Comparative study on karst ground water simulation between GM(1, 1)decomposition model and ARIMA model: A case study on discharge simulation of the Liulin Spring

The discharge of the Liulin spring is simulated with GM(1,1) decomposition model and ARIMA model respectively. According to the hydrological characteristics, the Liulin spring flow series could be divided into two periods. First, from 1957 to 1973 the spring flow was under natural state; second, from 1974 to 2009 the spring flow was impacted by both climate change and human activities. Using the data of first period, the spring flow under the natural state is fitted with GM(1,1) decomposition model and ARIMA model, and then the models are extrapolated to obtain the second periods’spring flow under the natural state. According the water balance principle, the spring flow decrement contributed by human activities is acquired by subtracting the observed discharge from simulated spring flow of the second period under the natural state. Thus, it is differentiated the effects of human activities from climate change. The simulated Liulin Springs’attenuation from 1970s to early 21st century is 2.26 m3/s by GM(1,1)decomposition model and 2.36 m3/s by ARIMA model with the relative error being 0.44% and 2.20% respectively, showing both GM(1, 1)decomposition mode land ARIMA model are suitable for spring flow simulation. Comparing the effects of human activities and climate change to the depletion of the Liulin Spring’s discharge, the authors find that the contribution of human activities is 8 to 9 times higher than that of the climate change. The empirical studies have shown that the GM (1,1) model is of high precision in simulating the exponential series. It can also improve accuracy by periodic amendment, when simulate the spring flow with periodic fluctuations. ARIMA model could reflect time-lag between precipitation and spring discharge and accurately simulate their quantitative relation.