Temporal Variations in Karst Spring Flow and Its Response to Climate Change in the Taihang Mountains, China

Abstract

The variation in flow for the Xiaonanhai spring and its main controlling factors were statistically examined using Mann-Kendall trend analysis, rescaled range (R/S) analysis, Morlet wavelet analysis, and cross-wavelet analysis. The annual spring flow has exhibited a continuous and significant downward trend since 1997, which varies temporally, with main periods of 14–17 years and 25–28 years. There is a positive correlation between the spring flow and precipitation (p<0.01). The results of the cross-wavelet analysis and the wavelet coherence analysis show that the main resonance periods of the precipitation and spring flow were present in 1983–1995 and 1998–2015, respectively. A stepwise regression model and a back propagation neural network model were established based on the analysis of the flow changes in order to predict the spring flow under the expected changes in precipitation. The simulation precision of the neural network structure is better than that of the stepwise regression model. Therefore, the neural network model can be used to predict the dynamic flow of the Xiaonanhai spring under the effects of future climate change.