【作者】 朱耀军；

【导师】 彭镇华；

【作者基本信息】 中国林业科学研究院 ， 森林培育， 2010， 博士

【摘要】 本论文采用广州市1985年、1995年和2007年1:10万土地利用图和2004年航片以及2005年1:1万基础地理信息等数据,通过在市域和建成区两个尺度上对广州市城市森林景观格局进行了相关的分析。市域尺度上主要研究了林地覆盖类型和景观边界的发生范围和变化趋势,并对森林覆盖的环境效应进行分析。建成区尺度上,根据高分辨率航片的目视解译,构建了基于树冠覆盖的城市森林基础数据,并对其分类景观格局的尺度效应、格局特征、特征梯度和景观连接度以及街区单元的城市森林可达性进行了分析,为更小尺度上城市森林景观格局优化提供参考。结果表明:(1)广州市林地覆盖总体上升,稳定林地覆盖占35.81%,是土地覆盖的主导类型。距离主干道路越远或道路密度越高,稳定林地覆盖比例升高,稳定非林地覆盖减少;变化类型中以1995年以后的退林和新造林类型为主,新造林类型多位于稳定林地与非林地的周边区域,而退林类型则多在稳定林地和非林地的内部区域,主要由于城市化对于林地的开发。(2)林地边界类型中,林-建类型分布以靠近主干道路区域为主,而林-耕型则远离主干道路分布,而林-水型在1995年以后向靠近主干道路的区域增加。1985-2007年,林地边界持续增加且以后期增加为主,其中林-耕类型减少,而林-水、林-建类型增加,且林-建类型增加显著,到2007年时达林地总边界的26.8%。林-建类型中,1995年以前以林-城类型的增长为主,之后以林-村和林-它类型增加为主且在远离主干道路的区域明显,其中林-它型在1995年以前以中心城区变化为主,之后以花都和从化南部区域相对显著。(3)研究区林地林龄与NDVI值呈正相关,以稳定林地对应的NDVI最高。全市尺度上NDVI与基于热红外波段计算的亮温值的相关性较差,但在受人类活动强烈影响的中心城区,考虑水体因素的影响,NDVI与亮温之间表现为显著的负相关关系。广州市热岛效应在中心城区有几个相对集中的区域。城市居民对于城市热环境的感知发生于更小尺度上,反映较大尺度上热量平衡的结果的区域下垫面性质的异质性形成的不同景观构成会通过局地小气候环流对小尺度的热环境及市民身体的舒适度产生不同程度的影响。(4)中心城区大于1hm2的斑块面积占78.83%,以生态片林为优势覆盖类型,主要分布于白云山景区及天河区、黄埔区的北部,多是原来大型林地斑块的破碎化成分;其次为小于0.1hm2以居所林占优的小斑块覆盖。树冠覆盖格局分析表明,白云区的斑块密度最小的,而最大斑块指数、平均斑块面积和连接度指数在研究区范围最大;芳村区没有形成较大面积的树冠连续覆盖,破碎化程度在各区中最明显。城市化程度高且处在老城区的东山区、荔湾区和越秀区的斑块密度最大,人为活动影响下,树冠斑块的破碎化严重。中心城区核心区(外环高速内区域)树冠总覆盖率为13.1%,以居所林和生产林覆盖为主,受城市化发展阶段等因素的影响,核心区各亚区之间树冠覆盖差异大,以南区的树冠覆盖占17.4%为最高。(5)从分类景观格局指数的尺度效应表明,多数景观指数与粒度变化之间都存在很强的正/负相关性,但因景观组成结构形成差异,多数响应曲线符合CUB、QUA或INV函数,随粒度增加特征值序列发生变化的可能性变大。核心区树冠覆盖景观格局分析表明,中区表现更为破碎,许多的生态过程可能发生于更小尺度,但中区的大型生态林斑块覆盖使其最大斑块指数最高。南区受“万亩果园”影响其斑块平均面积最大,面积标准差也最大。(6)基于移动窗口分析法对树冠覆盖类型景观选取了几个常用的景观指数,进行了尺度效应分析,在更小尺度上探讨了景观格局特征的梯度变化特征并进行了定量化显示表达,为更精细尺度上景观格局的空间量化和可视化提供了可能性。(7)核心区大于0.5hm2的树冠连续覆盖斑块景观连接度分析表明,整体景观连接度较低,以重要性高的斑块为中心形成的各连接组分,随着距离阈值增加迅速减少,在2km时研究区大型斑块之间能够互相连接。最大的组分出现在研究区中北部、东北部和东南部地区,也是最重要斑块相对集中的区域,大型组分之间随着距离阈值增加连接成大的组分,而研究区西南部区域的组分相对独立,与其它区域缺少连接。(8)能够在0.5km内能够步行到达大型城市森林的街区面积仅占核心区街区总面积的17.9%。越秀区在步行0.5km和1km范围内能够到达大型城市森林斑块的街区分别占52.3%和93.1%,为各区最好,其次为东山区、荔湾区(1km可达性街区分别有76.1%和69.8%)且明显好于其它四区(天河区、白云区、海珠区、芳村区)。核心区0.5km内不能到达大型城市森林斑块且树冠覆盖超过30%的街区中,多为单位专有绿地树冠覆盖类型;而以居住区占主要建筑类型的街区中,90%以上的街区树冠覆盖低于10%,其中64%的街区树冠覆盖低于5%。土地利用/覆盖变化类型和林地边界动态能反映林地覆盖/变化发生的区域分布和人类活动的干扰强度下的变化细节,受研究区主要道路影响域的而不同,对其深入分析对于较大尺度上的景观格局变化驱动力研究有重要意义。在快速城市化地区,发展以乔木为主体的城市森林,形成连续的树冠覆盖表面来形成小气候可提升城市舒适度,也是缓解城市热岛效应最有效的手段。在老城区改造或大型生态空间建设中,应结合生态学过程增加景观连接度的量度内容,提升“重要性节点斑块”的质量,通过在中心城区建成区增加中、小树冠斑块的密度,能够加强各类型城市森林斑块间的景观连接度,能发挥城市森林整体效应。重视高分辨率影像在生态学研究中的作用,结合社会经济的统计资料,将小尺度上的生态学过程与格局特征关联分析并进行空间化和可视化表达,有助于在细尺度上了解城市森林格局动态变化的驱动力,使城市森林景观的格局、过程和功能分析统一于不同尺度上的城市生态建设,对解决城市环境问题的更有实效性。可达性是反映城市森林服务功能的重要指标,注重发挥GIS空间分析在生态规划中的强效性,将自然要素和相关的社会经济因素叠加分析,更加科学有效地分析城市问题。更多还原

【Abstract】 As the main element of ecological landscape, urban forest has become the principal part of ecological construction in urban region. The high resolution aerial photo in 2004 and the TM images in 1985, 1995 and 2007, combined with fundamental geographic information(1:10 000)in 2005, were used to explore the landscape pattern of urban forest in Guangzhou on two research scales, including urban and built-up area. On urban scale, the changing extent and tendency of forestland cover type and landscape boundary were analyzed, and the environmental effects of urban forest as well. On built-up area scale, the fundamental data of urban forest was built based on tree-canopy cover derived from the visual interpretation on aerial photo. The scale effects and gradient of landscape pattern characteristics were discussed detailedly, as well as the landscape connectivity and accessibility of neighborhood block units considering the concerned large patches. The results acquired from high resolution imagery provided valuable information for the landscape pattern optimization of urban forest on small scale in Guangzhou. The results showed that:The forestland area of Guangzhou was increased from 1985 to 2007, and stable forestland was the dominant land cover type with the percentage of 35.81%. The proportion of stable forestland increased with the distance away from main roads and the density of roads net, and the stable no-forestland was contrary. The deforestation and afforestation were dominant in the change type in 1995, and the afforestation occurred mostly at the periphery of the stable forestland and non-forestland, while the deforestation dispersed within the boundary of stable types and mainly affected by the urbanized exploiting.The process and intensity of land use can be reflected clearly from the shift and direction of forestland boundary. The analyses were taken based on boundary characters classification and different landscape accessibility zones from the main roads in Guangzhou. Based on the results that both the forestland patch number and the whole lengths of forestland boundary were increased, the trend could be formed that boundary became more complex and landscape became more fragmental. The situation, which the adjacent landscape types of forestland were dominated by cropland, was broken with the increasing of building area and water, the lengths of forest-cropland boundary reduced rapidly and both forest-water and forest-building increased markedly. The decreasing speed of forest-cropland boundary became slow with the increasing distance from the key roads. The most acute zones where the forest-building increased were far away from the key roads gradually, and the forest-water boundaries were mostly distributed in the area which was 1km away from the roads. According to the interaction of landscape types both sides of the boundary, forest-building boundary was a kind of tensed type, and the ecological function would be affected by the increasing. A discussed method based on the analysis of the density of forestland boundary type was tried to maintain the stability of forestland resource in the rapid urbanization areas.Correlative analysis showed that there was a positive correlation between NDVI and the age of forestland, and stable forestland had the highest value of NDVI, while the corresponding value of afforestation recently was the lowest relatively. Although the effect of correlation between NDVI and brightness temperature from the thermal infrared band of TM imagery was unobvious relatively on urban scale, on built-up area scale which affected intensively by human activity, it showed a significant negative correlation. The heat island effect showed clearly in the central city of Guangzhou, and several heat island areas were concentratingrelatively. Tree canopy coverage landscape was analyzed based on the thought framework of function classification and scale system in the central city of Guangzhou. The results showed that the ratio of tree canopy coverage was 19.5%, and the ratio of Tianhe district was the highest (26.5%), while the ratio of Fangcun district was the lowest (6.6%). Ecological beneficial forests were dominant among the type groups of ecological benefits forests, public garden, residential forest stands, production forest stands and special protective forests. The coverage ratio of ecological beneficial forests was 55.4% of the total tree canopy, and secondly the ratio of production forest stands was 19.9%. Mean size of the ecological chunk forest was the largest, while the mean size of roadside trees was the smallest. The coverage ratio of extra large-sized patches larger than 1hm2 was 78.8%, the ratio of small-sized patches (<0.1hm2), medium-sized patches (0.1-0.5hm2) and large-sized patches (0.5-1hm2) was respectively 9.2%, 7.9% and 4.2%. The patch density of small-sized, medium-sized and large-sized patch in Dongshan district is the highest, but the density of extra large-sized patch in Huangpu was the highest, and the density of medium-sized and small-sized patch in Baiyun was the lowest, while the density of large-sized and extra large-sized patch in Fangcun district was the lowest. Extra large-sized patches were dominated by ecological beneficial forests and production forest stands, and residential forest stands were regnant in other sized patches. There were extra large-sized patches distributed mostly in the mountain area, which were located in the north of the central city and the east developing belt in Tianhe and Huangpu districts, and orchard fields in the south of Haizhu district as well. The number of small-sized patches which were planted artificially was the most and the patches of extra large-sized tree canopy coverage existed only in the parks of the old city area. Most of the large-sized patches between 0.5 and 1 hm2 were adjacent to the extra large-sized patches, the pattern might be caused by the fragment of extra large-sized patches in the process of urbanization. Continuous tree canopy is an effective approach for implementing the continuity of ecological progress and pattern. Roadside trees type, being the most at patch numbers and fragmental relatively, has the most potential for forming continuous coverage patch via spatial linkage. The results provided valuable information for related research of urban forest and the landscape pattern optimization at small scale in Guangzhou.The scale effects of classified landscape metrics showed that there were strong positive or negative correlations between grain size and landscape metrics, varying with the landscape composition. Most response curves of landscape metrics with the grain size could be modeled by the CUB, QUA, or INV function explicitly. The probability of changes of sequences of landscape metrics might be higher with the increasing grain size. The results derived from the landscape pattern analysis at core area of the central city in Guangzhou showed that the sub zone of C was more fragmental and some ecological process might be taken place in smaller scale, but the value of LPI (largest patch index) was the biggest among the five sub zones because of the existing large ecological forest. The sub zone of S affected by the large orchards area had the biggest MPS (mean patch size), as well as the standard deviation simutaneously.Several common landscape metrics were employed based on moving windows analysis in Fragstats3.3 software, and the results of scale effects of landscape metrics on grain size and extent were showed in explicit expression. Comparing with quantifying landscape pattern in the entire landscape previously, the explicit expression of partial quantification of landscape pattern could link the pattern and the ecological processes more effectively in smaller scale.The results of landscape connectivity, based on the large patches of the area larger than 0.5 hm2 , showed that the connectivity of core area of the central city in Guangzhou was not so well and the components which were consisted of important patches and other patches were directly related to the thresholds distance of the landscape connectivity. The importance of patches was inspected in different distance thresholds, and the importance values of larger patches for connectivity of tree-canopy cover landscape were more than other sized patches. The large components located in the areas where the important patches existed in the research region. All the patches could be connected each other in the 2km distance thresholds. Respectively, the components sited in the southwest of core area were isolated and could not connect well with others. The blocks were organic composition of urbanized area. Network analyst model in ArcGIS9.2 was used to analyze the accessibility between the blocks and urban forest patches within Outer Ring Highway in Guangzhou, and the service capacity of urban forest patches was examined as well. And then the difference among the blocks was analyzed, combined with tree canopy coverage from the interpretation of aerial photo. The results showed that 73% of the blocks could reach large green patch in 2km walk distance, and about 46% in 1km walk distance but only 18% of blocks could reach large green patch in 0.5km walk distance. Urban forest patches accessibility was the highest in Yuexiu District,while the provision and accessibility were the most inadequate and worst in Fangcun District and Haizhu District. Among blocks that could not reach large green patch in 0.5km walk distance but tree canopy coverage beyond 30%, the percent of institutions block was predominant, while the coverage of 90% of resident blocks was under 10%, and even 64% of those under 5%. The results acquired from Network Analyst model based on block units provided valuable information for the landscape pattern optimization of urban forest on small scale in Guangzhou.更多还原