This study compares various statistical downscaling methods for simulating daily precipitation in six U.S. regions using GCM data from the HadCM2 model. The methods include weather generators (WGEN and SPEL), vorticity-based models (B-Circ and C-Circ), and artificial neural networks (ANN1 and ANN2). The GCM used was forced by combined CO₂ and sulfate aerosol changes, with data spanning 1980–1999 (present) and 2080–2099 (future). The WGEN and SPEL models performed best in simulating daily precipitation, with WGEN and SPEL showing exact reproduction of wet-day occurrence statistics. The ANN models, however, performed poorly due to inadequate simulation of wet-day occurrence. The downscaling methods generally produced smaller changes in precipitation compared to direct GCM results, suggesting that changes in atmospheric circulation may not be the primary driver of precipitation changes. The vorticity-based models (B-Circ and C-Circ) showed good performance in simulating daily precipitation amounts, with errors less than 2.5%, and accurate wet-day probabilities. However, they captured monthly precipitation variability less accurately. The ANN models, while showing larger changes in some statistics, performed poorly in simulating present-day precipitation statistics, raising doubts about their reliability for future projections. The study highlights the importance of incorporating additional predictor variables beyond vorticity and circulation to improve downscaling accuracy. It also notes that the GCM's relationships between precipitation and circulation may not be fully consistent, and that downscaling methods should consider other factors to better simulate precipitation variability. The results suggest that future research should focus on improving downscaling models by incorporating more predictor variables and exploring stochastic approaches to better simulate future climate scenarios.This study compares various statistical downscaling methods for simulating daily precipitation in six U.S. regions using GCM data from the HadCM2 model. The methods include weather generators (WGEN and SPEL), vorticity-based models (B-Circ and C-Circ), and artificial neural networks (ANN1 and ANN2). The GCM used was forced by combined CO₂ and sulfate aerosol changes, with data spanning 1980–1999 (present) and 2080–2099 (future). The WGEN and SPEL models performed best in simulating daily precipitation, with WGEN and SPEL showing exact reproduction of wet-day occurrence statistics. The ANN models, however, performed poorly due to inadequate simulation of wet-day occurrence. The downscaling methods generally produced smaller changes in precipitation compared to direct GCM results, suggesting that changes in atmospheric circulation may not be the primary driver of precipitation changes. The vorticity-based models (B-Circ and C-Circ) showed good performance in simulating daily precipitation amounts, with errors less than 2.5%, and accurate wet-day probabilities. However, they captured monthly precipitation variability less accurately. The ANN models, while showing larger changes in some statistics, performed poorly in simulating present-day precipitation statistics, raising doubts about their reliability for future projections. The study highlights the importance of incorporating additional predictor variables beyond vorticity and circulation to improve downscaling accuracy. It also notes that the GCM's relationships between precipitation and circulation may not be fully consistent, and that downscaling methods should consider other factors to better simulate precipitation variability. The results suggest that future research should focus on improving downscaling models by incorporating more predictor variables and exploring stochastic approaches to better simulate future climate scenarios.