基于1993年和2003年采集并研制的15个天山云杉树轮宽度年表,重建了新疆天山北坡乌鲁木齐河357年的年径流总量。影响树木生长的水文气象因子对河水流量的影响具有3~5年的滞后,基于3个树轮差值年表的重建方程可解释器测流量资料总方差的69.4%,使我们可以重建出1633~1989年的年径流总量。重建流量序列具有11个丰水期和11个枯水期,以及59.5年,47.6年,3.72~3.84年和3.45~3.50年的变化准周期,且在1639年和1709年发生过从丰水到枯水的突变,持续最长的枯水期达13年。流量重建结果可为诊断现代极端的洪旱事件提供长期的流量变化背景资料,并使我们有可能发现对水资源管理系统具有重要影响的过去持续洪旱期。
We present tree-ring-based reconstruction of annual total runoff( 1633 - 1989A. D. ) for the Urtimqi River on the northern slope of Tien Shan Mountains, Xinjiang, China. At the study sites,Tien Shan spruce (Picea schrenkiana var. tiansclhanica)forests growing on hillsides beside valleys. All increment cores were taken from the living trees at the 15 sites of the Urtimqi River basin in 1993 and 2003. After air drying,mounting and sanding in the laboratory, annual ring widths were measured to the nearest 0. 001mm using a Velmex measuring system. The quality of the measurement and cross-dating of ring-widths was checked using the COFECHA program. After removed the biological growth trends while preserved the variations which were likely related to climate change,the 15 Tien Shan spruce chronologies were developed by the ARSTAN program. Because of the hydroclimatic forcing driving tree growth variability affected runoff with a three-to-four-year lag and being very slow,the transfer from precipitation to runoff based on the model between runoff ( Wt ) and tree-ring index ( It-4, It-3 , It-2, It-1 , It, It+1 , It+2, It+3 and It+4 ) . Thus,we chose the stepwise regression reconstruction method to reveal the potential of reconstructing runoff from the tree-ring width chronologies( STD, RES and ARS). The results indicated that the explained variance of the standard chronology to annual total runoff is 69.4%, the residual chronology is 80. 8% and the autoregressive chronology is 74.5%. Finally, based on three residual chronologies, a model was used to estimate the annual total runoff. The model accounted for 69.4% of the instrumental data variance and allowed us to reconstruct runoff for the period 1633 -1989. Cross-validation tests were employed to evaluate the statistical fidelity of our reconstruction model. The reduction of error is positive, which indicates significant skill in the tree-ring estimates. High runoff periods occurred during 1633 - 1638, 1666 - 1685, 1690 - 1701, 1723 - 1755, 1