The research on groundwater salinity dynamic plays an important role in understanding the rules of groundwater salinization and the reasonable plan, utilization and management of groundwater resource. A genetic programming (GP) model by using groundwater salinity time series was proposed to model groundwater salinity dynamic. The autoregressive integrated moving average (ARIMA) model was used to determine the input vectors of GP model. The GP model was applied to model groundwater salinity dynamic in the Yellow River Delta. Besides, the GP model was compared with ARIMA model to test the validity. Results indicated that the GP model of groundwater salinity dynamic was suitable for model groundwater salinity dynamic; and the modelling accuracy of GP model is higher than that of ARIMA model.