211 EFFECTS OF ORDERING ON SPATIAL ALLOCATION UNDER UNCERTAINTY: A DATADRIVEN EXPERIMENT BASED ON HIERARCHICAL SEQUENTIAL INDICATOR SIMULATION *D.B. Bahia1, J. F. C. L. Costa1, M. A. Bassani1, D. M. Marques1 1Laboratory of Mineral Research and Mine Planning (LPM), Federal University of Rio Grande do Sul (UFRGS), Brazil. (*Presenting author: dunasbahia@gmail.com) ABSTRACT To deliver minerals faster, smarter, and more responsibly, spatial modeling plays a critical role. Many of the questions a spatial model is expected to answer can be naturally formulated as binary propositions, expressed in yes/no terms. By structuring these propositions into a predefined hierarchical order, Hierarchical Sequential Indicator Simulation (HSIS) provides a for progressively allocating space within a decision-tree architecture and generating multiple realizations of equiprobable spatial scenarios. This work evaluates the effect of decision-tree ordering in a simplified HSIS experiment involving five mutually exclusive lithological categories. Two alternative hierarchical orderings are compared, with the decision tree order as the only varying component in the experiment. The results show that different hierarchical orderings yield distinct spatial allocations, even under otherwise identical modeling conditions. These findings highlight that responsible and efficient mineral development depends not only on the questions posed to spatial models, but also on how those questions are hierarchically organized. KEYWORDS Geoscience and Next-Gen Exploration; AI and Data-Driven Decision Making; Spatial Modeling; Geostatistics; Indicator Simulation; HSIS; Decision trees. 1. CONTEXT AND PROBLEM STATEMENT The representation of knowledge through binary propositions constitutes a fundamental abstraction across multiple scientific domains. In information theory, a binary proposition is encoded as a bit, the smallest unit of information, formally introduced by Claude Shannon (1948). A bit represents the outcome of a yes/no question and provides a minimal, lossless encoding of a logical proposition. Although originally developed in the context of communication systems, Shannon’s formulation established a general language for representing information, enabling complex systems to be described as collections of binary statements. Matheron (1963, 1971) established the theoretical foundations of geostatistics. Kriging, formulated as the Best Linear Unbiased Estimator (BLUE), provides a deterministic linear interpolation based on a weighted mean of nearby data, with weights determined by a continuity model. Within this framework, spatial modeling is primarily concerned with interpolation, yielding a single “best” representation of the unknown variable at unsampled locations. Although uncertainty is formally quantified through the kriging variance, it remains implicitly summarized around the estimated mean, and the spatial variability away from this mean is not explicitly reproduced. Journel and collaborators (Journel & Huijbregts, 1978; Journel, 1983; Journel & Alabert, 1989) developed a probabilistic framework for stochastic simulation, shifting the focus of geostatistical
RkJQdWJsaXNoZXIy MTM0Mzk2