Skip to content

In-situ observations

The importance of observations

In situ observations are the ground truth that turns data, models, and AI into trustworthy knowledge.

Observations are the foundation of environmental knowledge. Before data can be analysed, modelled, visualised or transformed into decisions, phenomena must first be observed and measured. In this sense, observations are not merely raw inputs: they are the empirical link between the real world and the scientific, technical and institutional processes that seek to understand it. This is particularly important for environmental systems, where changes are often local, dynamic and context-dependent.

In their article Time for in situ renaissance, Fekete et al. highlight the continuing importance of in situ observations in an era increasingly dominated by satellite data, numerical models and large-scale digital infrastructures. Remote sensing and modelling provide extraordinary spatial and temporal coverage, but they cannot replace direct measurements collected on the ground, in rivers, in soils, in the atmosphere, or within built and natural infrastructures. In situ observations remain essential for calibration, validation, uncertainty assessment and the detection of local processes that may otherwise remain invisible or poorly represented.

This role becomes even more critical in the age of Artificial Intelligence. AI systems can discover patterns, generate predictions and support decision-making, but their reliability ultimately depends on the quality, representativeness and contextual richness of the data on which they are trained and validated. In situ observations provide the ground truth needed to train models, assess their bias, evaluate their performance under real-world conditions and adapt them to local contexts. Without robust observational data, AI risks amplifying errors, producing plausible but unreliable outputs, or learning patterns that do not correspond to the actual behaviour of environmental systems.

A renewed attention to in situ observations is therefore not a return to the past, but a necessary condition for trustworthy digital environmental systems. Sensor networks, monitoring stations, field campaigns, citizen observations and IoT infrastructures can all contribute to richer and more reliable knowledge, provided that observations are properly documented, quality-controlled and made interoperable. For a workshop on observations, this means focusing not only on how data are collected, but also on how they are described, shared, preserved and reused. Good observations are the basis for reproducible science, reliable models, trustworthy AI and informed decisions.

Interoperability as a value

Missing interoperability leads to fragmented data, siloed knowledge, underexploited value and limited collaboration.

Interoperability is the ability of different systems, datasets and tools to work together seamlessly. In environmental science, interoperability is essential for enabling collaboration, data sharing, model integration and the reuse of knowledge across disciplines and domains. Without interoperability, data remain siloed, models cannot be combined, and insights cannot be shared effectively. Interoperability is not just a technical issue, but also a social and institutional one. It requires common standards, shared vocabularies, open data policies and a culture of collaboration. In the context of this workshop, interoperability means ensuring that observations are not only collected, but also described, documented and shared in ways that allow them to be discovered, accessed and reused by others. This involves using standards for data formats, metadata, APIs and ontologies, as well as adopting best practices for data management and sharing. Interoperability is the foundation for building a more connected, collaborative and impactful environmental science community.

In the IoT domain there are two drivers that led to a non-interoperable landscape lacking of catalogues and metadata.

  1. Specific business logic of sensor vendors
  2. Specific data package optimization approaches

The first driver is related to the fact that many sensor vendors have specific business logic that leads them to create proprietary data formats, APIs and platforms. This can be driven by the desire to lock customers into their ecosystem, to differentiate their products, or simply by a lack of awareness of interoperability standards.

The second driver is related to the fact that many sensor vendors optimize their data packages for specific use cases, such as low bandwidth, low power consumption or real-time processing. This can lead to the creation of custom data formats and protocols that are not compatible with existing standards.

Both drivers contribute to a fragmented landscape where data from different sensors cannot be easily integrated, shared or reused. To address this issue, it is important to promote the adoption of interoperability standards, to encourage open data policies and to foster a culture of collaboration among sensor vendors, researchers and practitioners in the environmental science community.