Abstract:Using precipitation data and the number of days without rainfall in arid dog days during near 40a (1969—2008) from 10 representative stations in Shaanxi major apple(pear) cultivating towns, which are similar in ecological environment , climate characteristics and the level of industrial development, index of arid dog days has been designed and calculated, and is divided into strongest, stronger, medium, weak and weaker five levels, on the basis of which ,conducting independent sample trial by using typical k-order autoregressive AR ( k) forecasting model. The basic result of this research is that index of arid dog days could objectively reflect the intensity of the summer drought, and all of them have different characteristics in spatial and temporal distribution, which is because of the different latitude, altitude, ecological environment and climate background.In addition, the accuracy and the basic accuracy of typical K-order autoregressive forecasting model is about 70%, indicating that the effect of this forecasting model is still good and has practical value.