Wind power - a renewable energy source increasingly attractive from an economic viewpoint - constitutes an electricity production alternative of growing relevance in current electric energy systems. However, wind power is an intermittent source that cannot be dispatched at the will of the producer. Modeling wind power production requires characterizing wind speed at the sites where the wind farms are located. The wind speed at a particular location can be described through a stochastic process that is spatially correlated with the stochastic processes describing wind speeds at other locations. This paper provides a methodology to characterize the stochastic processes pertaining to wind speed at different geographical locations via scenarios. Each one of these scenarios embodies time dependencies and is spatially dependent of the scenarios describing other wind stochastic processes. The scenarios generated by the proposed methodology are intended to be used within stochastic programming decision models to make informed decisions pertaining to wind power production. The methodology proposed is accurate in reproducing wind speed historical series as well as computationally efficient. A comprehensive case study is used to illustrate the capabilities of the proposed methodology. Appropriate conclusions are finally drawn. (C) 2009 Elsevier Ltd. All rights reserved.