Abstract
Recent developments in uncertainty quantification show that a full inversion of model parameters is not always necessary to forecast the range of uncertainty of a specific prediction in Earth sciences. Instead, Bayesian evidential learning (BEL) uses a set of prior models to derive a direct relationship between data and prediction. This recent technique has been mostly demonstrated for synthetic cases. This paper demonstrates the ability of BEL to predict the posterior distribution of temperature in an alluvial aquifer during a cyclic heat tracer push-pull test. The data set corresponds to another push-pull experiment with different characteristics (amplitude, duration, number of cycles). This experiment constitutes the first demonstration of BEL on real data in a hydrogeological context. It should open the range of future applications of the framework for both scientists and practitioners.
Résumé
Les développements récents en quantification de l’incertitude montrent que l’inversion complète des paramètres d’un modèle n’est pas toujours nécessaire pour prévoir l’intervalle d’incertitude d’une prédiction spécifique en Sciences de la Terre. Une alternative est l’utilisation de la technique Bayesian evidential learning (BEL) qui utilise une série de réalisations a priori pour générer une relation directe entre données et prédiction. Cette technique récente a été démontrée principalement dans des cas synthétiques. Ce papier démontre la capacité de BEL à prédire la distribution a posteriori de la température dans un aquifère alluvial lors d’un test d’injection-pompage de chaleur cyclique. Les données correspondent à un autre test d’injection-pompage avec des caractéristiques différentes (amplitude, durée, nombre de cycles). Cette expérience constitue la première validation de BEL sur des données réelles dans un contexte hydrogéologique. Cela devrait accroître le spectre d’applications futures de la méthode à la fois pour les scientifiques et les professionnels.
Resumen
La evolución reciente de la cuantificación de la incertidumbre muestra que no siempre es necesaria una inversión completa de los parámetros del modelo para prever el margen de incertidumbre de una predicción específica en las Ciencias de la Tierra. En cambio, el aprendizaje bayesiano (BEL) utiliza un conjunto de modelos anteriores para derivar una relación directa entre los datos y la predicción. Esta técnica reciente se ha demostrado principalmente en casos sintéticos. Este trabajo demuestra la capacidad de BEL para predecir la distribución posterior de la temperatura en un acuífero aluvial durante una prueba push-pull del trazador de calor cíclico. El conjunto de datos corresponde a otro experimento push-pull con características diferentes (amplitud, duración, número de ciclos). Este experimento constituye la primera demostración de BEL sobre datos reales en un contexto hidrogeológico. Ello debería abrir un abanico de aplicaciones futuras del marco de trabajo tanto para los científicos como para los profesionales.
摘要
不确定性量化的最新发展表明,模型参数的全反演并不总需要预测地球科学中具体预测的不确定性范围。相反,贝叶斯证据学习(BEL)使用一组先验模型来推导数据和预测之间的直接关系。这种最新技术主要被用于综合案例。本文证明了BEL在循环热示踪推拉试验中预测冲积含水层中温度后验分布的能力。该数据集对应于具有不同特征(振幅,持续时间,循环次数)的另一类推拉实验。该实验是BEL用于水文地质行业真实数据的首个案例。它为科学家和相关专业人员未来开展相关应用提供了参考。
Resumo
Desenvolvimentos recentes na quantificação da incerteza mostram que uma inversão completa dos parâmetros do modelo nem sempre é necessária para prever o intervalo de incerteza de uma previsão específica em Ciências da Terra. Em vez disso, a aprendizagem evidencial bayesiana (AEB) usa um conjunto de modelos anteriores para derivar uma relação direta entre dados e previsão. Esta técnica recente tem sido demonstrada principalmente para casos sintéticos. Este artigo demonstra a capacidade da AEB em predizer a distribuição posterior da temperatura em um aquífero aluvial durante um teste push-pull de traçador de calor cíclico. O conjunto de dados corresponde a outro experimento push-pull com características diferentes (amplitude, duração, número de ciclos). Este experimento constitui a primeira demonstração de AEB em dados reais em um contexto hidrogeológico. Deve abrir o leque de aplicações futuras do quadro para cientistas e profissionais.
Similar content being viewed by others
References
Aquilina L, Vergnaud-Ayraud V, Les Landes AA, Pauwels H, Davy P, Pételet-Giraud E, Labasque T, Roques C, Chatton E, Bour O, Ben Maamar S, Dufresne A, Khaska M, La Salle CLG, Barbecot F (2015) Impact of climate changes during the last 5 million years on groundwater in basement aquifers. Sci Rep 5:14132. https://doi.org/10.1038/srep14132
Bowman AW, Azzalini A (1997) Applied smoothing techniques for data analysis, Oxford Statistical Science Series. Oxford University Press, New York
Brouyère S (2001) Etude et modélisation du transport et du piégeage des solutés en milieu souterrain variablement saturé [Study and modelling of transport and retardation of solutes in variably saturated media]. PhD Thesis, University of Liege, Liege, Belgium
Chen M, Izady A, Abdalla OA, Amerjeed M (2018) A surrogate-based sensitivity quantification and Bayesian inversion of a regional groundwater flow model. J Hydrol 557:826–837. https://doi.org/10.1016/j.jhydrol.2017.12.071
Fenwick D, Scheidt C, Caers J (2014) Quantifying asymmetric parameter interactions in sensitivity analysis: application to reservoir modeling. Math Geosci 46:493–511. https://doi.org/10.1007/s11004-014-9530-5
Goovaerts P (1997) Geostatistics for natural resources evaluation. Applied Geostatistics Series. Oxford University Press, New York
Haggerty R, Schroth MH, Istok JD (1998) Simplified method of “push-pull” test data analysis for determining in situ reaction rate coefficients. Ground Water 36:314–324. https://doi.org/10.1111/j.1745-6584.1998.tb01097.x
Hermans T (2017) Prediction-focused approaches: an opportunity for hydrology. Groundwater 55:683–687
Hermans T, Irving J (2017) Facies discrimination with ERT using a probabilistic methodology: effect of sensitivity and regularization. Near Surf Geophys 15:13–25
Hermans T, Nguyen F, Caers J (2015a) Uncertainty in training image-based inversion of hydraulic head data constrained to ERT data: workflow and case study. Water Resour Res 51:5332–5352. https://doi.org/10.1002/2014WR016460
Hermans T, Wildemeersch S, Jamin P, Orban P, Brouyère S, Dassargues A, Nguyen F (2015b) Quantitative temperature monitoring of a heat tracing experiment using cross-borehole ERT. Geothermics 53:14–26. https://doi.org/10.1016/j.geothermics.2014.03.013
Hermans T, Oware EK, Caers J (2016) Direct prediction of spatially and temporally varying physical properties from time-lapse electrical resistance data. Water Resour Res 52:7262–7283
Hermans T, Nguyen F, Klepikova M, Dassargues A, Caers J (2018) Uncertainty quantification of medium-term heat storage from short-term geophysical experiments using Bayesian evidential learning. Water Resour Res 54:2931–2948. https://doi.org/10.1002/2017WR022135
Hou Z, Rubin Y (2005) On minimum relative entropy concepts and prior compatibility issues in vadose zone inverse and forward modeling. Water Resour Res 41:WR004082. https://doi.org/10.1029/2005WR004082
Jamin P, Brouyère S (2018) Monitoring transient groundwater fluxes using the finite volume point dilution method. J Contam Hydrol. https://doi.org/10.1016/j.jconhyd.2018.07.005
Kammen DM, Sunter DA (2016) City-integrated renewable energy for urban sustainability. Science 352:922–928
Klepikova M, Wildemeersch S, Hermans T, Jamin P, Orban P, Nguyen F, Brouyère S, Dassargues A (2016) Heat tracer test in an alluvial aquifer: field experiment and inverse modelling. J Hydrol 540:812–823. https://doi.org/10.1016/j.jhydrol.2016.06.066
Krzanowski WJ (2000) Principles of multivariate analysis: A user’s perspective, Oxford Statistical Series 22, Revised Edition. Oxford University Press, New York
Lesparre N, Robert T, Nguyen F, Boyle A, Hermans T (2019) 4D electrical resistivity tomography (ERT) for aquifer thermal energy storage monitoring. Geothermics 77:368–382
MacDonald AM, Bonsor HC, Ahmed KM, Burgess WG, Basharat M, Calow RC, Dixit A, Foster SSD, Gopal K, Lapworth DJ, Lark RM, Moench M, Mukherjee A, Rao MS, Shamsudduha M, Smith L, Taylor RG, Tucker J, van Steenbergen F, Yadav SK (2016) Groundwater quality and depletion in the Indo-Gangetic Basin mapped from in situ observations. Nat Geosci 9:762–766. https://doi.org/10.1038/ngeo2791
Paradis CJ, McKay LD, Perfect E, Istok JD, Hazen TC (2018) Push-pull tests for estimating effective porosity: expanded analytical solution and in situ application. Hydrogeol J 26:381–393. https://doi.org/10.1007/s10040-017-1672-3
Park J, Yang G, Satija A, Scheidt C, Caers J (2016) DGSA: A Matlab toolbox for distance-based generalized sensitivity analysis of geoscientific computer experiments. Comput Geosci 97:15–29. https://doi.org/10.1016/j.cageo.2016.08.021
Razavi S, Tolson BA, Burn DH (2012) Review of surrogate modeling in water resources. Water Resour Res 48:W07401. https://doi.org/10.1029/2011WR011527
Réseau National de Sites Hydrogéologiques (2019) Network of hydrogeological research sites: ENIGMA – Data Hermalle. http://hplus.ore.fr/en/enigma/data-hermalle. Accessed 30 June 2018
Satija A, Caers J (2015) Direct forecasting of subsurface flow response from non-linear dynamic data by linear least-squares in canonical functional principal component space. Adv Water Resour 77:69–81. https://doi.org/10.1016/j.advwatres.2015.01.002
Satija A, Scheidt C, Li L, Caers J (2017) Direct forecasting of reservoir performance using production data without history matching. Comput Geosci 21:315–333. https://doi.org/10.1007/s10596-017-9614-7
Scheidt C, Jeong C, Mukerji T, Caers J (2015a) Probabilistic falsification of prior geologic uncertainty with seismic amplitude data: application to a turbidite reservoir case. Geophysics 80:M89–M100. https://doi.org/10.1190/geo2015-0084.1
Scheidt C, Renard P, Caers J (2015b) Prediction-focused subsurface modeling: investigating the need for accuracy in flow-based inverse modeling. Math Geosci 47:173–191. https://doi.org/10.1007/s11004-014-9521-6
Scheidt C, Li L, Caers J (2018) Quantifying uncertainty in subsurface systems, Geophysical Monograph Series. Wiley, Hoboken, NJ and AGU, Washington, DC
Therrien R, McLaren R, Sudicky E, Panday S (2010) HydroGeoSphere: a three-dimensional numerical model describing fully-integrated subsurface and surface flow and solute transport. Groundwater Simulation Group, Waterloo, ON
Vandenbohede A, Louwyck A, Lebbe L (2009) Conservative solute versus heat transport in porous media during push-pull tests. Transp Porous Media 76:265–287. https://doi.org/10.1007/s11242-008-9246-4
Wildemeersch S, Jamin P, Orban P, Hermans T, Klepikova M, Nguyen F, Brouyère S, Dassargues A (2014) Coupling heat and chemical tracer experiments for estimating heat transfer parameters in shallow alluvial aquifers. J Contam Hydrol 169:90–99. https://doi.org/10.1016/j.jconhyd.2014.08.001
Acknowledgements
We thank Thomas Kremer, Maxime Evrard and Solomon Eshioke for their precious help in the field and Frédéric Nguyen and Jef Caers for fruitful discussions. We thank the associated editor and four anonymous reviewers for their constructive comments.
Funding
Field experiments were possible thanks to the F.R.S.-FNRS research credit 4D Thermography, grant number J.0045.16. N. Lesparre and G. De Schepper were supported by the project SUITE4D from the BEWARE Fellowships Academia Program (contract No. 1510466) and the SMARTMODEL project from the BEWARE Fellowships Industry Program (contract No. 1610056), respectively. Both programs are co-financed by the department of Research Programs of the Wallonia-Brussels Federation and Marie Skłodowska-Curie COFUND program of the European Union.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Hermans, T., Lesparre, N., De Schepper, G. et al. Bayesian evidential learning: a field validation using push-pull tests. Hydrogeol J 27, 1661–1672 (2019). https://doi.org/10.1007/s10040-019-01962-9
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10040-019-01962-9