Thèse Jumeaux Numériques des Systèmes Socio-Écologiques dans les Systèmes de Courants de Bord est Incertitudes et Mécanismes de l'Impact du Changement Climatique sur les Ressources Marines du Chil 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 : CECI - Climat, Environnement, Couplages et Incertitudes / CERFACS Direction de la thèse : Boris DEWITTE ORCID 0000000338178691 Début de la thèse : 2027-08-14 Date limite de candidature : 2026-11-23T23:59:59 Eastern Boundary Upwelling Systems (EBUS), such as the Humboldt Current System off Chile, are among the most productive marine ecosystems in the world and support valuable fisheries and coastal communities. They are particularly exposed to climate change and interacting stressors, including ocean warming, deoxygenation, and acidification. These stressors can interact in complex and nonlinear ways, affecting circulation, habitat availability, trophic interactions, species distributions, and marine-resource biomass. Yet major uncertainties remain in translating regional environmental changes into ecosystem and resource responses, partly because climate, biogeochemical, ecological, and socio-economic processes are traditionally studied separately.

This PhD will develop an integrated digital twin of the Northern/Central Chilean coastal ecosystem to bridge these domains and trace climate signals from ocean dynamics and biogeochemistry through ecosystem processes to marine-resource responses and management. The framework will couple the regional ocean model CROCO, its biogeochemical component BioEBUS, and Atlantis, a state-of-the-art deterministic, whole-of-ecosystem, end-to-end model integrating biophysical, ecological, economic, and social components. This approach will provide a process-based representation of the physical-biogeochemical-ecosystem cascade, enabling uncertainty propagation into ecosystem and biomass projections and the development of the first ensemble of regional projections of marine-resource dynamics for this highly productive region.

The research has three interconnected objectives. First, the PhD will develop, calibrate, and validate an Atlantis model for Northern/Central Chile and quantify how uncertainty from natural ocean variability, initial conditions, and ecosystem parameters propagates through the digital twin. Second, it will use extreme climate events, particularly El Niño, as storylines to investigate mechanisms linking physical and biogeochemical changes to trophic dynamics and marine-resource responses, with emphasis on nonlinear responses, ecological memory, hysteresis, and recovery. Third, the digital twin will be driven by regional climate projections to generate ensemble trajectories of bentho-pelagic ecosystem dynamics under different levels of global warming. These simulations will identify thresholds and potential regime shifts and enable controlled what-if experiments comparing alternative fisheries pressures and management strategies.

The project will combine advanced numerical modelling, high-performance computing, ecological diagnostics, and uncertainty quantification. Particular emphasis will be placed on distinguishing robust climate-driven ecosystem responses from uncertainty arising from natural variability and model formulation, while identifying mechanisms underlying changes in biomass, trophic structure, and ecosystem function.

The PhD will be conducted within an international collaborative environment involving CSIRO in Australia, the University of Toulouse/CECI-CERFACS, and the Universidad de Concepción, and will benefit from the CLAP and SAFESEA programmes under the auspices of the United Nations Decade of Ocean Science for Sustainable Development. Beyond its regional focus, the project will provide a benchmark framework for integrated climate-ecosystem projections in other Eastern Boundary Upwelling Systems. Its outcomes will support climate-resilient marine-resource management by linking environmental change to ecosystem and fisheries responses and informing science-based adaptation strategies. Overall, the project will train the next generation of researchers in the tools, concepts, and interdisciplinary skills needed to address complex sustainability challenges and advance integrated environmental and socio-ecological sciences. Eastern Boundary Upwelling Systems (EBUS), such as the Humboldt Current System, are among the most productive and dynamic marine ecosystems worldwide, sustaining high biodiversity and globally important fisheries. These systems are increasingly exposed to multiple stressors associated with climate change (CC), including ocean warming, deoxygenation, and acidification, often referred to as the 'deadly trio (see Figure 1 in the Appendix). These drivers are expected to interact in complex and potentially nonlinear ways, altering ecosystem structure, trophic dynamics, and species distributions (Sampaio & Rosa 2020). This is particularly critical in so-called oxygen minimum zones (OMZ), which are extended regions of acidic waters with very low-oxygen concentration, often found along the western coast of the continents, also known as Eastern Boundary Upwelling systems (Chavez & Messié 2009, Seibel 2010). One such system is the Northern Chilean Upwelling System (NCU, 18°-32°S. See Figure 2 in the Appendix), where coastal upwelling transports cold, nutrient rich waters to the euphotic zone (Escribano et al., 2003; Montecino et al., 2009), sustaining high biodiversity (Buchan et al., 2024), primary productivity and fisheries central to Chile's economy (e.g., anchovy, sardine, squid, hake). Furthermore, this region is strongly forced by the El Niño-Southern Oscillation (ENSO), the dominant mode of interannual climate variability, which drastically alters local environmental gradients and can amplify the ecosystem's vulnerability to global climate stressors.

At the global scale, climate-driven ecological responses are already widespread, but their rates and directions vary strongly among regions and taxa, demonstrating that biological exposure cannot be inferred from a single global mean trend (Burrows et al., 2011, Poloczanska et al., 2013, Gruber et al., 2021). Based on a variety of ecological metrics, previous studies have shown that CC-induced patterns of shifts vary across ocean regions. Although ocean pH is declining globally, high-latitude regions experience more rapid and severe acidification (Fabry et al., 2009, Jiang et al., 2023). Similarly, while global oxygen inventories are declining, certain localized regions have shown unexpected oxygenation trends due to complex changes in wind stress and ocean circulation (Levin, 2018; Oschlies et al., 2018). Furthermore, warming patterns are highly non-uniform, with coastal and marginal seas frequently warming at rates significantly higher than the global open ocean average (Lima & Wethey, 2012). In particular, a recent regional analysis using climate velocity, an indicator of the rate and direction of climatic change, shows that the NCU diverges from broader southeastern Pacific trends across all three components of the deadly trio (Yévenes-Vega et al., 2025). While most of the southeastern Pacific is projected to experience strong deoxygenation and pronounced pH decline, the NCU shows weaker deoxygenation signals, even localized oxygenation trends, and reduced rates of pH decline, together suggesting a potential climatic refugia for these biogeochemical variables. Warming, however, does not follow this buffered pattern: the NCU remains exposed to significant thermal stress, with projected warming trends likely associated with large-scale climate variability patterns resembling El Niño conditions. This contrast illustrates how the three stressors can respond divergently within the same system, and highlights that any apparent buffering capacity should not be interpreted as reduced overall vulnerability. At the same time, changes in the vertical and horizontal structure of the OMZ could significantly constrain habitat availability and drive species redistribution, altering competitive dynamics among taxa with differing hypoxia tolerance.

Despite progress in understanding ecosystem sensitivity to CC, major gaps still prevent a quantitative estimate of vulnerability. First, global climate models have severe biases (Concha et al., 2024, Almendra et al., 2024), partially related to their coarse resolution and their limited representation of coastal upwelling dynamics and key non-linear processes (Gruber et al., 2011), constituting a clear limitation to assess more direct effects on the ecosystems. How regional scale biogeochemical shifts translate into fine scale habitat change, and how that change propagates into coastal bento-pelagic food webs, fisheries, and coastal communities, remains poorly understood. This gap is particularly critical in coastal embayments, where fisheries have historically developed and which have become hotspots for socioeconomic development in coastal regions, such as the NCU, which will serve as the case study for this project. Integrated modeling tools, also referred to as digital twins (Bauer et al., 2021, Bronner et al., 2023), offer a promising approach to overcoming these limitations by providing a high-resolution representation of ocean circulation and a more realistic representation of interactions among downstream components, from biogeochemistry to population dynamics. They also provide a framework for investigating the sensitivity of complex socio-ecological systems by incorporating environmental forcing that is poorly represented in current-generation global coupled models, particularly the natural variability associated with submesoscale to mesoscale ocean dynamics. However, they introduce a new challenge: determining how uncertainty in the physical model propagates through the ecosystem models. This issue is critical for building confidence in projections with these new tools and remains largely undocumented (Payne et al., 2016).

To address this gap, we propose a process-based framework for understanding the propagation of uncertainty across the physical-biogeochemical-ecosystem cascade in a digital twin of a coastal marine system, from ocean dynamics to ecosystem dynamics. Addressing this challenge is essential for increasing confidence in projections of marine ecosystem change and providing local stakeholders with more robust, quantitative climate information to support effective adaptation. The project will focus on the coastal region of Northern/Central Chile, a biodiversity- and resource-rich hotspot exposed to ocean warming and deoxygenation. Methodologically, the project will provide a benchmark framework for uncertainty quantification in integrated digital twins of socio-ecological systems, while delivering end-to-end regional ocean projections for central Chile. These projections will provide actionable information to support decision-making and adaptation strategies in response to ongoing ocean warming, deoxygenation and acidification.

The project builds on cutting-edge community modeling developments, combining the regional ocean circulation model CROCO (https://www.croco-ocean.org) with the BioEBUS biogeochemical model (Gutknecht et al., 2013), specifically tuned for the Humboldt system (Montes et al., 2014), and the end-to-end Atlantis ecosystem model (Fulton et al., 2011), widely recognized as a state-of-the-art framework for marine ecosystem modeling (Geary et al., 2020). The project also benefits from a strong international context, including UN Ocean Decade actions and networks supporting the United Nations Sustainable Development Goals (SDGs), such as CLAP and OPDCC, as well as established collaborations with partners in France, Chile, CSIRO, and the Centre for Marine Socioecology at the University of Tasmania. These partnerships will provide valuable platforms for disseminating the project's outcomes across multiple scales, from local stakeholders to policy-makers. Overall objective:

To develop and apply a process-based digital twin of the Northern/Central Chilean coastal ecosystem to quantify how climate variability and climate change propagate through the physical-biogeochemical-ecosystem cascade, and to translate these impacts into information supporting climate-resilient marine-resource management.

Specific objectives:

1. Quantify and constrain uncertainty in the digital twin.

Develop, calibrate, and validate an end-to-end Atlantis model coupled to regional CROCO-BioEBUS simulations, and quantify how uncertainty arising from ocean variability, initial conditions, and ecosystem parameters propagates from physical forcing to ecosystem and biomass responses.

2. Understand ecosystem responses to climate variability and extreme events.

Use ENSO and decadal variability as climate storylines to identify the physical and biogeochemical mechanisms driving changes in trophic interactions, biomass, and ecosystem structure, including potential nonlinear responses, ecological memory, hysteresis, and recovery.

3. Project ecosystem and marine-resource responses under climate change and assess adaptation options.

Generate ensemble projections of bentho-pelagic ecosystem dynamics under different levels of global warming, identify nonlinear responses, thresholds and potential regime shifts, and use the digital twin as a virtual laboratory to compare fisheries-management strategies and support climate-resilient adaptation. The project combines advanced numerical modelling tools that form the foundation of the digital twin with analytical approaches and diagnostics developed in ecological science. This integration enables connecting physical and biogeochemical environmental forcing with ecosystem dynamics, quantifying changes in trophic interactions and biomass, and assessing the ecological consequences of multiple climate-related stressors.

1) Hydrodynamical and biogeochemical models

The hydrodynamical modeling is based on the community regional ocean model CROCO (https://www.croco-ocean.org) originally based on the Regional Ocean Modeling System (ROMS) (Shchepetkin & McWilliams 2005). The model is coupled to the biogeochemical model BioEBUS (Gutknecht et al., 2013), a nitrogen-based model specifically designed to investigate oxygen minimum zones'variability in EBUS. A carbon module is considered to access estimates of pH, besides oxygen. A configuration of this coupled model have been developed and validated for the Humboldt system extending from 12°N to 40°S with a 7 km horizontal resolution (Dewitte et al., 2012, Vergara et al., 2016, 2017, Pizarro-Koch et al., 2019, 2023, Ortiz-Castillo et al., 2025), which will serve as a benchmark for this project. To simulate future conditions, the pseudo-global-warming (PGW) approach (Brogli et al., 2023) will be used, which consists of adding the difference between future and present climatologies calculated over selected time slices to the boundary forcing of the model in an hindcast mode. A first version of this product is available using the outputs of the NCAR CESM1-LENS ensemble mean (Kay et al., 2015) as forcing boundary conditions (Llancaleo et al., 2026). This approach has demonstrated skill in correcting known biases in global Earth System Models in this region, resulting in a more physically consistent representation of projected climate change patterns (see Figure 3 in the Appendix). In this project, we will also generate regional projections using selected members of the CESM1 Large Ensemble to explore contrasting trajectories of environmental forcing. These simulations will be conducted within the framework of the recently funded Horizon Europe project SAFESEA, which will serve as a host programme for this PhD project and also focuses on the central Chile region as a case study for investigating ocean processes associated with long-term acidification. In addition, a higher-resolution (3 km) configuration of the biogeochemical model will be implemented for central Chile using a nested approach, allowing us to assess the sensitivity of the results to model resolution.

2) End-to-End Atlantis Model Setup

We will use the Atlantis model. Atlantis is a deterministic, whole-of-ecosystem, end-to-end model that dynamically couples biophysical, ecological, economic and social components to explore ecosystem dynamics and evaluate alternative management strategies (Geary et al., 2020). A spatially explicit, 3D configuration for the bento-pelagic ecosystem of the NCU (18°-32°S) will be developed, relying on previous works by Dr. Fulton (Fulton et al., 2011, Audzijonyte et al., 2019, see Short Bio of Dr. Fulton in Appendix). As a comprehensive whole-of-system framework, Atlantis enables a holistic and dynamic evaluation of marine ecosystems and provides a computational platform to project environmental and anthropogenic changes, from trophic food-web interactions to alternative fisheries management strategies. The physical environment from the ocean model (see above) will be dynamically coupled to the model using temperature, dissolved oxygen, pH and the 3-D velocity field. Parameterization of the biological and ecosystem models will be grounded in empirical observations from field surveys and long-term monitoring programs led by University of Concepcion (collaboration with Dr. Parada, see short Bio in Appendix). Existing datasets generated through research projects conducted by the University of Concepción and CECI/CERFACS will be compiled and integrated to derive biological, physiological, metabolic, trophic, and environmental parameters, providing observational constraints for the representation of key processes and functional relationships within the ecosystem models.
Functional groups and their predator-prey interaction matrix, together with consumption rates, maximum growth rates, assimilation efficiencies, linear mortality, and recruitment parameters, will be defined to characterize the ecosystem dynamics. Climate-sensitivity parameters-including thermal sensitivity, temperature-dependent respiration rates, pH sensitivity, and oxygen-limitation mortality-will be derived from available physiological datasets. These observations will be systematically compiled and harmonized, and complemented by data available through collaborative research networks associated with the CLAP programme. The resulting empirically constrained parameters will then be incorporated into the ecosystem modelling framework, enabling a more accurate calibration and fine-tuning of the digital twin.

3) Diagnostics

Climate signals will first be characterized from the regional ocean model simulations and climate projections to identify the magnitude, spatial distribution, and trajectories of environmental change across the study region. For key physical and biogeochemical variables, including temperature, dissolved oxygen, and pH, we will quantify changes in mean conditions and extreme events such as marine heatwaves, together with climate velocity (Loarie et al., 2009) and climate state (Burrows et al., 2014, Yévenes Vega et al., 2025). Climate velocity will quantify the rate and direction at which environmental conditions shift, expand, contract, or disappear in km yr¹, whereas climate state will classify individual grid cells according to the trajectories and residence times of climatic conditions into areas of climatic conservation, migration, or loss. Together, these metrics will provide a spatially explicit diagnosis of the climate signals used to force the ecosystem component of the digital twin.
The Atlantis ecosystem model will then be used as the biological component of the digital twin to translate these climate signals into ecosystem responses and to diagnose the individual and combined effects of warming, deoxygenation, and acidification. Ecosystem responses will be evaluated through complementary ecological indicators, including changes in relative biomass, trophic structure and diet composition, predator prey interaction strength, functional group distribution, average body size, fish condition, and recovery following extreme events. Response type and state of concern (Pethybridge et al., 2020) will be used to characterize the sensitivity of functional groups and food web components to climate forcing, while recovery time (Babcock et al., 2019) will quantify the persistence and propagation of perturbations through trophic pathways. Changes in body size and fish condition will provide additional diagnostics of climate mediated changes in trophic and foraging processes (Audzijonyte et al., 2019, Fulton et al., 2024).
Finally, we will explicitly quantify how uncertainty in the climate signal propagates from the physical ocean to biological and ecosystem responses within the digital twin. Ensemble experiments will account for two major sources of uncertainty: uncertainty in poorly constrained biological and ecological parameters and uncertainty associated with natural ocean variability and the initial ocean state. For the latter, simulations will be initialized under contrasting mesoscale states, thereby sampling variability arising from the chaotic nature of ocean circulation and its influence on the climate signals transmitted to the ecosystem model. The resulting ensembles will be used to quantify the spread and growth of uncertainty from physical and biogeochemical forcing to ecosystem responses (Payne et al., 2016), allowing us to distinguish robust climate driven ecosystem responses from those that are sensitive to model parameterization, initial conditions, or natural variability.

Numerical and Ecosystem Modelling: Proven experience in ecological, hydrodynamic, or biogeochemical modelling. Previous experience with end-to-end ecosystem models is highly desirable.

Data Analysis and Programming: Proficiency in programming languages such as R, Python, MATLAB, or C++ for processing large-scale climate and biogeochemical datasets (e.g., NetCDF files).

Oceanographic Knowledge: Solid understanding of physical oceanography and physical-biological interactions, particularly in Eastern Boundary Upwelling Systems (EBUS).

Transdisciplinary Approach: Ability to integrate biological, physical, and socio-economic information to assess socio-ecological vulnerability and evaluate fisheries management and climate-adaptation strategies.

Communication and Teamwork: Excellent written and oral communication skills, with the ability to present complex scientific concepts clearly to both specialist and non-specialist audiences. The candidate should be comfortable working in an interdisciplinary and international team, engaging constructively with researchers from different backgrounds.

Interpersonal Skills and Sociability: Strong interpersonal skills, an open and collaborative attitude, and the ability to build effective working relationships within a diverse research environment.

Application link : https://edd-projets.utoulouse.fr/

• Python
• C++
• MATLAB