Biases in simulating Amazon moisture transport: a convection-permitting vs regional climate model intercomparison

The dynamics of moisture transport in the Amazon are key to the South American water cycle. Convection-permitting regional climate models (CPRCMs) offer a valuable tool to better understand these processes. This study evaluates integrated water vapor transport (IVT) and precipitation in the Amazon Basin using two CPRCM simulations during austral summer (DJF) and winter (JJA). Simulations at 4.5 km resolution cover most of South America and are validated using ERA5 reanalysis. The CPRCM-ERA experiment (1998–2007) was driven by ERA-Interim data and downscaled to 25 km with an RCM. During DJF, precipitation was overestimated in the central Amazon (+ 1 to + 2 mm/day) and underestimated in the northeastern and western Amazon (–2.5 mm/day). The CPRCM-ERA simulation showed a weaker low-level jet (LLJ) and trade winds. IVT analysis indicated a negative bias in moisture transport over the LLJ region, more pronounced in CPRCM-ERA. This model underestimated moisture inflow by 6.1% and outflow by ~ 20%, while the RCM showed smaller biases (inflow + 0.8%, outflow + 7%). In JJA, biases differed: CPRCM-ERA underestimated moisture inflow by 4.7% and outflow by 4.6%, while RCM overestimated both (inflow + 8.2%, outflow + 7.5%). Despite these biases, differences from ERA5 were minor, suggesting CPRCMs provide useful insights into regional climate dynamics. Notably, CPRCM-ERA performed better than RCM during JJA, supporting the use of convection-permitting models to improve understanding of Amazonian moisture transport, despite their higher computational cost.