基于ANN模型的元江干热河谷生态脆弱区景观格局变化研究

发布时间：2018-03-09 06:19

本文选题：干热河谷　切入点：土地利用　出处：《昆明理工大学》2015年硕士论文　论文类型：学位论文

【摘要】：景观格局一直是景观生态学研究的重点,土地利用变化直接影响着地表的景观格局。元江干热河谷生态脆弱区域内土地利用不当使之存在潜在荒漠化的问题。本研究以景观生态学和人工神经网络为理论基础,选取元江干热河谷生态脆弱区部分区域作为研究区,采用遥感技术和地理信息系统技术,对研究区2006年和2012年两期卫星遥感影像进行解译,从而得到这两个时期的土地利用数据。在此基础上,采用土地利用类型转移矩阵分析模型对研究区土地利用的时空变化进行分析,进而解释研究区土地利用结构时空变化规律,再利用人工神经网络模型对研究区景观格局进行动态模拟和预测,定量分析演变特点。研究发现：研究区的土地利用类型主要以耕地、林地和稀树灌木草丛为主：2006年到2012年6年间,耕地、稀树灌木草丛、滩涂、居民用地、园地和工矿仓储用地面积增加,林地和水域面积减少；林地和耕地的面积变化最大,分别为-106.70 km2和81.57 km2。动态度最大的为滩涂,为9.90%/a,此外为工矿仓储用地和水域。耕地的贡献主要来自于林地和稀树灌木草丛；滩涂的贡献主要来自于水域和稀树灌木草丛；工矿仓储用地的贡献主要来自于稀树灌木草丛。林地主要转出为耕地和稀树灌木草丛：水域主要转出为园地和稀树灌木草丛。研究区受自然因素和人为因素的干扰,土地利用结构发生了明显变化,水域明显减少、滩涂明显增加,大量的林地被砍伐,并且开发了很多耕地。研究区林地和耕地具有较高的聚集度,斑块形状较复杂,滩涂和居民地斑块形状较简单；工矿仓储用地和水域斑块间平均最近距离较远；与稀疏灌木草丛、水域和园地这三类土地利用类型相邻的斑块类型较丰富,与工矿仓储用地和林地相邻的斑块类型比较单一,工矿仓储用地主要被稀树灌木草丛包围,林地主要被耕地包围;稀树灌木草丛主要分布在河谷两侧,聚集度高。从斑块水平上看,2006到2012年以来,大部分的斑块类型趋于破碎化,边缘复杂化。林地一直为斑块数量最大的景观地类；受人类活动影响,林地和耕地景观破碎化加重明显,且其受干扰的范围分布广且分散；稀树灌木草丛斑块间的连通性增强,斑块平均大小值较大。从景观水平上看,从2006至2012年,研究区总体景观格局呈现了一种多样化、破碎化的趋势。斑块数、斑块平均大小、面积加权平均形状因子、蔓延度和散布与并列指数等景观指数均表现出景观破碎化程度的提高。香农多样性指数与香农均度指数的持续增加表明其均匀度程度增加,从而揭示了研究区八类景观类型间存在均匀化分布的趋向。所建神经网络模型对测试集的预测值与实际值具有较好的拟合性,说明利用人工神经网络来研究干热河谷景观间接驱动因子对景观格局的影响是可行的。模型预测显示,随着林地、滩涂、居民用地、工矿仓储用地、耕地所占面积比增加,景观斑块密度增加；随着水域、稀树灌木草丛和园地所占面积比增加,景观斑块密度下降。景观多样性指数会随着居民用地和园地所占面积比的增加而增加,随着耕地所占面积比的增加而降低。景观聚集度会随着水域、稀树灌木草丛和园地所占面积的增加而增加；随着滩涂、居民用地和耕地所占面积比的增加而下降。随着林地、耕地、工矿仓储用地和滩涂所占面积比的增加,景观斑块形状更加复杂化,随着水域、稀树灌木草丛和园地所占面积比的增加,景观斑块形状趋于规范化。
[Abstract]:The landscape pattern has been the focus of research in landscape ecology, land use change has a direct impact on the landscape pattern of the surface. The Yuanjiang valley ecological fragile region land use improper to potential desertification problems. Based on landscape ecology and artificial neural network theory, selection of Yuanjiang river ecological fragile zone region as the research part area, using remote sensing and GIS technology, the study area in 2006 and 2012 satellite remote sensing image interpretation, so as to obtain the land use data of the two periods. On this basis, the analysis of land use type transfer matrix analysis model of spatial and temporal changes of land use in the study area, and then explain the variation of structure time and space of land use in the study area, using artificial neural network model for dynamic simulation and the landscape pattern of the study area Measurement, quantitative analysis of evolution characteristics. The study found: land use types in the study area is mainly farmland, woodland and savanna: 2006 to 2012 6 years, cultivated land, savanna, beaches, land, garden land and mining warehouse land area increased, woodland and water area decreased; and the changes of forest area arable land was the largest, respectively -106.70 km2 and 81.57 km2. dynamic degree is the largest beach, 9.90%/a, in addition to industrial storage land and waters. The main contribution of cultivated land from woodland and savanna; tidal contributions mainly from waters and savanna; mining warehouse land contribution mainly from savanna woodland mainly turn into cultivated land and savanna: waters mainly turn into garden and savanna. The study area affected by natural factors and human factors interference, land Significant changes occurred in the water use structure, significantly reduced, beach increased significantly, a lot of forest have been cut down, and developed a lot of cultivated land. The aggregation degree of woodland and cultivated land in the area is high, the patch shape is more complex, and the beach residents patch shape is simple; the industrial storage land and water patches between the average nearest distance far away; and sparse shrubs, waters and garden of these three types of land use patch type adjacent to the rich, and industrial use patches and woodland adjacent single, mining warehouse land is mainly surrounded by dilute tree shrubs, woodland is mainly surrounded by farmland; savanna is mainly distributed in the valley on both sides and a high degree of aggregation. From the patch level, from 2006 to 2012, most of the types of plaque fragmentation, edge has been complicated. Forest landscape types of the largest number of plaque; Affected by human activities, landscape fragmentation of woodland and cultivated land increased significantly, and the interference range widely distributed and decentralized; savanna connectivity between patches increased, mean patch size is larger. On the landscape level, from 2006 to 2012, the overall landscape pattern of the study area showing a diversification, fragmentation trend. Number of patches, mean patch size, area weighted average shape factor, spread and interspersion and juxtaposition index landscape index showed the degree of landscape fragmentation increased. Increasing the Shannon diversity index and Shannon's evenness index showed that the evenness degree increases, which reveals the tendency to homogenization distribution of the study area eight kinds of landscape types. The neural network model to predict the test set value is in good agreement with the actual value, that view to study the dry hot valley by using artificial neural network The concept of indirect driving factors of landscape pattern is feasible. The model prediction indicates that with the land, beaches, land, mining warehouse land, cultivated land area increased, landscape patch density increased; with water, savanna and garden area increased, patch density, landscape diversity decreased. The index will increase with the residential area and the garden area ratio increases, decreases with the cultivated land area ratio increasing. The degree of landscape aggregation with waters, savanna and garden area increases with the increase of residents; with beaches, land and cultivated land area increased decreased. With forest land, cultivated land, mining warehouse land and beach area ratio increased, patch shape is more complex, with waters, savanna and garden area increased, landscape patch shape The form tends to be standardized.

【学位授予单位】：昆明理工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：P901

【参考文献】