Spatially Constrained Variational Autoencoder for Geochemical Data Denoising and Uncertainty Quantification

Abstract

Geochemical survey data are essential across Earth Science disciplines but are often affected by noise, which can obscure important geological signals and compromise subsequent prediction and interpretation. Quantifying prediction uncertainty is hence crucial for robust geoscientific decision-making. This study proposes a novel deep learning framework, the Spatially Constrained Variational Autoencoder (SC-VAE), for denoising geochemical survey data with integrated uncertainty quantification. The SC-VAE incorporates spatial regularization, which enforces spatial coherence by modeling inter-sample relationships directly within the latent space. The performance of the SC-VAE was systematically evaluated against a standard Variational Autoencoder (VAE) using geochemical data from the gold polymetallic district in the northwestern part of Sichuan Province, China. Both models were optimized using Bayesian optimization, with objective functions specifically designed to maintain essential geostatistical characteristics. Evaluation metrics include variogram analysis, quantitative measures of spatial interpolation accuracy, visual assessment of de-noised maps, and statistical analysis of data distributions, as well as decomposition of uncertainties. Results show that the SC-VAE achieves superior noise suppression and better preservation of spatial structure compared to the standard VAE, as demonstrated by a significant reduction in the variogram nugget effect and an increased partial sill. The SC-VAE produces denoised maps with clearer anomaly delineation and more regularized data distributions, effectively mitigating outliers and reducing kurtosis. Additionally, it delivers improved interpolation accuracy and spatially explicit uncertainty estimates, facilitating more reliable and interpretable assessments of prediction confidence. The SC-VAE framework thus provides a robust, geostatistically informed solution for enhancing the quality and interpretability of geochemical data, with broad applicability in mineral exploration, environmental geochemistry, and other Earth Science domains.

Keywords

geochemical data denoising / spatially constrained variational autoencoder / geostatistics / bayesian optimization / uncertainty analysis / geochemistry

References

[1] Akiba T, Sano S, Yanase T. et al.. Optuna: A Next-Generation Hyperparameter Optimization Framework. Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, 20192623-2631.

[2] Blundell C, Cornebise J, Kavukcuoglu K. et al.. Weight Uncertainty in Neural Network. Proceedings of the 32nd International Conference on Machine Learning, PMLR, 20151613-162237

[3] Carranza E J MGeochemical Anomaly and Mineral Prospectivity Mapping in GIS, 2008New YorkElsevier

[4] Chen G X, Cheng Q M. Singularity Analysis Based on Wavelet Transform of Fractal Measures for Identifying Geochemical Anomaly in Mineral Exploration. Computers & Geosciences, 2016, 87: 56-66.

[5] Chen G X, Cheng Q M. Fractal-Based Wavelet Filter for Separating Geophysical or Geochemical Anomalies from Background. Mathematical Geosciences, 2018, 50(3): 249-272.

[6] Cheng Q M. Mapping Singularities with Stream Sediment Geochemical Data for Prediction of Undiscovered Mineral Deposits in Gejiu, Yunnan Province, China. Ore Geology Reviews, 2007, 32(1/2): 314-324.

[7] Cheng Q M, Xu Y G, Grunsky E. Integrated Spatial and Spectrum Method for Geochemical Anomaly Separation. Natural Resources Research, 2000, 9(1): 43-52.

[8] Chiles J P, Delfiner PGeostatistics: Modeling Spatial Uncertainty, 2012HobokenJohn Wiley & Sons.

[9] Doersch, C., 2016. Tutorial on Variational Autoencoders. arXiv: 1606.05908

[10] Erfanian N, Heydari A A, Feriz A M. et al.. Deep Learning Applications in Single-Cell Genomics and Transcriptomics Data Analysis. Biomedicine & Pharmacotherapy, 2023, 165: 115077.

[11] Frazier, P. I., 2018. A Tutorial on Bayesian Optimization. arXiv: 1807.02811. https://doi.org/10.48550/arXiv.1807.02811

[12] Gal Y, Ghahramani Z. Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning. Proceedings of the 33rd International Conference on Machine Learning, 2016

[13] Goodfellow I, Bengio Y, Courville A. et al.Deep Learning, 2016CambridgeMIT Press

[14] Goovaerts PGeostatistics for Natural Resources Evaluation, 1997OxfordUniversity Press.

[15] Grunsky E C, de Caritat P. State-of-the-Art Analysis of Geochemical Data for Mineral Exploration. Geochemistry: Exploration, Environment, Analysis, 2020, 20(2): 217-232

[16] Hengl T, Nussbaum M, Wright M N. et al.. Random Forest as a Generic Framework for Predictive Modeling of Spatial and Spatio-Temporal Variables. PeerJ, 2018, 6: e5518.

[17] Huang D Z, Zuo R G, Wang J. Geochemical Anomaly Identification and Uncertainty Quantification Using a Bayesian Convolutional Neural Network Model. Applied Geochemistry, 2022, 146: 105450.

[18] Isaaks E H, Srivastava R MApplied Geostatistics, 1989New YorkOxford University Press

[19] Kascenas A, Sanchez P, Schrempf P. et al.. The Role of Noise in Denoising Models for Anomaly Detection in Medical Images. Medical Image Analysis, 2023, 90: 102963.

[20] Kendall, A., Gal, Y., 2017. What Uncertainties do We Need in Bayesian Deep Learning for Computer Vision? Advances in Neural Information Processing Systems, arXiv: 1703.04977. https://doi.org/10.48550/arXiv.1703.04977

[21] Kingma, D. P., Welling, M., 2013. Auto-Encoding Variational Bayes. arXiv: 1312.6114. https://doi.org/10.48550/arXiv.1312.6114

[22] Lawley C J M, McCafferty A E, Graham G E. et al.. Data-Driven Prospectivity Modelling of Sediment-Hosted Zn-Pb Mineral Systems and Their Critical Raw Materials. Ore Geology Reviews, 2022, 141: 104635.

[23] Letham B, Karrer B, Ottoni G. et al.. Constrained Bayesian Optimization with Noisy Experiments. Bayesian Analysis, 2019, 14(2): 495-519.

[24] Li N, Deng J, Yang L Q. et al.. Paragenesis and Geochemistry of Ore Minerals in the Epizonal Gold Deposits of the Yangshan Gold Belt, West Qinling, China. Mineralium Deposita, 2014, 49(4): 427-449.

[25] Li N, Deng J, Yang L Q. et al.. Constraints on Depositional Conditions and Ore-Fluid Source for Orogenic Gold Districts in the West Qinling Orogen, China: Implications from Sulfide Assemblages and Their Trace-Element Geochemistry. Ore Geology Reviews, 2018, 102: 204-219.

[26] Liu GStudy on Characteristics and Its Prediction of Carlin-Type Gold Ore Belt in Sichuan-Shannxi-Gansu Region: [Dissertation], 2011ChongqingNortheastern University

[27] Luo Z J, Xiong Y H, Zuo R G. Recognition of Geochemical Anomalies Using a Deep Variational Autoencoder Network. Applied Geochemistry, 2020, 122: 104710.

[28] Mao J W, Qiu Y M, Goldfarb R J. et al.. Geology, Distribution, and Classification of Gold Deposits in the Western Qinling Belt, Central China. Mineralium Deposita, 2002, 37(3): 352-377.

[29] Pourgholam M M, Afzal P, Adib A. et al.. Recognition of REEs Anomalies Using an Image Fusion Fractal-Wavelet Model in Tarom Metallogenic Zone, NW Iran. Geochemistry, 2024, 84(2): 126093.

[30] Reichstein M, Camps-Valls G, Stevens B. et al.. Deep Learning and Process Understanding for Data-Driven Earth System Science. Nature, 2019, 566(7743): 195-204.

[31] Ridsdill-Smith T A, Dentith M C. The Wavelet Transform in Aeromagnetic Processing. Geophysics, 1999, 64(4): 1003-1013.

[32] Snoek, J., Larochelle, H., Adams, R. P., 2012. Practical Bayesian Optimization of Machine Learning Algorithms. Advances in Neural Information Processing Systems, 25

[33] Stanley C R. On the Special Application of Thompson – Howarth Error Analysis to Geochemical Variables Exhibiting a Nugget Effect. Geochemistry: Exploration, Environment, Analysis, 2006, 6(4): 357-368

[34] Stanley C R. Missed Hits or near Misses: determining how Many Samples Are Necessary to Confidently Detect Nugget-Borne Mineralization. Geochemistry: Exploration, Environment, Analysis, 2008, 8(2): 129-138

[35] Stanley C R, Lawie D. Average Relative Error in Geochemical Determinations: Clarification, Calculation, and a Plea for Consistency. Exploration and Mining Geology, 2007, 16(3/4): 267-275.

[36] Templ M, Filzmoser P, Reimann C. Cluster Analysis Applied to Regional Geochemical Data: Problems and Possibilities. Applied Geochemistry, 2008, 23(8): 2198-2213.

[37] Wang J, Zuo R G. Identification of Geochemical Anomalies through Combined Sequential Gaussian Simulation and Grid-Based Local Singularity Analysis. Computers & Geosciences, 2018, 118: 52-64.

[38] Wang J, Zuo R. Uncertainty Quantification in Geochemical Mapping: A Review and Recommendations. Geochemistry, Geophysics, Geosystems, 2024, 25(3): e2023GC011301.

[39] Wang X Q, Zhang Q, Zhou G H. National-Scale Geochemical Mapping Projects in China. Geostandards and Geoanalytical Research, 2007, 31(4): 311-320.

[40] Xie X J, Mu X Z, Ren T X. Geochemical Mapping in China. Journal of Geochemical Exploration, 1997, 60(1): 99-113.

[41] Xiong Y H, Zuo R G. Recognition of Geochemical Anomalies Using a Deep Autoencoder Network. Computers & Geosciences, 2016, 86: 75-82.

[42] Xu G M, Cheng Q M, Zuo R G. et al.. Application of Improved Bi-Dimensional Empirical Mode Decomposition (BEMD) Based on Perona-Malik to Identify Copper Anomaly Association in the Southwestern Fujian (China). Journal of Geochemical Exploration, 2016, 164: 65-74.

[43] Yang L Q, Deng J, Li N. et al.. Isotopic Characteristics of Gold Deposits in the Yangshan Gold Belt, West Qinling, Central China: Implications for Fluid and Metal Sources and Ore Genesis. Journal of Geochemical Exploration, 2016, 168: 103-118.

[44] Zhang GMetallogenic regularities and Ore-Prospecting Direction of Manaoke Gold Deposit in Sichuan: [Dissertation], 2012ChengduChengdu University of Technology(in Chinese with English Abstract)

[45] Zhang S E, Bourdeau J E, Nwaila G T. et al.. Denoising of Geochemical Data Using Deep Learning – Implications for Regional Surveys. Natural Resources Research, 2024, 33(2): 495-520.

[46] Zuo R G, Kreuzer O P, Wang J. et al.. Uncertainties in GIS-Based Mineral Prospectivity Mapping: Key Types, Potential Impacts and Possible Solutions. Natural Resources Research, 2021, 30(5): 3059-3079.

[47] Zuo R G, Luo Z J, Xiong Y H. et al.. A Geologically Constrained Variational Autoencoder for Mineral Prospectivity Mapping. Natural Resources Research, 2022, 31(3): 1121-1133.

[48] Zuo R G, Xia Q L, Wang H C. Compositional Data Analysis in the Study of Integrated Geochemical Anomalies Associated with Mineralization. Applied Geochemistry, 2013, 28: 202-211.

[49] Zuo R G, Xiong Y H, Wang J. et al.. Deep Learning and Its Application in Geochemical Mapping. Earth-Science Reviews, 2019, 192: 1-14.

RIGHTS & PERMISSIONS

China University of Geosciences (Wuhan) and Springer-Verlag GmbH Germany, Part of Springer Nature