Thèse Suivi Hydrologique Multi-Échelles par Satellite et In Situ dans les Deltas Tropicaux Afin de Quantifier les Risques Sociétaux Liés au Changement Climatique Mondial à l'Aide d'Une Intelligenc H/F - Doctorat.Gouv.Fr
- Toulouse - 31
- CDD
- Télétravail accepté
- Doctorat.Gouv.Fr
Les missions du poste
Établissement : Université de Toulouse École doctorale : SDU2E - Sciences de l'Univers, de l'Environnement et de l'Espace Laboratoire de recherche : GET - Geosciences Environnement Toulouse Direction de la thèse : José DARROZES ORCID 0000000317502819 Début de la thèse : 2027-09-01 Date limite de candidature : 2026-11-23T23:59:59 Ce projet de thèse vise à développer un cadre de surveillance multi-échelle intégré pour quantifier les ressources en eau continentales dans les deltas du Mékong et de la Rivière Rouge, en combinant des observations satellitaires, des données au sol issues de la réflectométrie GNSS (GNSS-R) et de stations inertielles (mesures du champ de gravité variable), des mesures in situ et des modèles hydrologiques. L'objectif scientifique principal est de relier les composantes du cycle de l'eau - humidité du sol en surface, humidité du sol dans la zone racinaire, stockage des eaux souterraines et stockage terrestre de l'eau (TWS) - au sein d'un cadre cohérent adapté aux environnements deltaïques tropicaux vulnérables.
En couplant la réflectivité de surface à haute fréquence, obtenues par la mission CYGNSS, aux anomalies du TWS, mesurées par GRACE/GRACE-FO, ainsi qu'aux données Sentinel-1/2, SMAP et météorologiques, le projet développera un réseau pilote GNSS-R/station inertielle au sol afin d'étalonner les produits satellitaires et d'évaluer la physique de la réflexion du sol. Un algorithme d'intelligence artificielle fondée sur la physique (PI-AI) appliquera les contraintes du bilan hydrique et de la conservation de la masse afin de reconstituer avec précision l'humidité de la zone racinaire, les anomalies des eaux souterraines et les déficits en eau douce pour les réservoirs de surface (mares, lacs, rivières, fleuves).
Lien direct avec l'impact sociétal et le changement global :
Au-delà de l'innovation en matière de télédétection, cette thèse aborde directement les conséquences sociétales du changement global - notamment la variabilité climatique, l'élévation du niveau de la mer, la construction anthropique de barrages en amont des zones de delta et la surexploitation des nappes phréatiques. Ces données provenant de sources multiples seront ensuite transformées en indicateurs opérationnels à haute résolution temporelles pour une gestion innovante des risques de sécheresse agricole et de salinisation des sols. Ce cadre permettra de quantifier directement les menaces pesant sur la sécurité alimentaire, les moyens de subsistance agricoles et la disponibilité en eau douce pour des millions d'habitants des deltas d'Asie du Sud-Est.
Le delta du Mékong sert de référence principale pour évaluer les risques combinés (sécheresse, salinisation et affaissement du sol induit par l'épuisement des nappes phréatiques), tandis que le delta du fleuve Rouge permet une validation croisée afin de tester la robustesse méthodologique et la transférabilité dans des contextes socio-écologiques contrastés. À terme, cette recherche jette les bases d'un observatoire hydrologique multiplateforme et d'un jumeau numérique hydrologique pour les deltas vietnamiens, faisant le lien entre l'hydrologie par satellite, l'intelligence artificielle et la modélisation physique afin de favoriser une prise de décision fondée sur des données probantes, une gestion durable de l'eau et des stratégies d'adaptation au changement climatique.
Scientific Context and Regional Framework
The Mekong and Red River deltas are two major hydrological systems in Southeast Asia, playing an essential role in food security, water resources, and the socio-economic development of Vietnam. Their functioning results from complex interactions between precipitation, river flows, groundwater, coastal exchanges, agricultural practices, and human activities. These interactions directly control freshwater availability (Rodell et al., 2018), agricultural productivity, and territorial resilience in the face of environmental changes.
Over recent decades, these deltas have become particularly vulnerable due to the combined effects of climate variability, increasing drought frequency, altered hydrological regimes, growing groundwater exploitation, and saline intrusion. These phenomena lead to a progressive decline in available water resources and the degradation of agricultural soils, particularly in the Mekong Delta , where anthropogenic pressure and climate change mutually reinforce each other. While presenting a different hydrological context, the Red River Delta is also highly exposed to agricultural drought and salinization risks in its coastal sectors.
However, understanding these processes remains limited by the lack of observational capabilities capable of simultaneously tracking the various components of the continental water cycle. In situ measurement networks provide precise but spatially limited information, whereas satellite missions each observe a specific part of the hydrological system with vastly different spatial and temporal resolutions (Bauer-Gottwein et al., 2015). This heterogeneity represents a major scientific bottleneck for achieving a coherent representation of the water status in large deltas.
Multi-Platform Observations and the Methodological Challenge
Within this context, missions like CYGNSS provide repetitive GNSS-reflectometry observations that characterize high-frequency surface soil moisture in tropical regions. Meanwhile, GRACE and GRACE-FO measure variations in total continental water storage-integrating soil moisture, surface water, and groundwater contributions-albeit at a much coarser spatial resolution. These complementary families of missions offer valuable insights into different compartments of the hydrological cycle.
Characterizing this water cycle at the delta scale can no longer rely on isolated data sources. Integrating altimetric, gravimetric, and radar missions provides coverage across the entire hydrological continuum:
Radar and optical missions (Sentinel-1/2) offer high spatial resolution essential for monitoring flooded surfaces, soil moisture, agricultural dynamics, and rice cropping systems (Tran et al., 2022).
GNSS Reflectometry (GNSS-R, via CYGNSS or ground networks) revolutionizes land surface characterization with high temporal resolution under vegetation cover.
Space gravimetry (GRACE / GRACE-FO) quantifies total water storage (TWS) and groundwater anomalies (Pham-Duc et al., 2019, for large watershed. Recent studies demonstrated the possibility to quantifies local watershed (Ramillien et al., 2021). These studies needs to be complemented by soil moisture or continental free water (lake, river...) from SMAP or CYGNSS to obtain information about groundwater.
Despite these advances, no unified methodological framework currently exists to bridge high-frequency surface observations from GNSS-reflectometry with continental storage variations from satellite gravimetry, while accounting for hydrological processes, field observations, and the physical constraints governing water transfers. Pure 'black-box' machine learning models also suffer from poor generalization in unobserved or extreme hydro-climatic contexts. To overcome this, contemporary research develops hybrid approaches that embed fundamental physical laws (mass conservation, water balance equations) into neural networks, ensuring that root-zone soil moisture, water fluxes, and subsurface estimates remain physically consistent. General Objective
To develop and validate a multi-platform hydrological observatory integrating satellite observations (CYGNSS, GRACE-FO, Sentinel, SMAP), a pilot network of ground-based GNSS-R stations, in situ measurements, hydrological models, and physics-informed artificial intelligence. This aims to reconstruct the hydrological continuum, monitor continental water resources, and develop robust tools for tracking agricultural drought, groundwater, and salinization risks in the Mekong and Red River deltas, contributing directly to the understanding of global changes and their impacts on societies.
Specific Objectives
Multi-Source Observation Framework: Design an observation acquisition and fusion framework integrating CYGNSS, GRACE-FO, Sentinel-1, Sentinel-2, SMAP, meteorological data, hydrological observations, in situ measurements, and a pilot network of ground-based GNSS-R stations to establish a coherent observatory of water cycle components.
Multi-Scale Estimation Methodology: Develop a multi-scale methodology to consistently estimate surface soil moisture, root-zone soil moisture, groundwater storage anomalies, and total continental water storage by combining satellite observations with hydrological balance constraints.
Physics-Informed Modeling: Design machine learning models that explicitly integrate the physical laws governing water transfers, notably mass conservation and the water balance, to enhance the robustness, interpretability, and generalization capacity of estimations.
Integrated Hydrological Indicators: Develop integrated hydrological indicators to characterize freshwater deficits, agricultural drought episodes, and soil salinization risks by combining satellite observations, field data, and derived hydrological variables.
Development of an operational decision-support tool based on hydrogeological, geophysical, and remote-sensing metrics, but also-and above all-on societal impact metrics.
Hydrological Digital Twin Prototype: Establish a hydrological Digital Twin prototype capable of integrating near-real-time satellite observations, ground-based GNSS-R stations, and hydrological models to ensure dynamic monitoring of water resources and generate scenarios of hydrological conditions.
Scientific Performance Evaluation: Assess the scientific performance of the developed models by analyzing their robustness, uncertainties, and transferability between the Mekong Delta and the Red River Delta under contrasting hydroclimatic, agricultural, and anthropogenic contexts. WP1: Multi-Platform Earth Observation & Ground Network Deployment
Establish a harmonized data pipeline for CYGNSS (surface moisture), GRACE-FO (gravity/TWS, Ramillien, Soane & Darrozes, 2021), Sentinel-1/2 (land use/vegetation/SAR), and SMAP.
Deploy a pilot network of ground GNSS-R stations (Ha et al, 2021 ; Roussel et al., 2016) along the Mekong transect (Vinh Long - Ben Tre - Tra Vinh - Soc Trang) equipped with GNSS receivers, soil moisture/EC sensors, and weather units for satellite calibration.
WP2: Physics-Informed AI for Hydrological Continuum Reconstruction
Develop neural network architectures constrained by mass balance equations: 'TWS'='SM'+'SW'+'CW'+'GWS' .
Disaggregate surface soil moisture down to the root-zone (RZSM) and extract Groundwater Storage Anomalies (GWSA) to quantify sub-surface water depletion, drought (Hoang et al., 2025)
WP3: Derivation of Societal Risk Metrics & Digital Twin Prototype
State of the art
Translate physical outputs into operational risk indicators:
Agricultural Drought Index: Measuring root-zone water stress during key crop growth phases.
Soil Salinization Risk Model: Combining GNSS-R dielectric responses, river salinity, and groundwater levels.
Build a prototype Hydrological Digital Twin integrating real-time satellite/ground data feeds for dynamic scenario simulation.
WP4: Inter-Regional Validation & Transferability Analysis
Test model robustness on coastal sectors of the Red River Delta (Nam Dinh, Thai Binh, Ninh Binh, Hai Phong).
Quantify model uncertainties and identify universal vs. locally calibrated parameters across contrasting hydro-climatic settings.
En couplant la réflectivité de surface à haute fréquence, obtenues par la mission CYGNSS, aux anomalies du TWS, mesurées par GRACE/GRACE-FO, ainsi qu'aux données Sentinel-1/2, SMAP et météorologiques, le projet développera un réseau pilote GNSS-R/station inertielle au sol afin d'étalonner les produits satellitaires et d'évaluer la physique de la réflexion du sol. Un algorithme d'intelligence artificielle fondée sur la physique (PI-AI) appliquera les contraintes du bilan hydrique et de la conservation de la masse afin de reconstituer avec précision l'humidité de la zone racinaire, les anomalies des eaux souterraines et les déficits en eau douce pour les réservoirs de surface (mares, lacs, rivières, fleuves).
Lien direct avec l'impact sociétal et le changement global :
Au-delà de l'innovation en matière de télédétection, cette thèse aborde directement les conséquences sociétales du changement global - notamment la variabilité climatique, l'élévation du niveau de la mer, la construction anthropique de barrages en amont des zones de delta et la surexploitation des nappes phréatiques. Ces données provenant de sources multiples seront ensuite transformées en indicateurs opérationnels à haute résolution temporelles pour une gestion innovante des risques de sécheresse agricole et de salinisation des sols. Ce cadre permettra de quantifier directement les menaces pesant sur la sécurité alimentaire, les moyens de subsistance agricoles et la disponibilité en eau douce pour des millions d'habitants des deltas d'Asie du Sud-Est.
Le delta du Mékong sert de référence principale pour évaluer les risques combinés (sécheresse, salinisation et affaissement du sol induit par l'épuisement des nappes phréatiques), tandis que le delta du fleuve Rouge permet une validation croisée afin de tester la robustesse méthodologique et la transférabilité dans des contextes socio-écologiques contrastés. À terme, cette recherche jette les bases d'un observatoire hydrologique multiplateforme et d'un jumeau numérique hydrologique pour les deltas vietnamiens, faisant le lien entre l'hydrologie par satellite, l'intelligence artificielle et la modélisation physique afin de favoriser une prise de décision fondée sur des données probantes, une gestion durable de l'eau et des stratégies d'adaptation au changement climatique.
Scientific Context and Regional Framework
The Mekong and Red River deltas are two major hydrological systems in Southeast Asia, playing an essential role in food security, water resources, and the socio-economic development of Vietnam. Their functioning results from complex interactions between precipitation, river flows, groundwater, coastal exchanges, agricultural practices, and human activities. These interactions directly control freshwater availability (Rodell et al., 2018), agricultural productivity, and territorial resilience in the face of environmental changes.
Over recent decades, these deltas have become particularly vulnerable due to the combined effects of climate variability, increasing drought frequency, altered hydrological regimes, growing groundwater exploitation, and saline intrusion. These phenomena lead to a progressive decline in available water resources and the degradation of agricultural soils, particularly in the Mekong Delta , where anthropogenic pressure and climate change mutually reinforce each other. While presenting a different hydrological context, the Red River Delta is also highly exposed to agricultural drought and salinization risks in its coastal sectors.
However, understanding these processes remains limited by the lack of observational capabilities capable of simultaneously tracking the various components of the continental water cycle. In situ measurement networks provide precise but spatially limited information, whereas satellite missions each observe a specific part of the hydrological system with vastly different spatial and temporal resolutions (Bauer-Gottwein et al., 2015). This heterogeneity represents a major scientific bottleneck for achieving a coherent representation of the water status in large deltas.
Multi-Platform Observations and the Methodological Challenge
Within this context, missions like CYGNSS provide repetitive GNSS-reflectometry observations that characterize high-frequency surface soil moisture in tropical regions. Meanwhile, GRACE and GRACE-FO measure variations in total continental water storage-integrating soil moisture, surface water, and groundwater contributions-albeit at a much coarser spatial resolution. These complementary families of missions offer valuable insights into different compartments of the hydrological cycle.
Characterizing this water cycle at the delta scale can no longer rely on isolated data sources. Integrating altimetric, gravimetric, and radar missions provides coverage across the entire hydrological continuum:
Radar and optical missions (Sentinel-1/2) offer high spatial resolution essential for monitoring flooded surfaces, soil moisture, agricultural dynamics, and rice cropping systems (Tran et al., 2022).
GNSS Reflectometry (GNSS-R, via CYGNSS or ground networks) revolutionizes land surface characterization with high temporal resolution under vegetation cover.
Space gravimetry (GRACE / GRACE-FO) quantifies total water storage (TWS) and groundwater anomalies (Pham-Duc et al., 2019, for large watershed. Recent studies demonstrated the possibility to quantifies local watershed (Ramillien et al., 2021). These studies needs to be complemented by soil moisture or continental free water (lake, river...) from SMAP or CYGNSS to obtain information about groundwater.
Despite these advances, no unified methodological framework currently exists to bridge high-frequency surface observations from GNSS-reflectometry with continental storage variations from satellite gravimetry, while accounting for hydrological processes, field observations, and the physical constraints governing water transfers. Pure 'black-box' machine learning models also suffer from poor generalization in unobserved or extreme hydro-climatic contexts. To overcome this, contemporary research develops hybrid approaches that embed fundamental physical laws (mass conservation, water balance equations) into neural networks, ensuring that root-zone soil moisture, water fluxes, and subsurface estimates remain physically consistent. General Objective
To develop and validate a multi-platform hydrological observatory integrating satellite observations (CYGNSS, GRACE-FO, Sentinel, SMAP), a pilot network of ground-based GNSS-R stations, in situ measurements, hydrological models, and physics-informed artificial intelligence. This aims to reconstruct the hydrological continuum, monitor continental water resources, and develop robust tools for tracking agricultural drought, groundwater, and salinization risks in the Mekong and Red River deltas, contributing directly to the understanding of global changes and their impacts on societies.
Specific Objectives
Multi-Source Observation Framework: Design an observation acquisition and fusion framework integrating CYGNSS, GRACE-FO, Sentinel-1, Sentinel-2, SMAP, meteorological data, hydrological observations, in situ measurements, and a pilot network of ground-based GNSS-R stations to establish a coherent observatory of water cycle components.
Multi-Scale Estimation Methodology: Develop a multi-scale methodology to consistently estimate surface soil moisture, root-zone soil moisture, groundwater storage anomalies, and total continental water storage by combining satellite observations with hydrological balance constraints.
Physics-Informed Modeling: Design machine learning models that explicitly integrate the physical laws governing water transfers, notably mass conservation and the water balance, to enhance the robustness, interpretability, and generalization capacity of estimations.
Integrated Hydrological Indicators: Develop integrated hydrological indicators to characterize freshwater deficits, agricultural drought episodes, and soil salinization risks by combining satellite observations, field data, and derived hydrological variables.
Development of an operational decision-support tool based on hydrogeological, geophysical, and remote-sensing metrics, but also-and above all-on societal impact metrics.
Hydrological Digital Twin Prototype: Establish a hydrological Digital Twin prototype capable of integrating near-real-time satellite observations, ground-based GNSS-R stations, and hydrological models to ensure dynamic monitoring of water resources and generate scenarios of hydrological conditions.
Scientific Performance Evaluation: Assess the scientific performance of the developed models by analyzing their robustness, uncertainties, and transferability between the Mekong Delta and the Red River Delta under contrasting hydroclimatic, agricultural, and anthropogenic contexts. WP1: Multi-Platform Earth Observation & Ground Network Deployment
Establish a harmonized data pipeline for CYGNSS (surface moisture), GRACE-FO (gravity/TWS, Ramillien, Soane & Darrozes, 2021), Sentinel-1/2 (land use/vegetation/SAR), and SMAP.
Deploy a pilot network of ground GNSS-R stations (Ha et al, 2021 ; Roussel et al., 2016) along the Mekong transect (Vinh Long - Ben Tre - Tra Vinh - Soc Trang) equipped with GNSS receivers, soil moisture/EC sensors, and weather units for satellite calibration.
WP2: Physics-Informed AI for Hydrological Continuum Reconstruction
Develop neural network architectures constrained by mass balance equations: 'TWS'='SM'+'SW'+'CW'+'GWS' .
Disaggregate surface soil moisture down to the root-zone (RZSM) and extract Groundwater Storage Anomalies (GWSA) to quantify sub-surface water depletion, drought (Hoang et al., 2025)
WP3: Derivation of Societal Risk Metrics & Digital Twin Prototype
State of the art
Translate physical outputs into operational risk indicators:
Agricultural Drought Index: Measuring root-zone water stress during key crop growth phases.
Soil Salinization Risk Model: Combining GNSS-R dielectric responses, river salinity, and groundwater levels.
Build a prototype Hydrological Digital Twin integrating real-time satellite/ground data feeds for dynamic scenario simulation.
WP4: Inter-Regional Validation & Transferability Analysis
Test model robustness on coastal sectors of the Red River Delta (Nam Dinh, Thai Binh, Ninh Binh, Hai Phong).
Quantify model uncertainties and identify universal vs. locally calibrated parameters across contrasting hydro-climatic settings.
Le profil recherché
Le candidat ou la candidate devra être titulaire d'un Master 2 ou d'un diplôme équivalent en télédétection, géosciences, hydrologie, géodésie spatiale, sciences de l'environnement, informatique géospatiale ou science des données appliquée à l'environnement.
Autre point fondamental un intêret ou une formation autour des enjeux et impacts sociétaux, et le développement durable est essentiel.
Une expérience préalable en GNSS-R, GRACE, télédétection radar ou modélisation
Compétences souhaitées :
traitement de données satellitaires ;
programmation en Python, MATLAB ou R ;
systèmes d'information géographique ;
analyse statistique et apprentissage automatique ;
notions d'hydrologie et de bilan d'eau ;
aptitude au travail interdisciplinaire ;
capacité de rédaction scientifique en anglais.
hydrologique constituera un atout.
Application link : https://edd-projets.utoulouse.fr/
Autre point fondamental un intêret ou une formation autour des enjeux et impacts sociétaux, et le développement durable est essentiel.
Une expérience préalable en GNSS-R, GRACE, télédétection radar ou modélisation
Compétences souhaitées :
traitement de données satellitaires ;
programmation en Python, MATLAB ou R ;
systèmes d'information géographique ;
analyse statistique et apprentissage automatique ;
notions d'hydrologie et de bilan d'eau ;
aptitude au travail interdisciplinaire ;
capacité de rédaction scientifique en anglais.
hydrologique constituera un atout.
Application link : https://edd-projets.utoulouse.fr/