长白山自然保护区历史森林景观的初步重建

本文在对环境因子(包括年均温、年降水量、海拔高度、坡度和坡向)进行空间表达的基础上,建立了长白山自然保护区当前森林景观垂直带(包括阔叶红松林、云冷杉林、岳桦林和苔原)与这些环境因子间的逻辑斯蒂回归模型,然后利用1997年各环境因子的空间数字面来反推1975年森林景观带的空间分布,并根据1975年MSS遥感影像计算机监督分类的结果,用Kappa指数对模型预测结果进行验证。结果表明,苔原、岳桦林、云冷杉林和阔叶红松林1975年的预测分布面积分别为7243.83hm2、6517.08hm2、125570.16hm2和34264.80hm2,与遥感分类结果相比,苔原和云冷杉林的面积分别减少了29.04%和3.05%;岳桦林和阔叶红松林的面积分别增加了246.45%和8.6%。用逻辑斯蒂回归模型可以较好地预测苔原在保护区内的分布范围,同时在一定程度上也能预测阔叶红松林的分布范围,但是对岳桦林和云冷杉林分布范围的预测结果较差。尽管如此,作为一种方法性的探讨,本文可以为森林景观空间分布和环境因子变化的关系以及森林景观对气候变化的反应研究提供一种新思路。

THE PRELIMINARY RECONSTRUCT OF HISTORICAL FOREST LANDSCAPES IN CHANGBAI MOUNTAIN NATURAL RESERVE

Based on the spatial parameters of environmental factors including annual mean temperature, precipitation, altitude, slope and aspect. The logistic regression models were established, which reflect the relationships between the current distribution of forest landscapes, consisting of Korean pine hardwood forest, spruce/fir forest, mountain birch forest and tundra, and these five abiotic factors. The spatial distribution of the four forest landscape belts were estimated according to the corresponding digital surfaces of the five abiotic factors in 1975. Finally, the MSS imagery in 1975 was supervised classified by maximum likelihood algorithm. The resulting landscape map was used to validate the forest landscapes distribution map produced by logistic regression models using Kappa index. Results show that the predicted area of tundra, mountain birch forest, spruce/fir forest and Korean pine hardwood forest are 7 243.83 hectares, 6 517.08 hectares, 125 570.16 hectares and 34 264.80 hectares respectively. Compared with the results of classification of MSS imagery, the predicted area of tundra and spruce/fir forest decreased by 29.04% and 3.05% respectively, and that of mountain birch forest and Korean pine hardwood forest increased by 246.45% and 8.6% respectively. The modeling result of tundra is favourable. And that of Korean pine hardwood forest is also acceptable in a certain extent. But agreement of mountain birch forest and spruce/fir forest is poor. However, as a methodological exploration, our study do provide a new idea for research on responses of forest landscape to climatic change as well as on the relationship between distribution of forest landscapes and changes in environmental factors.