Hydrogeology Journal

, Volume 20, Issue 4, pp 727–738 | Cite as

Sensitivity analysis of groundwater level in Jinci Spring Basin (China) based on artificial neural network modeling

  • Xian Li
  • Longcang Shu
  • Lihong Liu
  • Dan Yin
  • Jinmei Wen
Paper
First Online:
Received:
Accepted:

Abstract

Jinci Spring in Shanxi, north China, is a major local water source. It dried up in April 1994 due to groundwater overexploitation. The groundwater system is complex, involving many nonlinear and uncertain factors. Artificial neural network (ANN) models are statistical techniques to study parameter nonlinear relationships of groundwater systems. However, ANN models offer little explanatory insight into the mechanisms of prediction models. Sensitivity analysis can overcome this shortcoming. In this study, a back-propagation neural network model was built based on the relationship between groundwater level and its sensitivity factors in Jinci Spring Basin; these sensitivity factors included precipitation, river seepage, mining drainage, groundwater withdrawals and lateral discharge to the associated Quaternary aquifer. All the sensitivity factors were analyzed with Garson’s algorithm based on the connection weights of the neural network model. The concept of “sensitivity range” was proposed to describe the value range of the input variables to which the output variables are most sensitive. The sensitivity ranges were analyzed by a local sensitivity approach. The results showed that coal mining drainage is the most sensitive anthropogenic factor, having a large effect on groundwater level of the Jinci Spring Basin.

Keywords

Artificial neural networks (ANN) Sensitivity analysis China Groundwater flow dynamic Groundwater management 

Analyse de sensibilité des niveaux d’eau souterrains du Bassin de la Source Jinci (Chine) basée sur une modélisation par réseaux neuronaux artificiels

Résumé

La Source Jinci dans le Shanxi, nord de la Chine, est une importante ressource en eau locale. Elle s’est asséchée en avril 1994 du fait de la surexploitation des eaux souterraines. Le système aquifère est complexe, impliquant de nombreux facteurs non-linéaires et incertains. Les modèles par réseaux neuronaux artificiels (RNA) sont des techniques statistiques qui permettent d’étudier les paramètres des relations non-linéaires des systèmes aquifères. Toutefois, les modèles RNA n’offrent qu’une faible partie de la vision explicative des mécanismes des modèles prédictifs. Les analyses de sensibilité peuvent surmonter ce défaut. Dans cette étude, un modèle par réseaux neuronaux à rétro propagation a été construit à partir de la relation entre le niveau des eaux souterraines et de ses facteurs de sensibilité dans le Bassin de la Source Jinci. Ces facteurs comprennent : les précipitations, l’infiltration des eaux de la rivière, le drainage par les mines, les prélèvements dans la nappe et la vidange latérale vers l’aquifère quaternaire associé. Tous les facteurs de sensibilité ont été analysés avec l’algorithme de Garson basé sur les pondérations de connexion du réseau neuronal. Le concept de «plage de sensibilité» a été proposé pour décrire la plage de valeurs des variables d’entrées auxquelles les variables de sorties sont les plus sensibles. Les plages de sensibilités ont été étudiées par une analyse de sensibilité locale. Les résultats montrent que le drainage des mines de charbon est le facteur anthropique le plus sensible, ayant un impact fort sur les niveaux souterrains du Bassin de la Source Jinci.

Análisis de sensibilidad de niveles de agua subterránea en Jinci Spring Basin (China) basado en la modelación con redes neuronales artificiales

Resumen

El manantial Jinci en Shanxi, norte de China, es una fuente importante local de agua. Se secó en abril de 1994 debido a la sobreexplotación del agua subterránea. El sistema de agua subterránea es complejo, involucra muchos factores no lineares e inciertos. Los modelos de red neuronales artificiales (ANN) son técnicas estadísticas para estudiar relaciones no lineares paramétricas de los sistemas de agua subterránea. Sin embargo, los modelos ANN ofrecen poco visión explicativa de los mecanismos de los modelos predictivos. El análisis de sensibilidad puede superar esta deficiencia. En este estudio, se construyó un modelo de redes neuronales de retro-propagación basado en la región entre niveles de agua subterránea y sus factores de sensibilidad en la cuenca del manantial Jinci; estos factores de sensibilidad incluyeron a la precipitación, filtración del río, drenaje de minas, extracción de agua subterránea y descarga lateral para el acuífero Cuaternario asociado. Todos los factores de sensibilidad fueron analizados con el algoritmo de Garson basado en los pesos de conexión del modelo de la red neuronal. Se propuso el concepto de “intervalo de sensibilidad” para describir el intervalo de las variables de entrada a las cuales las variables de salida son más sensibles. Los intervalos de sensibilidad fueron analizados por una aproximación de sensibilidad local. Los resultados mostraron que el drenaje de una mina de carbón es el factor antropogénico más sensible, y que tiene un gran efecto sobre los niveles de agua subterránea de la cuenca del manantial Jinci.

基于人工神经网络模型的中国晋祠泉流域地下水位敏感性分析

摘要

中国华北山西省的晋祠泉是当地的主要水源。由于地下水的过量开采, 在1994年4月晋祠泉枯竭了。地下水系统是很复杂的, 受到许多非线性的和不确定的因素的影响。人工神经网络模型利用统计学的方法来研究地下水系统参数之间的非线性关系。但是, 人工神经网络模型对预测模型的机制揭示的并不多。敏感性分析可以克服这些不足。本次研究中, 基于晋祠泉流域地下水位和敏感性因素之间的关系建立了反向传播的神经网络模型, 这些敏感性因素包括降雨,河流渗漏, 矿区排水, 地下水开采和向有水力联系的第四系含水层的侧向排泄。所有这些敏感性因素都是利用基于神经网络模型连接权重的加森算法进行分析的。“敏感性范围”这个概念是用来描述输入变量值的变化范围的,输出变量对这些输入变量是最敏感的。敏感性范围是利用局部敏感性分析方法进行分析的。分析结果表明, 煤矿排水是最敏感的人为因素, 对晋祠泉流域地下水位的影响非常大。

Análise de sensibilidade dos níveis piezométricos na Bacia da Nascente de Jinci (China), baseada em modelação por redes neuronais artificiais

Resumo

A Nascente de Jinci, em Shanxi, no norte da China, é uma fonte principal de água doce local. Devido à sobreexploração de águas subterrâneas, essa fonte esgotou-se em Abril de 1994. O sistema hidrogeológico é complexo, envolvendo muitos fatores incertos e não lineares. As redes neuronais artificiais (RNA) são técnicas estatísticas que permitem estudar as relações não-lineares de sistemas hidrogeológicos. No entanto, os modelos RNA fornecem uma visão incompleta dos mecanismos dos modelos de previsão. A análise de sensibilidade pode ultrapassar este problema. Neste estudo, foi construída uma RNA de propagação retroativa (back-propagation), baseada na relação entre o nível piezométrico e os fatores de sensibilidade na Bacia da Nascente de Jinci; estes fatores de sensibilidade incluem a precipitação, o escoamento subsuperficial, a drenagem por explorações mineiras, a extração de água subterrânea e a descarga lateral para um aquífero Quaternário associado. Todos os fatores de sensibilidade foram analisados com o algoritmo de Garson, baseado em ponderadores de conexão do modelo de rede neuronal. É proposto o conceito de “faixa de sensibilidade” para descrever a gama de valores das variáveis de entrada para as quais as variáveis de saída são as mais sensíveis. As gamas de sensibilidade foram analisadas com base numa abordagem de sensibilidade local. Os resultados mostraram que a drenagem proveniente da exploração das minas de carvão é o fator antropogénico mais sensível, influenciando fortemente o nível de águas subterrâneas na Bacia da Nascente de Jinci.

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

Notes

Acknowledgements

The authors gratefully acknowledge Taiyuan Water Authority for providing the field data. Special thanks go to the reviewers for their useful comments which helped improve this paper. The present work was financially supported by a program for Changjiang Scholars and Innovative Research Team in University (PCSIRT), the “111” Project under Grant B08048 supported by the Ministry of Education and State Administration of Foreign Experts Affairs (PR China), and the National Natural Science Foundation of China (70711120407).

References

  1. Bakker M, Maas K et al (2007) Analytic modeling of groundwater dynamics with an approximate impulse response function for a real recharge. Adv Water Resour 30:493–504CrossRefGoogle Scholar
  2. Chaves P, Chang FJ (2008) Intelligent reservoir operation system based on evolving artificial neural networks. Adv Water Resour 31:926–936CrossRefGoogle Scholar
  3. Coulibaly P, Anctil F, Aravena R, Bobée B (2001a) Artificial neural network modeling of water table depth fluctuations. Water Resour Res 37(4):885–896CrossRefGoogle Scholar
  4. Coulibaly P, Bobee B, Anctil F (2001b) Improving extreme hydrologic events forecasting using a new criterion for artificial neural network selection. Hydrol Proce 15:1533–1536CrossRefGoogle Scholar
  5. Daliakopoulos IN, Coulibaly P, Ik T (2005) Groundwater level forecasting using artificial neural networks. J Hydrol 309:229–240CrossRefGoogle Scholar
  6. Du B (2010) Study on Fenhe River and groundwater interaction in Xishan karst region of Taiyuan City (in Chinese). J Taiyuan Univ Technol 41(3):272–277Google Scholar
  7. Erb RJ (1993) Introduction to backpropagation neural network computation. Pharm Res 10(2):165–170CrossRefGoogle Scholar
  8. Garson GD (1991) Interpreting neural network connection weights. Artificial Intell Expert 6:47–51Google Scholar
  9. Gevrey M, Dimopoulos IN, Lek S (2003) Review and comparison of methods to study the contribution of variables in artificial neural network models. Ecol Model 160:249–264CrossRefGoogle Scholar
  10. Goh ATC (1995) Back-propagation neural networks for modeling complex systems. Artif Intell Eng 9:143–151CrossRefGoogle Scholar
  11. Hao YH, Yeh TJ, Gao ZQ et al (2006) A gray system model for studying the response to climatic change: the Liulin karst springs, China. J Hydrol 328:668–676CrossRefGoogle Scholar
  12. He YB (1982) Some characteristics of karst water in north China (in Chinese). Carsologica Sinica 2:135–142Google Scholar
  13. He YB, Zou CJ (1996) Comparison of karst-water characteristics between southern and northern China (in Chinese). J Carsologica Sinca 15(3):259–268Google Scholar
  14. Hecht-Nielsen R (1987) Kolmogorov’s mapping neural network existence theorem. In: Proc. First IEEE Int. Joint Conf. Neural Networks, San Diego, CA, June 1987, pp III11–III14Google Scholar
  15. Huang HL (2003) Cessation of flow at Jinci Spring and protection of groundwater resources (in Chinese). Coal Geol China 15(2):26–28Google Scholar
  16. Johonson VM, Rogers L (2000) Accuracy of neural network approximators in simulation-optimization. J Water Resour Planning Manag 126(2):48–56CrossRefGoogle Scholar
  17. Lek S, Beland A et al (1995) Improved estimation, using neural networks, of the food consumption of fish populations. Mar Freshw Res 46(8):1229–1238CrossRefGoogle Scholar
  18. Lek S, Delacoste M et al (1996) Application of neural networks to modeling nonlinear relationships in ecology. Ecol Model 90:39–52CrossRefGoogle Scholar
  19. Li YG, Yang CB et al (1998) Modeling and management of the flow regime of the big spring in north of China (in Chinese). Adv Water Sci 9(3):275–281Google Scholar
  20. Liu LH, Shu LC, Chen XH et al (2010) The hydrologic function and behavior of the Houzhai underground river basin, Guizhou Province, southwestern China. Hydrogeol J 18:509–518CrossRefGoogle Scholar
  21. Loáiciga HA (2004) Residence time, groundwater age, and solute output in steady-state groundwater systems. Adv Water Resour 27:681–688CrossRefGoogle Scholar
  22. Ma T, Wang Y, Guo Q (2004) Response of carbonate aquifer to climate change in northern China: a case study at the Shentou springs. J Hydrol 297:274–284CrossRefGoogle Scholar
  23. Maier HR, Dandy GC (2000) Neural networks for the prediction and forecasting of water resources variables: a review of modeling issues and applications. Environ Model Softw 15:101–124CrossRefGoogle Scholar
  24. Maren AJ, Harston CT, Pap RM (1991) Handbook of neural computing application. Academic, San Diego, 239 ppGoogle Scholar
  25. Massada AB, Carmel Y (2008) Incorporating output variance in local sensitivity analysis for stochastic models. Ecol Model 213:463–467CrossRefGoogle Scholar
  26. Morshed J, Jagath J, Kaluarachchi (1998) Application of artificial neural network and genetic algorithm in flow and transport simulations. Adv Water Resour 22(2):145–158CrossRefGoogle Scholar
  27. Olden JD, Jackson DA (2002) Illuminating the “black box”: a randomization approach for understanding variable contributions in artificial neural networks. Ecol Model 154:135–150CrossRefGoogle Scholar
  28. Park YS, Rabinovich J et al (2007) Sensitivity analysis and stability patterns of two-species pest models using artificial neural networks. Ecol Model 204:427–438CrossRefGoogle Scholar
  29. Pastres R, Franco D et al (1997) Local sensitivity analysis of a distributed parameters water quality mode. Reliab Eng Syst Saf 57:21–30CrossRefGoogle Scholar
  30. Sahoo GB, Schladow SG, Reuter JE (2009) Forecasting stream water temperature using regression analysis, artificial neural network, and chaotic non-linear dynamic models. J Hydrol 378:325–342CrossRefGoogle Scholar
  31. Shu LC, Wang MM, Liu RG, Chen GH (2007) Sensitivity analysis of parameters in numerical simulation of groundwater. J Hohai Univ (Nat Sci) 35(5):491–495Google Scholar
  32. Sun CZ, Gong HL (2001) Research on resuming the flow of Jinci Spring using artificial neural networks technique in Shanxi Province (in Chinese). J Mt Sci 19(4):372–376Google Scholar
  33. Sun CZ, Wang JS, Lin XY (2001) Research on Jinci spring’s recovery after the use of water from the Yellow river as a municipal water supply (in Chinese). Carsolog Sin 20(1):11–16Google Scholar
  34. Wang W, Pieter HAJM et al (2006) Forecasting daily streamflow using hybrid ANN models. J Hydrol 324:383–399CrossRefGoogle Scholar
  35. White WB (1988) Geomorphology and hydrology of Karst Terrain. Oxford University Press, New York, 187 ppGoogle Scholar
  36. Yin D, Su XS, Li YG, Feng YM (2006) Use of an artificial neural networks technique in karst-water recoverable-reserve evaluation (in Chinese). J Jilin Univ (Earth Sci Edn) 36:55–59Google Scholar
  37. Yin D, Shu LC, Liu LH et al (2009) Research on reflowing condition of Jinci karstic spring, north China. In: Trends and Sustainability of Groundwater in Highly Stressed Aquifers. IAHS Publ. 329, IAHS, Wallingford, UK, pp 241–246Google Scholar
  38. Yuan DX (1994) Karstology of China (in Chinese). Geological Publishing House, Beijing, China, pp 149–164Google Scholar

Copyright information

© Springer-Verlag 2012

Authors and Affiliations

  • Xian Li
    • 1
  • Longcang Shu
    • 1
  • Lihong Liu
    • 2
  • Dan Yin
    • 3
  • Jinmei Wen
    • 4
  1. 1.State Key Laboratory of Hydrology-Water Resources and Hydraulic EngineeringHohai UniversityNanjingChina
  2. 2.College of Earth and EnvironmentAnhui University of Science and TechnologyHuainanChina
  3. 3.Water Conservancy and Hydropower Science Research Institute of Liaoning ProvinceShenyangChina
  4. 4.Department of Geology and MiningChongqing Exploration and Design Institute of Geological Hazard Control EngineeringChongqingChina

Personalised recommendations