Portail "HAL-Francophonie Afrique et Océan Indien"
Not a member yet
43052 research outputs found
Sort by
Toward Real‐Time Assessment of Infrasound Event Detection Capability Using Deep Learning‐Based Transmission Loss Estimation
No full textInternational audienceAccurate modeling of infrasound transmission loss is crucial for assessing the performance of the International Monitoring System, which monitors compliance with the Comprehensive Nuclear-Test-Ban Treaty by detecting atmospheric explosions. This modeling supports the design and maintenance of the operating monitoring network. State-of-the-art propagation modeling tools enable transmission loss to be finely simulated using atmospheric models. However, the computational cost prohibits the exploration of a large parameter space in operational monitoring applications. To address this, recent studies made use of a deep learning algorithm capable of making transmission loss predictions almost instantaneously. However, the use of nudged atmospheric models leads to an incomplete representation of the medium, and the absence of temperature as an input makes the algorithm incompatible with long-range propagation. In this study, we address these limitations by using both wind and temperature fields as inputs to a neural network, simulated up to 130 km altitude and 4,000 km distance. We exploit convolutional and recurrent layers to capture spatially and range-dependent features embedded in realistic atmospheric models, improving the overall performance. The neural network reaches an average error of 4 dB compared to full parabolic equation simulations and provides epistemic and data-related uncertainty estimates. Its evaluation on the 2022 Hunga Tonga-Hunga Ha'apai volcanic eruption demonstrates its prediction capability using atmospheric conditions and frequencies not included in the training. This represents a significant step toward near real-time assessment of International Monitoring System detection thresholds of explosive sources
Heavy potassium isotopes in carbonatites reveal oceanic crust subduction as the driver of deep carbon cycling
No full textInternational audienceCarbon cycling between surface and mantle reservoirs is pivotal in fostering habitability of Earth. A critical yet poorly constrained parameter is whether crustal carbon can “survive” devolatilization processes that accompany slab subduction and therefore influence deep carbon budgets. Carbonatites provide a key record to address this important topic. Here, we present high-precision potassium isotope data for a large set of carbonatite samples from both continental and oceanic settings, spanning from 2 billion years ago to the present. Modeling suggests that the heavy potassium isotopic compositions of carbonatites are inherited from their mantle sources, rather than resulting from magmatic and postmagmatic processes. Our results demonstrate a strong link between the subduction of oceanic crust and the recycling of carbonates into the mantle sources of carbonatites. These findings support the hypothesis that subduction of carbonate-bearing altered oceanic crust has been a critical mechanism for transferring carbon into the deep Earth through time
Human-Wildlife interactions in a tropical forest context: modeling, analysis and simulations
No full textAnthropisation and excessive hunting in tropical forests threaten biodiversity, ecosystem maintenance and human food security. In this article, we focus on the issue of coexistence between humans and wildlife in an anthropised environment. Assuming that the human population moves between its residential area and the surrounding forest to hunt, we study a resource-consumer model with consumer migration. A comprehensive analysis of the system is carried out using classical theory and monotone systems theory. We show that the possibilities for long-term coexistence between human populations and wildlife populations are determined by hunting rate thresholds. Depending on the level of anthropisation and the hunting rate, the system may converge towards a limit cycle or a co-existence equilibrium. However, the conditions for coexistence become more difficult as anthropisation increases. Numerical simulations are provided to illustrate the theoretical results.</div
First report of Tomato Mosaic Virus on tomato in Reunion Island
No full textSource Agritrop Cirad (https://agritrop.cirad.fr/615293/)International audienceIn May and October 2023, unusual virus-like disease symptoms of mosaic on leaves and brown spots on fruits were observed on cherry tomato (Solanum esculentum, cv. Tutti Frutti) and ancient tomato (cv. Coeur de boeuf) varieties in greenhouses located in the localities of St. Pierre and St. Louis in the South of Reunion Island. The recent description of emerging tobamoviruses on solanaceous plants in several parts of the world (Salem et al. 2023), with the first descriptions of tomato mottle mosaic virus on tomato in the sister island Mauritius (Maudarbaccus et al. 2021) and tobacco mild green mosaic virus on eggplant in Reunion (Lett et al. 2024), suggested the possible involvement of a tobamovirus. To identify the potential agent(s) causing these symptoms, leaf samples from 12 symptomatic plants were collected, dehydrated, and subjected to total RNA extraction using the RNeasy Plant Extraction Kit (Qiagen, France). The presence of tobamoviruses was tested by reverse transcription (RT)-PCR using generic primers designed to amplify a fragment of the RNA-dependent RNA polymerase (RdRp) of tobamoviruses (Li et al. 2018). RT-PCR amplicons of the expected size (880 bp) were obtained from the 12 samples and subjected to direct sequencing in both directions (Macrogen Europe, Amsterdam, The Netherlands). No PCR amplicons were obtained from the negative controls. BLASTn analysis showed that all 12 assembled nucleotide sequences (GenBank accession nos. PQ505139 to PQ505150) were very similar to each other (>99.2%) and shared 98.2 to 99.3% identity to tomato mosaic virus (ToMV) genome sequences (GenBank accession nos. OQ722320, ON987477, DQ873692, PP856214, and AB355139). To confirm this diagnostic, the specific primers ToMV-F5130 and ToMV-R5928 were designed on the reference sequences ON987477 and AB355139 for the amplification of a 743-bp product covering a region overlapping the movement protein and capsid protein regions. Amplicons of the expected size were obtained from the 12 samples and subjected to direct sequencing in both directions. BLASTn analysis confirmed that all 12 assembled nucleotide sequences (PQ505151 to PQ505162) were highly similar to each other (>99.7%) and shared 99.7 to 99.9% identity to ToMV genome sequences (OQ722320, ON987476, ON987477, and PP856214). To obtain a full genome sequence of a Reunion isolate of ToMV, ribodepletion and Illumina RNA sequencing (2 × 150 bp) were performed by Azenta (Leipzig, Germany). A total of ∼11.4 million paired reads was obtained. After de novo assembly using MEGAHIT version 1.2.9 (Li et al. 2015), a quasi-complete genome sequence of 6,361 nucleotides of ToMV (isolate Reunion-C23-576-St-Pierre-2023; PQ619469) with a mean coverage of 127,192× was obtained. The sequence shares the highest nucleotide identity (99.2%) with the isolate DSMZ PV-0143 of ToMV (PP856214) collected in the United States. To confirm that these diseased tomato plants were infected by ToMV, the 12 positive samples were inoculated into 2-week-old (2- to 3-leaf stage) healthy tomato plants (cv. Roma) using mechanical inoculation, according to Ogwok et al. (2010). Two weeks after inoculation, mild mosaic symptoms were observed on the leaves of the inoculated plants (44/44). The presence of ToMV genomic RNA was confirmed by RT-PCR and direct sequencing of the amplicons. To our knowledge, this is the first report of ToMV infecting tomato plants in Reunion. This new description represents a new warning signal about the risk of introducing tomato brown rugose fruit virus, which represents a major threat for many tomato growing regions of the world (Salem et al. 2023)
Effects of geodynamo priors and geomagnetic data on inverted core surface flows
No full textInternational audienceHigh quality continuous satellite magnetic data gives unprecedented insights into core surface motions but the inversion process remains under-determined, thus requiring additional information. We make use of the pygeodyn inversion tool, in which the core surface dynamics is forecasted with stochastic equations anchored to prior information from a geodynamo free run. There is no clear overview of the relative importance of the different 'ingredients' entering in the core flow inversion process, or of the robustness of the inferred flow features. We document the impact of three factors (or 'ingredients'): (1) choice of geodynamo prior, (2) choice of geomagnetic field model, and (3) choice of maximum spherical harmonic degree for secular variation. We compare three field models (CHAOS-7, COV-OBS.x2 and Kalmag) and three truncation degree for secular variation. The six dynamo priors allow us to analyse the influence of changing the input parameters towards Earthlike values, changing the dynamo set-up (isotropic boundary conditions or not, stratified or not), and the torques at play.We quantitatively show that changing the prior has a greater impact on the core surface flow than the choice of magnetic field model used as observations. This difference becomes less pronounced in the past due to the data quality decreasing and the field models showing more divergence. Furthermore, we find that the slowly evolving 'background' component of flow is more variable among models than the transient component. Transient flow motions appear very coherent for all considered dynamo priors and field models. The recovery from satellite measurements of subdecadal wave dynamics and of a growing eastward motion under the Pacific are robust and appear data driven. Finally, we quantify the efficiency of core flows for generating magnetic variations as a function of frequency. Normalized to their magnitude, rapid changes generate relatively more magnetic signal than slow variations. This is interpreted as the signature of a slow core dynamics approaching a state where induction is mitigated because of dynamical adjustment. As a consequence, a larger part of core surface motions enters the null-space of the kinematic core flow inverse problem towards long time-scales
Évaluation fonctionnelle des reconstructions mandibulaires par lambeaux libres osseux dans l’Océan Indien
No full textIntroduction: This study aimed to evaluate the functional outcomes of mandibular reconstruction with free bone flaps, focusing on swallowing, voice, and speech, in patients from Réunion Island.Materials and Methods: We conducted a retrospective study of 62 patients who underwent mandibular reconstruction with free bone flaps between 2009 and 2024 in Réunion. Functional outcomes were assessed at least six months postoperatively using a standardized questionnaire that collected patient characteristics, three self-assessment tools, and an evaluation by a speech therapist. The main outcomes included resumption of oral feeding, impact on swallowing and speech intelligibility (assessed via DHI, VHI, and SHI scores), and dependence on gastrostomy. Statistical analysis included univariate and bivariate analyses.Results: All 62 patients underwent mandibular reconstruction using fibula free flaps, with a primary success rate of 90%. Early complications were common. Fifteen patients received prosthetic or implant rehabilitation. Complete functional evaluation was available for 21 patients. Of these, 95% had resumed oral feeding, and speech intelligibility was rated as average or good in 90% of cases. No significant associations were found between radiotherapy, tongue resection, and swallowing function, likely due to limited statistical power.Conclusion: Mandibular reconstruction with free bone flaps enables most patients to resume oral feeding, with minimal impact on swallowing and speech. Optimizing surgical and rehabilitative practices may reduce complications and facilitate more systematic dental rehabilitation.Introduction : L'objectif de cette étude est d'évaluer les résultats fonctionnels de la reconstruction mandibulaire par lambeaux libres osseux en termes de déglutition, de voix et de parole à La Réunion.Matériel et Méthodes : Il s’agit d’une étude rétrospective incluant 62 patients opérés d’une reconstruction mandibulaire par lambeaux libres osseux entre 2009 et 2024 à La Réunion. L’évaluation fonctionnelle était réalisée dans un délai de 6 mois minimum post-opératoire par un questionnaire standardisé recueillant les caractéristiques des patients ainsi que 3 questionnaires d'auto-évaluation et une hétéro-évaluation par une orthophoniste. Les critères d'évaluation étaient la reprise de l'alimentation orale, l'impact sur la déglutition et sur l'intelligibilité de la parole avec les scores des DHI, VHI et SHI et la dépendance à une gastrostomie. L’analyse statistique consistait en une analyse univariée et bivariée. Résultats : 62 patients ont bénéficié d’une reconstruction mandibulaire par lambeau libre de fibula avec un taux de succès primaire de 90%. De nombreuses complications précoces ont été notées. 15 patients ont bénéficié d’une réhabilitation prothétique ou implantaire. 21 patients ont bénéficié d’une évaluation fonctionnelle complète. 95% d’entre eux ont repris une alimentation orale. L’intelligibilité de la parole était moyenne ou bonne chez 90% des patients. Aucune association significative entre la radiothérapie, l’exérèse linguale et la fonction de déglutition n’a pu être mise en évidence par manque de puissance. Conclusion : La reconstruction mandibulaire par lambeaux libres osseux permet une reprise alimentaire dans la majorité des cas avec des séquelles limitées sur la déglutition et l’élocution. L’amélioration des pratiques permettra une diminution des complications et une réhabilitation dentaire plus systématique
MODÈLE MARKOVIEN D’OCTROI DE CRÉDITBASÉ SUR LE PRINCIPE WIN-WIN ENMICROFINANCE
No full textThis dissertation develops a rigorous mathematical framework to model borrower trajectories in microĄnance using Markov chains with rewards. Thecentral objective is to design credit contracts that are simultaneously viable forlenders and borrowers Ů according to a built-in Şwin-winŤ principle encodedwithin the probabilistic structure.Three types of contracts are analyzed : an individual loyalty-based loan, a grouplending scheme with reversible exclusion, and a dual social sanction mechanism.For each, we provide a stationary behavioral analysis, an explicit computation ofthe expected discounted revenue (FRETA), and an economic interpretation. Theresult is an original decision-making framework, merging dynamic incentives withbehavioral constraints. This thesis stands out for unifying stochastic, economic, and ethical dimensionsinto a complete analytical model. As a forward-looking extension, we suggest incorporating Partially Observable Markov Decision Processes (POMDPs) and theChoquet integral to better capture ambiguity and subjective beliefs in risk evaluation.Cette thèse propose une approche mathématique rigoureuse pour modéliser les trajectoires des emprunteurs en microĄnance à travers des chaînes de Markov avec revenu. LŠobjectif central est de concevoir des contrats de crédit viablesà la fois pour lŠinstitution prêteuse et pour lŠemprunteur Ů selon un principewin-win intégré au cœur même de la structure probabiliste.Trois types de contrats sont développés : un contrat individuel à Ądélisation, unprêt solidaire à exclusion réversible, et un mécanisme à double sanction sociale.Pour chaque modèle, une analyse stationnaire des comportements, un calcul explicite du Ćux de revenu actualisé (FRETA), et une interprétation économiqueapprofondie sont proposés. LŠensemble forme un cadre décisionnel original, combinant incitations dynamiques et contraintes comportementales.Ce travail se distingue par sa capacité à uniĄer trois dimensions Ů stochastique,économique et éthique Ů dans une modélisation analytique complète. En guisedŠouverture, la thèse suggère une extension vers les POMDPs (Processus Décisionnels Markoviens Partiellement Observables) et lŠintégrale de Choquet pourcapter lŠambiguïté et la croyance subjective dans lŠévaluation du risque
Problématiques juridiques liées aux systèmes d’Intelligence Artificielle :Focus sur la question de responsabilité civile en RDC
No full textInternational audienceArtificial Intelligence (AI) systems raise unprecedented legal challenges, and the question of civil liability in the event of harm is a central pillar of these. In the Democratic Republic of Congo (DRC), as in many legal systems, civil liability is traditionally based on concepts such as fault, causality and injury. But the advent of AI is making these concepts considerably more complex. In the absence of specific legislation on AI, Congolese law will still have to rely on general principles of civil liability to deal with harm caused by AI. However, the application of these principles raises considerable challenges that argue in favour of adopting rules adapted to this new technological reality. In concrete terms, in addition to the general need to establish an effective legal and institutional framework commensurate with the problems associated with AI, it is at the same time particularly necessary for the Congolese legislator to establish a clear, fair and effective civil liability regime for harm caused by this technology. A strict liability regime for AI systems would therefore be appropriate and would guarantee the protection of victims while encouraging innovation and the responsible development of AI in the DRC.Les systèmes d'Intelligence Artificielle (IA) soulèvent des défis juridiques inédits, et la question de la responsabilité civile en cas de préjudice en est un pilier central. En République Démocratique du Congo (RDC), comme dans de nombreux systèmes juridiques, le régime de responsabilité civile repose traditionnellement sur des notions telles que la faute, le lien de causalité et le préjudice. Mais, l'arrivée de l'IA complexifie considérablement ces notions. En l'absence d'une législation spécifique sur l’IA, le Droit congolais devra encore s'appuyer sur des principes généraux de responsabilité civile pour traiter les préjudices causés par l'IA. Cependant, l'application de ces principes soulève des défis considérables qui plaident en faveur de l'adoption des règles adaptées à cette nouvelle réalité technologique. Concrètement, outre la nécessité d’instituer de manière générale un cadre juridique et institutionnel efficace à la hauteur des problématiques liées à l’IA, il est en même temps nécessaire et de manière particulière que le législateur congolais instaure un régime de responsabilité civile clair, équitable et efficace pour les préjudices causés par cette technologie. Et donc un régime de responsabilité objective du fait du système d’IA serait adéquat et permettrait de garantir la protection des victimes tout en encourageant l’innovation et le développement responsable de l'IA en RDC
A remote input of African dust to Last Glacial Europe
No full textInternational audienceDuring the Last Glacial Maximum global surface air temperatures were up to 6 °C lower than preindustrial levels and the mineral dust cycle was highly active, with global dust loading two to four times higher than during the Holocene. Loess deposits and Greenland ice cores show peak dust concentrations during this time. While Asian sources were traditionally seen as the main contributors to dust transported to Greenland, recent studies using geochemical methods suggest a mix of Asian, North African, and European sources. Europe experienced intense dust activity, with mineral particles largely emitted from regional sources. Here we present the trace elements, and strontium and lead isotopes from Last-Glacial Maximum samples collected at 15 sites across Europe. The results reveal that fine dust originated from remote sources, potentially northern Africa. Earth System model simulations support this finding, highlighting Northern Africa's substantial role in dust deposition during glacial periods across the Northern Hemisphere
