Abstract: This study presents the interaction of climate covariates and geo-location elements with wind potential in Ethiopia. We employ a dynamic spatial panel autoregressive random effects model with a spatial weight of inverse quartile separation distances of locations. This model is used to extrapolate potential spots of wind at unobserved spatial points. A meteorological data set observed at 60 stations and over 2000 to 2017 years was used for prediction. The result depicts that mean wind speed is stochastic and varies over longitude range and latitude span. It also changes with the influence of climate covariates, topographic conditions, and fluctuates over the months of a year. Particularly, the mean wind speed intensity is high along the central, eastern and northeastern regions of the country. It is relatively higher in the months of February, March, June, and July but lower in September and October. There is also evidence that mean wind speed is higher in summer and spring but relatively lower in winter and fall seasons. That means, higher mean wind speed is recorded mainly after the rainy season ends and before it starts. The model estimates also show that mean wind speed is significantly correlated across spatial spots and over temporal points. This shows the dependence of the mean wind speed across neighboring stations and over the months of a year. Besides, the mean wind speed increases with elevation and temperature whereas it decreases when precipitation increases. Sunshine fraction and relative humidity have negative effects, but their influence is insignificant.
Keywords: Bayesian inferences; Dynamic spatial panel autoregressive model; Prediction; spatial weights of inverse quartile separation distances; Stochastic process.