Towards Improving the Skill of Seasonal Rainfall Prediction over Rwanda
Abstract
Rainfall prediction exhibits spatial and temporal variability over Rwanda.
The improved seasonal rainfall prediction to reduce the climatic extreme events using dynamical
and statistical models with fewer uncertainties is important to the socio-economic growth of the
country.
Improving the skill of rainfall in seasonal forecast time scale over Rwanda has a huge
implication for provisions of food security and water resource planning. The intent of this work,
is to improve the skill of seasonal rainfall forecast over Rwanda by using statistical and
dynamical methods.
The data used in this were seasonal rainfall observations, gridded rainfall dataset, global sea
surface Temperatures (SSTs), and initialization data for weather research and forecasting,
environmental modelling system (WRF-EMS) obtained from Global Forecasting System (GFS).
Model simulation was done for the period 1981-2018 and the model output for the 30 stations
was used to determine the skill of the WRF-EMS model for March to May 2018 (MAM) and
September to December 2018 (SOND) over the entire country.
The Methodology employed for quality control analyses indicated that most of the data used in
the study were of acceptable. The Dynamical part involved assessing the skill and accuracy of
the Weather Research and Forecasting Model using measures like root mean square error
(RMSE), mean absolute error (MAE), correlation and regression analysis. The statistical part
involved identifying the appropriate predictors by using principal component regression, sea
surface temperatures anomalies sea surface temperature gradients, specific zonal wind, Indian
ocean dipole and El Nino Southern Oscillation and verification of the improved skill scores.
The results of the RMSE and MAE for the sample stations, which shows the capability of the
model to give the observed precipitation, which show the smaller value the better fit for the test.
The Absolute Mean Error for many stations were to 1, showing that the model had a high
accuracy in producing the rainfall observation during the wet seasons. also the result shows very
low Mean Absolute Square (MAE) and low Root Mean Square (RMSE) of less than one (1) for
all sample stations during the two (2) wet seasons, thus the capacity of the model to produce the
rainfall observation. The results from correlations analysis indicated that the use of sea surface
temperature gradient modes as predictors has advantage over grid point sea surface temperatures,
and has the potential to improve seasonal rainfall forecasts for both seasons especially for the
March-May season. The highest value of correlation observed with the season rainfall, and gridpoint
sea surface temperatures modes was 0.7 during September to December compared to 0.85
during March to May. The March-May rainfall continues to have higher potential of
predictability than September-December
The results from regression analysis indicated that the methods of using SST modes would
improve the prediction of rainfall season in the country especially for March to May rainfall
season, such linkage were statistically significant and can contribute in improving the prediction
skill of the long and short rainfall season in the country.
The results from WRF model indicate that the model simulate well the observed and the
predicted rainfall at both seasons and can be use among many seasonal predictions of rainfall in
Rwanda.
Results of the study show that the methods can be used to improve the skill of seasonal rainfall
forecast over Rwanda which is critical to the planning, sustainable and growth of development to
the socio-economic activities together with improvement of existing Early Warning System
(EWS) of extreme precipitation events and contribute to weather driven disaster preparedness
efforts in Rwanda.
Publisher
University of Nairobi
Rights
Attribution-NonCommercial-NoDerivs 3.0 United StatesUsage Rights
http://creativecommons.org/licenses/by-nc-nd/3.0/us/Collections
The following license files are associated with this item: