From FAIR Principles to Fitness-for-Purpose: Document-Driven, Application-Oriented Data Quality in Agrosystem Research

Making research data Findable, Accessible, Interoperable, and Reusable (FAIR) is widely recognised as essential for open and reproducible science. However, researchers often face a gap between FAIR-compliant datasets and data that are actually fit for specific scientific or operational applications. This gap arises because data quality is inherently application-dependent, while critical assumptions, limitations, and uncertainty characteristics are frequently documented only implicitly across publications, dataset metadata, and workflow descriptions. 

We present a document-driven, application-oriented approach to data quality assessment developed within the FAIRagro initiative. 
The method uses the \textbf{Application-Specific Data Quality Matrix (AS-DQM)}, which systematically captures reasoning linking documented data characteristics—such as spatial and temporal resolution, validation strategies, and known limitations—to application requirements and explicit fitness-for-Purpose statements (\href{https://zenodo.org/records/17981173}{FAIRagro resources}). Rather than computing new quality metrics, the AS-DQM formalizes existing knowledge already generated by research communities, reduces barriers to adoption, and supports responsible data reuse. 

The approach is illustrated using a Germany-wide phenology time series as a pilot example. By analysing dataset documentation together with a concrete phenology-based scientific studies, the AS-DQM demonstrates how application-specific quality requirements—such as acceptable temporal uncertainty, spatial aggregation assumptions, and suitability for regional-scale analyses—can be systematically extracted and made explicit. Comparing the resulting application-level quality profile with the dataset-level documentation shows how fitness-for-Purpose emerges from the interaction between data characteristics and application context, highlighting cases where datasets are conditionally suitable or explicitly unsuitable for specific analyses. 

We discuss strengths, limitations, and adoption challenges of document-driven, application-oriented data quality reasoning, emphasizing its broad relevance across Earth and environmental sciences and its role in fostering sustainable, community-driven FAIR data practices.