Interpolation spatiale avec un réseau de neurones génératif comme alternative au krigeage
En géosciences, les méthodes d'interpolation spatiale peuvent être divisées en géostatistiques, non-géostatistiques ou hybrides. Le krigeage est une méthode couramment utilisée en géostatistique, sous l'hypothèse d'une distribution normale des données. De plus, il peut être très gourmand en ressources lorsqu'il est utilisé pour réaliser une interpolation avec un volume de données conséquent. Les méthodes non-géostatistiques ont bénéficié des avancées récentes des Réseaux Antagonistes Génératifs (GAN), mais elles exigent une quantité importante de données étiquetées pour produire des résultats performants. Les approches hybrides sont limitées de part leurs dépendances aux contraintes associées aux approches géostatistiques. Dans cet article, nous proposons une nouvelle méthode d'interpolation spatiale non-géostatistique par apprentissage profond, en se basant sur une technique de reconstruction d'image sans entraînement au préalable, permettant ainsi de surmonter les limites des GAN. Notre méthode utilise des connexions résiduelles et un sur-échantillonnage bi-cubique dans le but d'adapter la technique de reconstruction d'image à notre application. Elle s'appuie sur un réseau de neurones convolutifs pour produire une carte à partir d'une carte de valeurs aléatoires, en réduisant la différence entre la carte générée et les valeurs observées. L'approche proposée est évaluée sur un jeu de données de modèle numérique de terrain selon deux méthodes d'échantillonnage différentes : régulière et aléatoire. Les résultats montrent des performances supérieures par rapport à l'état de l'art des méthodes l'interpolation.
(28/08/2023)
LITIS - App, LITIS, ULH, NU, UNIROUEN, NU, INSA Rouen Normandie, INSA, NU, EPOC, EPHE, PSL, UB, INSU - CNRS, CNRS
Uncertainty Quantification of Contaminated Soil Volume with Deep Neural Networks and Predictive Models
The estimation of the soil volume exceeding a contamination threshold over decommissioned industrial sites is critical for the design of remediation strategies. In practice, the volume calculation is mostly based on preliminary sampling surveys and the use of interpolation methods. However, if the volume is not estimated correctly, this can lead to environmental and economic risks. Geostatistical-oriented methodologies have been developed to better assess the volume using uncertainty ranges. In our study, we propose a methodology entitled “Evol” to better estimate the volume and reduce the uncertainty ranges with a combination of classic non-parametrical interpolation techniques and deep learning. Evol consists of generating a synthetic model from a real polluted site, extracting descriptive variables (features) from multiple sample sets, and evaluating the error in the volume calculation. A Deep Neural Network model is then trained with the features to estimate the volume and uncertainty range for any sample set. Our methodology demonstrated high accuracy in error estimation, as evidenced by a low RMSE of 0.008 across most sample sets. Additionally, the confidence volume intervals produced by our approach were narrower than those generated by classic techniques, resulting in interval size reductions of up to 89%.
(Environmental Modeling & Assessment, n° 1420-2026, 24/08/2023)
EPOC, EPHE, PSL, UB, INSU - CNRS, CNRS, BRGM
Assessment of the Presence of Transformation Products of Certain Pharmaceutical Products (Psychotropic Family) by Suspect and Non-Targeted HRMS Screening in Wastewater Treatment Plants
Aquatic environments are the final receptors of human emissions and are therefore contaminated by molecules, such as pharmaceuticals. After use, these compounds and their metabolites are discharged to wastewater treatment plants (WWTPs). During wastewater treatment, compounds may be eliminated or degraded into transformation products (TPs) or may be persistent. The aim of this study was to develop an analytical method based on high resolution mass spectrometry (HRMS) for the identification of six psychotropic drugs that are widely consumed in France and present in WWTPs, as well as their potential associated metabolites and TPs. Four out of six psychotropic drugs and between twenty-five and thirty-seven potential TPs were detected in wastewater, although this was based on full scan data. TPs not reported in the literature and specific to the study sites and therefore to the wastewater treatment processes were tentatively identified. For the selected drugs, most known and present TPs were identified, such as desmethylvenlafaxine or norcitalopram. Moreover, the short fragmentation study led rather to the identification of several TPs of carbamazepine as ubiquitous persistent TPs.
(Toxics. vol. 11, n° 2305-6304, pp. 713, 18/08/2023)
EPOC, EPHE, PSL, UB, INSU - CNRS, CNRS
muFlowReacT : A Library to Solve Multiphase Multicomponent Reactive Transport on Unstructured Meshes
In this paper we present a new reactive transport code for the efficient simulation of groundwater quality problems. The new code couples the two previously existing tools OpenFoam and PhreeqcRM. The major objective of the development was to transfer and expand the capabilities of the MODFLOW/MT3DMS‐family of codes, especially their outstanding ability to suppress numerical dispersion, to a versatile and computationally efficient code for unstructured grids. Owing to the numerous, previously existing transport solvers contained in OpenFoam, the newly developed code achieves this objective and provides a solid basis for future expansions of the code capabilities. The flexibility of the OpenFoam framework is illustrated by the addition of diffusional processes for gaseous compounds in the unsaturated zone and the advection of gases (multiphase transport). The code capabilities and accuracy are illustrated through several examples: (1) a simple 2D case for conservative solute transport under saturated conditions, (2) a gas diffusion case with reactions in the unsaturated zone, (3) a hydrogeologically complex 3D reactive transport problem, and finally (4) the injection of CO 2 into a deep aquifer with acidification being buffered by carbonate minerals.
(Groundwater, n° 0017-467X, 10/08/2023)
EPOC, EPHE, PSL, UB, INSU - CNRS, CNRS, UWA, UWA, CSIRO, BRGM
The role of ecotypic variation for plant facilitation in a metal-polluted system: Stress-intolerant target ecotypes are the best beneficiaries and stress-tolerant nurse ecotypes the best benefactors
(Science of the Total Environment. vol. 887, n° 0048-9697, pp. 164134, 01/08/2023)
EPOC, EPHE, PSL, UB, INSU - CNRS, CNRS
Comparison of Dynamic Cobble Berm Revetments with Differing Cobble Characteristics
Pressure on the coastline is escalating due to the impacts of climate change, this is leading to a rise in sea-levels and intensifying storminess. Consequently, many regions of the coast are at increased risk of erosion and flooding. Therefore coastal protection schemes will increase in cost and scale. In response there is a growing use of nature-based coastal protection which aim to be sustainable, effective and adaptable. An example of a nature-based solution is a dynamic cobble berm revetment: a berm constructed from cobble and other gravel sediments at the high tide wave runup limit. These structures limit wave excursion protecting the hinterland from inundation, stabilise the upper beach and adapt to changes in water level. Recent experiments and field applications have shown the suitability of these structures for coastal protection, however many of the processes and design considerations are poorly understood. This study directly compares two prototype scale laboratory experiments which tested dynamic cobble berm revetments constructed with approximately the same geometry but differing gravel characteristics; well-sorted rounded gravel (DynaRev1) and poorly-sorted angular gravel (DynaRev2). In both cases the structures were tested using identical wave forcing including incrementally increasing water level and erosive wave conditions. The results presented in this paper demonstrate that both designs responded to changing water level and wave conditions by approaching a dynamically stable state, where individual gravel is mobilised under wave action but the geometry remains approximately constant. Further, both structures acted to reduce swash excursions compared to a pure sand beach. However, their morphological behaviour is response to wave action varied considerably. Once overtopping of the designed crest occurred, the poorly-sorted revetment developed a peaked crest which grew in elevation as the water level or wave height increased, further limited overtopping. By comparison, the well-sorted revetment was characterised by a larger volume of submerged gravel and a lower elevation flat crest which responded less well to changes in conditions. This occurred due to two processes: (1) for the poorly-sorted case, gravel sorting processes moved small to medium gravel material (D) to the crest and (2) the angular nature of the poorly-sorted gravel material promoted increased interlocking. Both of these processes led to a gravel matrix that is more resistant to wave action and gravitational effects. Both revetments experienced some sinking due to sand erosion beneath the front slope. The rate of sinking for the well-sorted case was larger and continued throughout due to the large pore spaces within the gravel matrix. For the poorly sorted revetment in DynaRev2, sand erosion ceased after approximately 28 h due to the development of a filter layer of small gravel at the sand-gravel interface reducing porosity at this location, hence a larger volume of sand was preserved beneath the structure. Both designs present a low-cost and effective solution for protecting sandy coastlines but from an engineering viewpoint it appears better to avoid well-sorted gravel material and greater gravel angularity has been seen to increase crest stability.
(Coastal Engineering. vol. 183, n° 0378-3839, pp. 104312, 01/08/2023)
EPOC, EPHE, PSL, UB, INSU - CNRS, CNRS
The great shift: The non-indigenous species Ammonia confertitesta (Foraminifera, Rhizaria) outcompetes indigenous Ammonia species in the Gironde estuary (France)
(Estuarine, Coastal and Shelf Science. vol. 289, n° 0272-7714, pp. 108378, 01/08/2023)
LOG, INSU - CNRS, ULCO, CNRS, IRD [Ile-de-France], JAMSTEC, EPOC, EPHE, PSL, UB, INSU - CNRS, CNRS, LPG-ANGERS, LPG, UM, UA, INSU - CNRS, CNRS, Nantes univ - UFR ST, AD2M, CNRS, SBR, SU, CNRS
Quantifying source effects based on rainwater δ18O from 10-year monitoring records in Southwest China
The stable isotopic composition of rainwater (d(18)Op and dDp) has been considered an effective proxy in studying atmospheric circulation and hydrological cycle processes. However, the linkage between variabilities in moisture sources and d(18)Op remains poorly understood in Southwest China. Here, we utilized three long-term (10-year) d(18)Op records (Beibei, Furong, and Yangkou stations) in Chongqing, Southwest China, from 2010 to 2019 A.D., integrated with a cluster analysis based on the Hybrid Single-Particle Lagrangian Integrated Trajectory model (HYSPLIT) to quantitatively demonstrate the close remote coupling of moisture sources and d(18)Op on seasonal and interannual timescales. The clustering calculation results indicated that the percentage initial moisture source proportions of the westerlies and Asian summer monsoon have changed significantly at seasonal and interannual timescales, leading to the significant variation in d(18)Op in Southwest China. Domination of westerlies and inland moisture contribution resulted in the positive d(18)Op excursion in the dry season (November, December, January-April) while prevailing oceanic moisture led to the negative d(18)Op excursions in the rainy season (May-October). On the interannual scale, d(18)Op and d-excess exhibited a persistent positive/high trend, which is consistent with a decrease/increase in water vapor contributions from the ocean/westerly sources. Furthermore, d(18)Op showed a good correlation with the Indian Ocean Dipole index, El Nino-Southern Oscillation, and the South Asian monsoon index, indicating that large-scale circulation patterns may affect regional d(18)Op through their influence on moisture transport and convective activity in the source region. The long-term observation confirmed changes in moisture sources largely depend on the Westerly and Indian monsoon dy-namics controlled by ocean-atmosphere circulation, supporting the effects of moisture sources on the rainwater isotopic composition.
(Applied Geochemistry. vol. 155, n° 0883-2927, pp. 105706, 01/08/2023)
EPOC, EPHE, PSL, UB, INSU - CNRS, CNRS, PIK, HU Berlin, UCL, UMR ISPA, Bordeaux Sciences Agro, INRAE
Satellite-based shoreline detection along high-energy macrotidal coasts and influence of beach state
Earth observation coupled with novel image analysis techniques now present a unique and powerful tool for the historical study of shoreline change at local to global scale. However, satellite-derived shoreline (SDS) data is limited in certain areas and is associated with large uncertainties relating to environmental factors, tidal range, and wave action. We use 14 years of monthly topographic surveys at two macrotidal sites in the UK representing end members of beach type (reflective, dissipative) to investigate the influence of tidal elevation and wave action on SDS accuracy. We find that applying appropriate water level corrections can significantly improve SDS accuracy. Results show that a different approach is required for water level definition depending on beach type and reveal that ultimately SDS accuracy is primarily controlled by beach state (beach profile shape). Accounting for tidal elevation led to substantial accuracy improvement at both sites and formed the optimal SDS strategy for the reflective site (Slapton). At the dissipative site (Perranporth) considering wave-induced water level fluctuations (wave setup and/or runup), including wave shoaling, was critical for reducing the tidally corrected SDS RMSE by a third and the mean bias by three quarters. An important realization for areas with high cloud cover such as the UK, and/or low satellite coverage, was that critically low image availability restricts temporally the type of phenomenon that can be detected (e.g., seasonal/interannual variability) and may compromise computed longterm trends. Our results suggest that the optimal approach is site-specific and depends on the shoreline translation method used and is therefore different depending on the application. We propose optimal SDS strategies to increase confidence in SDS extraction in meso-macrotidal environments with potentially low satellite useability (i.e., high cloud cover and/or low satellite coverage) depending on the spatial scale of the intended application. Long-term trends derived using this approach can reproduce trends from ground-based surveys and therefore enable more accurate projections of future shoreline position to be made.
(Marine Geology. vol. 462, n° 0025-3227, 01/08/2023)
EPOC, EPHE, PSL, UB, INSU - CNRS, CNRS
[Dataset] Microcharcoal concentration, elongation, burnt vegetation types and age model of core MD96-2098 for the past 184 ka
(27/07/2023)
EPOC, EPHE, PSL, UB, INSU - CNRS, CNRS