【作者】 郑家文；

【导师】 丁平；

【作者基本信息】 浙江大学 ， 环境科学， 2006， 博士

【副题名】栖息地、分布、密度与集合种群

【摘要】 1999~2003年，在浙江省开化县境内调查了白颈长尾雉栖息地斑块及斑块内该雉的分布情况。研究白颈长尾雉栖息地丧失与片段化，以及斑块特征与种群分布、相对密度的关系。在此基础上，引入集合种群理论构建分析开化县白颈长尾雉集合种群，提出种群在斑块间的平均扩散距离估算方法，研究白颈长尾雉保护对策。主要的研究结果如下：1．利用遥感与地理信息系统技术，从开化县的TM遥感数据中识别1986年及2000年分别有284个和151个自颈长尾雉栖息地斑块。14年间该雉的栖息地面积丧失13309.74ha，平均邻近距离从416.18m增加到673.93m，片段化程度也相应加剧。由于很多小栖息地斑块的丧失及局部的封山育林，使得2000年的平均栖息地斑块面积218.32ha要比1986年大56.21ha。层次网络分析表明，1986年栖息地斑块网络结构中各层次间与各类簇内部斑块间的联系都比2000年紧密。2．以羽迹法实地调查开化县内151个栖息地斑块的占有情况，其中52个为白颈长尾雉占有。选取栖息地斑块的面积、形状指数、邻近距离、邻近指数、高程、坡度、坡向、植被指数、离最近公路距离、离最近居民点距离、离最近河流距离等11种景观参数，并分析它们对斑块占有的影响。t检验分析表明斑块的面积、形状指数、邻近距离、植被指数等四个特征对白颈长尾雉的栖息地斑块占有有极显著的影响。前向逐步选择的多元二值逻辑斯蒂回归分析得到白颈长尾雉栖息地斑块占有回归模型：其中P(y_i=1)斑块i内有白颈长尾雉出现的概率，HA_i、PR_i和VI_i分别是斑块i的面积、邻近指数和植被指数。3．在古田山自然保护区及周边，对27个栖息地斑块内的植被与景观特征以及白颈长尾雉的羽迹密度进行了调查。为减少数据冗余将总盖度、乔木层盖度、灌木层盖度、乔木层高度、灌木层高度、植被类型等6个栖息地斑块内的植被特征进行主成分分析，得到反映生态外貌特征的PC1和反映灌木层特征PC2。PC1与PC2连同斑块面积、海拔、坡度、连接度、干扰度、相对保护区位置等栖息地斑块特征被用于白颈长尾雉斑块占有影响分析，以及可预测模型的建立。t验与χ~2检验表明只有斑块面积显著影响斑块占有。采用逐步降元的多元线性回归分析建立模型：D_i=0.852+0.078PA_i-0.137PI_i-0.006PD_i其中：D_i为预测栖息地斑块i内局部种群的相对密度，PA_i为斑块i的面积，PI_i为斑块i的连接度，PD_i为斑块i内的干扰度。4．以估计的参数u=0.04、x=2、α=1、b=0.5及y=2，构建开化县白颈长尾雉概率函数集合种群模型。该模型预测开化县境内的白颈长尾雉种群可以长期续存，未来100年里平均的斑块占有率为18.0％。参数α、b及y敏感性分析结果表明：反映种群在斑块间的平均扩散距离的α参数对整个集合种群续存影响最大，当α=0.67时，开化县未来100年里平均的斑块占有率从18.0％扩大到34.0％。分析两种斑块间邻近距离对模型预测的影响表明斑块间边界邻近距离替代重心点间邻近距离对提高斑块占有率不显著。区域随机的引入，使集合种群的续存能力降低。当加入五十年一遇区域随机后，开化县未来100年里平均的斑块占有率从18.0％减少到8.0％。5．基于集合种群原理，通过空间等级层次聚类分析，搜寻已占有斑块与其它斑块相隔最远的距离确定种群在栖息地斑块系统内最长扩散距离，搜寻一个所有斑块都未灭绝的斑块族群，这一族群内斑块间最大的隔离距离确定为种群经常扩散距离，最长扩散距离与经常扩散距离的中间值即为种群的平均扩散距离。该斑块间扩散距离估算模型利用GIS控件实现了软件化，并通过实例集合种群的应用对比实际斑块间平均扩散距离，验证了该模型实用性，同时也说明了其局限性。6．选择四种较典型类型的白颈长尾雉栖息地斑块系统，在现状基础上设置斑块面积丧失50％与恢复20％、斑块间的隔离距离增加50％与缩短20％四种栖息地斑块变化情景，并分析各斑块系统内集合种群的生存能力。依据分析结果提出白颈长尾雉的保护原则：①应以整体区域复杂栖息地斑块系统的全局保护优先，②局部区域栖息地斑块系统内应以恢复栖息地面积优先。更多还原

【Abstract】 During 1999 to 2003, habitat patches and distribution of Elliot’s pheasant were investigated in Kaihua county, Zhejiang province, China. The current status of habitat loss and fragmentation, and the effects of habitat patch attributes on population distribution and relative population density were analyzed. Furthermore, the Elliot’s pheasant metapopulation in Kaihua county was constructed and analyzed. Also with metapopulation theory, a new method to evaluating average population dispersal distance between habitat patches was developed, and new conservation strategy on Elliot’s pheasant was presented. The major results are as follows.1. A total of 284 Elliot’s pheasant habitat patches in 1986 and 151 in 2000 were identified from the satellite image data using techniques of remote sensing and geographic information system. There had been lost 13309.74 ha habitat area and lengthened average nearest neighbor distance from 416.18 m to 673.93 m during 1986 to 2000, which made habitat fragmentation more seriously. Many small habitat patches had been removed and few local area had reforested, that made average habitat patches size increase from 162.11 ha in 1986 to 218.32 ha in 2000. The results of hierarchic network analysis showed that habitat patches were tighter in 1986 than in 2000 not only inter-clusters lever but also at intra-cluster lever.2. Using collecting molted feathers in line transects, a total of 52 habitat patches occupied by Elliot’s pheasant were recorded in all 151 habitat patches in Kaihua county. The patch area, shape index, distance to nearest neighbor, proximity index, altitude, slope degree, slope aspect, vegetation index, distance to the nearest road, distance to the nearest village, and distance to the nearest river were selected as variables to describe each habitat patch, and the effects of those patch attributes on Elliot’s pheasant were analyzed. The results t test showed that patch area, shape index, distance to nearest neighbor, and vegetation index significantly affect habitat patch occupancy. The predictable habitat patch occupancy of Elliot’s pheasant was generated by binary logistic regression using all variables in a forward stepwise selection method. Here is the model: where: P(y_i=1) is the probability of being occupied in patch i, HA_i is area of patch i, PR_i is proximity index of patch i, and VIi is vegetation index of patch i. 3. The number of molted feathers which could be calculated to express as relative local population size were collected with transaction line method, and the vegetation properties were surveyed in every one of 27 habitat patches in Gutian Mountain Nature Reserve and its vicinities. The vegetation properties of habitat patch, such as total cover, arbor cover, shrub cover, arbor height, shrub height, and vegetation type, were analyzed using principal component analysis to reduce data redundancy, and its PC1 present external ecological appearance and PC2 present shrub properties. In these attributes variables, such as PC1, PC2, area, elevation, slope degree, connectivity index, disturbance index, relative location, the patch area was the only variable that showed significant positive relationship to local relative population density in t test and x 2 test. The predictable local relative population density of Elliot’s pheasant was generated by multiple variables general linear model procedure in a backward stepwise selection method. Here is the model: D_i = 0.852 + 0.078PA_i - 0.137PI_i - 0.006PD_i where: D_i is local relative population density in patch i, PA_i is area of patch i; PI_i is disturbance index in patch i.4. With estimated parameters (u=0.04, x=2,α=1, b=0.5, y=2), the Elliot’s pheasant metapopulation in Kaihua county was constructed as incidence function metapopulation model. This metapopulation predicted 18.0% average patch occupancy in 100 years and its long time persistence. The results of parameters (α, b, and y) sensitivity analysis showed that metapopulation persistence was significantly affected withα, which is relative to the average population dispersal distance between habitat patches. Asαreduced to 0.67, the average patch occupancy would increase from 18.0% to 34.0% in 100 years. No significant effects on metapopulation persistence took place if center-to-center nearest neighbor distance replaced with the border-to-border nearest neighbor distance in metapopulation model. When regional stochasticity enclosed in model, metapopulation persistence would decrease. If there had one of fifty years encounter regional stochasticity, the average patch occupancy would decrease from 18.0% to 8.0% in 100 years.5. Based on metapopulation principles, using spatial hierarchic cluster analysis for patchy system, the longest population dispersal distance between habitat patches is the longest nearest neighbor distance of an occupied habitat patch, the frequenty population dispersal distance between habitat patches is the longest nearest neighbor distance in a cluster with all occupied habitat patches, the average population dispersal distance between habitat patches is average value of the longest population dispersal distance and the frequenty population dispersal distance. This average population dispersal distance estimate method had been carried out software program using GIS Active X Control. With the help of some metapopulation examples, the average population dispersal distance estimated out of this program could contrast with the known value, which could validate its practicability and limitation.6. Four typical habitat patch system were chosen, and made every patch system to encounter four suppositional patch change scenario, in which the first was to lose 50% patch area, the second was to recover 20% patch area, the third was to increase 50% distance to nearest neighbor, and the fourth was to decrease 20% distance to nearest neighbor. The Elliot’s pheasant metapopulation persistence in each scenario was analyzed. And according analysis results, the major principles of Elliot’s pheasant conservation are: the precedence protection action should manage the whole big area with complicated habitat patch system, and recover habitat area was very important action for habitat patch system in relative isolated local region.更多还原