Interval-Valued Kriging Models for Geostatistical Mapping with Imprecise Inputs

TL;DR

This paper introduces interval-valued kriging models that incorporate measurement imprecision in geostatistical mapping, improving uncertainty preservation and computational feasibility for practical applications.

Contribution

It develops novel interval-valued kriging models based on random set theory and a generalized L2 metric, addressing previous mathematical and computational challenges.

Findings

Enhanced uncertainty preservation in geostatistical mapping.

Improved computational efficiency with penalty-based optimization.

Successful application to snow load prediction in Utah.

Abstract

Many geosciences data are imprecise due to various limitations and uncertainties in the measuring process. One way to preserve this imprecision in a geostatistical mapping framework is to characterize the measurements as intervals rather than single values. To effectively analyze the interval-valued data, this paper proposes and develops interval-valued kriging models based on the theory of random sets and a generalized L2 metric. These models overcome the mathematical difficulties of a previous development and are computationally more feasible. Numerical implementation of our interval-valued kriging is provided using a penalty-based constrained optimization algorithm. An application to the prediction of design ground snow loads in Utah, USA, is presented that demonstrates the advantages of the proposed models in preserving crucial sources of uncertainty towards a more efficient risk-based designing scheme.