Abstract: A new nesting technique called three dimensions nesting (3DN) is used in a regional model which is nested in a circulation model based on the complex terrain with high resolution model (YH model). A long term integral is achieved steadily which provides a useful way to process the lateral boundary conditions for regional climate model. 3DN means not only keeping horizontal lateral boundary stable but also changing ways to use GSM data in vertical dimension. The primary results show that only with the appropriate high level data that the long term integral could obtain better outcomes in 3DN model. During the prophase integral, 3DN has an obviously positive impact just on the levels lower than 500 hPa, but not for the other levels. However, with the growing process of integral period, the good effect of 3DN is exhibited gradually. The variance of each level is smaller than the one derived from the two dimensions nesting (2DN) model, particularly for the high levels compared with the result from the prophase integral. As for the mean deviation of height field at 500 hPa level during the whole integral period, the variances of these two nesting technologies have a common character that is the value of deviation decreasing from the North to the South. In the eastern integral area, the deviation of 3DN is obviously smaller than the one of 2DN and more similar to the reference field. The mean variance of the u field for each level in the whole integral time shows that 3DN is better than 2DN atthe levels which are lower than 100 hPa; while the one of the v field progresses at all levels. At 100 hPa level, the improved percentage is nearly up to 5% and much better than that of the u field. In the long integral period, dynamic energy and geopotential height keep their stableness with a small swing range, which indicates that 3DN can be run steadily in long term integral numerical experiments. Large scale precipitation, vapor flux and so on are analyzed. 3DN is obviously better than 2DN for large scale precipitation, especially in the middle of the integral term, the result of 3DN is almost the same as that of the reference field. With the increase of the integral time, the errors of both nesting methods rise at a certain degree. During the long integral term, the nesting error of vapor prediction is relatively large at the upper boundary out-going level. However, within that area, the error extremum of vapor flux from 3DN is much smaller than that from 2DN. With regard to the whole integral term, the advantage of 3DN appears especially in the mid-phase and anaphase. It is necessary to point out that there is not much processing for the large scale model data, which is likely the cause for the less satisfaction at upper levels in the integral prophase. If potential nesting is considered, which means that a remittent area will be established when the regional model vertically absorbs data from the large scale model, better result may be obtained in the future study.