Publications

Zhao, F; Xu, N; Fang, ZG; Bai, SJ; Jiang, MJ; Zhu, YH (2025). A New Approach for Predicting NDVI of Winter Wheat Using Seasonal LSTM. JOURNAL OF THE INDIAN SOCIETY OF REMOTE SENSING.

Abstract
The prediction of the Normalized Difference Vegetation Index (NDVI) is closely related to crop growth and is essential for crop management decision-making, phenology, and yield prediction. However, a single Long Short-Term Memory (LSTM) method may not efficiently capture seasonal characteristics. Therefore, this study proposes a Seasonal Long Short-Term Memory (S-LSTM) model to predict NDVI for winter wheat. The historical MODIS time-series data used in this study is MOD09A1, which has a spatial resolution of 500 m and a temporal resolution of 8 days. First, the NDVI time series for winter wheat in the region is generated using MODIS time-series data and the distribution of winter wheat planting in Anhui Province. Second, the NDVI time series is decomposed into two parts: one component represents inter-annual variation, simulated using a seasonal periodic curve; while the other component approximates instantaneous changes through LSTM modeling. Finally, the two predictions are combined to obtain the final prediction. The results indicate that using the S-LSTM model for predicting NDVI of winter wheat in Anhui Province, China has high predictive performance. The Pearson correlation coefficient (r), Mean Absolute Error (MAE), Mean Average Percentage Error (MAPE), and Root Mean Square Error (RMSE) of the predicted results in the selected four regions are all better than the three methods LSTM, SVR and SVR-LSTM. Specifically, the Pearson correlation coefficient (r) is greater than 0.7, the MAE is less than 0.04, the MAPE is less than 10%, and the RMSE is less than 0.05. This method outperforms other methods in these evaluation metrics, demonstrating its effectiveness for crop management decision-making and yield prediction. This method enhances the robustness of S-LSTM in time series prediction and proves valuable for crop management decision-making and yield prediction.

DOI:
10.1007/s12524-025-02129-8

ISSN:
0974-3006