Validation
Validation is the bridge between map production and map credibility. Yet, it is not a one-size-fits-all procedure and the validation protocol must be designed in relation to the intended use of the map and integrated into project planning from the outset. The choice of validation approach, metrics, and reporting standards should therefore be defined before any map is produced. This is increasingly relevant with the growing accessibility of large-scale computing infrastructure and ML methods, which have dramatically lowered the barrier to generating new maps.
This section focuses on the principles and best practices of design-based map validation, which forms the foundation of current international guidelines for geospatial accuracy assessment [1], [2]. Under this framework, validation relies on reference data collected through probability sampling designs with known, non-zero inclusion probabilities, enabling statistically rigorous and unbiased estimation of map accuracy. While many of these concepts originated in land-cover validation, they remain broadly applicable across a wide range of geospatial products.
The discussion applies to both categorical and continuous map products. Categorical products include multi-class land-cover maps as well as thematic single-class products such as forest extent or forest change maps. Continuous products include variables such as biomass and forest height. Although the general validation principles are shared, specific aspects of response design and accuracy assessment may differ across product types.
We begin by clarifying several common misconceptions surrounding map validation, including the distinction between model evaluation, map uncertainty, map comparison, and map validation. We then present the practical workflow of design-based accuracy assessment, covering sampling design, response design and statistical analysis.
Design-based accuracy assessment relies on reference data collected through probability sampling, which enables statistically rigorous and unbiased estimation of map accuracy. However, probability sampling is not feasible for all mapping applications. In cases where validation must rely on existing non-probability reference data (e.g., opportunistic field campaigns or legacy datasets), alternative validation approaches are required. These are discussed in Section Validation beyond probability sampling, together with recommended best practices for situations in which probability-based, design-based accuracy assessment cannot be implemented.