Abstract:
In 2006 a Numerical Weather Prediction (NWP) model known as the Unified Model
(UM) from the United Kingdom Meteorological Office (UK Met Office) was installed at
the South African Weather Service (SAWS). Since then it has been used operationally
at SAWS, replacing the Eta model that was previously used. The research documented
in this dissertation was inspired by the need to verify the performance of the UM in
simulating and predicting weather over South Africa. To achieve this aim, three model
configurations of the UM were compared against each other and against observations.
Verification of rainfall as well as minimum and maximum temperature for the year 2008
was therefore done to achieve this. 2008 is the first year since installation, where all the
configurations of the UM used in the study are present. For rainfall verification the
model was subjectively verified using the eyeball verification for the entire domain of
South Africa, followed by objective verification of categorical forecasts for rainfall
regions grouped according to standardized monthly rainfall totals obtained by cluster
analysis and finally objective verification using continuous variables for selected stations
over South Africa. Minimum and maximum temperatures were subjectively verified
using the eyeball verification for the entire domain of South Africa, followed by objective
verification of continuous variables for selected stations over South Africa, grouped
according to different heights above mean sea level (AMSL). Both the subjective and
objective verification of the three model configurations of the UM (for both rainfall as
well as the minimum and maximum temperatures) suggests that 12km UM simulation
with DA gives better and reliable results than the 12km and 15km UM simulations
without DA. It was further shown that although there was no significant difference
between the model outputs from the 12km and the 15km UM without DA, the 15km UM
simulation without DA, proved to me more reliable and accurate than the 12km UM
simulation without DA in simulating minimum and maximum temperatures over South
Africa, on the other hand the 12km UM simulation without DA is more reliable and
accurate than the 15km UM simulation without DA in simulating rainfall over South
Africa.
