【作者】 乐通潮；

【导师】 张会儒；

【作者基本信息】 中国林业科学研究院 ， 森林经理学， 2014， 博士

【摘要】 红树林生态系统是处于海洋生态系统与陆地生态系统之间经长期演替形成的一种系统结构稳定、生产力极高的特殊森林生态系统，具有防浪护岸、促淤造陆、净化海水、维持生物多样性等生态系统功能。红树林健康经营的目的是为了维护生态系统功能完整、维持系统的健康稳定、保持功能的持续以及生产力高效。因此，红树林生态系统健康评价，调整红树林景观格局，优化红树林林分空间结构，是当前保护与恢复红树林工作亟待解决的问题。本文以福建漳江口红树林为研究对象，利用3期（2003、2008、2013年）多源高空间分辨率卫星影像（SPOT5和PLEIADES）研究了红树林湿地景观格局动态变化及演化机制，在不同的潮位带（高、中、低潮位）选择有代表性的红树林群落进行每木定位林分因子调查，开展基于GIS技术的红树林林分空间结构特征分析研究，运用压力-状态-响应（PSR）模型和层次分析法（AHP）对红树林生态系统健康进行评价。在景观层次上，提出红树林景观结构布局优化方案；在林分尺度上，建立了基于林分空间结构综合指数的红树林林分采伐和补植两个空间结构优化模型，应用优化模型对研究区红树林进行经营实践的模拟分析；从生态系统上，提出削除或减弱影响红树林生态系统健康的外界压力的红树林生态系统可持续利用管理措施。主要研究内容和结论如下：（1）基于SPOT5和PLEIADES影像高空间分辨率多源卫星数据，通过构建堤岸线、Band1/Band2、(Band4-Band3)/(Band4+Band3)、Band3/Band2等波段运算辅助信息，建立决策树分类的规则对研究区的红树林景观类型进行了信息提取。结果表明3期影像的分类总体精度分别达到97.24%、92.90%、97.86%，Kappa系数分别为0.97、0.91、0.97；与无增加辅助信息的监督分类的解译结果相比，总体精度分别提高了16.91%、7.90%、12.21%。（2）2003-2013年，研究区红树林面积由47.69hm2增加到64.24hm2，红树林的整体质心向东北方面略微移动，互花米草呈现先减少后快速增加的趋势，10年间增加了42.72hm2，侵占红树林面积13.38hm2，人工养殖塘几乎成倍增长，面积由148.69hm2增加到254.84hm2，将大量天然湿地转为人工湿地，并蚕食红树林面积2.11hm2，导致红树林生态系统衰退。红树林湿地景观格局指数分析表明研究区景观多样性持续增加，景观的破碎化程度增大，各类型斑块趋向均衡发展，景观的连续性与整体性变差。红树林湿地景观动态变化的主要因素为围塘养殖、堤岸维修、互花米草入侵等。利用马尔科夫模型对红树林湿地景观演变进行了预测，未来若无相应措施控制互花米草扩散以及遏制围塘养殖面积急剧增加，则天然湿地比重将显著下降，红树林生态系统将会进一步退化。（3）利用泰森多边形划分对象木与最邻近的竞争木构成的林分空间结构单元，利用GIS二次开发语言编程实现最邻近木编号的查询，并应用VBA程序实现林分空间结构指数的计算，可以大量减少外业调查工作。（4）树种隔离程度整体上较小，处于弱度混交类型，不同潮位带，混交度从大到小依次为高潮位、中潮位、低潮位，且处于低潮位带的林分接近于纯林；红树林种内竞争比种间竞争激烈，从大到小依次为桐花树>秋茄>白骨壤>木榄；调查林分平均大小比数差异小，各树种胸径大小分化不明显，生长优势状况较为均衡；不同林分的平均健康度差异不显著，林木之间的生长健康状况差别不大；不同潮位带红树林林木角尺度分布频率差别很小，各树种林木整体上属于均匀稍有向随机分布发展的趋势；不同树种聚集指数分布频率差异显著，桐花树和秋茄周边竞争木聚集分布占优势，白骨壤林木周边竞争木聚集、随机和均匀分布的机会均等，木榄林木周边竞争木随机分布的概率高。（5）运用压力-状态-响应（PSR）模型和层次分析法（AHP）对漳江口红树林生态系统健康进行评价，结果表明健康综合指数为0.66，处于健康状态。（6）根据生境条件对漳江口红树林的宜林地进行了区划，从互花米草除治和堤岸防护林布局方面提出红树林景观结构优化方案。优化后红树林面积比重增加了6.92%，互花米草面积比重减少了4.50%。红树林整体质心向东南方向偏移了1323m。（7）应用灰色关联度法对林分空间结构综合指数与其指数因子的关联度大小进行分析，结果表明：聚集指数（0.80）、竞争指数（0.80）、角尺度（0.74）、健康度（0.73）、混交度（0.71）、大小比数（0.67）。以林分空间结构综合指数为林分空间优化模型的目标函数，设置林木竞争、树种多样性和林木空间格局三个方面的6个约束条件，建立了基于GIS的红树林林分抚育采伐空间结构优化模型和林分补植空间结构优化模型。应用模型对研究区林分进行了模拟优化经营，结果表明，经营后的林分空间结构均得到了改善，向理想结构状态转化。通过上述研究，本文的主要创新点体现在：（1）通过构建堤岸线掩膜、Band1/Band2、(Band4-Band3)/(Band4+Band3)、Band3/Band2波段运算等辅助信息的决策树分类实现了基于高空间分辨率卫星数据的红树林湿地景观类型的有效区分，其解译结果比传统精度比监督分类法有很大提高；（2）将森林空间结构分析的方法应用于红树林这特殊生态系统，并采用泰森多边形法划分林分空间结构单元，利用GIS二次开发语言编程实现对象木的最邻近木编号的查询，应用VBA编程实现混交度、大小比数、竞争指数、健康度、角尺度和聚集指数等林木空间结构指数的计算；（3）建立基于GIS技术的红树林林分抚育采伐空间结构优化模型和林分补植空间结构优化模型。更多还原

【Abstract】 Mangrove ecosystem is formed through long terms of succession between marine andterrestrial ecosystems. It is a special forest ecosystem which has stable system structure andhigh productivity. Mangroves play an important role in protecting coastlines, creating land withsilt, purifying sea water, maintaining biodiversity. The object of mangrove healthy managementis to maintain complete, stable continuous system function and high productivity. Therefore, itis urgent work for mangrove protection and recovery to evaluate ecosystem health, adjustlandscape pattern and optimizate forest spatial structure.This paper taked Zhangjiangkou Estuary Mangrove National Natural Reserve, lacated inYunxiao County, Fujian Province, as the subject, and analyzed dynamic the changes oflandscape pattern and driving force in mangrove wetland based on three temporal differenthigh-resolution sensing images in2003,2008and2013. Every ojbcetive wood was located andthe stand description factors were surveyed in typical mangrove communities selected fromdifferent tidal zones. The characteristic of forest stand spatial structure was researched basedon GIS technology and mangrove ecosystem health was evaluated using PSR chain andanalytic hierarchy process. According to analysis results, optimation layout of mangrovelandscape structure was put forward on landscape level. The mangrove forest spaceoptimization models of selecting cutting and replanting were built based on forest stand spatialstructure comprehension index on the forest stand scale and applied the models to simulatemanagement in research area. Sustainable management measures were put forward to eliminateor weake influence of mangrove ecosystem health from outside pressure. The results showedthat:（1）Mangrove landscape informations were extracted accurately based on high-resolutionremote sensing images of SPOT5and PLEIADES. The auxiliary informations built embankment line, calculated by Band1/Band2,(Band4-Band3)/(Band4+Band3), Band3/Band2,were added to decision tree rules. The results showed that overall accuracy of three imageswere separately97.24%,92.90%,97.86%and their Kappa coefficients were0.97,0.91,0.97.Compared with supervise classification without auxiliary informations, the overall accuraciesraised separately16.91%,7.90%and12.21%.（2）Mangrove areas increased from47.69hm2to64.24hm2during the last ten years.Centre of mass of mangrove moved slightly to northeast. Spartina areas increased42.72hm2and encroached on mangrove of13.38hm2with the trend of rapidly increasing after falling.Artificial shrimp pond areas almost twice increased from148.69hm2to254.84hm2. Themangrove ecosystem was degradated gradually during a large amount of natural wetlands weretransformed to artificial wetlands and2.11hm2of mangrove were eliminated. The analysis ofmangrove wetland landscape pattern index showed that landscape diversity and fragmentationdegree increased continuously, all types of pacthes developed to equilibrium, and continuityand integrity of landscape appeared worse. The main factors of dynamic change of mangrovewetland landscape were impoundment for farming, bank maintenance and spartina invasion.The change of mangrove wetland landscape was predicted by Markov model in future, theproportion of natural wetland would be dropped significantly and mangrove ecosystem wouldbe further degradated if there are no measures to control the sharp increasement of spartina andartificial shrimp ponds.（3）The forest stand spatial structure unit made up objective wood and adjacentcompetited wood was devided using Thiessen polygon method. It will reduce a great deal ofwork of field survey to query the serial numbers of adjacent woods using secondarydevelopment program of GIS and calculate forest stand spatial structure indexes withapplication of VBA program. The difference of mingling degree among different tidal forestwas obvious and they were high tidal zone, middle tidal zone and low tidal zone in the order ofmingling degree, moreover, the low tidal forest was nearly pure stand. （4）Tree species composition was simple and diversity was low. The level of speciesisolation was low, which was weak mingling. In mangrove forests, the intraspecificcompetition is fiercer than the interspecific competition, and they were Aegiceras corniculatum,Kandelia candel, Avicennia marina and Bruguiera gymnorrhiza in the order of competitionindex. There was a small difference among the average neighbor-hood comparisons ofinvestigation forest stands. The difference of diameter at breast height was not obvious inspecies and the growth conditions of different trees were balanced. There were a littledifference among mangrove trees and the average healthy indexes were no significantdifference. The distribution frequency of uniform angle index was a little difference amongdifferent tidal mangrove trees, the overall distribution patterns were even distribution mainlyand inclined to be randomly slightly. The distribution frequency of aggregation index wassignificant difference among different species. The aggregation distribution was predominantin Aegiceras corniculatum and Kandelia candel, there were same distribution frequency ofaggregation, random and even in Avicennia marina, and the random distribution frequency ofBruguiera gymnorrhiza was higher than the others’.（5）There were significant difference among different species’height structure, the heightand under-branch height of Kandelia candel and Bruguiera gymnorrhiza were higher thanAvicennia marina’s and Aegiceras corniculatum’s. The diameter at breast heights were small,concentrated in3to7cm. There was a little difference among diameters of crown in differentspecies, Avicennia marina’s was largest and Aegiceras corniculatum’s was smallest.（6）The ecosystem health of mangrove was evaluated by PSR model and AHP method.The results showed comprehension healthy index was0.66, in healthy level.（7）Suitable forest land was divided to the habitat condition of mangrove. Mangrovelandscape structure was optimized from directions of Spartina treatment and layout ofprotective forests. After optimation, mangrove area ratio increased by6.92%, spartina arearatio reduced4.50%. Mangrove’s centre of mass moved1323m to southeast. （8）Applied gray correlation analysis method to research the relationship between forestspatial structure comprehension index and it’s factors, and it showed that aggregationindex(0.80), compitition index(0.80), uniform angle index (0.74), health index(0.73), minglingdegree(0.71), neighbor-hood comparison(0.67). The mangrove forest space optimizationmodels of selecting cutting and replanting were built based on GIS technology, according toforest stand spatial structure comprehension index as objective function, setting six indicesstand structure as constraints. Applying the two optimization models to guide forestmanagement, the spatial structure could be improved significantly which made spatial structureof mangrove stands transforming to ideal spatial structure.Overall, this study led to the following innovations:(1)The decision tree method withauxiliary informations built embankment line, calculated by Band1/Band2,(Band4-Band3)/(Band4+Band3), Band3/Band2, could greatly increase accuracy of mangrovewetland classification compared with traditional supervise classification method;(2)Forestspatial structure analysis method was applied to mangrove ecosystem. The forest spatialstructure unit was divided by Thiessen polygon method and serial numbers of adjacent woodwere found using GIS program, and the forest spatial structure indexes of mingling degree,neighbor-hood comparison, compitition index, health index, uniform angle index andaggregation index were calculated by VBA program.(3)Built mangrove forest spaceoptimization models of selecting cutting and replanting were built based on GIS technology.更多还原