Soil Water Prediction of Moving Dune Based on BP Neural Network Model in Northwest Liaoning Sandy Land

Abstract

With the moving dune in sandy land of Northwest Liaoning province as the research object, its water variation in soil was simulated and studied based on a BP Neural Network model. With principal meteorologic factors that affect soil water, such as precipitation and evaporation, as the input variables and the water content in soil as the output variable, a soil-water prediction model based on BP NN was built. Results show that the BP NN model achieved high precision, with mean absolute error of 0.35 and mean relative error of 11.53%. The BP NN prediction model for moving dune provides a new approach for the soil water acquisition.