Logo

Machine-Learning Demonstrator – Supporting water resilience in Europe

Destination Earth > DestinE Uses > Machine-Learning Demonstrator – Supporting water resilience in Europe

Main contractor: Deltares

What is it? A machine learning (ML) Demonstrator for Water Resilience to deliver a transformative leap in how stakeholder access, understand, and act on climate and water data for the Sava River Basin. It combines Large Language Models (LLMs) with physics-based hydrological and water quality models, supported by an emulator where appropriate to accelerate simulations and using storylines simulations from the Climate Change Adaptation Digital Twin (Climate DT) to assess future climate impacts and adaptation options.

Concrete applications examples: Users can explore the impacts of of future scenarios on water quantity and water quality with an AI-assisted demonstrator for the Sava River Basin.

Main target end users: International Sava River Basin Commission (ISRBC), water managers, consultants, public, or journalists

Conceptual framework of an AI-driven workflow for exploring the impacts of future scenarios on water quantity and water quality in the Sava River Basin. Users define management and climate scenarios through a natural-language interface. These scenarios are translated into model inputs and evaluated using hydrological, water quality and climate simulations. The demonstrator then generates indicators and visualizations to support decision-making.

Climate DT data for water resilience

Europe is facing a growing water crisis. Climate change is intensifying floods, droughts, and pollution, while water managers are under pressure to make faster, more inclusive, and more transparent decisions. This DestinE ML Demonstrator for Water Resilience is a bold response to this challenge: an AI-powered demonstrator that transforms how regional water authorities, environmental agencies, and policy makers plan for resilience. 

DestinE ML Demonstrator for Water Resilience combines the power of Large Language Models (LLMs) with trusted physics-based hydrological and water quality models, supported by an emulator where appropriate to accelerate simulations. This hybrid system enables users to interact with complex simulations using natural language, configure “what-if” scenarios, and receive tailored, interpretable outputs quickly and intuitively. 

The “what-if” scenarios—such as changes in land use, nutrient loads, pollutant emissions, or infrastructure will be translated into adjustments of model schematizations, model parameters, nutrient loads, and pollutant emissions. These adjustments will be used to simulate the effects on water availability, water quality (e.g., nitrogen, phosphorus, PFAS), water temperature, and/or sediment dynamics. The simulations will utilize multi-model climate change scenarios available from DestinE, enabling adaptive water resilience planning. Storyline simulations of present and future climate from the Climate Digital Twin will be used to develop storylines that demonstrate risks and adaptation options to improve water resilience.

The overall modelling framework follows the DPSIR (Drivers–Pressures–State–Impact–Response) framework, in which hydrological simulations provide inputs for pollutant emission modelling, which in turn drives the water quality model to assess environmental impacts and support management decisions. 

Overview of the integrated catchment modelling framework that supports the DPSIR (Drivers–Pressures–State–Impact–Response) concept. Global and local datasets – including topography, meteorological, land use, socio-economic factors, and biogeochemical information – are used to drive the model chain starting with the Wflow hydrological model; its output serves as hydrological input for the D-Emissions model for estimating pollutant loads. These emissions are subsequently used by the DELWAQ water quality model (D-Water Quality) to simulate the environmental state and assess impacts on the water system. The framework supports the evaluation of management responses and policy development. 

The demonstrator fills a critical gap in the current landscape: the ability to interactively explore and interpret climate, water availability and water quality data in a user-driven, scenario-based manner. It complements existing services such as Copernicus and national hydrological platforms by offering a flexible, user-centric interface that supports local decision-making. 

Natural-language interaction makes advanced climate and hydrological modelling accessible to non-technical users, while DestinE climate projections support long-term adaptation strategies in the Sava River Basin and provide a methodology that may inform future applications in other river basins. In addition, scenario-based storylines help communicate future risks and adaptation options to policymakers and the public while AI-assisted workflows reduce the effort required to assess multiple future scenarios 

Machine Learning (ML) Demonstrator contracts aim at showing the added value of the DestinE Digital Twins, ML tools and the wider DestinE System. The contracts include specific actions for machine learning and artificial intelligence-based solutions as part of the implementation of ML/AI techniques within the Destination Earth initiative of the European Commission, led by DG CNECT, and implemented by ECMWF, EUMETSAT, ESA and over 100 partner institutions across Europe.