Rethinking Variogram Modeling - 自动拟合驱动的建模工作流与并行应用

Aug 22, 2026

Upload by Dozel

Variogram modelling is a core part of resource modelling, providing insight into spatial continuity of a deposit and supporting Kriging-based estimation workflows. However, manual variogram fitting is time-consuming and requires specialized expertise. This paper introduces an auto-fitting approach and software that redesign the full workflow around algorithmic optimization, efficiently computing experimental variograms across many domains, grade attributes, and directions, then optimizing anisotropy, ranges, contributions, and nugget parameters.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages1

Reading Time3 min

AccessFree to download

Rethinking Variogram Modeling

自动拟合驱动的建模思路

计算实验变差函数与参数优化

模型可视化、人工校准与误差评估

现有模型更新

并行化部署（桌面与云）

Rethinking variogram modeling Boucher. A 1, Machado. P2, Minniakhmetov. I3, Androszczuk. P4, Rigby. A5, Wright. M6, Baudin. M7, van Dijken. B8, 1. Alexander Boucher, AR2Tech, Denver CO, USA . Email: [aboucher@ar2tech.com](aboucher@ar2tech.com) 2. Pericles Machado, AR2Tech, Denver, CO, USA, Email: [pericles@ar2tech.com](pericles@ar2tech.com) 3. Ilnur Minniakhmetov, BHP, Perth WA 6010. Email: [ilnur.minniakhmetov@bhp.com](ilnur.minniakhmetov@bhp.com) 4. Paul Androszczuk, BHP, Perth WA 6010. Email: [paul.androszczuk@bhp.com](paul.androszczuk@bhp.com) 5. Andrew Rigby, BHP, Perth WA 6010. Email: [andrew.n.rigby@bhp.com](andrew.n.rigby@bhp.com) 6. Michaela Wright, BHP, [Adelaide SA 5000. Email:](Adelaide SA 5000. Email: michaela.wright@bhp.com)[ ](Adelaide SA 5000. Email: michaela.wright@bhp.com)[michaela.wright@bhp.com](Adelaide SA 5000. Email: michaela.wright@bhp.com) 7. Murim Baudin, BHP, [Perth WA 6010. Email:](Perth WA 6010. Email: murim.baudin@bhp.com)[ ](Perth WA 6010. Email: murim.baudin@bhp.com)[murim.baudin@bhp.com](Perth WA 6010. Email: murim.baudin@bhp.com) 8. Brent van Dijken, BHP, Perth WA 6010. Email: [brent.vandijken@bhp.com](brent.vandijken@bhp.com) [Keywords: variography](Keywords: variography) , automation, optimization, geostatistics, estimation ABSTRACT Variogram modelling is a critical component of resource modeling, it provides an insight into the spatial continuity of a deposit as well as being a necessary component of all Kriging-based algorithms. Although important, it is a tedious, laborious, and difficult task which requires specialized knowledge from the modelers. This paper will present a novel approach and software to variogram modeling that is constructed around automatic fitting. Auto-fitting is an optimization process in which a variogram model that matches the data is found algorithmically. Instead of adding the autofitting algorithm to the traditional modeling workflow, our approach redesigns the entire modeling workflow around the autofitting requirements. First, the application can work with several domains and grade attributes at once, it efficiently computes the experimental variograms in hundreds of directions for each combination of domains and attributes. These experimental variograms are then passed to an optimization process, the auto-fitting, that will search for the anisotropy, the ranges, the contributions, and the nugget of the 3D model. Once completed, the models are then presented in a dashboard where each model has a fitting error associated with it. The modeler can then quickly review each model and manually refine any models that are deemed unsatisfactory through a rich and interactive user interface. Through that visualization, the modeler can inspect the model fit quality in hundreds of directions to ensure that the model properly captures the anisotropy of the data. The application also allows the updating of existing models when new data is added to a project. Amodeler can then quickly update all the variogram models whenever new data becomes available ensuring that the resource model reflects the sampled data. Finally, the application is completely parallelized and can be run on the desktop or natively on the cloud.

FAQs

为什么变差函数建模在资源建模中很关键？

变差函数建模用于刻画矿体/沉积体的空间连续性，并作为基于克里金（Kriging）的算法所必需的组成部分。

论文提出的自动拟合方法与传统流程的区别是什么？

不是简单把自动拟合算法叠加到传统工作流，而是围绕自动拟合的需求重构整个建模流程。

自动拟合会自动搜索哪些三维模型关键参数？

自动拟合会搜索各向异性、变程、贡献量以及3D模型的块金效应（nugget）。

系统如何帮助模型人员评估并改进模型质量？

模型会在仪表盘中展示，并为每个模型关联拟合误差；用户可通过交互式界面快速复核，并在不满意时进行手动精修，同时从大量方向检查各向异性是否被正确捕捉。

Aug 22, 2026

Upload by Dozel