitude in latest‐generation global climate models. When these anomalies occur, SIE takes decades to recover: this indicates that SIE may transition to a new, lower, state over the next few decades. Under internal variability alone, models are extremely unlikely to simulate these anomalies, with return period >1000 years for most models. The only models with return period <1000 years for these anomalies have likely unrealistically large interannual variability. Based on extreme value theory, the return period is reduced from 2650 years under internal variability to 580 years under a strong climate change forcing scenario.

Dynamical Tests of a Deep-Learning Weather Prediction Model

Abstract Global deep-learning weather prediction models have recently been shown to produce forecasts that rival those from physics-based models run at operational centers. It is unclear whether these models have encoded atmospheric dynamics, or simply pattern matching that produces the smallest forecast error. Answering this question is crucial to establishing the utility of these models as tools for basic science. Here we subject one such model, Pangu-Weather, to a set of four classical dynamical experiments that do not resemble the model training data. Localized perturbations to the model output and the initial conditions are added to steady time-averaged conditions, to assess the propagation speed and structural evolution of signals away from the local source. Perturbing the model physics by adding a steady tropical heat source results in a classical Matsuno–Gill response near the heating, and planetary waves that radiate into the extratropics. A localized disturbance on the winter-averaged North Pacific jet stream produces realistic extratropical cyclones and fronts, including the spontaneous emergence of polar lows. Perturbing the 500hPa height field alone yields adjustment from a state of rest to one of wind–pressure balance over ∼6 hours. Localized subtropical low pressure systems produce Atlantic hurricanes, provided the initial amplitude exceeds about 4 hPa, and setting the initial humidity to zero eliminates hurricane development. We conclude that the model encodes realistic physics in all experiments, and suggest it can be used as a tool for rapidly testing a wide range of hypotheses.

A global neuronal workspace model of functional neurological disorders.

We introduce here a general model of Functional Neurological Disorders based on the following hypothesis: a Functional Neurological Disorder could correspond to a consciously initiated voluntary top-down process causing involuntary lasting consequences that are consciously experienced and subjectively interpreted by the patient as involuntary. We develop this central hypothesis according to Global Neuronal Workspace theory of consciousness, that is particularly suited to describe interactions between conscious and non-conscious cognitive processes. We then present a list of predictions defining a research program aimed at empirically testing their validity. Finally, this general model leads us to reinterpret the long-debated links between hypnotic suggestion and functional neurological disorders. Driven by both scientific and therapeutic goals, this theoretical paper aims at bringing closer the psychiatric and neurological worlds of functional neurological disorders with the latest developments of cognitive neuroscience of consciousness.

A Lightweight Face Detector via Bi-Stream Convolutional Neural Network and Vision Transformer

Lightweight convolutional neural networks are widely used for face detection due to their ability to learn local representations through spatial induction bias and translational invariance. However, convolutional face detectors have limitations in detecting faces under challenging conditions like occlusion, blurring, or changes in facial poses, primarily attributed to fixed-size receptive fields and a lack of global modeling. Transformer-based models have advantages on learning global representations but are insensitive to capture local patterns. To address these limitations, we propose an efficient face detector that combines convolutional neural network and transformer architectures. We introduce a bi-stream structure that integrates convolutional neural network and transformer blocks within the backbone network, enabling the preservation of local pattern features and the extraction of global context. To further preserve the local details captured by convolutional neural networks, we propose a feature enhancement convolution block in a hierarchical backbone structure. Additionally, we devise a multiscale feature aggregation module to enhance obscured and blurred facial features. Experimental results demonstrate that our method has achieved improved lightweight face detection accuracy with an average precision of 95.30%, 94.20%, and 87.56% across the easy, medium, and hard subdatasets of WIDER FACE, respectively. Therefore, we believe our method will be a useful supplement to the collection of current artificial intelligence models and benefit the engineering applications of face detection.

Post‐processing output from ensembles with and without parametrised convection, to create accurate, blended, high‐fidelity rainfall forecasts

AbstractFlash flooding is a significant societal problem, but related precipitation forecasts are often poor. To address, one can try to use output from convection‐parametrising (global) ensembles, post‐processed to forecast at point‐scale, or convection‐resolving limited area ensembles. The new methodology described here combines both. We apply “ecPoint‐rainfall” post‐processing to the ECMWF global ensemble. Alongside we use 2.2 km COSMO LAM ensemble output (centred on Italy), and also post‐process that, using a scale‐selective neighbourhood approach to compensate for insufficient members and to preserve consistently forecast local details. The two resulting scale‐compatible components then undergo lead‐time‐weighted blending, to create the final probabilistic 6 h rainfall forecasts. Product creation for forecasters, in this way, constituted the “Italy Flash Flood use case” within the EU‐funded MISTRAL project; real‐time delivery of open access products is ongoing. One year of verification shows that, of the five components (2 raw, 2 post‐processed and blended), ecPoint is the most skilful. The post‐processed COSMO ensemble adds most value to summer convective events in the evening, when the global model has an underprediction bias. In two typical heavy rainfall case studies we observed underestimation of the largest point totals in the raw ECMWF ensemble, and overestimation in the raw COSMO ensemble. However, ecPoint elevated the ECMWF maxima and highlighted best the most affected areas and merged products seemed to be the most skilful of all. Even though our LAM post‐processing does not include (or arguably need) bias‐correction, this study still provides a unique blueprint for successfully combining ensemble rainfall forecasts from global and LAM systems around the world. It also has important implications for forecast products as global ensembles move ever closer to having convection‐permitting resolution.

TASK-3, two-pore potassium channels, contribute to circadian rhythms in the electrical properties of the suprachiasmatic nucleus and play a role in driving stable behavioural photic entrainment

ABSTRACT Stable and entrainable physiological circadian rhythms are crucial for overall health and well-being. The suprachiasmatic nucleus (SCN), the primary circadian pacemaker in mammals, consists of diverse neuron types that collectively generate a circadian profile of electrical activity. However, the mechanisms underlying the regulation of endogenous neuronal excitability in the SCN remain unclear. Two-pore domain potassium channels (K2P), including TASK-3, are known to play a significant role in maintaining SCN diurnal homeostasis by inhibiting neuronal activity at night. In this study, we investigated the role of TASK-3 in SCN circadian neuronal regulation and behavioural photoentrainment using a TASK-3 global knockout mouse model. Our findings demonstrate the importance of TASK-3 in maintaining SCN hyperpolarization during the night and establishing SCN sensitivity to glutamate. Specifically, we observed that TASK-3 knockout mice lacked diurnal variation in resting membrane potential and exhibited altered glutamate sensitivity both in vivo and in vitro. Interestingly, despite these changes, the mice lacking TASK-3 were still able to maintain relatively normal circadian behaviour.

Assessment of dynamical downscaling performance over cordex east Asia using MPAS-A global variable resolution model: climatology, seasonal cycle, and extreme events

Assessment of dynamical downscaling performance over cordex east Asia using MPAS-A global variable resolution model: climatology, seasonal cycle, and extreme events

Improved Diurnal Cycle of Precipitation on Land in a Global Non-Hydrostatic Model Using a Revised NSAS Deep Convective Scheme

Improved Diurnal Cycle of Precipitation on Land in a Global Non-Hydrostatic Model Using a Revised NSAS Deep Convective Scheme

Ocean Heat Convergence and North Atlantic multidecadal heat content variability

Abstract We construct an upper ocean (0-1000m) North Atlantic heat budget (26°-67°N) for the period 1950-2020 using multiple observational datasets and an eddy-permitting global ocean model. On multidecadal timescales ocean heat transport convergence controls ocean heat content (OHC) tendency in most regions of the North Atlantic with little role for diffusive processes. In the subpolar North Atlantic (45°N-67°N) heat transport convergence is explained by geostrophic currents whereas ageostrophic currents make a significant contribution in the subtropics (26°N-45°N). The geostrophic contribution in all regions is dominated by anomalous advection across the time-mean temperature gradient although other processes make a significant contribution particularly in the subtropics. The timescale and spatial distribution of the anomalous geostrophic currents are consistent with a simple model of basin scale thermal Rossby waves propagating westwards/northwestwards in the subpolar gyre and multidecadal variations in regional OHC are explained by geostrophic currents periodically coming into alignment with the mean temperature gradient as the Rossby wave passes through. The global ocean model simulation shows that multidecadal variations in the Atlantic Meridional Overturning Circulation are synchronized with the ocean heat transport convergence consistent with modulation of the west-east pressure gradient by the propagating Rossby wave.

Automatic breast ultrasound (ABUS) tumor segmentation based on global and local feature fusion.

Accurate segmentation of tumor regions in automated breast ultrasound (ABUS) images is of paramount importance in computer-aided diagnosis (CAD) system. However, the inherent diversity of tumors and the imaging interference pose great challenges to ABUS tumor segmentation. In this paper, we propose a global and local feature interaction model combined with graph fusion (GLGM), for 3D ABUS tumor segmentation. In GLGM, we construct a dual branch encoder-decoder, where both local and global features can be extracted. Besides, a global and local feature fusion (GLFF) module is designed, which employs the deepest semantic interaction to facilitate information exchange between local and global features. Additionally, to improve the segmentation performance for small tumors, a graph convolution-based shallow feature fusion module (SFFGC) is designed. It exploits the shallow feature to enhance the feature expression of small tumors in both local and global domains. The proposed method is evaluated on a private ABUS dataset and a public ABUS dataset. For the private ABUS dataset, the small tumors (volume smaller than 1 cm³) account for over 50% of the entire dataset. Experimental results show that the proposed GLGM model outperforms several state-of-the-art segmentation models in 3D ABUS tumor segmentation, particularly in segmenting small tumors.

Carbon Cycle Models Quantify for a Green and Low-Carbon Economy

CO2 emissions from the combustion of fossil fuels represent the largest anthropogenic disruption of the natural global carbon cycle and lead to an undesirable increase in atmospheric CO2 levels. The so-called "carbon-neutral" renewable energy source biomass appears to be a promising replacement for fossil fuels. The "Combined Energy and Biosphere Model" (CEBM), was developed as a method for calculating the atmospheric CO2 content after intensive use of biomass. It includes (i) a biosphere part computing the annual cycle of carbon in the biosphere on 2,433 grid elements covering the continents’ plant mass and soil, as well as (ii) an energy economic part for calculating the fossil CO2 emissions of 119 countries. As the history of understanding the carbon cycle (including quantification of plant growth and decay by formulae, inclusion of deforestation, fossil fuel use and the fertilizer effect) shows, such a global carbon cycle model (as the CEBM) provides the needed methodology to assess the net effects of large-scale biomass energy use on the plant’s atmosphere. Results show that biomass (in a systemic view) is “only half as carbon neutral” as previously thought, even on a principal level. The conclusions indicate that (even if biomass is a valuable mitigation strategy) its global potential is limited. When interpreting CEBM scenarios, as a very first priority, reducing the current annual increase in energy demand is most highly needed for preserving climate.

Projection of mean and extreme precipitation and air temperature over India: a CMIP6 analysis

ABSTRACT This study used fifteen Coupled Model Intercomparison Project Phase 6 (CMIP6) models to analyze the mean and extreme projection precipitation during the Indian summer monsoon (ISM) season, i.e., June to September (JJAS). A continuous rise in precipitation and its associated unpredictability were predicted by global climate models using the CMIP6. The extreme values are estimated on the basis of generalized extreme value (GEV) distribution. Four socioeconomic paths are used as SSP126, SSP245, SSP370 and SSP585 to understand the impact of low-emission to high-emission scenarios. In all the scenarios, it is seen that all the models predict a significant rise in JJAS mean and extreme precipitation. Moreover, a strong correlation is obtained between the historical (1995–2014) and near-emission future simulation (2021–2040), mid-emission future simulation (2041–2060), high-emission future simulations (2081–2100) of air temperature and precipitation. In extreme emission scenarios, ensemble mean of CMIP6 models shows an increasing amplitude in surface air temperature and precipitation over India in the near future (2021–2040), mid future (2041–2060) and the end of the century (2081–2100). The CMIP6 simulations mainly support the results of all models, but they also demonstrate improved robustness across models.

Large-Scale Climate Modes Drive Low-Frequency Regional Arctic Sea Ice Variability

Abstract Summer Arctic sea ice is declining rapidly but with superimposed variability on multiple timescales that introduces large uncertainties into projections of future sea ice loss. To better understand what drives at least part of this variability, we show how a simple linear model can link dominant modes of climate variability to low-frequency regional Arctic sea ice concentration (SIC) anomalies. Focusing on September, we find skillful projections from global climate models (GCMs) from the Coupled Model Intercomparison Project Phase 6 (CMIP6) at lead times of 4-20 years, with up to 60% of observed low-frequency variability explained at a 5-year lead time. The dominant driver of low-frequency SIC variability is the Interdecadal Pacific Oscillation (IPO) which is positively correlated with SIC anomalies in all regions up to a lead time of 15 years, but with large uncertainty between GCMs and internal variability realization. The Niño 3.4 Index and Atlantic Multidecadal Oscillation have better agreement between GCMs of being positively and negatively related, respectively, with low-frequency SIC anomalies for at least 10-year lead times. The large variation between GCMs and between members within large ensembles indicate the diverse simulation of teleconnections between the tropics and Arctic sea ice, and the dependence on initial climate state. Further, the influence of the Niño 3.4 Index was found to be sensitive to the background climate. Our results suggest that, based on the 2022 phases of dominant climate variability modes, enhanced loss of sea ice area across the Arctic is likely during the next decade.

FedCMD: A Federated Cross-modal Knowledge Distillation for Drivers’ Emotion Recognition

Emotion recognition has attracted a lot of interest in recent years in various application areas such as healthcare and autonomous driving. Existing approaches to emotion recognition are based on visual, speech, or psychophysiological signals. However, recent studies are looking at multimodal techniques that combine different modalities for emotion recognition. In this work, we address the problem of recognizing the user’s emotion as a driver from unlabeled videos using multimodal techniques. We propose a collaborative training method based on cross-modal distillation, i.e., “FedCMD” (Federated Cross-Modal Distillation). Federated Learning (FL) is an emerging collaborative decentralized learning technique that allows each participant to train their model locally to build a better generalized global model without sharing their data. The main advantage of FL is that only local data is used for training, thus maintaining privacy and providing a secure and efficient emotion recognition system. The local model in FL is trained for each vehicle device with unlabeled video data by using sensor data as a proxy. Specifically, for each local model, we show how driver emotional annotations can be transferred from the sensor domain to the visual domain by using cross-modal distillation. The key idea is based on the observation that a driver’s emotional state indicated by a sensor correlates with facial expressions shown in videos. The proposed “FedCMD” approach is tested on the multimodal dataset “BioVid Emo DB” and achieves state-of-the-art performance. Experimental results show that our approach is robust to non-identically distributed data, achieving 96.67% and 90.83% accuracy in classifying five different emotions with IID (independently and identically distributed) and non-IID data, respectively. Moreover, our model is much more robust to overfitting, resulting in better generalization than the other existing methods.