复杂类型海洋环境监测数据的空间抽样方法优化

发布时间：2019-05-09 16:01

【摘要】：“空、天、地、底”立体监测网的建立,为海洋资源开发与经济发展奠定了较好的海洋数据基础,但受海洋数据管理、数据分析以及数据应用等软/硬平台的限制,海洋领域“大数据,小知识”的现象日益凸显,因此,如何从海量海洋数据中快速获取信息为智能决策提供服务是当前研究热点之一。抽样调查能够快速地从海量数据中获取有效关键数据,适用于大规模大范围数据,其周期短,费用低,是解决大数据快速应用的主要途径。而目前抽样方法仍较多地依赖于传统概率抽样理论来建立,在面向多应用需求的驱动下,如何抽取有效可靠的数据使之快速生成有效信息,是现有抽样方法面临的挑战。海洋环境监测数据主要指通过浮标、调查船以及人工监测等手段,获取的海洋环境态势监测数据,其数据特性及抽样问题可归纳为:(1)海量性,海洋环境监测数据正以空前的速度增加和加快,动态更新频繁,多源,数据在空间尺度和时间尺度上都显现出存在冗余问题;(2)空间相关性,数据具有空间属性特征,距离近的数据相似度高,容易造成空间关联失效或样本重叠,抽样精度下降;(3)空间异质性,数据涵盖信息复杂,空间分布不均匀,给数据复用及后期处理带来困难。因此,考虑海洋环境监测数据的特征,设计优化空间抽样方法来帮助有效使用数据是值得研究的问题。在设计抽样和估计时,过度降低抽样成本会产生估计精度出现偏差、抽样结果失真等问题,而过高的样本量会增大数据冗余,从而提高成本,因此,如何能够平衡抽样精度与成本是设计优化空间抽样方法的主旨。本文主要研究内容如下:(1)分析海洋环境监测数据特征,总结其呈现出的多模态、高维度、多属性等特性给抽样方法带来的挑战,综述现有抽样方法的研究现状,分析这些方法运用于海洋环境监测数据时产生的问题。(2)提出一种系统空间抽样优化方法,考虑数据空间相关性,将半变异函数引入空间抽样方法设计,兼顾样本点在研究海域内分布均匀的同时,又保证在其抽样精度的前提下减少信息的冗余。(3)面向该批海洋环境监测数据的应用需求,考虑抽样对象的多属性联系,通过计算各属性权重对方法做进一步扩展,设计能够满足海洋环境监测数据多属性综合评估需求的空间抽样方法,用于更全面,更经济的抽样估计。(4)以某海域空间数据为实验对象,通过方差、抽样比率和趋势面分析,对本文设计的复杂类型海洋环境监测数据的空间抽样方法与传统抽样方法进行比较,结果表明,通过使用本方法计算得到的步长进行抽样能够有效减少数据量,同时保证一定的抽样精度,比较好的反应总体特征趋势。
[Abstract]:The establishment of "space, sky, earth and bottom" three-dimensional monitoring network has laid a good foundation for marine data development and economic development, but is limited by soft / hard platforms such as marine data management, data analysis and data application. The phenomenon of "big data, little knowledge" in marine field is becoming more and more prominent. Therefore, how to quickly obtain information from massive marine data to provide services for intelligent decision-making is one of the current research hotspots. Sampling survey can quickly obtain effective key data from massive data, which is suitable for large-scale and large-scale data, and its cycle is short and the cost is low, which is the main way to solve the problem of big data's rapid application. At present, sampling methods still rely on the traditional probability sampling theory. Driven by multi-application requirements, how to extract effective and reliable data to generate effective information quickly is a challenge for existing sampling methods. Marine environmental monitoring data mainly refers to the marine environmental situation monitoring data obtained by means of buoys, survey ships and manual monitoring. The data characteristics and sampling problems can be summarized as follows: (1) oceanography. The monitoring data of marine environment are increasing and accelerating at an unprecedented rate, and the dynamic updates are frequent and multi-source. There are redundant problems in the spatial scale and time scale of the data. (2) Spatial correlation, the data has spatial attribute characteristics, the similarity of close data is high, which is easy to cause spatial association failure or sample overlap, and the sampling accuracy decreases; (3) Spatial heterogeneity, complex data coverage and uneven spatial distribution, which bring difficulties to data reuse and post-processing. Therefore, considering the characteristics of marine environment monitoring data, the design of optimal spatial sampling method to help the effective use of data is a problem worthy of study. In the design of sampling and estimation, excessive reduction of sampling cost will lead to errors in estimation accuracy and distortion of sampling results, while excessive sample size will increase data redundancy and thus increase cost. How to balance the sampling precision with the cost is the main theme of the design optimization spatial sampling method. The main contents of this paper are as follows: (1) analyzing the characteristics of marine environmental monitoring data, summarizing the challenges brought by multi-modal, high-dimensional and multi-attribute characteristics to sampling methods, and summarizing the research status of existing sampling methods. The problems arising from the application of these methods to marine environmental monitoring data are analyzed. (2) A systematic spatial sampling optimization method is proposed, considering the spatial correlation of data, the semi-variant function is introduced into the design of spatial sampling method. At the same time, the sampling point can reduce the redundancy of information under the premise of sampling accuracy. (3) considering the application demand of the marine environmental monitoring data, considering the multi-attribute relation of sampling object, the sampling point is not only uniformly distributed in the sea area, but also guaranteed to reduce the redundancy of information under the premise of sampling accuracy. By further expanding the method by calculating the weight of each attribute, a spatial sampling method which can meet the needs of multi-attribute comprehensive assessment of marine environment monitoring data is designed for more comprehensive use. More economical sampling estimation. (4) taking the spatial data of a certain sea area as the experimental object, through the analysis of variance, sampling ratio and trend surface, The spatial sampling method of complex marine environment monitoring data designed in this paper is compared with the traditional sampling method. The results show that the amount of data can be effectively reduced by using the step size calculated by this method. At the same time, to ensure a certain sampling accuracy, a better response to the overall characteristic trend.
【学位授予单位】：上海海洋大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：P717

【相似文献】