Precipitation in southern Africa: dynamical downscaling of CAM using WRF

Abstract

Regional climate modeling has become an important tool for downscaling global climate models (GCMs), to assess high impact regional climate pro-jections. Southern Africa is one of the most exposed regions to the effects of climate change, as it is highly depended on rain fed agriculture. Nevertheless, there are only a handful of regional climate studies previously performed in southern Africa. In this thesis, the Weather and Research Forecast (WRF) model’s performance as a dynamical downscaler for the Community Atmospheric Model (CAM) in southern Africa is investigated, focusing on precipitation patterns over the time period of 1990-2009. The domain covers southern Africa from 5◦ -38◦ S and 8◦ -53◦ E with a resolution of 27 km x 27 km and 36 vertical layers. WRF is initialised by CAM and Community Land Model (CLM) data, and forced every 6hours by SSTs and lateral boundary conditions from CAM. Seasonal, annual, interannual and extreme events of precipitation in a historical run with WRF are outlined in this thesis. Additionally, a preliminary study of downscaling a future time slice (2050-2069) is performed. The results from the historical run have been validated with satellite observational data.

WRF reproduces the mean seasonal and annual precipitation cycle satis- factorily, although overestimating in the summer months in the Indian Ocean and over the southern African plateau. The latter is probably caused by a too strong low-level convergence in WRF, leading to a stronger Walker circulation and positive precipitation feedback mechanism over the southern African plateau. Compared to satellite observations, WRF generally provides slightly better results concerning mean precipitation than CAM. Testing different physical scheme combinations (cumulus parametrization and planetary boundary layer) in shorter runs imply that in some areas the the overestimation of precipitation can be reduced with alternative scheme options, although attaining biases in other regions. The correlations between observed seasonal means and CAM/WRF are computed over selected regions, showing that the interannual variability is generally poorly captured by both models, although somewhat better over a few regions in CAM. The results suggest that SST anomalies are not the governing driving force behind interannual variability over most parts of southern Africa. WRF’s applicability for computing extreme precipitation changes over a time period is briefly discussed, and WRF seems to adequately reproduce the precipitation changes compared to TRMM (for 1998-2009).

To the degree this study can be compared with previous, the results are well in line with former work. As in previous studies the wet summer precipitation bias in the RCM (WRF) is similar to the GCM (CAM) over southern Africa. However, WRF generally reproduces seasonal precipitation somewhat closer to observed values compared to CAM, and sensitivity studies suggest the biases might be further diminished by changing the physical parametrization through alternative scheme options.

Thus this study seems to be a step in the right direction for dynamically downscaling of GCMs over southern Africa and motivates for further research.