Sensors and data collection -

Sensors and data collection

To be relevant, reliable, and useful, the models developed in the other themes must rely on high-quality data, available in sufficient quantity for their design, calibration, and validation. However, in many contexts—particularly in countries of the Global South—such data are scarce, incomplete, or difficult to access due to economic, technical, or institutional constraints. The “Sensors and Data Collection” theme addresses this context by focusing on data production, which has become a major scientific challenge. It explores the opportunities offered by the emergence of connected, low-cost sensors that can be deployed at large scale, making it possible to design data collection strategies better adapted to local realities. The objective is to co-develop, with local partners, robust, sustainable, and accessible data acquisition systems, closely aligned with modeling needs. This theme therefore aims to bring together model design and measurement systems within an integrated approach, from data to model.

Scientific objectives

Development of sensors adapted to specific research and modeling needs in the Global South

Deployment of sensors

Training and appropriation

Data storage, sharing, and access

Development of sensors adapted to specific research and modeling needs in the Global South

Deployment of sensors

Training and appropriation

Data storage, sharing, and access

Scientific challenges

Developing open-source, low-cost sensors while ensuring reliability and reproducibility of measurements

The development of low-cost, open sensors faces a major challenge: guaranteeing the quality, reliability and reproducibility of measurements. This requires the definition of calibration protocols, validation methods and tools for fault detection or calibration, possibly in real time.
The challenge is to reconcile technological accessibility with scientific excellence.

Developing data assimilation techniques to integrate sensor data into simulation models in real time

Integrating sensor data into simulation models is a central challenge. It involves transforming local, one-off measurements into information that can be used on the scale of the models, which is often spatialized and continuous.
The theme develops data assimilation methods for dynamically coupling observations and simulations, with applications in crisis management and environmental monitoring.

Designing and integrating embedded models within sensors

The rise of embedded AI is opening up new prospects, but also posing major challenges. We need to design sensors capable of processing data locally (filtering, event detection, synthesis), in order to reduce the volume of data transmitted and improve its relevance.
This means developing efficient, computationally frugal models adapted to constrained devices.

Applications

Air quality and health

Urban ecology.

African languages

Optimized irrigation

Biosignals and health

Underwater bioacoustics