Abstract: Due to the abundant primordial forest and good grassland resources in the stream section，the Kaiduhe River plays a crucial role in the development of agriculture and grassland farming in South Xinjiang，China. To better understand the climatic regimes of the Kaiduhe River watershed，it is important to obtain the climatic variation information in the past for this region，because it enables the recent observed changes to be viewed from a long-term perspective. The abundant forest resource in the Kaiduhe River watershed provides an excellent opportunity for dendroclimatological studies. Tree-ring datum have been used widely as important archives in paleoclimatic research because of their precise dating，annual resolution，and comparability with instrumental meteorological data. In this study，a total of 89 cores from 46 spruce trees [（Picea schrenkiana）] were obtained for 2 individual sites，named as Haerhatigou No.1 （HE1） and Haerhatigou No.2 （HE2），ranging from the the upper and lower forest limits in the middle section of the Kaiduhe River watershed in August 2011. All of tree-ring cores were natural dried，mounted，surfaced and cross-dated following standard dendrochronological procedures. Each ring-width was subsequently measured with a resolution of 0.001 mm by dendrometer. The quality control of cross-dating was carried out using the COFECHA program. The ARSTAN program was used for de-trending and developing tree-ring chronologies. The age- and size-related variation of the tree-ring widths was removed by a single detrending process applied on each tree-ring measurements: a negative exponential curve，a kind of conservative function. After all these processes，three kinds of chronologies from two sample sites were developed: standard chronology，residual chronology，and Arstan chronology. The standard chronologies for two sample sites were used in our study，which contains the common variations among the individual tree core series and retains low through high-frequency common variance，presumably in response to climate.Compared among statistical characteristics of the chronologies，HE1’s standard deviation，mean sensitivity，mean correlation between all series，mean correlation within a tree，mean correlation between trees，and signal-to-noise ratio except expressed population signal were all higher than HE2’s. The result means the chronology at the upper orest limits may contain more climatic signals. Compared with two chronologies in the original，high-and low-frequency domain，the results show that there are significant correlations. The results of tree growth-climate responses show that two chronologies were found to be positively correlated with the monthly mean temperature from the previous September to current March at the 95% confidence level. The simulation results of tree-ring width chronologies with climatic factors showed that the monthly mean temperature from the previous September，the previous December，the current February，the current March and the current April may simulate the chronologies at the the upper and lower forest limits well. The similar long-term trends among two chronologies and three winter NAO indexes reveal that the tree-ring width growing variation in the study area has strong teleconnections with the oscillations of land-atmospheric-ocean circulation systems.

Key words: Kaiduhe River, Picea schrenkiana, tree-ring width, climate response