• Homepage
  • >
  • Development and Evaluation of a Vector Autoregression Precipitation Prediction Model for the Blue Nile Basin, Ethiopia

Development and Evaluation of a Vector Autoregression Precipitation Prediction Model for the Blue Nile Basin, Ethiopia

Background
Ethiopia receives most of its precipitation in JJAS, which is highly correlated with simultaneous tropical SST anomalies. However, no significant correlation has been identified when SST leads precipitation. Vector autoregression (VAR) models have been shown to be successful in forecasting SSTs and precipitation in many regions.

Objective
Develop and evaluate VAR models to predict monthly precipitation across June-September at various scales in the Blue Nile Basin.

Initial Results
Variables evaluated include the first two EOF modes of precipitation, Tropical Pacific SST, Tropical Atlantic SST, Indian Ocean SST, North Pacific SST, North Atlantic SST, and global (60S-60N,0-360) Geopotential Heights at levels of 200hPa and 850hPa. EOFs are calculated for each month from May to September. The related principal components are then used as the state vectors for VAR models. 

Dissemination: Paper in preparation

Figure 2. EOFs of May-September precipitation

 

Figure 1. Hindcast performance using a 2nd order VAR model at basin scale

Figure 3. Spatial distribution of hindcast skill with a 2nd order VAR model at basin scale

Research Team

Sarah Alexander, Graduate Research Assistant, Civil and Environmental Engineering, University of Wisconsin - Madison.

Paul Block, Assistant Professor, Civil and Environmental Engineering, University of Wisconsin-Madison.

Shu Wu, Assistant Researcher, Center for Climatic Research, University of Wisconsin - Madison.