HAL-Ecole des Ponts ParisTech
Not a member yet
42743 research outputs found
Sort by
Characterizing coastal aquifer heterogeneity from a single piezometer head chronicle
No full textInternational audienceWe propose a new method for identifying the hydraulic properties of coastal aquifers based on their response to continental and marine influences observed from isolated piezometers displaying both tidal and seasonal fluctuations. From a single piezometer, a first approximation of hydraulic conductivity and porosity can be derived analytically from the mean hydraulic head and its tidal fluctuations. If this approximation also allows a correct simulation of seasonal head variations, the aquifer can be modelled as homogeneous. Otherwise, the seasonal head variations provide constraints on the heterogeneity of hydraulic properties. This new method is applied to a coastal aquifer in Normandy, formed by a relatively flat, kilometer-wide coastal strip of Quaternary sands followed by a foothill of higher Brioverian shales. Hydraulic head variations monitored daily for 8 years in a piezometer located in the Quaternary sand formation enable us to constrain their hydraulic conductivity (5–20 m.d−1) and porosity (7–10 %) within restricted ranges. They also suggest the existence of a significantly more porous formation (15–30 %) upstream of the observation piezometer, which could consist of more porous sands or fractured shales. In the case of limited surface–subsurface interactions, the identification method relies on a step-by-step combination of analytical solutions and simplified 1D numerical models. When more important surface–subsurface interactions are involved, a spatialized 2D model is used, that allows the analysis of seepage areas. Based on this case study, we gradually introduce 1D and 2D approaches, and discuss their applicability. We finally illustrate one major application of this approach by analyzing the interest of aquifer properties investigation on groundwater-induced flooding vulnerability
Modelling molecular composition of SOA from toluene photo-oxidation at urban and street scales
No full textInternational audienceNear-explicit chemical mechanisms representing toluene SOA formation are reduced using the GENOA algorithm and used in 3D simulations of air quality over Greater Paris and in the streets of a district near Paris. The SOA concentrations formed by the toluene photo-oxidation are found to mostly originate from molecular rearrangement with ring opening of a bicyclic peroxy radical (BPR) with an O–O bridge (45%), followed by OH-addition on the aromatic ring (22%), Highly Oxygenated organic Molecules (HOM) formation without ring opening (13%), condensation of methylnitrocatechol (8%), irreversible formation of SOA from methylglyoxal (6%), and ring-opening pathway (3%). The concentrations simulated using the most comprehensive reduced chemical scheme (rdc. Mech. 3) are also compared to those simulated with a SOA scheme based on chamber measurements, and one reduced from the Master Chemical Mechanism. Using rdc. Mech 3 leads to between 50% and 75% more toluene SOA concentrations than the other schemes, mostly because of molecular rearrangement. The SOA compounds from rdc. Mech. 3 are more oxidized and less volatile, with molecules of different functional groups. Concentrations of methylbenzoquinones, which may be of particular health interest, represent about 0.5% of the toluene SOA concentrations. Those are slightly higher in streets than in the urban background (by 2%)
Exploring the links between wellbeing and low carbon mobility practices
No full textInternational audienc
Three Pillars improving Vision Foundation Model Distillation for Lidar
No full textInternational audienceSelf-supervised image backbones can be used to address complex 2D tasks (e.g., semantic segmentation, object discovery) very efficiently and with little or no downstream supervision. Ideally, 3D backbones for lidar should be able to inherit these properties after distillation of these powerful 2D features. The most recent methods for image-to-lidar distillation on autonomous driving data show promising results, obtained thanks to distillation methods that keep improving. Yet, we still notice a large performance gap when measuring the quality of distilled and fully supervised features by linear probing. In this work, instead of focusing only on the distillation method, we study the effect of three pillars for distillation: the 3D backbone, the pretrained 2D backbones, and the pretraining dataset. In particular, thanks to our scalable distillation method named ScaLR, we show that scaling the 2D and 3D backbones and pretraining on diverse datasets leads to a substantial improvement of the feature quality. This allows us to significantly reduce the gap between the quality of distilled and fully-supervised 3D features, and to improve the robustness of the pretrained backbones to domain gaps and perturbations
Off-the-grid prediction and testing for linear combination of translated features
No full textWe consider a model where a signal (discrete or continuous) is observed with an additive Gaussian noise process. The signal is issued from a linear combination of a finite but increasing number of translated features. The features are continuously parameterized by their location and depend on some scale parameter. First, we extend previous prediction results for off-the-grid estimators by taking into account here that the scale parameter may vary. The prediction bounds are analogous, but we improve the minimal distance between two consecutive features locations in order to achieve these bounds. Next, we propose a goodness-of-fit test for the model and givenon-asymptotic upper bounds of the testing risk and of theminimax separation rate between two distinguishable signals. Inparticular, our test encompasses the signal detectionframework. We deduce upper bounds on the minimal energy,expressed as the -norm of the linear coefficients, tosuccessfully detect a signal in presence of noise. The generalmodel considered in this paper is a non-linear extension of theclassical high-dimensional regression model. It turns out that,in this framework, our upper bound on the minimax separationrate matches (up to a logarithmic factor) the lower bound on theminimax separation rate for signal detection in the highdimensional linear model associated to a fixed dictionary offeatures. We also propose a procedure to test whether thefeatures of the observed signal belong to a given finitecollection under the assumption that the linear coefficients mayvary, but have prescribed signs under the nullhypothesis. A non-asymptotic upper bound on the testing risk isgiven.We illustrate our results on the spikes deconvolution model with Gaussian features on the real line and with the Dirichlet kernel, frequently used in the compressed sensing literature, on the torus
Transport et Mobilité, Les Carnets du GREC francilien, Groupe régional d’expertise sur le changement climatique et la transition écologique en Île-de-France
No full textThis document deals at transport and mobility policies in the Île-de-France region and the sustainability objectives set by the master plans (SDRIF, PDUIF). Socio-ecological transformations call for an overhaul of regional strategies, going far beyond transport policies alone. In particular, we believe that only by coordinating affordable housing, regional planning and mobility policies, while taking into account socio-spatial inequalities, will we be able to take an ambitious and proactive approach to the levers of sobriety. Reinforcing the logic of proximity, including in outlying areas, and regulating the movement of goods (in line with the explosion in e-commerce) are two major objectives to be taken into account in this reflection
OmniSat: Self-Supervised Modality Fusion for Earth Observation
No full textInternational audienceThe diversity and complementarity of sensors available for Earth Observations (EO) calls for developing bespoke self-supervised multimodal learning approaches.However, current multimodal EO datasets and models typically focus on a single data type, either mono-date images or time series, which limits their impact. To address this issue, we introduce OmniSat, a novel architecture able to merge diverse EO modalities into expressive features without labels by exploiting their alignment.To demonstrate the advantages of our approach, we create two new multimodal datasets by augmenting {existing ones} with new modalities. As demonstrated for three downstream tasks---forestry, land cover classification, and crop mapping---OmniSat can learn rich representations without supervision, leading to state-of-the-art performances in semi- and fully supervised settings. Furthermore, our multimodal pretraining scheme improves performance even when only one modality is available for inference. The code and dataset are available at The diversity and complementarity of sensors available for Earth Observations (EO) calls for developing bespoke self-supervised multimodal learning approaches.However, current multimodal EO datasets and models typically focus on a single data type, either mono-date images or time series, which limits their impact. To address this issue, we introduce OmniSat, a novel architecture able to merge diverse EO modalities into expressive features without labels by exploiting their alignment.To demonstrate the advantages of our approach, we create two new multimodal datasets by augmenting {existing ones} with new modalities. As demonstrated for three downstream tasks—forestry, land cover classification, and crop mapping—OmniSat can learn rich representations without supervision, leading to state-of-the-art performances in semi- and fully supervised settings. Furthermore, our multimodal pretraining scheme improves performance even when only one modality is available for inference. The code and dataset are available at https://github.com/gastruc/OmniSat
Do wages underestimate the inequality in workers' rewards? The joint distribution of job quality and wages across occupations
No full textInternational audienceInformation on both wages and job quality is needed in order to understand the occupational dispersion of wellbeing. We analyse subjective wellbeing in a large UK sample to construct a measure of ‘overall reward’, the sum of wages and the value of job quality, in 90 different occupations. If only wages are included, then labour market inequality is underestimated: the dispersion of overall rewards is one‐third larger than the dispersion of wages. Our findings are similar, and stronger, in data on US workers. We find a positive correlation between job quality and wages in all specifications, both between individuals in the cross‐section and within individuals in panel data. The gender and ethnic gaps in the labour market are larger than those in wages alone, and the overall rewards to education on the labour market are underestimated by earnings differentials alone
Construyendo nuevas miradas, buscando un cambio en nuestras ciudades. La experiencia de Río Urbano
No full textInternational audienceThis essay seeks to share the experience of the Rio Urbano organization in San José (Costa Rica), which began as an individual initiative and today is a reference in the environmental management of the country's rivers. Río Urbano's objective is to transform the relationship between urban dwellers and the rivers that flow through them. The present reflection proposes a long-term transversal reading of the actions carried out at the beginning of the collective, the postulates on which they were based and theactivities carried out today. Thereby, the aim is to share its extensive trajectory and contribute to a more global reflection on the scope of this type of initiative to build more sustainable cities.Este ensayo busca compartir la experiencia de la organización Río Urbano en San José (Costa Rica), la cual empezó como iniciativa individual y hoy es una referencia en materia de gestión ambiental de los ríos del país. El objetivo de Río Urbano es transformar el relacionamiento de las personas que habitan las urbes con los ríos que las atraviesan. La presente reflexión propone una lectura transversal, a largo plazo, de las acciones realizadas en los inicios del colectivo, los postulados que las fundamentaron y las actividades efectuadas hoy. De esta manera se pretende compartir su extensa trayectoria y aportar a una reflexión más global sobre el alcance de este tipo de iniciativas para construir ciudades más sostenibles