Thèse Dryland Water Resources Under Global Change Lake And Reservoir Storage Dynamics Drivers And Water Management H/F Doctorat.Gouv.Fr

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 : Manuela GRIPPA ORCID 0000000248897975 Début de la thèse : 2027-09-01 Date limite de candidature : 2026-11-23T23:59:59 See English below


Water availability is one of the major challenges facing dryland regions worldwide (An et al. 2021). These environments are characterized by strong and intensifying climatic variability, high levels of aridity and water stress, and limited and unevenly distributed freshwater resources (Wang et al. 2018). At the same time, drylands support large and growing populations and sustain important agricultural and economic activities, placing increasing pressure on already scarce water resources.

Among these resources, lakes and reservoirs represent major surface-water stocks and play a critical role in water supply and regulation (Wang et al. 2025). They provide readily available water for domestic consumption, agriculture, livestock, industry, and ecosystem services, while reservoirs also allow societies to regulate water availability across seasons and during periods of drought (Messager et al. 2016, Li et al. 2023). However, changes in precipitation and evaporation, together with increasing water demand and intensified human water use and management, can substantially alter the amount and timing of water stored in lakes and reservoirs (Zhao et al. 2025, de Fleury et al. 2022).

Anticipating the future evolution of these water resources remains difficult because the spatiotemporal variability of surface water storage is still poorly characterized at the global scale. This is particularly the case in drylands, where water storage can vary considerably in space and time, and in-situ measurements are often scarce or inaccessible (An et al. 2021, Wang et al. 2018). In addition, hydrological responses to climatic and anthropogenic forcings can exhibit complex and non-linear behaviour in these regions, posing challenges for conventional modelling approaches (Peugeot et al., 2026).

Recent advances in remote sensing have recently provided new insights into lake water dynamics at the global scale (i.e. Xu et al. 2022, Lou et al. 2022, Yao et al. 2023, Zhao et al. 2025). However, the extent to which observed changes in water storage are driven by long-term climate change, shorter-term hydroclimate variability, and human activities remains inadequately resolved. Indeed, existing studies can be difficult to compare as they span different time periods, examine different parameters and metrics, report contrasting trends, and generally focus on a relatively limited number of medium- to large-sized lakes and reservoirs. Furthermore, the conventional binary approach of representing lakes as natural water bodies and reservoirs as managed systems has important limitations. Not all reservoirs are actively operated or managed, while some natural lakes are subject to substantial water withdrawals (e.g., de Fleury et al. 2022). In addition, existing lake and reservoir inventories remain incomplete or outdated, and some reservoirs-particularly in drylands-can be misclassified as lakes or absent from existing inventories due to recent construction (Wang et al. 2022; Lehner et al., 2024). These limitations introduce additional uncertainties when assessing the respective roles of lakes and reservoirs in regional water-resource dynamics.

The Surface Water and Ocean Topography (SWOT) satellite mission, launched in December 2022, offers an unprecedented opportunity to improve our understanding of these systems. By providing high-resolution, spatially extensive, and synchronous observations of both surface-water elevation and areas, SWOT enables new approaches for monitoring the dynamics of lakes, reservoirs, and rivers, including the vast number of medium-sized and small water bodies that have historically been difficult to observe systematically (Fu et al 2024). Initial studies using SWOT observations have already demonstrated promising capabilities for assessing water levels and storage changes across a variety of dryland environments (Buzzanga et al. 2025, Girard et al. 2025b, Masoumi and Tourian 2026, Peng et al. 2026).
SWOT data can also help characterize hydrological connectivity and hydrodynamic interactions between lakes and rivers, and estimate inflows to and outflows from lakes and reservoirs (Riggs et al. 2026, He et al. 2026).

When combined with complementary remote-sensing observations and environmental indicators, such as precipitation, evapotranspiration, land use, vegetation dynamics, SWOT can provide a more comprehensive and integrated view of dryland hydrological systems. Such a perspective is essential for understanding current water storage variability and broader hydrological behaviours, disentangling their climatic and anthropogenic drivers, and anticipating how freshwater resources may evolve under future global changes.
The main objective of this proposed thesis is to assess the impact of global change, including climate change, hydroclimate variability, and human water management, on surface water storage and fluxes across the Earth's drylands, where water scarcity, high hydrological variability, and growing pressure on freshwater resources pose major challenges to ecosystems, agriculture, and human societies.

To achieve this objective, the thesis will pursue two main research aims:

1 - Develop new approaches to quantify water storage variability and hydrological dynamics
We will combine frontier Earth-observation technologies, particularly those enabled by the recently launched Surface Water and Ocean Topography (SWOT) mission, with artificial intelligence and machine-learning methodologies to quantify changes in water stocks in lakes and reservoirs and characterize hydrological dynamics across dryland regions at the global scale.

2 - Identify the main drivers of hydrological variability
We will quantify the main climatic, environmental, and anthropogenic drivers responsible for observed hydrological behaviours across drylands. This will be achieved by jointly analysing climate variability; watershed and basin characteristics; and water management, including streamflow regulation, reservoir operation, irrigation, and other forms of freshwater use.
The proposed research adopts an integrated methodology that combines new-generation remote-sensing observations, statistical and ML approaches, and socio-hydrological analysis to quantify water storage variability in lakes and reservoirs, identify the main drivers of their hydrological behaviour, and offer new scientific support for monitoring surface water resources and informing sustainable water management across critical dryland regions worldwide.

The major methodological steps are described below:

1. Assessment of Surface Water Storage Variability Using SWOT

The first step is to develop a global-scale assessment of water-volume variations in lakes and reservoirs across the world's drylands using observations from the SWOT mission.
SWOT observations will be processed to derive accurate water-surface elevations and characterize temporal variations in lake and reservoir water levels. For lakes, the vector-based single-pass product, LakeSP, will be utilized. LakeSP organizes SWOT-observed water-surface elevations and associated diagnostic uncertainties into individual lakes larger than 0.01 km2 (1 ha), as in inventoried in the mission's Prior Lake Database (Wang et al. 2025). Because LakeSP can contain inherent measurement errors and outliers, we will employ the recently developed heuristic filtering approach of Trudel et al. (2026), to balance effective noise removal with preservation of seasonal hydrological dynamics.

To supplement LakeSP, SWOT Pixel Cloud (PIXC) data, which retain pixel-based measurements at their native high spatial resolution, will also be used following Girard et al (2025a), to characterize water bodies that are absent from the PLD and therefore not included in the SWOT standard vector lake products. Characterizing small water bodies that are absent from the standard vector lake products is particularly important in drylands. Lakes in these regions can be highly intermittent or ephemeral because of aridity and dynamic hydrological connectivity with groundwater and drainage networks (Pekel et al. 2016), while reservoirs may experience large operational fluctuations or may have been constructed only recently (Fan et al. 2024). These characteristics can result in water bodies being absent from the standard SWOT LakeSP products.

Following Girard et al (2025a), SWOT water-level observations will be complemented with multispectral optical and synthetic aperture radar (SAR) images to derive water surface areas. A growing suite of remote-sensing methods is available for this purpose, including the CNN U-net algorithm developed at GET for optical imagery (de Fleury et al. 2025); the optical lake area mapping method developed by Yao et al. (2019), which effectively recovers inundation areas obscured by cloud contamination; the size-adaptive Optical-SAR Pond Object Mapper (OptiSAR-POM), which is particularly suitable for charactering small-lake area dynamics (Liu et al., 2024); and more recent algorithms developed by the French THEIA expertise group for Sentinel-2 and Sentinel-1 imagery.

Combining water level and water-surface area will allow the development of elevation-area-volume hypsometric relationships and the estimation of temporal variations in water storage.

2. Water-Balance Assessment

The second step will focus on quantifying the main components of the water balance of lakes and reservoirs. Changes in water storage will be compared with estimates of major water fluxes, including inflows and outflows, precipitation, and evaporation.

Lake and reservoir inflows and outflows are critical for understanding the roles of lakes in buffering hydroclimatic variability and of reservoirs in regulating river discharge. We will estimate both inflows and outflows for lakes and reservoirs intersected by SWOT-observable rivers using the recently developed LakeFlow algorithm (Riggs et al. 2023, 2026). LakeFlow combines lake-river mass conservation with SWOT observations of (1) lake storage variations and (2) hydraulic variables of the associated upstream and downstream river reaches (e.g., river width, elevation, and slope) to optimize flow-law parameters, including baseflow cross-sectional area and channel roughness. These optimized flow-law parameters enable the estimation of the discharge entering and leaving individual lakes or reservoirs. Compared with conventional modeling approaches, LakeFlow is more directly grounded in satellite observations and has been tested and validated across lakes and reservoirs in the conterminous US (Riggs et al. 2026), demonstrating strong potential for application in dryland environment.

Precipitation estimates will be derived from satellite products such as GPM, while evaporation will be estimated using established remote-sensing methodologies and approaches developed in previous studies (i.e. de Fleury et al 2022).
This water-balance framework will provide an integrated assessment of the processes controlling changes in lake and reservoir storage.

3. Characterization of Hydrological Behaviour

The temporal evolution of lake and reservoir water storage will then be analysed to identify hydrological signatures and characterize distinct patterns of hydrological behaviour. These signatures will capture seasonal dynamics, including wet- and dry-season responses, as well as interannual variability and longer-term trends. de Fleury et al. (2022) applied a similar approach to a relatively small region in the central Sahel and were able to highlight water withdrawals from reservoirs and reveal hydrological connections among lakes, river systems, and associated groundwater tables.

Machine-learning and neural-network-based clustering approaches will then be employed to identify groups of lakes and reservoirs exhibiting similar hydrological behaviours. This resulting typology will provide a basis for comparing hydrological regimes across different dryland environments.

4. Identification of the Drivers of Hydrological Behaviour

Three main categories of drivers will be investigated:
- Hydroclimate drivers: precipitation, evaporation, temperature and drought.
- Watershed characteristics: river intermittency, land-use and land-cover characteristics, soil properties, watershed morphology, and drainage-network characteristics.
- Human regulation and water management: dams and reservoirs, water withdrawals, irrigation, flow regulation, upstream management, and other anthropogenic modifications of hydrological connectivity.

For the first two categories, particular attention will be given to characteristics that are especially relevant to dryland hydrology. In terms of climate, these include pronounced wet-dry seasonality and prolonged dry periods, which strongly regulate the timing and magnitude of hydrological processes. Flow intermittency represents another key characteristic of dryland systems (Fovet et al., 2022), with important implications for hydrological connectivity, biogeochemical processes, and ecosystem functioning. Finally, land-cover changes changes in soil properties at small spatial scales have been shown to strongly influence hydrological processes at the water basin scale (Pegeout et al., 2026).

The third category, human regulation and water management, is particularly important for understanding dryland water resources, but also presents a persistent challenge. Rather than relying on the conventional assumption that lakes represent natural systems whereas reservoirs represent regulated systems, the proposed research will explicitly evaluate multiple indicators and processes of human influence. This distinction is important and novel because natural lakes may also be strongly affected by water withdrawals, diversions, and upstream reservoir regulation. In other words, our proposed approach conceptualizes lakes and reservoirs not as isolated water bodies, but as integral components of connected catchment drainage systems.

To quantify the influence of human regulation and water management, we will integrate four complementary lines of data evidence and methodologies.

First, we will explore the recently developed Global Registry of Agricultural Irrigation Networks (GRAIN) (Suresh et al. 2026) to characterize the spatial distribution and intensity of irrigation infrastructure across global drylands. Irrigation canals play a very important role in sustaining agriculture in water-limited regions, including extensively irrigated basins such as the Amu Darya and Syr Darya. Canal network density and extent will therefore provide spatial indicators of irrigation intensity, surface water diversion, and potential consumptive water use.

Second, by leveraging expertise and collaboration between the two proposing institutions, we will explore the capability of SWOT to observe the hydraulic characteristics of major irrigation canals and estimate canal discharge as a direct measure of water diversion and withdrawal. Where canal dimensions and SWOT sampling permit, observations of water surface elevation, width, and longitudinal water-surface gradient will be used to investigate the potential for estimating canal discharge and its temporal variability, building on a recent study demonstrating this capability (Sharma et al. 2026). Such observations could provide direct constraints on the magnitude and timing of water diverted from natural drainage networks. For canals that are too narrow or otherwise insufficiently sampled by SWOT, canal density and connectivity will instead be incorporated as contextual indicators of irrigation and diversion intensity within the broader socio-hydrological analysis.

Third, we will incorporate estimates of human water use from integrated hydrological modelling frameworks, such as the Inter-Sectoral Impact Model Intercomparison Project (ISIMIP) (e.g., Frieler et al. 2024). These modelling products account for hydroclimatic conditions, irrigated cropland and crop characteristics, and other water-demand processes, and will be used to characterize the spatial and temporal patterns of water withdrawals and consumptive use across dryland drainage networks. In combination with the infrastructure-based indicators above, these estimates will allow us to examine how upstream water consumption and management propagate through drainage systems and influence downstream lake and reservoir storage dynamics.

Fourth, we will exploit the lake and reservoir water budget components derived in Method 2, including inflow, outflow, storage variation, and evaporative losses, to provide a process-based assessment of climatic and human controls on individual water bodies. Examining how these components covary through time will help distinguish different mechanisms of storage change. For instance, storage decrease associated primarily with reduced inflow or enhanced evaporation may indicate strong hydroclimatic control, whereas changes in outflow that are largely decoupled from inflow and lake-surface evaporation may reveal active regulation. Storage declines that cannot be explained by these water budget components may, in turn, indicate direct withdrawals or diversions from the lake or reservoir.

When combined with independent information on irrigation infrastructure and modeled water use, these water budget diagnostics will provide a complementary basis for attributing observed lake and reservoir dynamics to hydroclimatic forcing, consumptive water use, and direct water regulation.

5. Regional Evaluation and Validation

The methodology will be evaluated in selected well-documented dryland basins and benchmark sites, where independent observations and ancillary information are available for validation and verification.

- West African drylands, including sites associated with the AMMA-CATCH observatory in Senegal, Niger, and Benin (Galle et al 2018), as well as SWOT calibration/validation sites in, Burkina Faso and Niger (Girard et al 2025b), and the on-going SCO project VOQUALISE in Senegal, will provide valuable in-situ observations for validating SWOT-derived water storage and hydrological variables. Moreover, ongoingn collaborations with several collegues, particularly Roland Yonaba (2IE), will provide local knowledge and information on water management and operations at selected reservoirs.

- Central Asia drylands, particularly the Syr Darya and Amu Darya basins, will provide an important test bed for disentangling hydroclimatic variability from intensive human water management. Existing collaborations and research networks with Green University and TIIAME National Research University in Uzbekistan, the National Center of Space Research and Technology in Kazakhstan, the National Academy of Sciences of Tajikistan, and the SCO Central Asia Hydrology from Space (CAHYSPA) project may facilitate access to complementary information, potentially including water use datasets, in situ measurements of lake and river hydraulic variables, and/or socioeconomic and water-management information. These datasets and information will be valuable for assessing and validating the effects of irrigation, water diversion, and reservoir regulation on surface water dynamics.

- Western North American drylands, including the Great Salt Lake Basin and the Colorado River Basin, will provide complementary test regions where hydrometric observations and information on water regulation and use are comparatively abundant and openly accessible. Data from the U.S. Geological Survey (USGS), U.S. Army Corps of Engineers (USACE), and regional water management agencies can provide independent constraints on reservoir water level and storage, river discharge, reservoir operations, and water withdrawals.

These observations can be further complemented by modelling and data-assimilation products and toolkits, including NOAA's National Water Model (NWM; Cosgrove et al. 2024), which together provide a particularly data-rich environment for evaluating the proposed methodology and attribution framework.

Overall, the research will bring together, remote sensing, hydrology, machine learning, and the analysis of human water management. This combination will move beyond the characterization of water storage variability alone toward a socio-hydrological understanding of dryland water systems and will provide a step forward towards identifying the factors that control the resilience and vulnerability of dryland water systems.