Abstract: Forest arithmetic mean height measurement plays a quite important role in the study of ecological factors and forest resource inventory, however , which is usually thought to not only cost too much labor and time, but be low accuracy. A new sampling technique introduced by Yojiro Ueno is that without knowing the exact number of trees , it can efficiently and easily estimate the arithmetic mean height . Based on computer simulation, the paper applied this technique into plots from Larix plantations and Mongolian oak forest in Changbaishan mountains. The results from different repeat times showed that 1) the differences of mean relative error as [AKP-] and variation coefficient as CV between different repeat times tended fo decrease as the number of sampling points increased. This sampling technique was efficient by doing only one time when the number of sampling points was big enough. 2)[AKP-] and CV had similar decreasing trends as the number of sampling points was increasing , which showed that they decreased remarkably before the number of sampling points reached a certain value , after that the speed of decreasing slowed down .Their fluctuation degrees also decreased as the number of sampling points increased. 3) This technique used in Larix plantations was better than in Mongolian oak forest. It was demonstrated that the lower CV, the larger the number of all trees , the better effectivity this sampling technique had.