Three ring-width chronologies were developed from Qilian Juniper （Sabina przewalskii Kom.） at the upper treeline along a west-east gradient in the Anyemaqen Mountains. Most chronological statistics, except for mean sensitivity （MS）, decreased from west to east. The first principal component （PC1） Ioadings indicated that stands in a similar climate condition were most important to the variability of radial growth. PC2 Ioadings decreased from west to east, suggesting the difference of tree-growth between eastern and western Anyemaqen Mountains. Correlations between standard chronologies and climatic factors revealed different climatic influences on radial growth along a west-east gradient in the study area. Temperature of warm season （July-August） was important to the radial growth at the upper treeline in the whole study area. Precipitation of current May was an important limiting factor of tree growth only in the western （drier） upper treeline, whereas precipitation of current September limited tree growth in the eastern （wetter） upper treeline. Response function analysis results showed that there were regional differences between tree growth and climatic factors in various sampling sites of the whole study area. Temperature and precipitation were the important factors influencing tree growth in western （drier） upper treeline. However, tree growth was greatly limited by temperature at the upper treeline in the middle area, and was more limited by precipitation than temperature in the eastern （wetter） upper treeline.
We established a Juniperus przewalski tree ring width chronology, based on tree ring cores collected from the A’nyêmaqên Mountains. Statistical analysis showed that the chronology was highly correlated with instrumental streamflow records from previous August to current July from the Tangnaihai Station in the upper reaches of the Yellow River, with a correlation coefficient of 0.656. Streamflow for the upper reaches of the Yellow River was reconstructed for the past 1234 years. Low flow periods for the 11-year averaged streamflow reconstruction were definite as lower than mean plus 1 standard deviation, and high flow periods were higher than mean minus 1 standard deviation. Over the past 1234 years, high flows occurred 18 times, and low flows occurred 12 times. The main low flow periods were identified as AD 1140–1156, AD 1295–1309, AD 1473–1500, and AD 1820–1847, and the main high flow periods were identified as AD 846–873, and AD 1375–1400. Extremely low streamflow over the reconstruction period was seen during the late 15th century, coinciding with a widespread drought phenomenon, which took place in the northeastern Tibetan Plateau over the same period. Reconstructed streamflow shows significant low-frequency variability, which is in line with drought variability of neighboring regions, as inferred from tree rings and other proxies. Multi-taper spectral analysis suggests the existence of significant periods of 2–5, 22, 35–38, 55–62, and 114–227 years, particularly significant for cyclic variations of years 159 and 36.
