【作者】 曹杨；

【导师】 尚士友；

【作者基本信息】 内蒙古农业大学 ， 农业生物环境与能源工程， 2010， 博士

【摘要】 湿地水生植物群落动态变化过程可以反映湿地生态系统健康程度及其生态服务功能。建立挺水植物群落时空扩散模型,定量分析挺水植物群落扩散与湿地环境变化之间的响应关系,探索湿地复杂非线性演化规律、剖析演化机制,可以为湿地可持续发展提供科学依据与决策支持。本文以乌梁素海湿地为研究区,在遥感、地理信息系统的支持下,通过景观指数定量研究了1986～2008年湿地景观类型格局及动态变化特征,测算了挺水植物群落重心位置及其移动轨迹,分析了挺水植物扩散的影响因子。以地理元胞自动机(Geo-CA)为模型框架,分别基于空间统计学和人工智能方法构建了乌梁素海湿地挺水植物扩散时空动态模型,模拟了湿地挺水植物群落扩散过程。主要研究结论如下:(1)基于多源遥感数据提取乌梁素海湿地类型信息,定量分析了湿地类型区景观指数变化,结果表明:1986～2008年间,乌梁素海湿地挺水植物面积由120.89km2上升至167.43 km2,斑块个数由1138个降至690个,斑块密度由1.2335个/km2降至0.7479个/km2,最大斑块面积比例由6.7002%升高至17.3879%,形状指数由36.2997下降至33.8494。挺水植物侵吞明水区和浅水沼泽区,成为湿地优势景观类型,斑块破碎度降低,边缘结构趋于简单。湿地挺水植物群落发展具有不平衡性,22年间挺水植物群落的重心偏移基本可分三个阶段:1986～1997年重心位置由(E 108°50′55.35″,N 40°58′53.78″)移动至(E 108°50′25.26″,N 40°58′37.57″),向西南偏移了863.03m;1997～2000年重心由(E 108°50′25.26″,N 40°58′37.57″)移至(E 108°50′35.33″,N 40°58′28.32″),向东南偏移了369.76m;2000～2008年重心位置由(E 108°50′50.12″,N 40°58′39.11″)移动至(E 108°50′35.33″,N 40°58′28.32″),向东北偏移487.73m。(2)在实测数据的基础上进行挺水植物群落扩散影响因子相关分析,通过逐步回归分析进一步确定了水深、TN是乌梁素海挺水植物扩散的主要影响因子,回归系数分别为-0.46和0.17。其中,水深超过1.5m挺水植物扩散停止,检测挺水植物体内N / P≤14,TN仍对乌梁素海挺水植物扩散起促进作用。通过地统计学插值方法模拟了乌梁素海不同时期水深、TN的时空分布。(3)基于空间统计学原理,计算不同时期乌梁素海挺水植物扩散动态度(LC)变化,获取挺水植物扩散速度拐点,进而将乌梁素海挺水植物群落扩散分为:快速扩散期(1986～2002年)和滞缓期(2002～2008年),挺水植物扩散动态度分别为2.1%和0.61%。建立动态度最优分段的Logistic-CA模型模拟湿地挺水植物时空扩散过程,结果表明:1997、2002和2008年模拟总体精度分别为74.01%、72. 29%和74.36%,Kappa系数分别为0.6222、0.6103和0.6370,实际Moran I系数分别为0.628、0.686和0.709,模拟结果的Moran I系数分别为0.641、0.712和0.720。动态度校正后的逻辑回归CA模型在精度、空间形态一致性定量检验及控制模拟精度下降方面均优于常规Logistic-CA模型。(4)将案例推理(CBR)引入湿地挺水植物扩散Geo-CA模型,解决湿地挺水植物扩散模拟面积大、时间长、影响因素复杂、非线性转化规则难以获取等问题。构建基于案例推理的乌梁素海挺水植物扩散元胞自动机模型(CBR-CA),依据地理特征和空间特征的相似度,搜索动态历史案例库中k个最近邻,隐含获取Geo-CA转化规则,模拟挺水植物群落扩散过程,结果表明:1997、2002和2008年模拟总体精度分别为76.41%、73.05%和75.91%,Kappa系数分别为0.6402、0.6382和0.6393,实际Moran I系数分别为0.628、0.686和0.709,模拟结果的Moran I系数分别为0.640、0.710和0.718。CBR-CA模型能适应湿地环境快速变化,并体现出复杂系统自适应特点。(5)分别基于LC-Logistic-CA模型和CBR-CA模型预测了2015年乌梁素海湿地挺水植物扩散状况,预测结果显示:挺水植物面积分别为174.34km2和172.92 km2,斑块个数分别为634个和646个,斑块密度分别为0.7287个/km2和0.7311个/km2,最大斑块所占景观面积的比例分别为18.0019%和17.9558%,形状指数分别为32.8392和33.1556,斑块破碎度降低,边缘结构趋于简单。更多还原

【Abstract】 The dynamic variation of aquatic plant communities in wetlands may reflect the health degree of wetland ecosystems and ecological services. The establishment of Spatio-TemPora distribution model of emergent plants communities, the quantitative analysis of response relationship between wetland environmental change and the distribution of emergent plant communities,the exploration of complex non-linear law of wetlands evolution and the analysis of evolutionary mechanism may provide the sustainable development of wetlands with a scientific basis and decision support.With the support of remote sensing and geographic information system, the pattern and the dynamic variation characteristics of all kinds of wetland landscape in Wuliangsuhai wetland from 1986 to 2008 were studied quantitatively by landscape index. Furthermore, the location of center of gravity of the emergent plant communities and their moving tracks were calculated, and impact factors of the spread of emergent plants were analyzed. Taking cellular automata as a model framework, the Spatio-TemPora distribution dynamic model of emergent plants was established and the diffusion process of wetland emergent plant communities was simulated in Wuliangsuhai wetland on the basis of spatial statistics and the method of artificial intelligence respectively. The main research results are as follows:(1)Based on multi-source remote sensing data, the information of the type of Wuliangsuhai wetlands was extracted, and the index variation of wetlands landscape was analyzed quantitatively. The results showed that from the year 1986 to 2008, the area of emergent plants in the wetland grew from 120.89km2 to 167.43 km2; the number of patches declined from 1138 to 690; patches density decreased from 1.2335/km2 down to 0.7479个/km2; the ratio of the gest patches area increased from 6.7002% to 17.3879% ; the shape index fell from 36.2997 to 33.8494. Emergent plants swallowed clear water areas and shallow marshes areas, which became a dominant type of wetland landscape. The degree of fragmentation of patches was decreased and the edge structure became simple. Wetland emergent plant communities developed with imbalance, in 22 years, the barycenter of emergent plant communities has shifted through three basic stages. From 1986 to1997, the location of barycenter moved from the (E108°50’55 .35 ", N40°58’53 .78") to (E 108°50’25 .26 ", N 40°58’37 .57"), shifting 863.03m toward southwest; From 1997 to 2000, the location of barycenter moved from the (E 108°50′25.26″,N 40°58′37.57″) to (E 108°50′35.33″,N 40°58′28.32″), shifting 369.76m toward southeast; From2000 to 2008, the location of barycenter moved from the (E 108°50′50.12″,N 40°58′39.11″)to (E 108°50′35.33″,N 40°58′28.32″), shifting 487.73m toward northeast.(2)Correlation analysis of impact factors of emergent plant communities diffusion were carried out on the basis of the measured data. By stepwise regression analysis, the water depth and TN were further identified to be the main impact factors of diffusion of emergent plants in Wuliangsuhai, and their regression coefficients were -0.46 and 0.17 respectively. Where the water depth was more than 1.5m, diffusion of emergent plants stopped and N/P was not more than 14 in emergent plants. TN still promoted the diffusion of emergent plants in Wuliangsuhai. Temporal and spatial distribution of the water depth and TN at different stages were simulated with the method of geostatistical interpolation .(3)Spatial Statistics theory was used to calculate the dynamic degree of emergent plants diffusion(LC)in Wuliangsuhai at different times, so that the turning point of diffusion speed of emergent plants can be obtained . And then the diffusion process of emergent plants communities in Wuliangsuhai was divided into the diffusion period (1986-2002) and the stagnant period (2002-2008), in which the dynamic degrees of emergent plants diffusion were 2.1% and 0.61% respectively. Logistic-CA model of sub-optimal dynamic degree was established to simulate the process of spatio-tempora distribution of emergent plants. The results showed that the overall accuracies of simulation were 74.01%, 72.29% and 74.36%, Kappa coefficients were 0.6222, 0.6103 and 0.6370, and actual Moran I coefficient were 0.641, 0.712 and 0.720 in 1997,2002 and 2008 respectively. Accuracy and quantitative test on consistency of spatial form and control of the decline of the simulation accuracy of Logistic regression CA model whose dynamic degree was proofread were all better than those of conventional one.(4)That case-based reasoning (CBR) was led in Geo-CA model of wetland emergent plants diffusion solved many problems, including that simulation area of wetland emergent plants diffusion is large, simulation time is long, impact factors are complex, non-linear transformation rules are difficult to obtain and so on. Cellular automaton model (CBR-CA) of large emergent plants diffusion in Wuliangsuhai was constructed on the basis of case-based reasoning. According to the similarity of geographical features and spatial features, k nearest neighbors were searched in dynamic historical case library, conversion rules of Geo-CA were obtained implicitly, and the diffusion process of emergent plants communities was simulated. The results suggested that the overall accuracies of simulation were 76.41%, 73.05% and 75.91%, Kappa coefficients were 0.6402, 0.6382 and 0.6393, actual Moran I coefficients were 0.628, 0.686 and 0.709, and Moran I coefficients that simulated the results were 0.640, 0.710and 0.718 in 1997,2002 and 2008 respectively. CBR-CA model can adapt to rapid changes of wetland environments and reflect the Self-adaptability of complex systems.(5)Based on LC -Logistic-CA model and CBR-CA model respectively, the diffusion condition of the emergent plants in Wuliangsuhai wetlands in 2015 was predicted and the results showed that the area of the emergent plants were 174.34km2 and 172.92 km2, the number of patches were 634 and 646, patches density were 0. 7287/km2 and 0.7311/km2, the ratio of the gest patches area were 18.0019% and 17.9558%, the shape index were 32.8392 and 33.1556 and the degree of fragmentation of patches decreased, moreover, the edge structure became simple.更多还原