Skip to main content

Advertisement

Log in

Performance comparison of physical process-based and data-driven models: a case study on the Edwards Aquifer, USA

Comparaison des performances des modèles basés sur les processus physiques et ceux sur les données: une étude de cas sur l’aquifère Edwards, États-unis d’Amérique

Comparación del rendimiento de los modelos basados en procesos físicos y datos: un estudio de caso del Acuífero Edwards, EEUU

基于物理过程的模型和数据驱动的模型比较:Edwards含水层的一个实例研究

Comparação de desempenho de modelos físicos e baseados em dados: um estudo de caso no Aquífero Edwards, EUA

  • Paper
  • Published:
Hydrogeology Journal Aims and scope Submit manuscript

Abstract

Physical process-based groundwater flow models are the major tools for studying fluid-flow behavior and for simulating the hydrological responses of water levels and spring discharge to human- and/or nature-induced triggers such as pumping and recharge. Such models are built with deep understanding of the physical processes and are based on geological models, developed by integrating data from geology, geophysics, and geochemistry. However, data-driven models can be built with limited data, eliminating the need for a detailed understanding of the physics. In this research, a data-driven model is built for simulating hydraulic responses (both groundwater levels and spring discharges) in a complex groundwater flow system of the Edwards Aquifer in Central Texas, USA, with the recurrent neural network (RNN) technique. The model is first trained and validated with the observation data of four targets—water levels from two index wells, and spring flow rates from San Marcos and Comal springs—from 2001 through 2015. The model is then used to predict the hydrological responses for the drought of record (1947–1958). The performance of the RNN model for the training, validation and prediction period is then quantitatively compared to that of the physical process-based MODFLOW model in terms of four statistical measures. The statistical measures suggest that the RNN model performs almost as well as the MODFLOW model. With further improvements, a data-driven model may be a surrogate to (or integrate with) a physical process-based model for simulating hydrological responses in the Edwards Aquifer.

Résumé

Les modèles d’écoulement des eaux souterraines basés sur les processus physiques sont les principaux outils permettant d’étudier le comportement de l’écoulement des fluides et de simuler les réponses hydrologiques des niveaux d’eau et du débit des sources à des éléments déclencheurs induits par l’homme et/ou la nature, tels que le pompage et la recharge. Ces modèles sont construits grâce à une compréhension profonde des processus physiques et sont basés sur des modèles géologiques, développés en intégrant des données de géologie, de géophysique et de géochimie. Cependant, les modèles basés sur des données peuvent être construits avec des données limitées, ce qui élimine la nécessité d’une compréhension détaillée de la physique. Dans cette recherche, un modèle basé sur les données est construit pour simuler les réponses hydrauliques (à la fois les niveaux des eaux souterraines et les débits de sources) dans un système complexe d’écoulement des eaux souterraines de l’aquifère Edwards dans le centre du Texas, aux États-Unis d’Amérique, avec la technique du réseau neuronal récurrent (RNR). Le modèle est d’abord entrainé et validé avec les données d’observation de quatre cibles (niveaux d’eau de deux puits, et débits des sources de San Marcos et Comal) de 2001 à 2015. Le modèle est ensuite utilisé pour prédire les réponses hydrologiques pour la sécheresse enregistrée (1947–1958). La performance du modèle RNR pour la période d’entrainement, de validation et de prévision est ensuite comparée quantitativement à celle du modèle MODFLOW basé sur les processus physiques en termes de quatre mesures statistiques. Les mesures statistiques suggèrent que le modèle RNR fonctionne presque aussi bien que le modèle MODFLOW. Avec de nouvelles améliorations, un modèle basé sur les données peut être un substitut (ou s’intégrer) à un modèle basé sur les processus physiques pour simuler les réponses hydrologiques dans l’aquifère d’Edwards.

Resumen

Los modelos de flujo de aguas subterráneas basados en procesos físicos son las principales herramientas para estudiar el comportamiento del flujo de los fluidos y para simular las respuestas hidrológicas de los niveles de agua y la descarga de los manantiales a los activadores inducidos por el hombre y/o la naturaleza, como el bombeo y la recarga. Esos modelos se construyen con un profundo conocimiento de los procesos físicos y se basan en modelos geológicos, desarrollados mediante la integración de datos de geología, geofísica y geoquímica. Sin embargo, los modelos basados en datos pueden construirse con datos limitados, eliminando la necesidad de una comprensión detallada de la física. En esta investigación se construye un modelo basado en datos para simular las respuestas hidráulicas (tanto los niveles de las aguas subterráneas como las descargas de los manantiales) en un complejo sistema de flujo de aguas subterráneas del Acuífero Edwards en el centro de Texas (EEUU), con la técnica de la red neural recurrente (RNN). El modelo se entrena y valida primero con los datos de observación de cuatro objetivos (niveles de agua de dos pozos de índice, y caudales de manantial de los manantiales de San Marcos y Comal) desde 2001 hasta 2015. El modelo se utiliza luego para predecir las respuestas hidrológicas a la sequía registrada (1947–1958). El rendimiento del modelo RNN para el período de entrenamiento, validación y predicción se compara entonces cuantitativamente con el del modelo MODFLOW basado en el proceso físico en términos de cuatro medidas estadísticas. Las medidas estadísticas sugieren que el modelo RNN funciona casi tan bien como el modelo MODFLOW. Con otras mejoras, un modelo basado en datos puede ser un sustituto de un modelo basado en procesos físicos (o integrarse en él) para simular las respuestas hidrológicas en el acuífero Edwards.

摘要

基于物理过程的地下水流模型是研究地下水流以及模拟受人为或自然因素影响(例如开采和补给)对地下水位及泉流量响应的主要工具。这类地下水流模型是建立于地质模型的基础上,需要对地下水流物理过程的深入理解,通过对地质、地球物理以及地球化学的数据集成开发而成的。但是,数据驱动模型可以建立在有限的数据基础上,而不需要详细理解物理过程。本次研究利用循环神经网络技术 (RNN)建立了一个数据驱动模型用于模拟美国Edwards含水层复杂地下水系统的水利响应 (包括地下水位及泉流量)。首先,利用2001年至2015年的四个目标数据(两个地下水位观测井的地下水数据及San Marcos泉和Comal泉的流量观测数据)来训练和验证RNN模型。然后,这个RNN模型用于预测干旱期(1947–1958)的地下水响应。四个统计参数指标用于对比RNN模型和物理过程的MODFLOW模型的训练、验证和预测。统计参数指标表明RNN模型可以获得和MODFLOW模型一样的结果。通过进一步改进,数据驱动模型可以作为替代(或集成于)物理过程的模型用于模拟Edwards含水层的水力响应。

Resumo

Os modelos físicos de fluxo de água subterrânea baseados no processo são as principais ferramentas para estudar o comportamento do fluxo de fluido e para simular as respostas hidrológicas dos níveis de água e descarga de nascentes a gatilhos induzidos pela natureza e/ou pela natureza, como bombeamento e recarga. Tais modelos são construídos com profundo conhecimento dos processos físicos e são baseados em modelos geológicos, desenvolvidos pela integração de dados de geologia, geofísica e geoquímica. No entanto, modelos orientados a dados podem ser construídos com dados limitados, eliminando a necessidade de um entendimento detalhado da física. Nesta pesquisa, um modelo orientado a dados é construído para simular respostas hidráulicas (níveis de águas subterrâneas e descargas de nascentes) em um complexo sistema de fluxo de águas subterrâneas do Aquífero Edwards no Texas Central, EUA, com a técnica de rede neural recorrente (RNR). O modelo é primeiro treinado e validado com os dados de observação de quatro alvos (níveis de água de dois poços índices e vazões de nascentes das nascentes San Marcos e Comal) de 2001 a 2015. O modelo é então usado para prever as respostas hidrológicas para a seca recorde (1947–1958). O desempenho do modelo de RNR para o período de treinamento, validação e previsão é então comparado quantitativamente ao desempenho do modelo MODFLOW baseado em processos físicos em termos de quatro medidas estatísticas. As medidas estatísticas sugerem que o modelo de RNR funciona quase tão bem quanto o modelo MODFLOW. Com outras melhorias, um modelo orientado a dados pode ser um substituto para (ou integrar-se a) um modelo físico baseado em processo para simular respostas hidrológicas no aquífero Edwards.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5
Fig. 6
Fig. 7

Similar content being viewed by others

References

  • Afzaal H, Farooque AA, Abbas F, Acharya B, Esau T (2019) Groundwater estimation from major physical hydrology components using artificial neural networks and deep learning. Water 12:5. https://doi.org/10.3390/w12010005

    Article  Google Scholar 

  • Asher MJ, Croke BFW, Jakeman AJ, Peeters LJM (2015) A review of surrogate models and their application to groundwater modeling. Water Resour Res 51:5957–5973. https://doi.org/10.1002/2015WR016967

    Article  Google Scholar 

  • Barker RA, Ardis AF (1992) Configuration of the base of the Edwards-Trinity aquifer system and hydrogeology of the underlying pre-Cretaceous rocks, west-central Texas. US Geol Surv Water Resour Invest Rep 91-4071

  • Bear J (1972) Dynamics of fluid in porous media. Elsevier, New York

    Google Scholar 

  • Bianchi FM, Maiorino E, Kampffmeyer MC, Rizzi A, Jenssen R (2017) Recurrent neural networks for short-term load forecasting. SpringerBriefs Comput Sci. https://doi.org/10.1007/978-3-319-70338-1

  • Campana ME, Mahin DA (1985) Model-derived estimates of groundwater mean ages, recharge rates, effective porosities and storage in a limestone aquifer. J Hydrol 76:247–264. https://doi.org/10.1016/0022-1694(85)90136-2

    Article  Google Scholar 

  • Chang H, Zhang D (2019) Machine learning subsurface flow equations from data. Comput Geosci 23:895–910. https://doi.org/10.1007/s10596-019-09847-2

    Article  Google Scholar 

  • Chang L-C, Chang F-J, Chiang Y-M (2004) A two-step-ahead recurrent neural network for stream-flow forecasting. Hydrol Process 18:81–92. https://doi.org/10.1002/hyp.1313

    Article  Google Scholar 

  • Chitsazan M, Rahmani G, Neyamadpour A (2015) Forecasting groundwater level by artificial neural networks as an alternative approach to groundwater modeling. J Geol Soc India 85:98–106

  • Collins EW, Hovorka SD (1997) Structure map of the San Antonio segment of the Edwards Aquifer and Balcones Fault Zone, South-Central Texas. Structural framework of a major limestone aquifer: Kinney, Uvalde, Median, Bexar, Comal, and Hays counties, Austin, TX. University of Texas Bureau of Economic Geology, Austin, TX

  • Coulibaly P, Anctil F, Aravena R, Bobée B (2001) Artificial neural network modeling of water table depth fluctuations. Water Resour Res 37:885–896. https://doi.org/10.1029/2000WR900368

    Article  Google Scholar 

  • Dash NB, Panda SN, Remesan R, Sahoo N (2010) Hybrid neural modeling for groundwater level prediction. Neural Comput Appl 19:1251–1263. https://doi.org/10.1007/s00521-010-0360-1

    Article  Google Scholar 

  • Doherty J (2004) PEST: model-independent parameter estimation user manual, 5th edn. Watermark, Brisbane, Australia

  • Fratesi SB, Green RT, Bertetti FP, McGinnis RN, Toll N, Başağaoğlu H, Gergen L, Winterle JR (2015) Development of a finite-element method groundwater flow model for the Edwards Aquifer. SwRIProject no. 20-17344, Southwest Research Institute, San Antonio, TX

  • Ghose DK, Panda SS, Swain PC (2010) Prediction of water table depth in western region, Orissa using BPNN and RBFN neural networks. J Hydrol 394:296–304. https://doi.org/10.1016/j.jhydrol.2010.09.003

    Article  Google Scholar 

  • Gong Y, Zhang Y, Lan S, Wang H (2016) A comparative study of artificial neural networks, support vector machines and adaptive neuro fuzzy inference system for forecasting groundwater levels near Lake Okeechobee. Florida Water Resour Manag 30:375–391. https://doi.org/10.1007/s11269-015-1167-8

    Article  Google Scholar 

  • Harbaugh AW, Banta ER, Hill MC, McDonald MG, Geological S (2000) MODFLOW-2000. The U.S. geological survey modular ground-water model - user guide to modularization concepts and the ground-water flow process. Open-file Report

  • Hunt AG, Landis GP, Faith JR (2016) Groundwater ages from the freshwater zone of the Edwards Aquifer, Uvalde County, Texas: insights into groundwater flow and recharge. US Geol Surv Sci Invest Rep 2015-5163. https://doi.org/10.3133/sir20155163

  • Klemt WB, Knowles TR, Edler GR, Sieh TW (1979) Ground-water resources and model applications for the Edwards (Balcones fault zone) aquifer in the San Antonio region. Texas Water Development Board, Austin, TX. https://www.twdb.texas.gov/publications/reports/other_reports/doc/Edwards%20BFZ.pdf. Accessed April 2020

  • Kuniansky EL, Spangler LE (2017) US Geological Survey Karst Interest Group Proceedings, San Antonio, Texas, May 16–18, 2017. US Geol Surv Sci Invest Rep 2017-5023. https://doi.org/10.3133/sir20175023

  • Land LS, Prezbindowski DR (1981) The origin and evolution of saline formation water, lower Cretaceous carbonates, South-Central Texas, USA. J Hydrol 54:51–74

    Article  Google Scholar 

  • Lindgren RJ, Dutton AR, Hovorka SD, Worthington SRH, Painter S (2004) Conceptualization and simulation of the Edwards Aquifer, San Antonio region, Texas. US Geol Surv Sci Invest Rep 2004-5277

  • Lindgren RJ, Taylor CJ, Houston NA (2009) Description and evaluation of numerical groundwater flow models for the Edwards Aquifer, South-Central Texas. US Geol Surv Sci Invest Rep 2009-5183

  • Liu, A, Troshanov N, Winterle J, Zhang A, Eason S (2017) Updates to the MODFLOW groundwater model of the San Antonio segment of the Edwards Aquifer. Edwards Aquifer Authority, San Antonio, TX

  • Lynch DR (1984) Mass conservation in finite element groundwater models. Adv Water Resour 7:67–75. https://doi.org/10.1016/0309-1708(84)90003-4

    Article  Google Scholar 

  • Maclay RW, Land LF (1987) Simulation of flow in the Edwards Aquifer, San Antonio region, Texas, and refinement of storage and flow concepts. US Geol Surv Open-File Rep 2336-A

  • McCarl BA, Dillon CR, Keplinger KO, Williams RL (1999) Limiting pumping from the Edwards Aquifer: an economic investigation of proposals, water markets, and spring flow guarantees. Water Resour Res 35:1257–1268. https://doi.org/10.1029/1998WR900116

    Article  Google Scholar 

  • Mekonnen BA, Nazemi A, Mazurek KA, Elshorbagy A, Putz G (2015) Hybrid modelling approach to prairie hydrology: fusing data-driven and process-based hydrological models. Hydrol Sci J 60:1473–1489. https://doi.org/10.1080/02626667.2014.935778

    Article  Google Scholar 

  • Mo S, Zhu Y, Zabaras N, Shi X, Wu J (2019) Deep convolutional encoder-decoder networks for uncertainty quantification of dynamic multiphase flow in heterogeneous media. Water Resour Res 55:703–728. https://doi.org/10.1029/2018WR023528

    Article  Google Scholar 

  • Mohammadi K (2009) Groundwater table estimation using MODFLOW and artificial neural networks. In: Practical hydroinformatics: computational intelligence and technological developments in water applications. Water Science and Technology Library book series, Springer, Heidelberg, Germany, pp 127–138

  • Mohanty S, Jha MK, Kumar A, Panda DK (2013) Comparative evaluation of numerical model and artificial neural network for simulating groundwater flow in Kathajodi–Surua inter-basin of Odisha, India. J Hydrol 495:38–51. https://doi.org/10.1016/j.jhydrol.2013.04.041

    Article  Google Scholar 

  • Nourani V, Mousavi S, Sadikoglu F, Singh VP (2017) Experimental and AI-based numerical modeling of contaminant transport in porous media. J Contam Hydrol 205:78–95. https://doi.org/10.1016/j.jconhyd.2017.09.006

    Article  Google Scholar 

  • Pearson Jr. FJ, Rettman PL (1976) Geochemical and isotopic analyses of waters associated with the Edwards Limestone aquifer, Central Texas. Edwards Underground Water District, San Antonio, TX

  • Puente C (1975) Relation of precipitation to annual ground-water recharge in the Edwards Aquifer, San Antonio area, Texas. US Geol Surv Open-File Rep 75-298

  • Puente C (1978) Method of estimating natural recharge to the Edwards Aquifer in the San Antonio area, Texas. Water-Resources Investigations Report. Austin, TX

  • Rajaee T, Ebrahimi H, Nourani V (2019) A review of the artificial intelligence methods in groundwater level modeling. J Hydrol 572:336–351. https://doi.org/10.1016/j.jhydrol.2018.12.037

    Article  Google Scholar 

  • Sahoo BB, Jha R, Singh A, Kumar D (2019) Long short-term memory (LSTM) recurrent neural network for low-flow hydrological time series forecasting. Acta Geophys 67:1471–1481. https://doi.org/10.1007/s11600-019-00330-1

    Article  Google Scholar 

  • Sahoo S, Russo TA, Elliott J, Foster I (2017) Machine learning algorithms for modeling groundwater level changes in agricultural regions of the U.S. Water Resour Res 53:3878–3895. https://doi.org/10.1002/2016WR019933

    Article  Google Scholar 

  • Sanford W (2002) Recharge and groundwater models: an overview. Hydrogeol J 10:110–120. https://doi.org/10.1007/s10040-001-0173-5

    Article  Google Scholar 

  • Scanlon BR, Mace RE, Barrett ME, Smith B (2003) Can we simulate regional groundwater flow in a karst system using equivalent porous media models? Case study, Barton Springs Edwards Aquifer, USA. J Hydrol 276:137–158. https://doi.org/10.1016/S0022-1694(03)00064-7

    Article  Google Scholar 

  • Sharp JMJ, Banner JL (1997) The Edwards Aquifer: a resource in conflict. GSA Today 7:1–8

    Google Scholar 

  • Shen C, Laloy E, Elshorbagy A, Albert A, Bales J, Chang F-J, Ganguly S, Hsu K-L, Kifer D, Fang Z, Fang K, Li D, Li X, Tsai W-P (2018) HESS opinions: incubating deep-learning-powered hydrologic science advances as a community. Hydrol Earth Syst Sci 22:5639–5656. https://doi.org/10.5194/hess-22-5639-2018

    Article  Google Scholar 

  • Smiley D (2012) PyNeurGen: Python Neural Genetic Algorithm Hybrids. https://sourceforge.net/projects/pyneurgen/. Accessed April 2020

  • Taver V, Johannet A, Borrell-Estupina V, Pistre S (2015) Feed-forward vs recurrent neural network models for non-stationarity modelling using data assimilation and adaptivity. Hydrol Sci J 60:1242–1265. https://doi.org/10.1080/02626667.2014.967696

    Article  Google Scholar 

  • Thorkildsen D, McElhaney PD (1992) Model refinement and applications for the Edwards (Balcones Fault Zone) Aquifer in the San Antonio Region, Texas. Texas Water Development Board, Austin, TX. https://texashistory.unt.edu/ark:/67531/metapth576336/. Accessed April 2020

  • Todd Engineers (1999) Optimization technical studies in support of the Edwards Aquifer Optimization Program. Todd Engineers, Emeryville, CA

  • Trichakis IC, Nikolos IK, Karatzas GP (2011) Artificial neural network (ANN) based modeling for karstic groundwater level simulation. Water Resour Manag 25:1143–1152. https://doi.org/10.1007/s11269-010-9628-6

    Article  Google Scholar 

  • Wunsch A, Liesch T, Broda S (2018) Forecasting groundwater levels using nonlinear autoregressive networks with exogenous input (NARX). J Hydrol 567:743–758. https://doi.org/10.1016/j.jhydrol.2018.01.045

    Article  Google Scholar 

  • Xu T, Valocchi AJ (2015) Data-driven methods to improve baseflow prediction of a regional groundwater model. Comput Geosci 85:124–136. https://doi.org/10.1016/j.cageo.2015.05.016

    Article  Google Scholar 

  • Yang C, Samper J, Molinero J, Bonilla M (2007) Modelling geochemical and microbial consumption of dissolved oxygen after backfilling a high level radioactive waste repository. J Contam Hydrol 93:130–148

    Article  Google Scholar 

  • Zhang J, Zhu Y, Zhang X, Ye M, Yang J (2018) Developing a long short-term memory (LSTM) based model for predicting water table depth in agricultural areas. J Hydrol 561:918–929. https://doi.org/10.1016/j.jhydrol.2018.04.065

    Article  Google Scholar 

  • Zhou T, Wang F, Yang Z (2017) Comparative analysis of ANN and SVM models combined with wavelet preprocess for groundwater depth prediction. Water 9:1–21. https://doi.org/10.3390/w9100781

    Article  Google Scholar 

  • Zhou Y, Li W (2011) A review of regional groundwater flow modeling. Geosci Front 2:205–214. https://doi.org/10.1016/j.gsf.2011.03.003

    Article  Google Scholar 

Download references

Acknowledgements

The authors would like to thank the editors and the anonymous reviewers for their useful comments, which helped to improve this paper.

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Changbing Yang.

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Zhang, A., Winterle, J. & Yang, C. Performance comparison of physical process-based and data-driven models: a case study on the Edwards Aquifer, USA. Hydrogeol J 28, 2025–2037 (2020). https://doi.org/10.1007/s10040-020-02169-z

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s10040-020-02169-z

Keywords

Navigation