Fattahi et al. (2026) Deep Learning LSTM-Based Model for Predicting SPI and SPEI Drought Indices

⚠️ Warning: This summary was generated from the abstract only, as the full text was not available.

Identification

• Journal: Natural Hazards Review
• Year: 2026
• Date: 2026-07-13
• Authors: Mohammad Hadi Fattahi, Tajeddin Derikvand, Farhang Rahmani
• DOI: 10.1061/nhrefo.nheng-2561

Research Groups

Not specified in the provided text.

Short Summary

The study develops a deep learning LSTM-based model to predict the Standard Precipitation Index (SPI) and Standardized Precipitation-Evapotranspiration Index (SPEI) in two Iranian watersheds, demonstrating superior accuracy and dynamic property preservation compared to traditional time-series models.

Objective

• To propose and evaluate a Long Short-Term Memory (LSTM) deep learning model for the accurate prediction of drought indices (SPI and SPEI).

Study Configuration

• Spatial Scale: Two watersheds in Iran.
• Temporal Scale: 1972 to 2020.

Methodology and Data

• Models used: Long Short-Term Memory (LSTM), ARIMA, and various linear and nonlinear control models.
• Data sources: Statistical time-series data from 1972 to 2020.

Main Results

• The LSTM model outperformed control models in accuracy, achieving a Mean Absolute Error (MAE) of 0.057, Root Mean Squared Error (RMSE) of 0.079, and R-squared (RS) of 0.0053.
• Nonlinear dynamic analysis using the Lyapunov exponent (LE) and approximate entropy (ApEn) confirmed the LSTM model's superiority over standard time-series models.
• The LSTM model better conserved the dynamic properties of the base time-series compared to the ARIMA model.

Contributions

• Provides a high-accuracy deep learning framework for drought index prediction that maintains the nonlinear dynamic characteristics of hydrological time-series better than traditional linear models.

Funding

Not specified in the provided text.

Citation

@article{Fattahi2026Deep,
  author = {Fattahi, Mohammad Hadi and Derikvand, Tajeddin and Rahmani, Farhang},
  title = {Deep Learning LSTM-Based Model for Predicting SPI and SPEI Drought Indices},
  journal = {Natural Hazards Review},
  year = {2026},
  doi = {10.1061/nhrefo.nheng-2561},
  url = {https://doi.org/10.1061/nhrefo.nheng-2561}
}