Quality Engineering is not the methods to solve a real problem, in this article a long-term forcasting, but it is the methods to evaluate the error variance of prediction from available sources of informations. The most difficult part of a longterm-prediction is to define Mahalanobis space. When we define the space of all normal years for some prediction such as total rain falls during summer time, it is possible to predictonly whether it would be normal or unusual but impossible to predict under or upper than normal.
Those problems are explained in this article with a new method.