HAL-Ecole des Ponts ParisTech
Not a member yet
42743 research outputs found
Sort by
Collaboration in Technology and Multinational Production
No full textHow does technology sharing affect global production choices? To answer this question, I incorporate technology choices into a structural multinational production model and allow for collaboration in specialized assembly assets across firms when choosing the optimal production locations for their varieties. I find that both the technology choice itself and the potential sharing of it have important effects on the expected cost and profits of firms. Conditional on technology choices, the median firm's cost of serving a market increases by 24.65% compared to traditional models that do not model input technology. On the other hand, allowing for collaboration reduces their cost by 2.9%, with large firm heterogeneity. Importantly, the model allows for analysis not only of trade policy shocks but also of industrial policies. While restricting technology access has limited global effects, it creates significant production consequences for specific firms and countries, highlighting industrial policy as a more targeted tool than trade policy
On and off-the-record correction practices: A survey-based study of how chemistry researchers react to errors
No full textInternational audienceThis survey-based study (982 participants) explores chemistry researchers' practices and motivations in correcting errors in scientific publications. While respondents believe errors should be corrected in principle, practical challenges arise due to scientific, social, and pragmatic factors. These include the perceived seriousness of the error, its scientific impact, the age of the publication, and the time required. Difficulties also stem from criticizing others, especially senior colleagues. Despite these challenges, researchers are motivated to correct errors to limit their spread, contribute to the common good, and advance their own work. Researchers prefer informal error correction through private correspondence, discussions with colleagues, or teaching situations, over formal corrections to the scholarly record. The peer-review stage is crucial for detecting and correcting errors, but it is criticized for its deficiencies, including lack of professionalism among reviewers and editors. Some authors yield to reviewer pressure knowingly introducing changes that are clearly wrong. While the low participation rate (2%) does not allow generalization, the study shows that science correction is complex involving a continuum of practices. To improve science correction, the study suggests that online platforms and repositories can facilitate the transition from off-the-record discussions to on-the-record initiatives, ultimately feeding into the public record
The pressure-robust virtual element methods with reconstruction for the Stokes equation
No full textBased on the idea of reconstruction, we develop a unified theoretical framework on the pressure-robust virtual element methods for Stokes equation. The reconstruction can be done on any shape-regular simplicial subdivision of polytopal elements and consists of two steps: the interpolation onto H(div)-conforming finite element space and the solution of a local problem element by element. By using this reconstruction we only modify the right-hand side of original virtual element methods in order to obtain the pressure-robust schemes with optimal convergence. This framework is applied to the conforming and nonconforming virtual element methods of arbitrary polynomial degree in two and three dimensions. In particular, we present the pressure-robust scheme for the nonconforming virtual element method and give some explanations on developing the pressure-robust conforming virtual element method by using this reconstruction. Numerical experiments confirm the optimal convergence rate for this nonconforming pressure-robust scheme and show that the velocity error is independent of the dynamic viscosity parameter as well as of large pressure gradients
Ramparts around Lakes on Titan Impact Winds and Methane Evaporation
No full textInternational audienceUnderstanding the lake–atmosphere interactions on Titan is crucial to answering some of Titan’s long-standing scientific questions, such as the unexpectedly high methane abundance close to the surface, or the “magic islands” that appear on one of the seas. Here we focus on specific lakes called Sharp-Edged Depressions, which are surrounded by high-elevation terrains called ramparts with unique structural and spectral properties. We aim at understanding how topography and surface properties around the lakes affect air–surface interactions. To study the effect of ramparts, we utilize the mtWRF model configured in 2D. We model a variety of lake morphologies: those surrounded by hills (300 m high, 30 km wide), those with no surrounding topography, and those at the bottom of a depression. To encapsulate the effects of surface properties, we vary the surface roughness, albedo, and thermal inertia. We also consider the effect of varying the background wind and the season. Our model indicates that the addition of topography creates weaker winds over the lake (≈ −30%) and reduces the quantity of methane evaporated (≈ −10%) compared to the no-topography case. However, topography results in deeper vertical transport of methane and deepens the marine layer (from a few meters to ≈ 50 m). Higher roughness (from 0.4 to 40 cm) of the ramparts tends to decrease the wind speed (≈ −50%) and reduces the horizontal extension of the lake breeze (≈ −20%). Albedo and thermal inertia variations have a negligible effect. Seasons have a strong effect on the evaporation of methane, with evaporation rates nearly 3 times higher in summer than in winter
Comparative Analysis of Tropospheric Water Isotope Distributions on Mars and Earth: Insights into Ice Cloud Microphysical Processes and Storm Dynamics
No full textInternational audienceIsotopic analysis is a critical tool for understanding planetary water cycles and quantifying the role of distinct atmospheric processes. This study investigates the spatio-temporal distribution and controlling factors of the HDO/H₂O ratio in water vapor within the tropospheres of Earth and Mars, highlighting the similarities and differences in water vapor transport and phase changes on both planets.We found significant isotopic enrichment from ice sublimation on both planets, with a stronger effect observed on Mars due to longer ice-crystal residence times, lower atmospheric pressures, and substantial temperature fluctuations. In contrast, Earth's near-surface oceans buffer these isotopic variations. Quantifying ice sublimation effects through observational data could help improve the microphysical parameterization in atmospheric models.Moreover, during Mars's global dust storm, the D/H ratio markedly increased and propagated upward due to reduced condensation and the absence of liquid precipitation. In contrast, Earth-based observations during typhoon events indicate isotopic depletion propagating northward from tropical regions, driven primarily by raindrop evaporation within convective systems. Thus, storm events lead to opposite isotopic responses on Earth (depletion) and Mars (enrichment). Consequently, isotopic signals have considerable potential as proxies for reconstructing storm history and intensity across planetary environments.This comparative analysis underscores both shared and planet-specific aspects of tropospheric water cycling, supporting a unified conceptual framework that effectively explains isotopic distributions under differing planetary conditions. Our results may enhance climate and weather models by improving representations of cloud microphysics and atmospheric water transport, while offering new tools for interpreting past climate events based on isotopic evidence
Model and Observation for surface–atmosphere interactions over heterogeneous landscape: MOSAI project
No full textInternational audienceThe land surface impacts the atmosphere from daily to seasonal time scales. An accurate assessment of the land–atmosphere exchanges, and their correct representation in surface schemes, are therefore essential for weather and climate forecasts. However, Earth System and Numerical Weather Prediction models often have large biases in their representation of surface–atmosphere fluxes when compared to observations. The Model and Observation for Surface–Atmosphere Interactions over heterogeneous landscape (MOSAI) project aims at improving modeling and estimating surface fluxes, with a focus on the impact of surface heterogeneity.First of all, a fair evaluation of the simulated land–atmosphere interactions in the models needs dedicated measurements and appropriate methods. Thus, the first MOSAI objective is to establish the uncertainty and representativity of land–atmosphere exchanges measured over a heterogeneous landscape. The second MOSAI objective addresses some simplifications and hypotheses in the coupling between land and atmospheric models, and their impacts on the simulated land–atmosphere exchanges. Then, the third scientific objective is to propose and test observation–model comparison methods that go further than point-to-point, time, or case-study comparisons.The MOSAI strategy leans on permanent surface energy balance stations belonging to research infrastructure and on devoted experimental campaigns in different heterogeneous landscapes. A wide range of models are also involved, from large-eddy simulation to climate models.This article will start with an overview of the impact of heterogeneous surfaces on the atmospheric boundary layer, on the difficulty of measuring surface fluxes over heterogeneous landscape, and on the simplifications of surface–atmosphere interactions over heterogeneous landscape employed by numerical models. Then, the objectives and strategy of the MOSAI project will be presented, illustrated by preliminary results
Projections of coral reef carbonate production from a global climate–coral reef coupled model
No full textInternational audienceCoral reefs are under threat due to climate change and ocean acidification. However, large uncertainties remain concerning future carbon dioxide emissions, climate change and the associated impacts on coral reefs. While most previous studies have used climate model outputs to compute future coral reef carbonate production, we use a coral reef carbonate production module embedded in a global carbonclimate model. This enables the simulation of the response of coral reefs to projected changes in physical and chemical conditions at finer temporal resolution. The use of a fastintermediate complexity model also permits the simulation of a large range of possible futures by considering different greenhouse gas concentration scenarios (Shared Socioeconomic Pathways (SSPs)) and different climate sensitivities (hence different levels of warming for a given level of acidification), as well as the possibility of corals adapting their thermal bleaching thresholds. We show that without thermal adaptation, global coral reef carbonate production decreases to less than 25 % of historical values in most scenarios over the 21st century, with limited further declines between 2100 and 2300 irrespective of the climate sensitivity. With thermal adaptation, there is far greater scenario variability in projections of reef carbonate production. Under high-emission scenarios the rate of 21st century declines is attenuated, with some global carbonate production declines delayed until the 22nd century. Under high-mitigation sce-narios, however, global coral reef carbonate production can recover in the 21st and 22nd centuries and thereafter persist at 50 %-90 % of historical values, provided that the climate sensitivity is moderate
Mobile-based deep learning system for early detection of diabetic retinopathy
No full textInternational audienceDiabetic retinopathy (DR) is a major cause of vision loss worldwide, particularly in regions where access to eye care is limited. Early detection is crucial to prevent irreversible damage. In this study, we propose an Assisted Mobile Diagnostic (AMD) system, a portable, real-time, deep learning-based solution for detecting DR. In contrast to conventional AI-based systems that rely on high-performance computers and tabletop fundus cameras, our approach integrates an optimized deep learning model for mobile execution into a mobile system combining a non-mydriatic retinal camera and a mobile device. The captured retinal images are preprocessed using techniques, including blurring and contrast enhancement, before being analyzed by a fine-tuned DenseNet-121 model. A private dataset for retinal images and two large public datasets, APTOS (Asia Pacific Tele-Ophthalmology Society) and EyePACS (Eye Picture Archive Communication System), were used to train and evaluate different deep transfer learning models. Our approach with DenseNet-121 achieves an accuracy of 97.38% on APTOS, 90.90% on EyePACS, and 98.61% on the private dataset. The system ensures real-time inference on mobile equipment, with an average processing time of 162.52 ms, enabling effective screening for DR screening. The MAD system facilitates tele-ophthalmology applications, seamlessly integrating into a multi-platform framework for remote diagnosis. The deep learning-based application is considered part of a multi-platform framework for tele-ophthalmology, integrating both a digital tablet and a desktop application
Constraints on the possible atmospheres on TRAPPIST-1 b: insights from 3D climate modeling
No full textInternational audienceContext. JWST observations of the secondary eclipse of TRAPPIST-1 b at 12.8 and 15 µm revealed a very bright dayside. These measurements are consistent with an absence of atmosphere. Previous 1D atmospheric modeling also excludes – at first sight – CO2-rich atmospheres. However, only a subset of the possible atmosphere types has been explored, and ruled out, to date. Recently, a full thermal phase curve of the planet at 15 µm with JWST has also been observed, allowing for more information on the thermal structure of the planet.Aims. We first looked for atmospheres capable of producing a dayside emission compatible with secondary eclipse observations. We then tried to determine which of these are compatible with the observed thermal phase curve.Methods. We used a 1D radiative-convective model and a 3D global climate model (GCM) to simulate a wide range of atmospheric compositions and surface pressures. We then produced observables from these simulations and compared them to available emission observations.Results. We found several families of atmospheres compatible at 2σ with the eclipse observations. Among them, some feature a flat phase curve and can be ruled out with the observation, and some produce a phase curve still compatible with the data (i.e., thin N2 –CO2 atmospheres, and CO2 atmospheres rich in hazes). We also highlight different 3D effects that could not be predicted from 1D studies (redistribution efficiency, atmospheric collapse).Conclusions. The available observations of TRAPPIST-1 b are consistent with an airless planet, which is the most likely scenario. A second possibility is a thin CO2-poor residual atmosphere. However, our study shows that different atmospheric scenarios can result in a high eclipse depth at 15 µm. It may therefore be hazardous, in general, to conclude on the presence of an atmosphere from a single photometric point
The Storage Location Assignment and Picker Routing Problem: A Generic Branch-Cut-and-Price Algorithm
No full textInternational audienceThe Storage Location Assignment Problem (SLAP) and the Picker Routing Problem (PRP) have received significant attention in the literature due to their pivotal role in the performance of the Order Picking (OP) activity, the most resource-intensive process of warehousing logistics. The two problems are traditionally considered at different decision-making levels: tactical for the SLAP, and operational for the PRP. However, this paradigm has been challenged by the emergence of modern practices in e-commerce warehouses, where decisions are more dynamic. This shift makes the integrated problem, called the Storage Location Assignment and Picker Routing Problem (SLAPRP), pertinent to consider. Scholars have investigated several variants of the SLAPRP, including different warehouse layouts and routing policies. Nevertheless, the available computational results suggest that each variant requires an ad-hoc formulation. Moreover, achieving a complete integration of the two problems, where the routing is solved optimally, remains out of reach for commercial solvers, even on trivial instances. In this paper, we propose an exact solution framework that addresses a broad class of variants of the SLAPRP, including all the previously existing ones. This paper proposes a Branch-Cut-and-Price framework based on a novel formulation with an exponential number of variables, which is strengthened with a novel family of non-robust valid inequalities. We have developed an ad-hoc branching scheme to break symmetries and maintain the size of the enumeration tree manageable. Computational experiments show that our framework can effectively solve medium-sized instances of several SLAPRP variants and outperforms the state-of-the-art methods from the literature