The relationship between climate factors and tree growth are very complex. Traditional standardization methods still face challenges in retaining the climate signal while removing age-related growth trends and non-climatic disturbances. In this paper, we introduce an Empirical Mode Decomposition (EMD) based standardization method to develop the ring-width chronology, which has the advantages of excluding non-climatic disturbances in individual tree-ring series. First, tree-ring measurements were decomposed by the EMD to produce a number of Intrinsic Mode Functions (IMFs) with different physically-meaningful frequency band. Second, the lowest frequency IMF band, mainly representing age-dependent growth trend, was used to calculate tree-ring index divided by the original treering width records. Based on this method, a new tree-ring chronology was developed from 44 Spruce trees (Picea crassifolia Kom.) at Kong-gang-mu (Kgm) Forest of Sunan County of Qinghai Province in the Qilian Mountains, and the elevations ranged from 2620m to 2770m, which covered the period 1723A.D. to 2000A.D. However, the chronology was more reliable before 1845A.D. when the EPS was greater than 0. 85. Compared two ring-width chronologies from EMD method and traditional method, we can found that the EMD method has the significant advantages in extracting the age-dependent growth trend,especially for improving fitting on the early " slow-growth and the late flat-growth periods. It is noted that two chronologies exhibited the coincident significant correlation with meteorological factors of Sunan Meteorological Station (38°50'N, 99°37'E;2311m a.s.l.) in observation period (1957 ~ 2000),showed the highest correlation coefficient (r = 0.664, p<0.01) with the total precipitation of the year from the previous August to current July, which indicated that the EMD method successfully preserved climatic signals. Although the EMD method appears to offer both practical and conceptual advantages, we emphasize that it is premature at this stage to conclude that the EMD-based method is better than the traditional methods. However, EMD method provides an additional option to mitigate bias in low-frequency variations longer than mean segment lengths of tree rings,which is inevitable with traditional methods. The spectral analysis showed three main wavebands of this chronology : less than 10-yrs period, 10 ~ 30yrs, 30 ~60yrs, which would be integrated to three components of IMFs of all original ring width series derived from EMD method. The 10-yrs scale of IMF component was relatively stable in the time-frequency domain, representing a major portion of the high-frequency precipitation variations, with the maximum amplitude around the 1910s ~ 1920s. The 10 ~30yrs scale of IMF component displayed the obvious lowest amplitude during the 1925 ~1933,corresponding to an anomalously dry period at that time of Gansu Province. And the 30 ~ 60yrs scale of IMF component showed that three dry-wet fluctuation. The integration components of the tree-ring width series with different main periods revealed the temporal distribution and detailed fluctuations of the major periods by their corresponding amplitudes in the time-frequency domain, which provide more information of frequency-domains of reconstructed climate series at multiple time-scales.