Water Level Variation and Prediction of the Pingshan Sinkhole in Guizhou, Southwestern China

摘要

The study area is located in the Houzhai Basin, Guizhou Province, in southwest China. The aquifer is a highly heterogeneous triple porosity karst aquifer with diverse hydraulic properties. The Pingshan Sinkhole serves as a perfectly placed well whose water level records the water level in the aquifer. The rainfall and water level time series from 1990 to 1995 have been analyzed and the results indicate that the observed water level variation can be divided into three stages, i.e. low water level stages, water level decreasing-rising stages and high water level stages. A Pingshan Sinkhole water level prediction model has been developed using partial least squares (PLS) regressions and artificial neural network (ANN) methods. The rainfall in the previous 12 hours and the antecedent water level are the main factors influencing the sinkhole water level in the rainy season. The predicted results obtained by the ANN or PLS regression combined with the ANN method are better than the results of PLS regression method alone. The study of the variation and prediction of water level in the Pingshan Sinkhole could benefit the understanding of the hydrological cycle and the hydrogeological conditions in this highly heterogeneous triple porosity karst aquifer.