Forecasting simulation for the discharge of epikarst spring on the basis of multiple linear regression

In light of the spring discharge and the precipitation date in Chenqi drainage area in Puding County, Anshun City, Guizhou Province, the auto-correlation and partial auto-correlation of the epikarst spring discharge as well as cross-correlation of the discharge-precipitation series are analyzed. And then, the lag time of the discharge series and the precipitation-discharge lag time are determined, and the multiple linear regression model established. The output results of the calibration period fit the measured discharge series well, the nash efficiency coefficient (NEC) is 0.996, the root-mean-square error (RMSE) is 3.0×10-7 m3/s and the mean relative error (MRE) is 2.12%. The NEC of the validation period is 0.985, the RMSE is 3.96×10-7 m3/s, and the MRE is 5.36%. The results demonstrate that along with the increase of the prediction period, the prediction error of the spring discharge in subsiding phase increases. The model established in this paper is feasible for simulating the spring discharge in 10 hours, and the average value of the relative errors of the calculated discharge in subsiding phase after6 rainfalls is less than 5%.