Abstract: Optimal Regional Averaging Method is an effective method to reconstruct temperature variations of the past. It has the advantage of obtaining a reliable regional-scale temperature reconstruction using a limited dataset. This method has been successfully applied in reconstructing average temperature on a given target area in the recent years. However, it should be noted that this method only can calculate average temperature over a local area, and there is still no continental-or hemispheric-scale temperature reconstructions reported. In this paper, we use spherical harmonic function to improve this method.To validate reliability of this method, we calculate average temperature series over the Northern Hemisphere during the period 1951~2013 using temperature data from Climatic Research Unit(CRU), and compare it with the result derived from the simple average method. The results show that the improved method is better than the simple digital method while using the same dataset, and it can calculate global average temperature series in the case of discontinuous data.We also use 15 regional Maximum Latewood Density (MXD) series from Schneider et al. (2015) to reconstruct summer temperature during the past 1400 years using the improved Optimal Regional Averaging Method. The reconstruction results show that there are obvious warm period in the Medieval Period(800~1100A.D.), low temperatures during the Little Ice Age(1700~1900A.D.), and the warming trend in the 20 century. These results support that the improved Optimal Regional Averaging Method is an effective method to reconstruct temperature variations of the past.