Can geodiversity be predicted from space?

Abstract

Identification of geodiversity at regional and national scales with traditional surveying techniques remains a logistically difficult and often financially prohibitive task. Earth observation data in combination with statistical modeling may provide a unique approach for mapping and analyzing geodiversity cost-efficiently. In this study, the possibility to map geodiversity using digital elevation models (DEMs) and remote sensing (RS) data was scrutinized in Finland, Northern Europe at a landscape scale. The main methods in the evaluation of spatial prediction ability of geodiversity and the relative contributions of DEM and RS data were generalized additive modeling (GAM) and variation partitioning. Based on the results, geodiversity was mainly determined by polygenetic bedrock topography and fluvial activity, and correlated most strongly with variables describing high potential energy and topographical heterogeneity. The spatial patterns of geodiversity were robustly predicted across the study areas using multivariate GAMs. The higher predictive performance of the DEM- versus RS-based variables was evident across the study areas. In conclusion, we propose that statistically-based spatial prediction offers an opportunity for characterizing geodiversity of a large area in a systematic, repeatable, and spatially exhaustive manner.