Abstract:
Reliable seasonal forecasting of water resources variability may be of great value for agriculture and
energy management in Ethiopia. This work aims to develop statistical forecasting of seasonal total water
storage (TWS) anomalies in Ethiopia using sea-surface temperature and sea-level pressure indices.
Because of the spatial and temporal variability of TWS over the country, Ethiopia is divided into four
regions each having similar TWS dynamics. Periods of long-term water deficit observed in GRACE TWS
products for the region are found to coincide with periods of meteorological drought. Multiple linear
regression is employed to generate seasonal forecasting models for each region. We find that the skill of
the resulting models varies from region to region, with R
2
from 0.33 to 0.73 and correlation from 0.27 to
0.77 between predicted and observed values (using leave-one-out cross-validation). The skill of the
models is better than the climatology in all regions.
