HAL-Ecole des Ponts ParisTech
Not a member yet
42743 research outputs found
Sort by
Evaluating Weather and Chemical Transport Models at High Latitudes using MAGIC2021 Airborne Measurements
No full textInternational audienceHigh latitude wetland emissions of methane (CH 4 ) remain a significant source of uncertainty in global methane budgets. At these latitudes, flux estimation approaches, such as atmospheric inversions, are challenged by complex meteorological conditions, limited observational coverage, and uncertainties in atmospheric transport modelling. This study evaluates the performance of various atmospheric transport models and reanalysis datasets using meteorological and CH 4 in-situ measurements collected during the MAGIC2021 campaign near Kiruna, Sweden. Over six days of measurements in August 2021, the ERA5 reanalysis, produced by the European Centre for Medium-Range Weather Forecasts (ECMWF) and providing global atmospheric data, showed better agreement with observations compared to the mesoscale Weather Research and Forecasting (WRF) model, though WRF provided valuable insights into local atmospheric dynamics. Among global simulations of CH 4 mixing ratios, inversion-optimised models which adjust emissions to match observations, achieved the best performance overall particularly when constrained by surface measurements. Regional simulations from WRF coupled with chemistry (WRF-Chem) revealed biases in CH 4 mixing ratios in the boundary layer, suggesting an overestimation of emissions by wetland models. All chemistry-transport models exhibited a positive bias in the stratosphere. Simulations with higher vertical resolution demonstrated an improved representation of vertical CH 4 profiles in the upper layers of the atmosphere. Despite the limited spatio-temporal coverage of the observations, we were able to identify the best performing transport models and to evaluate fluxes from different biogeochemical model parameterisations using the MAGIC2021 highresolution dataset, demonstrating the utility of in-situ vertical profile datasets for transport and flux model evaluation.</div
Deep learning in the abyss: a stratified Physics Informed Neural Network for data assimilation
No full textThe reconstruction of deep ocean currents is a major challenge in data assimilation due to the scarcity of interior data. In this work, we present a proof of concept for deep ocean flow reconstruction using a Physics-Informed Neural Network (PINN), a machine learning approach that offers an alternative to traditional data assimilation methods. We introduce an efficient algorithm called StrAssPINN (for Stratified Assimilation PINNs), which assigns a separate network to each layer of the ocean model while allowing them to interact during training. The neural network takes spatiotemporal coordinates as input and predicts the velocity field at those points. Using a SIREN architecture (a multilayer perceptron with sine activation functions), which has proven effective in various contexts, the network is trained using both available observational data and dynamical priors enforced at several collocation points. We apply this method to pseudo-observed ocean data generated from a 3-layer quasi-geostrophic model, where the pseudo-observations include surface-level data akin to SWOT observations of sea surface height, interior data similar to ARGO floats, and a limited number of deep ARGO-like measurements in the lower layers. Our approach successfully reconstructs ocean flows in both the interior and surface layers, demonstrating a strong ability to resolve key ocean mesoscale features, including vortex rings, eastward jets associated with potential vorticity fronts, and smoother Rossby waves. This work serves as a prelude to applying StrAssPINN to real-world observational data
Switching Macroeconomic Growth and Volatility: Evidence from a Mean-Variance Markov-Switching Dynamic Factor Model
No full textAs illustrated by the Great Recession, the COVID-19 pandemic and the global decline in GDP growth since the mid-2000s, economists need to account for sudden and deep recessions, shifts in macroeco- nomic volatility, and longer-term fluctuations in GDP growth. This paper puts forward a Mean-Variance Markov-Switching Dynamic Factor Model (MV-MS-DFM) that accounts for these stylised facts by allow- ing the mean and the volatility of macroeconomic variables to switch abruptly, and trend GDP growth to vary smoothly over time. We show that allowing for different volatility regimes improves the detection of turning points in the U.S. business cycle, that the Great Recession and the COVID-19 pandemic only led to temporary increases in volatility, and that the U.S. trend GDP growth has declined by around 1 percentage point since the early 2000s. Information criteria and marginal likelihood comparisons support our model specification. The model provides a unified framework connecting the literature on turning- point detection to the more recent literature on Growth-at-Risk in macroeconomic forecasting. While tightening financial conditions are shown to increase the probability of falling in a recession, the model can generate left-skewed density forecasts without including any financial variable in the information set. The paper finally discusses how to adjust the model estimation strategy to deal with the COVID-19 period
Coupled chemo-thermo-hydromechanical simulation for integrity assessment of CO2 injection wells
No full textInternational audienceCoupled chemo-thermo-hydromechanical simulation for integrity assessment of CO2 injection well
Removing Behavioral Barriers to Energy Renovation: A Discrete Choice Experiment
No full textThrough a discrete choice experiment conducted among French homeowners, we determine whether easing financial constraints through two financing programmes (third-party financing and Energy Efficient Mortgages) leads to an increase in the adoption rate of energy-efficient renovations. We examine whether the introduction of a contractual mechanism that intrinsically promotes trust between the parties (one-stop-shop) can increase the adoption rate.Among other things, our study reveals certain preferences among households with regard to financing energy-efficient home renovations and assesses certain cognitive biases. We find that both financing programmes increase the probability of choosing an energy renovation over the opt-out option. Nevertheless, participants exposed to the mortgage programme have a preference for the status quo (not renovating), unlike those exposed to third-party financing.With the latter financing programme, the effectiveness of the work seems guaranteed, unlike with the mortgage. This explains why the group offered the energy-efficient mortgage is more inclined to choose renovation scenarios that include an administrative facilitator, highlighting the importance of procedural support and risk transfer in the decision-making process.</p
Maximal number of subword occurrences in a word
No full textInternational audienceWe consider the number of occurrences of subwords (non-consecutive sub-sequences) in a given word. We first define the notion of subword entropy of a given word that measures the maximal number of occurrences among all possible subwords. We then give upper and lower bounds of minimal subword entropy for words of fixed length in a fixed alphabet, and also showing that minimal subword entropy per letter has a limit value. A better upper bound of minimal subword entropy for a binary alphabet is then given by looking at certain families of periodic words. We also give some conjectures based on experimental observations
Spatialized and Prospective LCA of Agrivoltaism in the Mediterranean basin under Climate Change
No full textInternational audienceMediterranean agriculture will face significant challenges from climate change throughout the 21st century, necessitating adaptation tosustain production. Without such measures, reduced efficiency and yields will further increase agriculture’s environmental footprint tomaintain output levels. Among the options for adaptation and mitigation, agrivoltaics—the integration on the same land of photovoltaicenergy production and crop cultivation—offers a promising solution by mitigating climate impacts on crops while promoting renewableenergy production.However, the environmental performance of agrivoltaic systems depends on complex interactions among technological, agronomic, andmeteorological factors, all of which are expected to evolve. Anticipating these systems’ future performances is essential to identify contextsand scenarios where agrivoltaics maximize environmental benefits. Such projection also helps prevent suboptimal deployment, whereconventional photovoltaic and agricultural systems might outperform agrivoltaics.Our study thus supports decision-making by conducting a prospective, consequential life cycle assessment (LCA) of agrivoltaics, whetherdriven by crop or energy demand. We integrate a parameterized LCA model with prospective databases using PREMISE and simulateyields and environmental flows with the ORCHIDEE land surface model, tailored for agrivoltaic and conventional systems. The analysisfocuses on the Mediterranean coasts of France, Spain, and Italy, considering local soil carbon and water dynamics. Substantial attention isgiven to addressing the uncertainty and the sensitivity of the results in relation to the three components of the model, i.e., foreground,background and land surface model.Our work thus provides an insightful mapping of where, when, to what extent and under which scenarios agrivoltaic impacts outperform theconventional ones by studying a large scope of possibilities for socio-economic patways (SSP), climate trajectories, technological and methodological choices. This comprehensive approach provides actionable insights for planning agrivoltaic deployment under diversescenarios for the 21st century
Dynamic Network Formation with Farsighted Players and Limited Capacities
No full textDocuments de travail du Centre d’Économie de la Sorbonne 2025.19 - ISSN : 1955-611X - eISSN : 2968-6687We investigate a T -stage dynamic network formation game with linear-quadratic payoffs. Players interact through network which they create as a result of their actions. We study two versions of the dynamic game and provide the equilibrium analysis. First, we assume that players sequentially propose links to others with whom they want to connect and choose the levels of contribution for their links. The players have limited total contributions or capacities for forming links at every stage which can differ among players and over time. They cannot delete links, but the principle of natural elimination of links with no contribution is adopted. Next, we assume that the players simultaneously and independently propose links to other players and have overall limited capacities for the whole game, and not for each stage. This means that every player can redistribute the capacity not only over links, but also over time. The equilibrium concept for the first version of the dynamic game is subgame perfect equilibrium, while it is the Nash equilibrium in open-loop strategies for the second version. Both models are illustrated with numerical examples
Uniform minorization condition and convergence bounds for discretizations of kinetic Langevin dynamics
No full textInternational audienceWe study the convergence in total variation and -norm of discretization schemes of the underdamped Langevin dynamics. Such algorithms are very popular and commonly used in molecular dynamics and computational statistics to approximatively sample from a target distribution of interest. We show first that, for a very large class of schemes, a minorization condition uniform in the stepsize holds. This class encompasses popular methods such as the Euler-Maruyama scheme and the schemes based on splitting strategies. Second, we provide mild conditions ensuring that the class of schemes that we consider satisfies a geometric Foster--Lyapunov drift condition, again uniform in the stepsize. This allows us to derive geometric convergence bounds, with a convergence rate scaling linearly with the stepsize. This kind of result is of prime interest to obtain estimates on norms of solutions to Poisson equations associated with a given numerical method
Combinatorics of the Quantum Symmetric Simple Exclusion Process, associahedra and free cumulants
No full textInternational audienceThe quantum symmetric simple exclusion process (QSSEP) is a model of quantum particles hopping on a finite interval and satisfying the exclusion principle. Recently, Bernard and Jin have studied the fluctuations of the invariant measure for this process, when the number of sites goes to infinity. These fluctuations are encoded into polynomials, for which they have given equations and proved that these equations determine the polynomials completely. In this paper, we give an explicit combinatorial formula for these polynomials, in terms of Schröder trees. We also show that, quite surprisingly, these polynomials can be interpreted as free cumulants of a family of commuting random variables