A new statistical model for solar irradiance prediction is described. The method makes use of the atmospheric parameterizations as well as a time-series model to forecast a sequence of global irradiance in the 3-10 min time frame. A general literature survey of some of the prominent research of the recent past reveals a definite lack of irradiance models that approach subhourly intervals, especially in the range mentioned. In this article, accurate parameterizations of atmospheric phenomena are used in a prewhitening process so that a time-series model may be used effectively to forecast irradiance components up to an hour in advance in the 3-10 min time intervals. Input requirements are not very restrictive; the model requires only previous global horizontal irradiance measurement at a site. Results show that when compared with actual data at two locations in the southeastern United States, the forecasts are quite accurate, and the model is site-independent. Under some instances, forecasts may be inaccurate when there are sudden transitional changes in the cloud cover moving across the sun. In order for the proposed irradiance model to predict such transitional changes correctly, frequent forecast updates become necessary.