Nikoo (2026) Integrating deep learning into future hydrological modeling under climate change scenarios in an arid region

Identification

• Journal: Journal of Arid Environments
• Year: 2026
• Date: 2026-02-24
• Authors: Mohammad Reza Nikoo
• DOI: 10.1016/j.jaridenv.2026.105577

Research Groups

• Department of Civil and Architectural Engineering, Sultan Qaboos University, Muscat, Oman

Short Summary

This study developed a hybrid deep learning framework for downscaling climate projections and a hybrid HEC-HMS–LSTM model for streamflow simulation to assess climate change impacts on an arid region in Oman, revealing significant future changes in precipitation, temperature, and streamflow under different SSP scenarios.

Objective

• To apply a hybrid deep learning framework for downscaling and a hybrid HEC-HMS–Long Short-Term Memory (LSTM) approach for streamflow simulation to assess climate change impacts in an arid region.
• To evaluate projected streamflow changes at Oman's largest dam using advanced deep learning models (LSTM, TCN, GRU) driven by CMIP6 climate projections under Shared Socioeconomic Pathways (SSP1-2.6, SSP2-4.5, and SSP5-8.5).

Study Configuration

• Spatial Scale: An arid catchment in Oman, specifically focusing on the area around Oman's largest dam.
• Temporal Scale: Future climate projections for the period between 2080 and 2099. Historical periods were used for model calibration and validation.

Methodology and Data

• Models used:
  • Hydrological model: HEC-HMS
  • Deep learning models: Long Short-Term Memory (LSTM), Temporal Convolutional Network (TCN), Gated Recurrent Unit (GRU)
  • Hybrid models: HEC-HMS–LSTM (for streamflow simulation), LSTM–Transformer (for downscaling)
• Data sources:
  • Climate projections: Coupled Model Intercomparison Project Phase 6 (CMIP6) under Shared Socioeconomic Pathways (SSP1-2.6, SSP2-4.5, SSP5-8.5).
  • Observational data for HEC-HMS calibration/validation (implied).

Main Results

• The HEC-HMS model showed strong performance (Nash-Sutcliffe Efficiency, NSE = 0.842 for calibration, 0.811 for validation) but struggled with simulating peak events.
• Hybridization of HEC-HMS with LSTM significantly improved accuracy, reducing the peak flow error to 3.60%.
• Downscaling using a hybrid LSTM–Transformer model outperformed other models, leading to improved precipitation projections.
• Future climate projections (2080-2099):
  • Precipitation: Decreasing annual peak precipitation under SSP1-2.6 (23.82 mm), but increasing extremes under SSP5-8.5 (36.65 mm).
  • Temperature: Rose across all scenarios, with summer temperatures reaching 37.4 °C under SSP5-8.5.
  • Streamflow: Projected an 80% decline under SSP1-2.6, but a near-baseline return under SSP5-8.5 (243.37 m³/s).

Contributions

• Developed and applied a novel hybrid deep learning framework for downscaling CMIP6 climate projections in an arid region.
• Introduced a hybrid HEC-HMS–LSTM approach that significantly improved the simulation of hydrological extremes, particularly peak flow events, which traditional hydrological models often fail to capture accurately.
• Provided comprehensive projections of future precipitation, temperature, and streamflow changes under multiple SSP scenarios (SSP1-2.6, SSP2-4.5, SSP5-8.5) for a critical water resource in an arid environment.
• Demonstrated the superior performance of advanced deep learning models (LSTM, TCN, GRU, LSTM-Transformer) in hydrological modeling and climate downscaling for arid regions.

Funding

• Not explicitly mentioned in the provided text.

Citation

@article{Nikoo2026Integrating,
  author = {Nikoo, Mohammad Reza},
  title = {Integrating deep learning into future hydrological modeling under climate change scenarios in an arid region},
  journal = {Journal of Arid Environments},
  year = {2026},
  doi = {10.1016/j.jaridenv.2026.105577},
  url = {https://doi.org/10.1016/j.jaridenv.2026.105577}
}