1. What is a digital twin?
A digital twin is a virtual representation of a physical object or system that reflects its evolution throughout its lifecycle. It is updated using real-time data and combines simulation, machine learning and reasoning to support decision-making.
Digital twins can be used to:
- Monitor real-world systems;
- Understand how they function;
- Simulate future scenarios;
- Support informed decision-making.
Key characteristics
1. Representing the real world
A digital twin can represent an individual object (such as a sensor or infrastructure), a system (hydrological, atmospheric or industrial), a geographical area or an entire environment.
2. A time-based perspective
A digital twin follows the evolution of a real-world system throughout its lifecycle, accounting for changes in how it is used, transformations in its environment and impacts related to water, energy, emissions or waste.
3. Continuous updates
Digital twins can be fed by in situ sensors, automated monitoring networks, satellite observations and historical environmental databases.
4. Simulation and machine learning
They combine physics-based models, numerical simulations and artificial intelligence techniques, including machine learning and neural networks.
2. How does a digital twin work?
A digital twin operates through a continuous feedback loop between the physical and digital worlds.
1. Data collection
Data are gathered from environmental sensors, field surveys, satellites and institutional databases.
2. System modelling
The real-world system is represented using physics-based models, scientific knowledge, and statistical or AI-based approaches.
3. Data assimilation
New observations are incorporated into the model to update its representation of the system, correct discrepancies with real-world conditions and account for uncertainties.
4. Simulation and scenario analysis
The digital twin can simulate different scenarios, such as climate change, pollution or land-use planning, and generate indicators to assess potential outcomes.
5. Feedback loop
Simulation results are compared with real-world observations. These comparisons help refine the models and inform future updates.
3. Where do the data come from?
Data are the foundation of any digital twin.
Raw data
These are collected directly from sensors measuring variables such as temperature, humidity, pollution or water flow, as well as from field instruments and automated monitoring networks.
Processed data
Raw data are cleaned, harmonised and prepared for use in models. They can then be used in simulations and AI-based analyses.
Main data sources
- Scientific field surveys and measurement campaigns;
- Environmental sensor networks and research infrastructures, such as PIVOTS;
- Satellite observations, including data from the European Copernicus programme;
- Public and institutional databases.
4. When does a model become a digital twin?
A model becomes a digital twin when it is continuously connected to the physical system it represents.
1. Connection to real-world data
The model receives continuous streams of observations and measurements.
2. Dynamic updates
It evolves as new information becomes available, rather than remaining a static representation.
3. Interaction with the real world
It can simulate scenarios, generate indicators and provide insights to support decision-making.
An intermediate concept is the digital shadow: a digital model that is automatically synchronised with its physical counterpart but does not yet provide a complete feedback and interaction loop.
5. What are environmental digital twins used for?
Environmental digital twins help us understand complex natural systems, anticipate how they may change and explore different scenarios without directly intervening in the real world.
They can be used to monitor environmental conditions, analyse interactions between natural processes and human activities, and support decisions about environmental management.
- Improving water management and flood risk prevention
Digital twins are increasingly being developed to represent rivers and river basins. In France, the Jumeaux numériques du fleuve (River Digital Twins) project aims to improve our understanding of river dynamics and support land and water management in the Gironde estuary.
Internationally, initiatives such as China's Digital Twin River Basins use simulation to model water flows, anticipate floods, optimise water resource management and improve risk prevention.
- Understanding and anticipating climate change
At a global scale, digital twins can help represent and simulate the complex processes that shape the Earth system.
The European Destination Earth (DestinE) initiative is developing digital twins to simulate climate processes and improve our ability to anticipate extreme events such as droughts, floods and heatwaves.
Research projects such as Digital Twin Antarctica explore interactions between climate, ice, oceans and geophysical processes, helping scientists better understand Antarctica's future evolution and its potential global consequences.
- Supporting the transition towards more sustainable agriculture
In agriculture, international organisations such as the FAO (Food and Agriculture Organization of the United Nations) and CGIAR, a global agricultural research network, are developing modelling and simulation tools that combine satellite observations, field sensors, and agricultural and climate models.
These approaches can help improve farming practices, manage natural resources more efficiently, strengthen food security and adapt agricultural systems to changing climatic conditions.
Towards integrated environmental systems
Beyond these individual applications, many initiatives aim to develop digital systems that can represent water, soils, the atmosphere, ecosystems and human activities together.
The goal is to move beyond a fragmented understanding of the environment towards a more comprehensive view of the interactions between human societies and natural systems.
However, one major challenge remains: interoperability. Developing these integrated systems requires data, models and tools to be compatible, so that information can be shared and used across different platforms and scales.
6. What are the main challenges?
Developing and operating digital twins involves several significant challenges.
Diverse data sources: combining information from multiple sources, often with different formats, levels of accuracy and measurement uncertainties.
System complexity: representing interconnected processes, non-linear interactions and phenomena occurring across different spatial and temporal scales.
Continuous updates: integrating new observations, managing changes in model performance over time and maintaining stable machine learning systems.
Costs and environmental impact: managing computational demands, energy consumption and the environmental footprint of digital technologies.
Accessibility and practical use: making results understandable to decision-makers and turning complex data into useful information that can guide action.
7. Resource-efficient AI: why and how?
Why does it matter?
AI models used in digital twins can require substantial computing power, consume large amounts of energy and depend on extensive datasets.
What is the goal?
To develop AI approaches that are more resource-efficient, effective and better suited to environmental constraints.
How can this be achieved?
- Simplifying models;
- Selecting only the most relevant data;
- Optimising computational processes;
- Reducing unnecessary duplication.
What does this mean for JUNON and digital twins?
By adopting more resource-efficient AI approaches, JUNON aims to reduce the environmental footprint of digital twins, make simulations more efficient and improve access to these technologies.