Zhongping Yang

Application of Back-Propagation Artificial Neural Network Models for Prediction of Groundwater Levels: Case study in Western Jilin Province, China

Evaluation and forecast of groundwater levels through specific model helps in forecasting of groundwater resources. Among the different robust tools available, the back-propagation artificial neural network (BPANN) model is commonly used to empirically forecast hydrological variables. Here, we discuss the modeling process and accuracy of this method based on the root mean squared error (RMSE), the mean absolute error (MAE) and coefficient of efficiency (R2). The arid and semi-arid areas of western Jilin province (China) were chosen as study area owing to the decline of groundwater levels during the past decade mainly due to over exploitation. The simulations results indicated that BPANN is accurate in reproducing (fitting) and forecasting the groundwater levels time series based on the R2 are 0.97 and 0.74, respectively. The RMSE, MAE for BPANN model in the predicting stage are 0.08, 0.066, respectively. It is evident that the BPANN is able to predict the groundwater levels reasonable well.

Application of muti-quadric interpolation in numerical simulation of 2D groundwater flow

Multi-quadric (MQ) interpolation is a numerical method to solve partial differential equations (PDEs). The purpose of this paper was to discuss the effectiveness of the method applied in groundwater flow simulation, and the influence of shape parameters, support domain, node spacing, and time steps on results. We used MQ method for the numerical solution of four cases of 2D groundwater flow in confined aquifer. Case one was a flow in a confined aquifer with constant transmissivity (T), while case two was with T varying continuously. Case three represented the flow to a fully penetrated pumping well in a confined aquifer. Case four was a transient flow in confined aquifer with constant transmissivity. The method was found to have good agreement with analytical solution of the four cases. The error was affected by the combination of shape parameters, size of support-domain, and node spacing of MQ method. Good accuracy could be obtained under comparatively big time step, when space and time discretizations were dealt with MQ method and the finite difference method, respectively.