Hybrid deep learning downscaling of GCMs for climate impact assessment and future projections in Oman

Accurate downscaling of global circulation models (GCMs) is critical for assessing the impacts of climate change and water resources management. In this research, Fourteen GCMs were evaluated through a Taylor diagram, including EC-Earth3-CC, ACCESS-CM2, AWI-ESM-1-1-LR, BCC-ESM1, CanESM5, IITM-ESM, MPI ESM1-2HR, INM-CM5-0, IPSL-CM5A2-INCA, KIOST-ESM, NorCPM1, NorESM2-MM, TaiESM1, and ACCESS-ESM1-5. IITM-ESM showed the best performance, making it the preferred model for future climate studies. To downscale the selected GCM, a novel hybrid deep learning method was employed, combining a sequence-to-sequence model with a Temporal Convolutional Network (TCN) as the encoder and a Transformer as the decoder. This approach was compared to Quantile Mapping, Random Forest, long short-term memory (LSTM), and TCN models, with optimization using the Particle Swarm Optimization (PSO) algorithm. The proposed model outperformed others, achieving an NSE of 0.907, RMSE of 2.10, BIAS of 0.63, and a relative error of 21.96%. Then, an HEC-HMS model was constructed for the Wadi Dayqah basin, utilizing data from 1992 to 2006 for calibration and data from 2007 to 2011 for validation. Precipitation and temperature were downscaled for the near (2030-2039), mid (2040-2049), and far future (2040-2049) periods. Hydrological modeling was conducted for future climate scenarios SSP126, SSP245, and SSP585, revealing notable changes. SSP126 and SSP245 project substantial declines in precipitation, especially in spring and summer, while SSP585 forecasts more extreme variability and precipitation events. Temperature increases are relatively modest under SSP126, with a 5.4% rise in June, while SSP245 shows a 19.2% increase in July, and SSP585, the most extreme, predicts a 24.6% rise in June. Maximum annual streamflow is expected to decrease significantly under SSP126 and SSP245, whereas SSP585 predicts extreme peak flows up to seven times the historical average. These results underscore adaptive water management's importance in addressing the impacts of climate change.