Assessment Of The Skill Of The High Resolution Regional Model In The Simulation Of Airflow And Rainfall Over East Africa
Abstract
Weather is very important on human activities over East Africa just like the rest of the world.
It affects our lives in different ways, including the food and water resources. East Africa
depends on agriculture for food and socio-economic developments. The agricultural
production greatly depends on the cumulative effects of weather characteristics. Accurate
weather forecasting would therefore assist in planning and management of the agricultural
production, water resource management, aviation safety, sea navigation, communication,
environmental protection, and many others. Improvement on the accuracy of weather
predictions through verification of the numerical models is one of the major subjects on
studies. The demand for verification of forecasting systems to ascertain their strengths and
weaknesses is increasing dramatically as models evolve more rapidly. Wind fields and
precipitation forecasts have always been of great interest to forecasters because they
influence daily life. The period September to December marks the short rain season for the
East Africa region.
This study was on assessment of the skill of High Resolution regional Model (HRM) in
predicting airflow and precipitation over East Africa. The overall objective of this study was
to assess the performance of the HRM in simulating the precipitation and the circulation
pattern over the East Africa region. The study period was from September to November
2005.
The data used in the study consisted of observed rainfall data from synoptic stations over the
region. These data was obtained from the Kenya Meteorological Department and
'0·
Intergovernmental Authority on Development climate Prediction and Application Center
Nairobi. The initial and boundary conditions data required for integrating the model was
acquired from Deutscher Wetterdienst (DWD), Germany. The HRM was run daily (24 hour
forecast) from 1st September to 30th November 2005 at the Kenya Meteorological
Department. The HRM was run with the 00 UTC initial conditions with a resolution of 14
and 28 km.
The predicted airflow verified against the observed characteristic wind fields. The
precipitation forecast was verified against the observed precipitation. The following
verification procedures were undertaken; measures of relationship between a forecast and
corresponding observations using the mean absolute error (MAE), root mean square error
(RMSE), simple correlation analysis and degrees of freedom in significance estimates of
correlation coefficients. Besides, the determination of the skill scores using the Frequency
x
Bias Index (FBI), Equitable Threat Score and True Skill Statistic or Hansen-Kuipers
discriminant was examined. In addition to assessing the model skill an analysis of the
seasonal performance was carried out.
The September-November 2005 period experienced depressed rainfall. Most stations
recorded rainfall less than 50 percent of their long term mean. Positive Sea Surface
temperature anomalies persisted over southern Atlantic, tropical equatorial pacific and Indian
Oceans during this period. This situation led to weaker than normal pressures over the St.
Helena and the Mascarene areas and hence slightly weaker convergence over most parts of
the region. However, some areas of the central highlands, coast and the lake basin had some
good rains, which was associated with the southward shift of the zonal arm of the inter
Tropical Convergence Zone (ITCZ) coupled up with the westward progression of easterly
waves from the Indian Ocean. The ITCZ was however diffuse for most of the period leading
to depressed rainfall over the region.
The verification approaches show that the model simulates well airflow and precipitation
with some cases of overestimation and underestimation. The HRM simulates well the spatial
distribution of precipitation over the East Africa region. High rainfall received over parts of
East Africa was captured in the model forecast. The results showed low root mean square
errors and mean absolute errors and high correlation coefficients in most parts of the region,
which is an indicator of good performance of the model. The experiments indicate that finer
model resolutions produced statistically significant improvements in performance for HRM,
especially for lighter thresholds (0.1-4 mm). Performance characteristics change for the
heavier amounts (8 and 16 mm) when HRM with-the two different resolutions are compared.
The land-lake breeze diurnal cycle is evident from !he study. The early morning and late
night precipitation over the lake and afternoon and evening showers were well simulated in
the study. The results show a line of confluence running from north to south due to the effect
of the Great Rift Valley escarpment around 35.9 degrees east. The channelling effect of the
Turkana region is also evident from the results.
Overall, HRM performed reasonably well over the region of study and hence it could be
adopted for operational purposes. Numerical model forecasts can be used in weather
dependent activities such aviation, tourism and recreation industry, fishing industry and
wildlife management among others, and will help the government to minimize the severe
socio-economical implications ofthe adverse weather events. However there is need for some
studies to be carried out on the model physics to understand and improve.
Publisher
Department of Meteorology