【作者】 黎治华；

【导师】 高炜； 高志强；

【作者基本信息】 华东师范大学 ， 自然地理学， 2011， 博士

【摘要】 本论文重点研究了四个内容：1)基于MODIS数据反演和重构时间序列数据集的技术、方法和程序；2)长江三角洲地区生态环境演变的时空分布、状态及其动态；3)研究区生态环境演变趋势的建模和预测；4)生态环境在长时间序列数据上的年际扰动和季节性突变分析。归纳起来,所做研究工作主要是MODIS时间序列数据集的反演和重构、长江三角洲地区生态环境的演变两个方面,解决区域(或全球)生态环境发展演变过程和生态环境要素扰动与突变的监测、评估与预测问题,具体内容如下：(一)遥感数据时间序列数据集的反演与构建MODIS数据的预处理、反演和时序数据集重构是本研究工作的基础,主要包括：1)由8d MODIS数据合成16d的过程中,尝试使用“最大值”法合成了16d的LST,使用“最小值”法合成了16d的MOD09A1数据；同时,尝试了增加检测点对BISE算法(最佳指数斜率提取)进行改进,将改进后的BISE和Savizky-Golay滤波器结合使用,以提高时序数据集重构的质量；2)将光谱角制图(SAM)和最小距离算法(MDM)各自的优势通过程序进行“融合”,形成了综合性分类算法(SAM-MDM),并尝试应用到遥感长时间序列数据的土地覆盖分类上来。对2005年分类数据进行精度评估,结果是Kappa系数和总体精度分别达到0.71265和83.70%。3)根据已有水分指数算法基于MODIS进行反演和对比,实验表明地表含水指数(SWCI)具有较好的性能,与地面降水观测数据进行相关分析,以验证SWCI描述地表水分含量的能力。4)以往研究表明EVI与NDVI相比具有描述高密植被能力强、不饱和的特征,以此参照以往NDVI反演植被覆盖率(FVC)的算法,尝试使用EVI来反演FVC,并用于研究区11a植被覆盖状况演变的过程和特征的评估。(二)长江三角洲地区生态环境演变的过程与扰动长江三角洲地区生态环境演变是本文重点研究的对象,主要工作是应用MODIS数据反演出植被覆盖率(FVC)、地表温度(LST)、地表含水指数(SWCI)和土地覆盖(LULC)等,以分析生态环境的演变过程、趋势和扰动。主要工作包括：1)使用LST、SWCI、FVC和土地覆盖研究生态环境状况及其演变是可行的。通过相关分析表明遥感反演出的生态要素指标与地面观测数据具有较高的相关程度,如LST与气温观测值具有较高相关程度,年内SWCI与降水量相关程度较高。2) LST、SWCI、EVI和LULC的时空动态演变分析表明生态环境要素的时空分布是不平衡的,沿“吴淞口-长江-大别山”(相当于1200mm等降水量一线)是研究区生态要素南北差异的分界线。植被覆盖率空间分布上呈现“东南高西北低、山高平低”的总体特征；LST空间分布的总体特征是“南高北低、低覆盖地高水域低、平原高山丘低”；SWCI空间分布总体特征表现为“南低北高、低覆盖地低高覆盖高”；耕地主要集中在北边,林地主要集中在南边。3)生态环境要素的时空动态分析揭示了环境演变是一种时空轮动消长的“波浪式”模式。从时间上看,2000-2010年11 a来,LST略有升高、SWCI略有变“湿”、植被覆盖率有增加趋势。研究区林草地明显减少,城市等低覆盖地明显增加,耕地略有增加,水域变化轻微,长江三角洲核心区域受人类影响最为强烈,耕地、草地、林地的转移概率较大。这种结果是波浪式的年际轮动过程所形成的,通过LUCC分析建立了土地覆盖类型转移矩阵,尝试使用马尔科夫(Markov)链对土地覆盖动态变化未来的趋势进行了预测。4)在研究区应用扰动指数(DI)分析了生态环境的扰动,尝试季节变化指数(SVI)模型的构建,以此分析2010年生态环境的年际扰动、季节性突变的程度和空间分布,并用当年地面观测记录文献和相关研究报告对结果进行验证,表明DI、SVI对灾害发生的区域检测具有一定的可行性。本论文的选题和研究内容是一个十分综合的、复杂的大课题,这里虽然做了一些研究工作,但并非完满,还需要完善和深入研究的问题有：1)生态环境要素关系模型的研究和完善,以进一步揭示生态环境要素间的演变关系；2)生态环境要素演变状态的定量分析,定量回答生态环境演变质与量的关系问题；3)时间序列趋势预测模型还需要进一步完善,重点通过时间和空间融合建模以提高预测的能力；4)生态环境演变的扰动和季节突变模型的进一步深化,提高生态环境扰动和季节突变检测的精度,并建立与生态环境各类灾害相应的评估体系。更多还原

【Abstract】 Long term (over 10 a) time series data from satellite-based sensors promise to improve remote sensing data observation capabilities and increase our understanding of land use/land cover changes and eco-environment evolutions and tendencies. MODIS(Moderate-resolution Imaging spectro radiometer) provides an opportunity to construct some new analysis models based on time series with contemporaneous measurements of the same spatial and temporal scales. In this thesis, the MODIS remote sensing imagery was used to study the eco-environment evolution and tendency of the Yangtze River delta area from 2000 to 2010.1) The technology, methodology and process which used to retrieve and reconstruct the time series data of eco-environment based on MODIS were proposed.2) The spatial and temporal processes, conditions and dynamic characteristics of the eco-environment evolutions of the Yangtze River delta area were analyzed based on remote sensing time series datasets.3) On the basis above, the time series was used to analyze the eco-environment evolution trends of Yangtze River delta area.4) Using DI and SVI model, a study on interannual disturbance and seasonal variation mutations of the eco-environment was performed on 2010 based on the long time series datasets.The MODIS data preprocessing, retrieval and reconstruction of the time series datasets are the basis of this thesis. The major works are as follows:(1) According to the characteristics, contents and practical scope of MODIS data, three data products were chose in the research, including MOD13Q1, MOD11A2 and MOD09A1. (2) Because the spatial and temporal resolutions, projection and coordinate are different among MOD13Q1, MOD09A2 and MOD09A1, all the data products are required to simulated to the unified spatial resolution, temporal resolution and projection. The "maximum value" and "minimum vaule" synthetic methods were used to generate 8-day LST and 16-day MOD09A1 products respectively, which had greatly improved data quality. (3) The BISE model was improved and combined with the Savizky-Golay filter for reconstruction of time series, the reconstruction quality had improved. (4) A new classification method named SAM-MDM was introduced with the advantages of both SAM (Spectral Angle Mapper) and MDM (Minimum Distance Method), and performed on the EVI classification, the Kappa coefficient and overall accuracy of 2000’s LULC classification were 0.72506 and 84.09% respectively, while in 2005, they were 0.71265 and 83.70%.Many indices, such as the fraction of vegetation cover (FVC), land surface temperature (LST), surface water capacity index (SWCI), land use and land cover (LULC) and so on were derived from MODIS data, they were ecological elements used to analyze the eco-environment evolution of the Yangtze River Delta.1) Studies showed that it is feasible to apply EVI instead of NDVI to produce FVC, the characteristics of temporal-spatial distributions and changes of plants were studied using the FVC retrieved from EVI.2) the temporal-spatial change progress of LST during 2000 to 2010 was analyzed. The temporal-spatial eco-environment heat distribution and its dynamic change characteristics were studied from the contrast among different areas and different years.3) The temporal and spatial distribution of LULC and its dynamic changes were analyzed using dynamic degree model of LULC. It indicated that these transfer patterns of LULC work as wave-like evolutions. The author also built the transition probability matrix for predicting the trend of LUCC with Markov process.4) The temporal and spatial characteristics of the evolution of SWCI was explained by analyzing the temporal and spatial distribution, results revealed the wave-like features of inter-annual alternation and spatial round action for eco-environment evolution.5) the spatial and temporal characteristics of the subinterval eco-environments and its elements were released by analyzing SWCI, such as the differences between Northern China and Southern China for vegetation coverage, land, and surface water, the difference between hilly areas and plain areas, the difference between urban and rural.In the study of eco-environment elements’evolution and their relationships, the Correlation Analysis, Multivariate Regression Analysis, Transition Probability Matrix and Time Series Analysis were used.1) Three models—the eco-environment elements correlation, multiple regression analysis model, time series self-correlation model—were built, using the 11a time-series data to analyze the extent of temporal and spatial variation of the eco-environmental evolutional elements.2) The correlation between LST, SWCI, EVI and ground observed temperature and precipitation was calculated, the results showed a high relevance.3) The result indicated that the Yangtze River-the Dabie Mountain line (1200mm precipitation line) is the dividing line of Northern China and Southern China to study the distribution and evolution of various eco-environment elements.4) The result also showed that the core area of the Yangtze River was affected evidently by human activity and Shanghai was the most affected area. From 2000 to 2010, the forest and grassland reduced significantly, the low-coverage land increased evidently, the cultivated land increased little, and water-body changed slightly, those tendencies had a positive correlation with the human activity. It could figure out that the farmland, grassland and woodland had larger transition probability.5) Using disturbance index (DI), an analysis on spatial distribution of eco-environment disturbance was performed on 2010, and had tried to build the seasonal variation index (SVI) to analyze the spatial distribution of the seansonal mutation of eco-environment in 2010, the both result validations with ground-based observations, the current literature and research reports showed it got a certain effect.The study on the eco-environment evolutions and its trends such as disturbance beased on remote sensing time series datasets is a very comprehensive and complex issue, this thesis got certain results, but not complete success. The further research issues are as follows. First, to improve the eco-environmental factor models to reveal the evolving relationship better; Second, to perform the quantitative analysis on the states of eco-environmental factors; Third, the time series trend forecasting models need further improvement; Fourth, the study on disturbance models needs to improve for detection and assessment precision of eco-environment evolution.更多还原