Accounting for a spatial trend in fine-scale ground-penetrating radar data: a comparative case study

Abstract

In geostatistics, one major challenge is addressing spatial trend effects in high‑resolution data. This study compares three kriging algorithms—ordinary kriging (OK), universal kriging (UK), and intrinsic random function of order k (IRF‑k)—applied to densely sampled ground‑penetrating radar (GPR) data over a laterite‑type bauxite deposit. Structural variogram modeling and cross‑validation were performed to assess predictive accuracy. Although IRF‑k slightly outperformed the others, all methods yielded similar and satisfactory estimates: mean squared error (MSE) values were 0.1267 for IRF‑k, 0.1322 for OK, and 0.1349 for UK. The high performance uniformity is attributed to the dense GPR dataset and chosen kriging neighborhood parameters