This work addresses the problem of the cosimulation of cross-correlated variables with inequality constraints. A hierarchical sequential Gaussian cosimulation algorithm is proposed to address this problem, based on establishing a multicollocated cokriging paradigm; the integration of this algorithm with the acceptance–rejection sampling technique entails that the simulated values first reproduce the bivariate inequality constraint between the variables and then reproduce the original statistical parameters, such as the global distribution and variogram. In addition, a robust regression analysis is developed to derive the coefficients of the linear function that introduces the desired inequality constraint. The proposed algorithm is applied to cosimulate Silica and Iron in an Iron deposit, where the two variables exhibit different marginal distributions and a sharp inequality constraint in the bivariate relation. To investigate the benefits of the proposed approach, the Silica and Iron are cosimulated by other cosimulation algorithms, and the results are compared. It is shown that conventional cosimulation approaches are not able to take into account and reproduce the linearity constraint characteristics, which are part of the nature of the dataset. In contrast, the proposed hierarchical cosimulation algorithm perfectly reproduces these complex characteristics and is more suited to the actual dataset.
One of the most challenging aspects of multivariate geostatistics is dealing with complex relationships between variables. Geostatistical co-simulation and spatial decorrelation methods, commonly used for modelling multiple variables, are ineffective in the presence of multivariate complexities. On the other hand, multi-Gaussian transforms are designed to deal with complex multivariate relationships, such as non-linearity, heteroscedasticity and geological constraints. These methods transform the variables into independent multi-Gaussian factors that can be individually simulated. This study compares the performance of the following multi-Gaussian transforms: rotation based iterative Gaussianisation, projection pursuit multivariate transform and flow transformation. Case studies with bivariate complexities are used to evaluate and compare the realisations of the transformed values. For this purpose, commonly used geostatistical validation metrics are applied, including multivariate normality tests, reproduction of bivariate relationships, and histogram and variogram validation. Based on most of the metrics, all three methods produced results of similar quality. The most obvious difference is the execution speed for forward and back transformation, for which flow transformation is much slower.
A hierarchical sequential Gaussian cosimulation method is applied in this study for modeling the variables with an inequality constraint in the bivariate relationship. An algorithm is improved by embedding an inverse transform sampling technique in the second simulation to reproduce bivariate complexity and accelerate the process of cosimulation. A heterotopic simple cokriging (SCK) is also proposed, which introduces two moving neighborhoods: single and multiple searching strategies in both steps of the hierarchical process. The proposed algorithm is tested over a real case study from an iron deposit where iron and aluminum oxide shows a strong bivariate dependency as well as a sharp inequality constraint. The results showed that the proposed hierarchical cosimulation with a multiple searching strategy provides satisfying results compared to the case when a single searching strategy is employed. Moreover, the proposed algorithm is compared to the conventional hierarchical cosimulation, which does not implement the inverse transform sampling integrated into the second simulation. The proposed methodology successfully reproduces inequality constraint, while conventional hierarchical cosimulation fails in this regard. However, it is demonstrated that the proposed methodology requires further improvement for better reproduction of global statistics (i.e., mean and standard deviation).
Traditional geostatistical simulation techniques rely on the assumption of multi-Gaussianity. Although the normal score transform is widely used to convert data to a Gaussian distribution, it only guarantees that the normal scores will be univariate Gaussian and the variables may still have complex multivariate relationships. For this reason, multi-Gaussian transforms became popular for simplifying multivariate geostatistical modelling. This study evaluates three multi-Gaussian transforms: flow transformation, projection pursuit multivariate transform, and rotation based iterative Gaussianisation. Three two-dimensional synthetic case studies were designed with complex multivariate relationships to make it difficult to produce good multivariate Gaussian distributions. The quality of the fitted transforms, the forward transformation of data from the same population and the back transformation from a standard multivariate Gaussian distribution were assessed based on statistical indices and visual inspection. The methods were also evaluated using a real case study with eight variables from the Prominent Hill copper deposit in South Australia. The effects of multi-Gaussian transforms on the reproduction of variograms, univariate and bivariate statistics were qualitatively and quantitatively investigated.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.