中文摘要：

Constructing a mathematic transfer function of pollen and climatic factors is one of the most important approaches in the quantitative reconstruction of paleoclimate elements.In the function building,the key point is the response sensitivity to the climate change for different pollen families and categories. As an example in this paper,the pollen samples from the surface and stratum in Dajiuhu basin,Shennongjia are used to estimate the sensitivity of pollen factor-temperature in the transfer function by the EOF analysis,multiple regression and stepwise regression techniques.Thereafter,the selection methods of pollen factors are discussed and compared with other results from different researchers. The results show that in the pollen samples,the quantity of woody plants is larger than others,but the woody plants have relatively lower sensitivity and slow responses to the climate evolution and abrupt climate change.While the pollens from lower grade plants(herb and fern)have a relatively high sensi- tivity to the temperature change and fast response to the abrupt climate changes.Therefore,the pollen of herb and fern may significantly record the extreme events in the climate change.Also in different regions,the pollen samples have different sensitivity and optimum combination in the transfer functions.Final conclusion is that the stepwise regression is one of the best methods for transfer function building since it can obtain the maximum multiple correlation coefficients and optimum combination of sensitive pollen factors.

英文摘要：

Constructing a mathematic transfer function of pollen and climatic factors is one of the most important approaches in the quantitative reconstruction of paleoclimate elements. In the function building, the key point is the response sensitivity to the climate change for different pollen families and categories. As an example in this paper, the pollen samples from the surface and stratum in Dajiuhu basin, Shennongjia are used to estimate the sensitivity of pollen factor-temperature in the transfer function by the EOF analysis, multiple regression and stepwise regression techniques. Thereafter, the selection methods of pollen factors are discussed and compared with other results from different researchers. The results show that in the pollen samples, the quantity of woody plants is larger than others, but the woody plants have relatively lower sensitivity and slow responses to the climate evolution and abrupt climate change. While the pollens from lower grade plants (herb and fern) have a relatively high sensitivity to the temperature change and fast response to the abrupt climate changes. Therefore, the pollen of herb and fern may significantly record the extreme events in the climate change. Also in different regions, the pollen samples have different sensitivity and optimum combination in the transfer functions. Final conclusion is that the stepwise regression is one of the best methods for transfer function building since it can obtain the maximum multiple correlation coefficients and optimum combination of sensitive pollen factors.