Sampling for non-destructive spectroscopy with a particular focus on agriculture, food and feed

Sampling plays a pivotal role in the analytical process, particularly when employing non-destructive spectroscopic sensors (NDSS). This review bridges Theory of Sampling (ToS) and Design of Experiments (DoE) to address sampling challenges in NDSS agri-food applications. Sampling quality is the primary driver of overall uncertainty, often significantly surpassing laboratory and instrumental errors. Non-destructive spectroscopic setups inherently sample through their optical configurations. We highlight the importance of replication methods to determine sources of variance, particularly in physical sampling procedures, and provide practical guidelines for achieving representative sampling. Additionally, the review briefly discusses computational augmentation and resampling techniques. Practical considerations and case studies from food and feed applications illustrate the constraints and solutions for effective sampling, providing insights for researchers and industry aiming to optimize NDSS measurements. • Sampling related both to Theory of Sampling and Design of Experiments. • Sampling quality (or its lack) is main contributor to overall uncertainty. • Non-destructive spectroscopic setups sample by their optical configuration. • Replication plans for distinguishing sources of variance uncertainty, such as physical sampling procedures. • Focus on particular practical considerations related to sampling for non-destructive spectroscopic measurements.