Wind energy is the fastest growing energy generation source in the world today. As the penetration rate (the ratio of wind energy to other generation) grows, the need for accurate wind energy forecasts becomes paramount: for system reliability, scheduling, and long-range planning. The use of geographically dispersed meteorological observations in conjunction with advanced statistical space-time forecast methodologies results in significant improvements to short-range forecasts. The statistical method that we propose depends on spatially dispersed observations of wind and other atmospheric variables. Since changes with wind often propagate with the wind it is possible to use upstream (upwind) observations to detect precursors to wind energy output at a wind energy site. The algorithm incorporates observations from locations away from the wind farm and is demonstrably better than algorithms that do not include the remote data.