Neural circuit policy and hybrid deep learning models for enhanced meteorological drought forecasting performance
Applied Soft Computing, cilt.193, 2026 (SCI-Expanded, Scopus)
- Yayın Türü: Makale / Tam Makale
- Cilt numarası: 193
- Basım Tarihi: 2026
- Doi Numarası: 10.1016/j.asoc.2026.114844
- Dergi Adı: Applied Soft Computing
- Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Applied Science & Technology Source, Compendex, INSPEC
- Anahtar Kelimeler: Deep learning, Drought forecasting, Hybrid models, Machine learning, Neural circuit policy
- Kırklareli Üniversitesi Adresli: Evet
Özet
Drought poses significant risks to communities, agriculture, and water systems, making accurate forecasting critical for sustainable water management. The Standardized Precipitation Index (SPI) is widely recognized as a key indicator of meteorological drought, yet its forecasting remains challenging due to the inherent complexities of drought. Recent advancements in artificial intelligence, particularly Machine Learning (ML) and Deep Learning (DL) techniques, have shown promise in improving drought forecasting. This study introduces the Neural Circuit Policy (NCP), a novel DL model for forecasting SPI at 3-, 6-, 9-, and 12-month time scales. Evaluated through a case study in a drought-prone basin in Türkiye, the NCP model is compared with Random Forest (RF) and eXtreme Gradient Boosting (XGB) as ML models, Long Short-Term Memory (LSTM) as a DL model, and a hybrid model of a Convolutional Neural Network (CNN) and LSTM. The study also investigates the potential for improving drought forecasting accuracy through hybrid models that integrate NCP with RF, XGB, CNN, and LSTM. In addition, the models’ accuracy in classifying drought categories is analyzed as a complementary measure of performance. The results demonstrate that NCP achieves high forecasting performance across all SPI time scales (KGE > 0.92, NSE > 0.93, RMSE < 0.23, MAE < 0.15, R2 > 0.94). The Borda Count confirms that NCP consistently ranks higher than individual models. Moreover, NCP-based hybrids, particularly LSTM-NCP, achieve superior drought category classification (weighted F1 > 0.89). These findings highlight the strong capability of NCP and its hybrid models for enhanced SPI-based meteorological drought forecasting.