Advertisement

Hydrogeology Journal

, Volume 24, Issue 1, pp 211–229 | Cite as

Toward real-time three-dimensional mapping of surficial aquifers using a hybrid modeling approach

  • Michael J. Friedel
  • Akbar Esfahani
  • Fabio Iwashita
Paper

Abstract

A hybrid modeling approach is proposed for near real-time three-dimensional (3D) mapping of surficial aquifers. First, airborne frequency-domain electromagnetic (FDEM) measurements are numerically inverted to obtain subsurface resistivities. Second, a machine-learning (ML) algorithm is trained using the FDEM measurements and inverted resistivity profiles, and borehole geophysical and hydrogeologic data. Third, the trained ML algorithm is used together with independent FDEM measurements to map the spatial distribution of the aquifer system. Efficacy of the hybrid approach is demonstrated for mapping a heterogeneous surficial aquifer and confining unit in northwestern Nebraska, USA. For this case, independent performance testing reveals that aquifer mapping is unbiased with a strong correlation (0.94) among numerically inverted and ML-estimated binary (clay-silt or sand-gravel) layer resistivities (5–20 ohm-m or 21–5,000 ohm-m), and an intermediate correlation (0.74) for heterogeneous (clay, silt, sand, gravel) layer resistivities (5–5,000 ohm-m). Reduced correlation for the heterogeneous model is attributed to over-estimating the under-sampled high-resistivity gravels (about 0.5 % of the training data), and when removed the correlation increases (0.87). Independent analysis of the numerically inverted and ML-estimated resistivities finds that the hybrid procedure preserves both univariate and spatial statistics for each layer. Following training, the algorithms can map 3D surficial aquifers as fast as leveled FDEM measurements are presented to the ML network.

Keywords

Airborne geophysics Geostatistics Heterogeneity Machine-learning USA 

Vers une cartographie tridimensionnelle en temps réel des aquifères superficiels à partir d’une approche de modélisation hybride

Résumé

Une approche de modélisation hybride est proposée pour une cartographie tridimensionnelle (3D) en quasi temps réel des aquifères superficiels. Tout d’abord, les mesures électromagnétiques aéroportées dans le domaine fréquentiel (MEDF) sont inversées pour obtenir les résistivités de subsurface. Puis un algorithme d’apprentissage supervisé (AS) est appliqué en utilisant les mesures MEDF, les profils de résistivités inversées, et les données géophysiques et hydrogéologiques des forages. Troisièmement, l’algorithme d’apprentissage supervisé AS est utilisé avec les mesures indépendantes MEDF pour cartographier la distribution spatiale du système aquifère. L’efficacité de l’approche hybride est démontrée par la cartographie d’un aquifère superficiel hétérogène et de son unité encaissante au Nord-Ouest du Nebraska, USA. Dans ce cas, des tests indépendants de performance révèlent que la cartographie de l’aquifère est non biaisée avec une forte corrélation (0.94) pour les résistivités inversées de strates binaires (argile-silt ou sable-gravier) estimées (5–20 ohm-m ou 21–5000 ohm-m) et une corrélation intermédiaire (0.74) pour les résistivités inversées de strates hétérogènes (argile, silt, sable, gravier) estimées (5–5000 ohm-m). La faible corrélation du modèle hétérogène est attribuée à une surestimation de la résistivité de graviers sous-échantillonnés (environ 0.5 % des données utilisées avec l’algorithme d’apprentissage) et lorsque corrigée, la corrélation augmente (0.87). Une analyse indépendante des résistivités inversées numériquement et des valeurs AS estimées montre que la procédure hybride préserve aussi bien les caractéristiques statistiques univariantes que les caractéristiques spatiales de chaque strate. Après la phase d’apprentissage, l’algorithme peut cartographier en 3D les aquifères superficiels d’autant plus rapidement que les mesures MEDF sont rattachées au réseau AS.

Hacia el mapeo tridimensional en tiempo real de los acuíferos superficiales usando un enfoque de híbrido de modelación

Resumen

Se propone un enfoque de modelación híbrido para un mapeo en tres dimensiones (3D) en tiempo real próximo a los acuíferos superficiales. En primer lugar, las mediciones electromagnéticas aéreas en el dominio de la frecuencia (FDEM) son numéricamente invertidas para obtener las resistividades del subsuelo. En segundo lugar, se entrena un algoritmo de aprendizaje automático (ML) usando las medidas FDEM y los perfiles invertidos de resistividad, y datos hidrogeológicos y geofísicos de la perforación. En tercer lugar, el algoritmo entrenado ML se utiliza junto con las mediciones independientes FDEM para mapear la distribución espacial del sistema de acuíferos. La eficacia del enfoque híbrido se demuestra para mapear un acuífero superficial heterogéneo y una unidad confinante en el noroeste de Nebraska, EE.UU. Para este caso, las pruebas de rendimiento independientes revelan que el mapeo del acuífero no está sesgado con una fuerte correlación (0.94) entre las resistividades de las capas invertidas numéricamente (5–20 ohm-m ó 21–5000 ohm-m) y la estimación de ML binaria (arcilla-limo o arena-grava), y una correlación intermedia (0.74) para resistividades (5–5 mil ohm-m) de capas de heterogéneas (arcilla, limo, arena, grava). La correlación reducida para el modelo heterogéneo se atribuye a una sobre-estimación de las muestras de gravas de alta resistividad (aproximadamente 0.5 % de los datos de entrenamiento), y cuando se lo elimina se incrementa la correlación (0.87). El análisis independiente de las resistividades estimadas por inversión numérica y de ML encuentra que el procedimiento híbrido preserva tanto las estadísticas univariadas como las espaciales para cada capa. Siguiendo del entrenamiento, los algoritmos pueden mapear acuíferos superficiales 3D tan rápidamente como las medidas FDEM niveladas se presenten a la red ML.

采用混合模拟方法进行表层含水层实时三维绘图

摘要

本文介绍了混合模拟方法进行表层含水层近实时三维绘图。首先,把空载频率域电磁测量结果进行数字上反转,以获取地表以下的电阻率。第二,采用频率域电磁测量结果和反转的电阻率剖面以及钻孔地球物理和水文地质资料演示机学算法。第三,一起利用所述的机学算法和独立的频率域电磁测量结果绘制含水层系统的空间分布。本文展示了混合方法在绘制美国内布拉斯加州异质表层含水层和承压单元上的效力。在这个案例中,独立的性能测试显示,含水层绘图非常准确,数字上反转的和机学估算的二元(粘土-粉砂或砂-砂砾)层电阻率(5–20 ohm-m或21–5000 ohm-m)间有很强的关联度,异质(粘土、粉砂砂和砂砾)层电阻率(5–5000 ohm-m)间关联度一般。异质模型的关联度降低是由于高估了采样不多的高电阻率砂砾(大约演示资料的0.5 %),以及模拟后关联度增加(0.87)。数值反转的和机学估算的电阻率独立分析发现,混合程序保存着每个地层单变量的和空间统计资料。培训演示之后,算法可以就象频率域电磁测量结果出现在机学网络那样快地绘制3D表层含水层。

Em direção a um mapeamento tridimensional tempo real de aquíferos superficiais, utilizando uma abordagem de modelagem híbrida

Resumo

Uma abordagem de modelagem híbrida é proposta para mapeamento de aquíferos superficiais em tempo próximo ao real em três dimensões (3D). Em primeiro lugar, as medições eletromagnéticas aeroembarcadas no domínio da frequência (FDEM) são numericamente invertidas para obter resistividades de subsuperfície. Em segundo lugar, um algoritmo de aprendizado de máquina (ML) é treinado usando as medidas FDEM e perfis de resistividade invertidos, e geofísica do poço e dados hidrogeológicos. Em terceiro lugar, o algoritmo ML treinado é utilizado em conjunto com as medições FDEM independentes para mapear a distribuição espacial do sistema aquífero. A eficácia da abordagem híbrida é demonstrada para mapear um aquífero superficial heterogêneo e unidade confinante no noroeste de Nebraska, EUA. Para este caso, testes de desempenho independentes revelam que o mapeamento do aquífero é não tendencioso com uma forte correlação (0.94) entre a resistividade da camada numericamente invertida e estimada binariamente por ML (argila-silte ou areia e cascalho) (5–20 ohm-m ou 21–5000 ohm-M), e uma correlação intermediária (0.74) para resistividades em camadas heterogéneas (argila, sedimentos, areia, cascalho) (5–5000 ohm-m). Correlação reduzida para o modelo heterogêneo é atribuída a sobre-estimavas dos cascalhos de alta resistividade sub amostrados (cerca de 0.5 % dos dados de treinamento), que quando retirados aumentam a correlação (0.87). Análise independente das resistividades numericamente invertidas e ML-estimadas conclui que o procedimento híbrido preserva ambas as estatísticas univariadas e espaciais para cada camada. Após o treinamento, os algoritmos podem mapear aquíferos superficiais em 3D tão rápido quanto medições FDEM levantadas são apresentados à rede ML.

Notes

Acknowledgements

The authors thank Jared Abraham of Exploration Resources International (formerly US Geological Survey) for providing the original data sets used in this study, and B. J. Minsley of the US Geological Survey and two anonymous reviewers for their detailed and constructive comments.

Supplementary material

10040_2015_1318_MOESM1_ESM.pdf (322 kb)
ESM 1 (PDF 322 kb)

References

  1. Abraham JD, Cannia JC, (2011) Airborne electromagnetic surveys for 3D geological mapping (2011) Paper 505, USGS, Reston, VA. http://digitalcommons.unl.edu/usgsstaffpub/506. Accessed on 15 August 2015
  2. Abraham JD, Cannia JC, Bedrosian PA, Johnson MR, Ball LB, Sibray SS (2012a) Airborne electromagnetic mapping of the base of aquifer in areas of western Nebraska. US Geol Surv Sci Invest Rep 2011–5219, 38 ppGoogle Scholar
  3. Abraham JD, Bedrosian PA, Asch TH, Ball LB, Cannia JC, Phillips JD, Lackey S (2012b) Evaluation of geophysical techniques for the detection of paleochannels in the Oakland area of eastern Nebraska as part of the Eastern Nebraska Water Resource Assessment. US Geol Surv Sci Invest Rep 2011–5228, 40 ppGoogle Scholar
  4. Akaike H (1974) A new look at the statistical model identification. IEEE Trans Autom Control 19(6):716–723CrossRefGoogle Scholar
  5. Auken E, Christiansen AV, Jacobsen BH, Foged N (2005) Piecewise 1-D laterally constrained inversion of resistivity data. Geophys Prospect 53:497–506CrossRefGoogle Scholar
  6. Avdeev DB, Kuvshinov AV, Pankratov OV, Newman GA (1998) Three-dimensional frequency domain modelling of airborne electromagnetic responses. Explor Geophys 29:111–119CrossRefGoogle Scholar
  7. Benavides A, Everett ME, Pierce C (2009) Unexploded ordnance discrimination using time-domain electromagnetic induction and self-organizing maps. Stoch Environ Res Risk Assess 23:169–179CrossRefGoogle Scholar
  8. Breiman L (1996) Bagging predictors. Mach Learn 24:123–140Google Scholar
  9. Carneiro CC, Fraser SJ, Crosta AP, Silva AM, Barros CE (2012) Semiautomated geologic mapping using self-organizing maps and airborne geophysics in the Brazilian Amazon. Geophysics 77(4):K17–K24. doi: 10.1190/geo2011-0302.1 CrossRefGoogle Scholar
  10. Cooper GF, Herskovits E (1992) A Bayesian method for the induction of probabilistic networks from data. Mach Learn 9:309–347Google Scholar
  11. Dickson BL, Giblin AM (2007) An evaluation of methods for imputation of missing trace element data in groundwaters. Geochem Explor Environ Anal 7(2):173–178CrossRefGoogle Scholar
  12. Farquharson CG (2000) Background for program EM1DFM. University of British Columbia Geophysical Inversion Facility, Vancouver, 20 pp. http://www.eos.ubc.ca/ubcgif/iag/sftwrdocs/em1dfm/bg.pdf. Accessed June 2015
  13. Fessant F, Midenet S (2002) Self-organizing map for data imputation and correction in surveys. Neural Comput Appl 10:300–310CrossRefGoogle Scholar
  14. Fitterman DV, Deszcz-Pan M (1998) Helicopter EM mapping of saltwater intrusion in Everglades National Park, Florida. Explor Geophys 29:240–243CrossRefGoogle Scholar
  15. Fraser D (1983) Airborne electromagnetic surveys of the Cascade range, western United States. US Geol Surv Open File Rep 83–92, 35 ppGoogle Scholar
  16. Friedel MJ (2011) Modeling hydrologic and geomorphologic responses across post-fire landscapes using a self-organizing map approach. Environ Model Softw 26(12):1660–1674CrossRefGoogle Scholar
  17. Friedel MJ (2012a) Data-driven modeling of surface temperature anomaly and solar activity trends. Environ Model Softw 37:217–232CrossRefGoogle Scholar
  18. Friedel MJ (2012b) Hybrid modeling to predict the economic feasibility of mining undiscovered porphyry copper deposits: an alternative hybrid approach. Appl Soft Comput 13:1016–1103CrossRefGoogle Scholar
  19. Friedel MJ, Iwashita F (2013) Hybrid modeling of spatial continuity for application to numerical inverse problems. Environ Model Softw 43:60–79CrossRefGoogle Scholar
  20. Friedel MJ, Asch T, Oden C (2012b) Hybrid analysis of multiaxis electromagnetic data for discrimination of munitions and explosives of concern. Geophys J Int 190(2):960–980CrossRefGoogle Scholar
  21. Goovaerts P (1997) Geostatistics for natural resources evaluation. Oxford University Press, New YorkGoogle Scholar
  22. Gunnik JL, Boshch JHA, Siemon B, Roth B, Auken E (2012) Combining ground-based and airborne EM through artificial neural networks for modelling hydrogeological units under saline groundwater conditions. J Hydrol Earth Syst Sci 9:3269–3309CrossRefGoogle Scholar
  23. Hall, M, Frank E, Holmes G, Pfahringer B, Reutemann P, Witten IH (2009) The WEKA Data Mining Software: an update. SIGKDD Explorations 11(1), 18 pp. http://www.cs.waikato.ac.nz/~ml/weka/citing.html. Accessed 15 August 2015
  24. He X, Koch J, Sonnenborg TO, Jørgensen F, Schamper C, Refsgaard JC (2014) Transition probability-based stochastic geological modeling using airborne geophysical data and borehole data. Water Resour Res 50(4):3147–3169CrossRefGoogle Scholar
  25. Iwashita F, Friedel MJ, de Souza CR, Filho FSJ (2011) Hillslope chemical weathering across Paraná, Brazil: a data mining-GIS hybrid approach. Geomorphology 132(3–4):167–175CrossRefGoogle Scholar
  26. Kalteh AM, Berndtsson R (2007) Interpolating monthly precipitation by self-organizing map (SOM) and multilayer perceptron (MLP). Hydrol Sci J 52(2):305–317CrossRefGoogle Scholar
  27. Kalteh AM, Hjorth P (2009) Imputation of missing values in precipitation-runoff process database. Hydrol Res 40(4):420–432CrossRefGoogle Scholar
  28. Kalteh AM, Hjorth P, Berndtsson R (2008) Review of the self-organizing map (SOM) approach in water resources: analysis, modelling and application. Environ Model Softw 23:835–845CrossRefGoogle Scholar
  29. Koch J, He X, Jensen KH, Refsgaard JC (2014) Challenges in conditioning a stochastic geological model of a heterogeneous glacial aquifer to a comprehensive soft data set. Hydrol Earth Syst Sci 18(8):2907–2923CrossRefGoogle Scholar
  30. Kohler A, Ohrnberger M, Scherbaum F (2010) Unsupervised pattern recognition in continuous seismic wavefield records using self-organizing maps. Geophys J Int 182:1619–1630CrossRefGoogle Scholar
  31. Kohonen T (2001) Self-organizing maps, 3rd edn. Springer Series in Information Sciences 30, Springer, Heidelberg, Germany, 253 ppGoogle Scholar
  32. Lipinski BA, Sams JI, Smith BD, Harbert W (2008) Using HEM surveys to evaluate disposal of by-product water from CBNG development in the Powder River Basin, Wyoming. Geophysics 73(3):B77–B84CrossRefGoogle Scholar
  33. Malek MA, Harun S, Shamsuddin SM, Mohamad I (2008) Imputation of time series data via Kohonen self organizing maps in the presence of missing data. Eng Technol 41:501–506Google Scholar
  34. Minsely BJ (2011) A trans-dimensional Bayesian Markov chain Monte Carlo algorithm for model assessment using frequency-domain electromagnetic data. Geophys J Int 187:252–272CrossRefGoogle Scholar
  35. Minsley BJ, Abraham JD, Smith BD, Cannia JC, Voss CI, Jorgenson MT, Walvoord MA, Wylie BK, Anderson L, Ball LB, Deszcz-Pan M, Wellman TP (2012) Airborne electromagnetic imaging of discontinuous permafrost. Geophys Res Lett 39, L02503CrossRefGoogle Scholar
  36. Paine JG, Minty BRS (2005) Airborne hydrogeophysics. In: Hydrogeophysics. Springer, Dordrecht, The Netherlands, pp 333–360Google Scholar
  37. Pearl J (1985) A model of self-activated memory for evidential reasoning. In: Proceedings of the 7th Conference of the Cognitive Science Society. University of California, Irvine, CA, 1985, pp 329–334Google Scholar
  38. Rallo R, Ferre-Gine J, Arenas A, Giralt F (2002) Neural virtual sensor for the inferential prediction of product quality form process variables. Comput Chem Eng 26(12):1735–1754CrossRefGoogle Scholar
  39. Refsgaard JC, Aukien E, Bamberg CA, Christensen BSB, Clausen T, Dalgaard E, Efferso F, Ernstsen V, Gertz F, Hansen AL, He X, Jacobsen BH, Jensen KH, Jorgensen F, Jorgensen LF, Koch J, Nilsson B, Petersen C, De Schepper G, Schamper C, Sorensen KI, Therrien R, Thirup C, Viezzoli A (2014) Nitrate reduction in geologically heterogeneous catchments: a framework for assessing the scale of predictive capability of hydrological models. Sci Total Environ 468–469:1278–1288CrossRefGoogle Scholar
  40. Riggelsen C, Ohrnberger M, Scherbaum F (2007) Dynamic Bayesian networks for real-time classification of seismic signals. In: Kok JN, Koronacki J, Lopez de Mantaras R, Matwin S, Mladenia D, Skowron A (eds) Proceedings: 11th European Conference on Principles and Practice of Knowledge Discovery in Databases, Warsaw, Poland, September 17–21, 2007. Lecture Notes 4702, Springer, Berlin, pp 565–572Google Scholar
  41. Siemon B (2006) Electromagnetic methods: frequency domain airborne techniques. In: Kirsch R (ed) Groundwater geophysics: a tool for hydrogeology. Springer, Berlin, pp 155–170Google Scholar
  42. Smith BD, Abraham JA, Cannia JC, Minsely BJ, Deszcz-Pan M, Ball LB (2009) Helicopter electromagnetic and magnetic geophysical survey data, portions of the North Platte and South Platte Natural Resource Districts, Western Nebraska, May 2009. US Geol Surv Open-File Rep 2010–1259, 33 ppGoogle Scholar
  43. Szidarovszky F, Duckstein L, Bogardi I (1984) Multiobjective management of mining under water hazard by game theory. Eur J Oper Res 15:251–258CrossRefGoogle Scholar
  44. Vesanto J, Alhoniemi E (2000) Clustering of the self-organizing map. IEEE Trans Neural Networks 11:586–600CrossRefGoogle Scholar
  45. Wang S (2003) Application of self-organising maps for data mining with incomplete data sets. Neural Comput Applic 12:42–48CrossRefGoogle Scholar
  46. Witten IH, Eibe F, Hall MA (2011) Data mining: practical machine learning tools and techniques. Elsevier, London, 629 ppGoogle Scholar
  47. Zhu K-G, Ma M-Y, Che H-W, Yang E-W, Yu S-B (2012) PC-based artificial neural network inversion for airborne time-domain electromagnetic data. Appl Geophys 9(1):1–8CrossRefGoogle Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2015

Authors and Affiliations

  • Michael J. Friedel
    • 1
    • 2
  • Akbar Esfahani
    • 3
  • Fabio Iwashita
    • 4
  1. 1.Department of HydrogeologyGNS ScienceLower HuttNew Zealand
  2. 2.Mathematical and Statistical SciencesUniversity of ColoradoDenverUSA
  3. 3.Center for Health Policy ResearchUniversity of CaliforniaLos AngelesUSA
  4. 4.Earth Sciences DepartmentUniversity of FlorenceFlorence, 5Italy

Personalised recommendations