A geostatistical approach to data harmonization - Application to radioactivity exposure data

When data from different networks are merged for mapping purpose, biases may appear and lead to unrealistic results. In this paper we define a geostatistical model that generalizes the universal kriging model such that it can handle heterogeneous data. Multiple bias sources can be treated simultaneously through the notion of bias factors. The associated best linear unbiased estimation and prediction (BLUE and BLUP) formulas lead to both bias estimation and harmonized kriging. We illustrate the methodology with an example of country bias estimation in a radioactivity exposure assessment procedure. The application leads to a discussion on multicollinearity problems in data harmonization. This is an important issue that may impair results particularly when bias factors and natural drifts are correlated. Solutions for handling multicollinearity are suggested and further investigation directions proposed.