Simulation of rainfall-underground outflow responses of a karstic watershed in Southwest China with an artificial neural network

• Main Authors: Chen Xi, Chen Cai, Hao Qingqing, Zhang Zhicai, Shi Peng
• Format: Article
• Language: English
• Published: Elsevier 2008-06-01
• Series: Water Science and Engineering
• Subjects: karst; underground channel; correlation analysis; artificial neural network
• Online Access: http://www.sciencedirect.com/science/article/pii/S1674237015300028

Karstic aquifers in Southwest China are largely located in mountainous areas and groundwater level observation data are usually absent. Therefore, numerical groundwater models are inappropriate for simulation of groundwater flow and rainfall-underground outflow responses. In this study, an artificial neural network (ANN) model was developed to simulate underground stream discharge. The ANN model was applied to the Houzhai subterranean drainage in Guizhou Province of Southwest China, which is representative of karstic geomorphology in the humid areas of China. Correlation analysis between daily rainfall and the outflow series was used to determine the model inputs and time lags. The ANN model was trained using an error backpropagation algorithm and validated at three hydrological stations with different karstic features. Study results show that the ANN model performs well in the modeling of highly non-linear karstic aquifers.