From Natural Resources Evaluation to Spatial Epidemiology: 25 Years in the Making

When in the winter of 1994, under the supervision of my post-doc adviser André Journel, I started writing "Geostatistics for Natural Resources Evaluation" in the bedroom of a tiny Palo Alto apartment, little did I know that 25 years later I would be conducting NIH-funded research on medical geostatistics from a lakefront office nestled in the Irish Hills of Michigan. The professional and personal path that led me to trade the mapping of heavy metal concentrations in the topsoil of the Swiss Jura for the geostatistical analysis of cancer data was anything but planned, yet André's help and guidance were instrumental early on. Looking back, shifting scientific interest from the characterization of contaminated sites to human health made sense as the field of epidemiology is increasingly concerned with the concept of exposome, which comprises all environmental exposures (e.g., air, soil, drinking water) that a person experiences from conception throughout the life course. Although both environmental and epidemiological data exhibit space-time variability, the latter has specific characteristics that required the adaptation of traditional geostatistical tools, such as semivariogram and kriging. Challenges include: (i) the heteroscedasticity of disease rate data (i.e., larger uncertainty of disease rates computed from small populations), (ii) their uneven spatial support (e.g., rates recorded for administrative units of different size and shape), and (iii) the limitations of Euclidean metrics to embody proximity when dealing with data that pertain to human mobility. Most of these challenges were addressed by borrowing concepts developed in adjacent fields, stressing the value of interdisciplinary research and intellectual curiosity, something I learned as a fresh PhD in agronomical sciences joining André's research group at the Stanford Center for Reservoir Forecasting in the early nineties.