A new approach to predict normalized difference vegetation index using time-delay neural network in the arid and semi-arid grassland

Abstract

ABSTRACT Normalized difference vegetation index (NDVI) has been used to conduct important research on plant growth and vegetation productivity. In this paper, a new approach to predict NDVI based on precipitation in the grass-growing season for the arid and semi-arid grassland is proposed using time-delay neural network (TDNN). To intuitively know the ability of TDNN to learn the relationship between NDVI and precipitation and to predict NDVI, the performance of the TDNN model is compared with back propagation neural network (BPNN) trained with the same data. The results indicate that TDNN works well to predict precipitation. Moreover, the relationship between precipitation and NDVI can be predicted accurately by the proposed TDNN model. The results show the goodness-of-fit between the observed NDVI and predicted NDVI measured by the determination coefficient of R 2 being 0.999 from the TDNN model, with the mean absolute percentage error, mean absolute error, and root-mean-square error to be 0.23%, 0.20, and 0.19, respectively. The analysis shows that the proposed method can result in an accurate NDVI prediction. Thus, the algorithm is applied to predict the NDVI during the grass-growing season for the validation of the approach. This validation results suggest the potential application of this approach for reduction of overgrazing pressure and vegetation restoration in the arid and semi-arid grassland.