A novel type of fuzzy information granule involving a time-dependent (non-stationary) membership function is proposed. The Hausdorff distance for the new proposed type of granules can be expressed in a simple way. A long-term prediction method based on fuzzy inference system and the new type of granules along with their distance measurement is proposed. Prediction results via the new proposed type of granules are able to summarize the underlying nature of the future data to a significant extent.