After years of research, our paper “Sampling Strategies for Uncertainty Reduction in Categorical Random Fields” has been published in Mathematical Geosciences.

A core piece of the doctoral puzzle

This paper represents one of the central chapters of my PhD thesis, and seeing it finally published brings a deep sense of relief and pride. The work formalizes the optimal sampling problem for categorical spatial models using information-theoretic measures – specifically, how to decide where to place new measurements so that uncertainty about the underlying geology is reduced as efficiently as possible.

The key contribution is a rigorous comparison between information-driven sampling strategies and the more conventional random or regular grid approaches. We show that by leveraging entropy and mutual information concepts from information theory, one can design sampling campaigns that extract significantly more knowledge per sample than standard approaches.

This is not just a theoretical exercise. In mining and geosciences, every drill hole or measurement point has a real cost. Being able to demonstrate, formally, that an adaptive information-driven strategy outperforms brute-force regular grids means tangible savings and better geological models.

Why it matters to me

This paper took years of iterative work – formulating the problem, running geostatistical simulations, debugging code, and refining the mathematical framework. There were moments where the path forward was unclear. But the result is something I am genuinely proud of: a solid contribution to the intersection of geostatistics and information theory that I hope will be useful to others working on spatial sampling problems.