Research Information System
Stochastic Environmental Research and Risk Assessment, 2025 (SCI-Expanded)
Understanding and analyzing vegetation change is crucial for climate change, ecological change, and drought monitoring. As drought conditions become catastrophic, there is a need for accurate and timely information on vegetation conditions and health and its possible evolution in the near future to minimize risk. In recent years, artificial intelligence methods such as machine learning and deep learning have become increasingly important owing to their relatively good performance in analyzing various data, including remote sensing data. This study proposes an integrated artificial intelligence (AI) and deep learning (DL) approach for predicting the Normalized Difference Vegetation Index (NDVI) using satellite-derived data. The research focuses on the agricultural croplands of Maragheh, Iran, utilizing MODIS and TRMM satellite data. Three AI methods Decision Tree-Classification and Regression Tree (DT-CART), Support Vector Regression (SVR), and Recurrent Neural Networks (RNN) with Long Short-Term Memory (LSTM) and Bidirectional LSTM (BiLSTM) architectures were employed. The input variables included TRMM rainfall (5-month lag), evapotranspiration, gross primary productivity (GPP), and MODIS-derived daytime/nighttime surface temperatures. The model performance was evaluated using the correlation coefficient (R), root mean square error (RMSE), and Nash–Sutcliffe efficiency (NSE). The results demonstrated that BiLSTM outperformed the other models (R = 0.94, RMSE = 0.061), with rainfall, temperature, and evapotranspiration identified as critical predictors. This study highlights the potential of AI-DL frameworks for vegetation monitoring and drought assessment in data-limited regions, offering valuable insights for agricultural planning and resource management.