Convection in Central Amazonia investigated by WRF high resolution simulation: shallow-to-deep convection transition and the role of shallow cumulus clouds under pristine conditions

In this study we evaluate the skill of WRF model to simulate the actual diurnal cycle of convection in the Amazon basin . Models typically are not capable to simulate the well documented cycle that yields in rain at afternoon. First, a shallow cumulus field is established in the morning, with light or no precipitation. Shallow cumulus are followed by a towering process around noon. It evolves from shallow to deep convection, and rain occurs around 14h (LT). Models typically make precipitation take place in the early morning. The fail in models is explained by the typical size of shallow cumulus (~0.5 – 2.0 km), whereas models usually are set with coarser resolution (> 20 km), and employ cumulus parametrization. To check the ability of the model to correct simulate the diurnal cycle of convection in this study we employed high spatial resolution (x = 0.625 km) to reach the shallow cumulus scale. The simulations correspond to a dynamical downscaling of ERA-Interim from 25 to 28 February 2013 with 40 vertical levels, 30 minutes outputs, and three nested grids (10 km, 2.5 km, 0.625 km). Improved vegetation (USGS + PROVEG), albedo and greenfrac (computed from MODIS-NDVI + LEAF-2 land surface parametrization), as well as pseudo analysis of soil moisture were used as input data sets, resulting in more realistic precipitation fields when compared to observations in sensitivity tests. Convective parametrization was switched off for the 2.5/0.625 km grids, where cloud formation was solely resolved by the microphysics module (WSM6 scheme, which provided better results). Results showed a significant improved capability of the model to simulate diurnal cycle. Shallow cumulus begin to appear in the first hours in the morning. They were followed by a towering process that culminates with precipitation in the early afternoon, which is the behavior well described by observations but rarely obtained in models. Rain volumes were also realistic (~20 mm for single events) when compared to typical events during the period, which is in the core of the wet season. Cloud fields evolution also differed with respect to Amazonas River bank, which is a clear evidence of the interaction between river breeze and large scale circulation. This results will be particularly interesting for campaigns occurring in Amazonia aiming to detect aerosol effects, once this simulations represent the baseline condition of clouds without anthropogenic aerosol influence.