Data Quality Characteristics for Improved Metrology in Sensor Networks

Traditional calibration methods are often impractical due to physical inaccessibility and cost constraints, making it difficult to guarantee measurement traceability and uncertainty. In our recent paper, presented at the 2025 IMEKO TC-6 International Conference on Metrology and Digital Transformation (M4DConf) in Benevento, Italy, we address these challenges by proposing the use of well-known data quality characteristics in a metrologically sound manner for assessing data quality in sensor networks.

The main objective is to enable robust data quality assessment in sensor networks, even when reference data or traditional calibration is unavailable. Our approach draws on recent standards – specifically ISO/IEC 5259-2 and ISO 8000-210 – to identify and apply the most relevant data quality characteristics from a metrological perspective. 

After reviewing the standards and literature, we identified six key data quality characteristics as most relevant for metrology in sensor networks:

• Accuracy: How closely data reflects the real world, even in the absence of reference sensors.
• Traceability: The ability to track the origin and modifications of data, supported by comprehensive metadata. Also utilizing alternative calibration methods to estimate uncertainty with some degree of traceability.
• Completeness: The extent to which all required data points are present, as missing data can severely impact reliability.
• Consistency: The degree to which data remains stable over time and across sensors, helping to detect drift or faults.
• Precision: Both measurement and representational precision, ensuring that data is recorded and stored with appropriate detail.
• Timeliness: The availability of up-to-date data, with correct timestamping being crucial for subsequent analyses.

These characteristics form the foundation for a metrologically sound assessment of sensor data quality, especially in networks where traditional calibration is not feasible.

Another central theme of the paper is the importance of metadata. By systematically recording sensor specifications, measurement methods, uncertainty, and data provenance, metadata can compensate for the lack of direct traceability and support automated, efficient quality assessments.

Not only is it necessary with novel methods for processing sensor network data to estimate the data quality. It is also necessary to develop guidelines for how to use and what metadata is needed in different use cases. Furthermore, alternative calibration methods can be used to improve trustworthiness as well as alternative approach to collecting sensor data. 

FunSNM includes five uses cases for applying developed methods to real world sensor networks addressing sensor network data quality in different ways. We hope these insights will support the development of more reliable, transparent, and efficient sensor networks across diverse applications.