QuadMap: Variable resolution maps to better represent spatial uncertainty
José Padarian, Alex B. McBratney
Abstract
Uncertainty assessment is an integral component of spatial modelling not only from a analytical point of view but also as a communication tool. However, end users find uncertainty maps difficult to perceive alongside the prediction map. A common misconception is that finer resolution maps necessarily have higher precision. Here, we present an approach to take advantage of users’ perceptions of the resolution-quality relationship by incorporating prediction and uncertainty into a single, variable resolution digital map where the uncertainty is encoded as the pixel size. We use the quadtree algorithm to recursively partition the original map, aggregating pixels with uncertainty greater than a target threshold. In the resulting maps, users can immediately see where the uncertainty is large since it corresponds to coarser “pixelated” areas. This approach is not only useful to visualise and communicate uncertainty but it could be extended to integrate the quadtree into analytical workflows.