利用树轮资料重建长白山区过去气候变化

RECONSTRUCTION OF CLIMATE CHANGE ON CHANGBAI MOUNTAIN, NORTHEAST CHINA USING TREE-RING DATA

本文以树轮气候学中常用的树轮宽指数资料作为过去气候的代用资料，重建了长白山区1655年以来1～4月月平均最高气温的变化。重建中的校准方程稳定性较好，并可解释重建变量方差的57．4％。在1833年以来重建可靠性较高的时段中，1974～1979年为低温期，1951～1963年为高温期。而结束于1861和结束于1897年的两个高温期持续时间最长。本文的研究结果展示了利用树轮资料重建我国长白山区过去气候变化的巨大潜力。

Tree--ring data were sampled from the north slope of Changbai Mountain, Northeast China. Three spedes from five sites growing along altitudinal gradient were coiled. For each site the tree cores were cross--dated using skeleton plot and measured to the nearest 0.01 mm using standard techniques. The dating and measurement were validated using an improved version of the COFECHA program. Final chronologies were developed using the program ARSTND. The age trend of trees was modeled by a cubic spline. Using a bi--weight robust mean method, de--trend indias were averaged to form standard chronologies (STD). To minimize the variations that reflect competitive interactions between trees, further standardization was done by fitting an autoregression model to the de--trend index and residual chronologies (RES) were calculated using the same average method.Common interval analysis indicates that RES chronologies are with more common yearto --year variations and express more population signal. Therefore, RES chronologies are suitable for dendroclimatic research in this study. Comparisons of climatic quality among five RES chronologies indicate that trees growing near their upper elevational hot are more suitable for dendroclimatic study than those growing at the upper elevational limit and at the center of the distribution. Climatic data for Tianchi, Antu and Changbai meteorological stations were employed and four climatic elements, monthly mean minimum temperatUre, monthly mean temperature, monthly mean maximum temperature and monthly total precipitation, were involved in this study. All data were tested for homogeneity.Response function, correlation function, scatter diagram and single--year analysis were utilized to stUdy the response of tree radial growth to climatic elements. The results showed that ma studied were mainly affected by climatic elements. From 63% to 78% percent of growth variance can be accounted for by monthly mean air temperature and monthly total precipitation. And the response to temperature is stronger than to precipitation. While,for three temperature elements, the response to monthly mean maximum temperature is the strongest, especially, in winter and spring seasons. As a result, monthly mean maximum temperature was selected to reconstruct.Multiple linear regression analyses were used to perform calibration analysis.Regression equations were developed by regressing monthly mean maximum temperature for various combinations of blocks of month against ring-width data from all sites. To reconstruct a seasonal series of climate, January to April was selected as the reconstructing season, and an average record of January to April mean maximum temperature was formed from the three stations providing a regional record. This regional record allows to reconstruct past climate of Changbai Mountain.Regressions were calculated between January ~ April mean maximum temperature and ring --width data for various combinations of tree- ring sites. Band on the equation statistics such as the percentage of variance expressed by the equation, two regression equations were selected, one is for reconstruction before 1833, and the other for after. house a chrOnology from ANPK site was included in the second equation, the statistics of the equation were improved. The stability of there two equations was tested by cross--validation procedure using verification statistics including sign test, product mean and reduction error. All of these statistics indicated that the selected equations were of the predicative capability,and that they can be used as the transfer functions.According to the transfer functions, January ~ April mean maximum temperatures of Changbai Mountain were reconstructed year by year, and the reconstruction was smoothed by applying a resistant smoother. It was found that the reconstructed low--frequency variations were more reliable than the year--to--year variations. Therefore, only the low--frequency variations were discussed. From the 339 years' low--frequency variations, it was found that the departures oscill