黄渤海叶绿素a卫星遥感数据重构及时空分布规律研究

	黄渤海叶绿素a卫星遥感数据重构及时空分布规律研究
	王跃启
学位类型	博士
导师	刘东艳
	2014-05
学位授予单位	中国科学院研究生院
学位授予地点	北京
学位专业	环境科学
关键词	黄、渤海海表chl-a浓度卫星遥感数据重构 Dineof 时空分布规律
摘要	海洋浮游植物是海洋初级生产力的主要贡献者，海洋叶绿素a（Chl-a）浓度是海洋浮游植物现存量的重要表征参数之一。因此，全面系统的了解海洋Chl-a浓度的时空变化规律及与海洋环境变化的关系，对阐明海洋生态系统的变化规律有着重要作用。近年来，已有许多不同类型、不同技术参数的海洋水色遥感传感器投入运行，获得的大量遥感数据为海洋生态系统的实时监测提供了广阔的平台。相对于船舶、浮标等现场测定方法，卫星遥感技术对于海洋的观测范围大、周期长并且具有连续性，而且费用也较低。然而，水色遥感数据本身也存在着不容忽视的技术问题：（1）由于海洋上空云层覆盖、传感器扫描参数的设计等原因，使依赖于可见光和近红外波段反演的海洋水色遥感数据存在着较大比例的数据缺失；（2）由于难以检测的薄云或者海洋水体中多种水色物质相互的影响，使目标反演产品中面临着一定的数据精度问题；（3）多平台传感器的在轨运行为海洋环境的监测提供了持续的海量数据，如何有效的融合多源遥感数据，提高的数据的分析精度，也是亟待解决的问题。因此，针对特定的研究区域和具体的研究问题，需要从方法学上提高遥感数据使用的精度。中国的黄海和渤海（黄渤海，31-41⁰N，117-127⁰E）是典型的半封闭的陆架浅海，由于其复杂的水体特征和多云雾等气候条件的限制，海洋水色遥感数据在该海区的应用上面临着精度较低和覆盖率不足等现实问题。因此，本研究立足于该海域的海洋水色遥感应用的数据质量问题，利用MODIS和SeaWiFS数据，针对黄渤海区域1997-2013年期间的海洋水色遥感数据，进行了有效的数据重构和和分析，重建1997年-2013年期间黄渤海区域完整的海表Chl-a浓度数据库；并且利用重建后的完整的时间序列Chl-a浓度遥感数据库对黄渤海区域表层Chl-a浓度的十多年来的时空分布规律进行了详细、系统的研究。首先，利用有限的实测数据，结合前人的研究成果，对SeaWiFS、MODIS和MERIS三种海表Chl-a浓度标准算法产品进行了对比验证。结果显示，在黄渤海区域，虽然三种标准算法产品在绝对值上与实测数据相差较大，但是所反映的Chl-a浓度的相对时空分布形势上具有合理性，因此，可以利用标准产品进行黄渤海表层Chl-a浓度的相对时空变化规律分析。同时，亦对黄渤海区域SeaWiFS、MODIS和MERIS三种标准海表Chl-a浓度产品进行了数据一致性比较，结果显示：三种数据产品之间的均方根误差（RMSE）为0.16-0.18（log10(Chl-a)mg m^-3），平均绝对偏差（MAD）为0.10-0.12（log10(Chl-a)mg m^-3），相关系数（R）为0.87-0.89，三种数据产品之间的偏差远小于标准算法的设计误差（RMSE=0.22），呈现了很好的一致性。其次，针对海表Chl-a浓度卫星遥感资料的数据缺失和多源数据的融合问题，将近年来国外学者发展的基于经验正交分解的数据重构方法（DINEOF）引入黄渤海水色遥感资料的完整性重构研究中。针对该海区表层Chl-a浓度遥感数据的特点和实际的应用需要，对传统的DINEOF方法进行了有效的改进，具体体现在：（1）将深度分区方案加入到重构过程中，一方面可以有效的提高数据的重构精度（RMSE降低约10%），另一方面分区后的数据可以有效的支持并行计算，缩短算法的运行时间，提高运行效率。（2）基于回归残差理论的异常值检测方法可以有效的检测和消除数据集中的薄云等引起的“噪声”值，增加数据的可信度，提高DINEOF的重构精度（RMSE降低约30%）。（3）通过标准变量和辅助变量的划分对多变量DINEOF组合重构方法（M-DINEOF）进行了有效改进，并应用到多源海表Chl-a浓度的融合重构试验中，由于最佳模态数的减少，组合重构结果的统计误差会有一定程度的增加，但是相对于直接的替代方法，其长时间序列数据的重建结果却更为合理，融合重构后的长时间序列完整数据集可以有效的支持Chl-a浓度的长时间序列的时空规律研究。最后，利用重建后完整的海表Chl-a浓度标准数据产品，系统的分析了黄渤海区域Chl-a浓度1997-2013年期间的空间分布、季节变化和年际变化分布规律。（1）从空间分布特征上来看，多年平均Chl-a浓度与水深呈现明显的负相关关系（R=-0.87, P<0.001），多年变化的标准差与水深之间呈现正相关关系(R=0.69, P<0.001)；在水深较浅的近岸海区Chl-a浓度平均值较高（3.0-5.0 mg m^-3），但是标准差较低（<0.3 mg m^-3），说明近岸地区的Chl-a浓度常年维持在较高的水平；而水深较深的开阔海域（黄海中部）Chl-a浓度平均值较低（0.5-2.0 mg m^-3），但是标准差较高（>0.8 mg m^-3），说明该区域Chl-a浓度存在较大的时间变异。（2）黄渤海区域海表Chl-a浓度的季节变化存在明显的空间差异，水深较浅的近岸海区（水深< 20 m）Chl-a浓度的季节波动幅度（3.8 mg m^-3- 4.4 mg m^-3）较小，呈现出冬末春初季（2月、3月）高、夏季（6月、7月）低，秋季出现弱高峰的变化特征；在水深20 m-50 m的海区，Chl-a浓度在早春的3月份出现最高值，最低值出现在夏季的7月、8月份；而水深大于50m的开阔，Chl-a浓度的最高值出现在春季的4月份，最小值出现在夏季的7月、8月份；长江口邻近海区受入海流量的影响，Chl-a浓度夏季出现最高值，个别年份夏季月均值高于10 mg m^-3；而黄河口邻近海区，海表Chl-a浓度的季节变化特征与近岸地区相似（水深< 20 m）。（3）在年际变化尺度上，整个研究区平均海表Chl-a浓度呈现显著的增加趋势（约1.0 % Year^-1），并且增长率有随着水深增加而增加的态势，在近岸地区增长率小于0.3% Year^-1，在水深较大的开阔海域增长率大于1.3% Year^-1，增长率最高的海区为北黄海和南黄海的中西部区域，增幅约为2%-3% Year^-1。通过与黄渤海海区已有研究结果的比较和对影响Chl-a浓度的主要海洋环境因子的对比分析发现，Chl-a浓度的空间分异受水深的影响较大，近岸水和远岸水之间不同的性质和博弈关系，影响了Chl-a浓度的空间变化规律。而海表风场和海表温度等海洋环境的季节变化在不同水深条件下对表层营养盐的调控作用导致了海表Chl-a浓度季节变化的空间差异。近岸和离岸海区Chl-a浓度不同的年际增长幅度，则是体现了Chl-a浓度对营养盐的逐年增加的响应在不同深度海区的差异性。
其他摘要	Phytoplankton is the major contributor of marine primary productivity, Chl-a (Chl-a) is one of the important estimators of phytoplankton biomass. Therefore, it is important to make an comprehensive study of the variation of Chl-a concentration and its relationship with the marine environment, and it is crucial to interpret the evolution of marine ecosystem. The rapid development of satellite remote sensing technology has provided a useful complementary approach for us to acquire variety of satellite-derived ocean color data and to quantify the real-time characteristics of the marine ecosystem. Satellite remote sensing observation has wider viewing, longer continuous time series and lower cost compare to the traditional ship-based and buoy-based in-situ sampling methods. Over the last decades, an increasing large number of satellite sensors have been measuring ocean color parameters that provide a global view of the state of the marine ecosystem and its long-term evolution. However, satellite ocean color remote sensing technology also has some un-ignored restrictions: (1) the satellite ocean color datasets obtained by visible and infrared bands commonly present with large-scale missing data due to cloud coverage, or malfunctions in the satellite sensors; (2) The satellite ocean color products may have low accuracies caused by other optical constituents suspended in waters and/or undetected cloud; (3) These multi-sensor observations of ocean color have the potential to better detect long-term trend due to their greatly improved temporal frequency and almost global coverage but are limited by their relatively short duration and uncertainties associated with instrument calibration and data processing algorithms. Therefore, there are demands of effective statistical methods to improve the accuracy of remote sensing products, according to local regions and specific problems. Bohai and Yellow Sea are (31-41⁰N, 117-127⁰E) typical shallow epicontinental sea. In the area, the satellite ocean color datasets have more serious problems of accuracy and missing data coverage result from shallow water depth and complex ocean dynamics. The applications of remote sensing ocean color products in Bohai Sea and Yellow Sea is still in exploration phase, especially in the study of long-term temporal and spatial patterns of Chl-a concentration. This study focus on the key techniques of ocean color remote sensing applications in Bohai Sea and Yellow Sea, constructed an long-term cloud-free satellite sea surface Chl-a concentration dataset using effective data reconstruction and merging method based on SeaWiFS and MODIS datasets. Moreover, we gave an comprehensive analyses of temporal and spatial patterns used these long-term reconstructed satellite datasets. Firstly, this study made evaluations of standard algorithm derived Chl-a product from SeaWiFS, MODIS and MERIS sensors, using in situ data collected during cruises. The satellite remote sensing Chl-a values can reflect reasonable relatively temporal and spatial patterns, although their absolute values have large biases with in situ measurement. We also present an extensive comparative analysis of standard Chl-a products obtained from the three sensors. Based on regional statistics, the three Chl-a records appear relatively consistent. The root mean square error (RMSE), the mean absolute difference (MAD) and the correlation coefficient (R) of the pairwise comparisons amount to 0.1574-0.1821 (log10(Chl-a)mg m^-3), 0.1016-0.1175 (log10(Chl-a)mg m^-3) and 0.8701-0.8929, respectively. Considering the inherent uncertainties in the OC4v4 algorithm (RMSE=0.22 log10(Chl-a)mg m^-3), the large dynamics range of Chl-a concentration, and a large portion of coastal waters in this region, these three products present no large systematic error in their absolute values as well as in the spatial patterns. It is therefore possible to use multi-sensor datasets to constructed long-term Chl-a data records. Secondly, In response to the serious missing data coverage and multi-sensor data merging demand, we present an application of the DINEOF method to reconstruct the missing data in a long-term satellite Chl-a dataset over Bohai Sea and Yellow Sea. In this study, we modified the ordinary DINEOF method based on the dataset characteristic and our application objectives to improve the overall performance of this interpolating technique. The depth subdivision scheme used in the DINEOF reconstruction made significant improvement to the accuracy of the reconstruction (the RMSE decreased about 10%), and it is possible to parallelize the computational load to speed up the reconstruction process substantially. A new outlier detection method base on standardized residuals theory can efficiently detect and eliminate spurious values, and significantly improve the accuracy of DINEOF reconstruction (the RMSE decreased about 30%). We successfully applied a modified multivariate DINEOF method, to combine the processes of missing data filling and multi-sensors merging, to reconstruct a long-term complete satellite Chl-a data records. Thirdly, We make an comprehensive analyses of the spatial distribution, seasonal and interannual variability of sea surface Chl-a concentration in Bohai Sea and Yellow Sea over the period from September 1997 to December 2013, based on the reconstructed long-term satellite ocean color datasets. (1) General spatial characteristics are that the Chl-a concentrations in coastal waters concentrations were higher than in offshore waters. Linear statistical analysis between Chl-a concentrations and water depths displayed a significant negative correlation (R=-0.87, P<0.001), but there is a significant positive correlation (R=0.69, P<0.001) between (standard deviation) STD and water depths. These results indicated a significant spatial correlation between water depth and Chl-a concentrations. The high Chl-a values (3.0-5.0 mg m^-3) and the low variability (STD<0.3 mg m^-3) are observed in coastal waters, indicating that these areas present persistent high Chl-a values throughout the study period. The most central area shows low Chl-a values (0.5-2.0 mg m^-3) and high STD (STD > 0.8 mg m^-3), indicating that this area exhibits high amplitude of Chl-a variability during the study period.(2) The seasonal patterns of Chl-a concentration show significant spatial heterogeneity in the study area. In the shallow depth water, the Chl-a concentration present small seasonal amplitude (3.8 mg m^-3- 4.4 mg m^-3), with high values in winter (January, February) and low values in earlier summer (June, July). In the moderate depth (20-50 m) water, the Chl-a concentration present high values in March and low values in July and August. In the waters with depth above 50m, the Chl-a concentration present a prominent high value in April, which is evidence of the spring bloom in the central Yellow Sea. For the large river mouths, a different seasonal pattern was noted between Yangtze and Yellow Rivers. The Yangtze River mouth displayed a unique feature with a Chl-a maxima from May to August and attributed it to the maximum freshwater discharge in summer; The Yellow River mouth displayed a similar seasonal pattern with coastal area, due to the influence of seasonal hydrodynamic. (3) On the interannual scales, the Chl-a concentration in most of the study area show increased tendency, with the average increasing rate is 1.0 % Year^-1, The increasing rate show different magnitude along the water depth. The increasing rate of coastal area is below 0.3% Year^-1 and the increasing rate of the deeper water is about 1.3% Year^-1, the western and central area of the Yellow Sea present highest increasing rate (2%-3% Year^-1). According to the previous studies and the analyze of the marine environment parameters (temperature, wind and nutrient), we discussed the mechanisms of Chl-a spatial and temporal patterns. The spatial pattern is mainly controlled by the features of bathymetry. The interactions between coastal and open sea waters control Chl-a spatial patterns. The different nutrient supplying between coastal and open sea water contribute to different seasonal variability in Chl-a concentration. The different magnitude of increasing multi-year trend indicate different responds to increasing nutrient supplying in coastal and open sea waters.
语种	中文
文献类型	学位论文
条目标识符	http://ir.yic.ac.cn/handle/133337/7078
专题	中国科学院烟台海岸带研究所知识产出_学位论文
推荐引用方式 GB/T 7714	王跃启. 黄渤海叶绿素a卫星遥感数据重构及时空分布规律研究[D]. 北京. 中国科学院研究生院,2014.

条目包含的文件
文件名称/大小	文献类型	版本类型	开放类型	使用许可
毕业论文-王跃启-.pdf（19143KB）			开放获取	CC BY-NC-SA	请求全文