What do we want to learn from spatial data? Asking the right questions – challenges in radon mapping

Exposure to indoor radon is recognized as a health hazard, which is why regulation aimed to its reduction has been developed. One important tool of radon (Rn) policy are Rn maps. They serve (1) for visualizing the geographical distribution of the hazard and (2) as data base for regionalized legal measures, i.e. action which should be appropriate for the Rn situation at a place.

(1) has as objective information of stakeholders (the public, administrations, legislators, Rn professionals) about magnitude and location of the problem;

(2) Objective is decision support for identification of regions in which certain measures should be applied, to comply with Rn regulation.

These two objectives correspond to asking different questions and imply different maps, in general. For visualizing, isopleth or choropleth maps are usually considered adequate, the latter for assigning hazard scores to geographical units such as municipalities. On the other hand, identification of areas where certain action applies, amounts to classification of areas according to the necessity of that action.

While sharing certain steps, these two type of maps entail different technical challenges. They basically origin in the high spatial variability of the Rn hazard, usually quantified by indoor Rn concentration in buildings, its probability to exceed a threshold, or the collective hazard (i.e. sum over affected persons). Due to the multitude and different nature of physical control factors, the scale of variability extends from small-scale local to continental.

Level maps (objective 1) raises the question of resolution (a) wanted by the stakeholders and (b) achievable with data; this acts back to data acquisition, i.e. Rn surveying. Resolution is related to the appearance of maps in terms of roughness and noise. For class maps (objective 2), the critical question is, in addition, reliability of defining an area as target of certain action, in terms of sensitivity and specificity (or likewise of 1^st and 2^nd kind error probabilities) of a decision.

In this presentation, we shall give real-world examples of the objectives and resulting Rn maps. Further we shall describe estimation methodology suited to create maps that comply with the quality targets addressed above.