The share of wind energy in total installed power capacity from wind power is increasing worldwide. Short-term wind forecasts together with a quantification of uncertainties are required for the reliable operation of power systems with significant wind power penetration. A challenge for utilizing wind power as a source of energy is the intermittent and hardly predictable nature of wind. We build and evaluate new statistical techniques for producing forecasts at multiple locations and lead times using spatio-temporal information. We show that borrowing information by utilizing the spatial correlation among individual sites reduces the errors in forecasts and provides the option of generating predictions at locations that are not within the observation samples. To meet the computational requirements, we adopt a Bayesian framework and obtain forecasts that are accurate and reliable, not only at locations where recent data are available but also at locations without observations. The methodologies and results are relevant for wind forecasts across the globe and significantly add to the existing literature on spatio-temporal modeling.