珊瑚礁遥感监测方法研究

【作者】 徐兵；

【导师】 张鹰；

【作者基本信息】 南京师范大学 ， 海洋地理学， 2013， 博士

【摘要】 在气候环境变化和人类活动的双重影响下,全球珊瑚礁正面临着严重的危机,有可能成为第一个因全球变暖而消失的生态系统。遥感技术可以快速、大面积、周期性地获取全球范围内珊瑚礁的各类信息,已经发展成为珊瑚礁调查和监测的主要手段。为了更有效地保护珊瑚礁生态系统,研究适用于我国珊瑚礁遥感监测的技术与方法,具有重要的现实意义。本文以广东徐闻珊瑚礁保护区为研究区域,开展珊瑚礁水下地形遥感反演、底质光谱特征分析、珊瑚信息提取、健康状态变化检测等相关技术研究。主要结果如下：(1)基于研究区TM遥感影像和水深数据,建立了珊瑚礁区水下地形的遥感反演模型。基于水深反演专题图,结合珊瑚礁生态地貌特点,初步划分了珊瑚礁水下生态带。通过遥感解译从更宏观的角度描述了珊瑚礁水下地貌概况,表明了水深遥感方法在珊瑚礁区水下地形反演中的可行性。(2)通过对珊瑚礁不同底质的光谱测量与特征分析,研究发现各类底质光谱反射率总体上呈现出一个高反射区(400～580nm)、快速衰减区(580～740nm)和一个低反射区(740～900nm),不同的底质光谱之间存在特征差异,且底质光谱反射率受到水深和水质环境的影响较大。对比野外实验采集的五种珊瑚优势种的光谱反射率,研究发现不同珊瑚优势种之间光谱特征存在明显差异。同一种珊瑚在健康与白化两种状态下光谱之间也存在明显差异,通过光谱的一阶微分处理可以更清晰地发现这些差异。而对于同属不同种的珊瑚而言,其光谱差异非常细微,光谱可分性较差。通过对ETM+、QuickBird、IKONOS和SP0T5等几类常用的传感器进行光谱响应分析,结果表明：对于简单的底质粗分类,这几种传感器都能够有效发挥作用；而对于更细致的底质精分类,则需要借助高光谱资料进行更深入的分析。光谱测量与分析为遥感监测工作的后续开展提供了光谱学基础,也为遥感数据源选择和底质分类提供了依据。(3)通过研究珊瑚信息提取方法,在水深反演基础上提出利用礁盘水深识别模型来提取大型礁盘信息。结果表明：应用该模型能够提取出潜在的大型礁盘信息,但是受模型精度的影响对一些礁盘存在误判情况,结合野外调查数据进一步检验,能够更准确地识别大型礁盘。基于底质光谱特征,应用高空间分辨率QuickBird影像对珊瑚礁底质进行粗分类研究,采用监督和非监督分类分别提取出珊瑚信息,并作对比分析。分析结果表明：提取的珊瑚主要分布于近岸lkm附近的区域,水深2m以浅范围；监督分类提取珊瑚信息的精度约为71%,高于非监督分类的精度。受水深和遥感数据影响,底质信息提取的精度还有待进一步提高,研究工作初步建立起了珊瑚礁浅海底质信息提取的思路和方法,为后续的研究提供了参考。(4)基于珊瑚礁生态系统的空间特征,研究基于多时相影像空间统计的变化检测方法,利用该方法识别出珊瑚礁健康状态发生变化的区域。论文对研究区2004、2007和2010年三景TM影像资料进行空间局部自相关统计分析,结果表明：三个不同年份局部区域表现出了较为明显的空间统计变化特征,空间低值聚集区和高值聚集区变化较为明显,反映出了2004～2010年研究区范围内生态系统的总体变化趋势。通过对2004～2010年间的4个野外调查样方资料进行分析,对其空间统计上的变化趋势进行了遥感解译,结果证实了变化检测方法的有效性。此外,受连续多年的强台风影响,近岸海域覆盖了大量的死亡珊瑚碎片和泥砂混合物,空间统计分析上近岸区域表现为明显的同质性增强趋势,表明了珊瑚礁生态系统受台风灾害影响较明显。该变化检测方法具有对深度、水体等不太敏感的优点,降低了实际应用难度,为珊瑚礁遥感动态监测提供了一个新的思路。本文研究了珊瑚礁遥感监测方法,讨论了遥感技术在珊瑚礁资源调查与动态监测方面的可行性,对一些技术和方法的具体应用进行了尝试。珊瑚礁生态系统具有较强的区域特征,虽然以徐闻保护区为例取得了初步的结果,但是未来还需要在水下光谱、物理模型、水体影响、多源遥感相结合等方面开展深入研究。更多还原

【Abstract】 As one of the world’s ecosystems, coral reefs are facing a major crisis under the influence of global climate change and human activities, and it may become the first extinct ecosystem over the word on account of global warming. Remote sensing techniques, as a rapid and periodic investigation and monitoring tool, has being widely applied in coral reefs research. It is of great significance to establish a remote sensing method on coral reefs investigation and monitoring to protect the ecosystem effectively.It takes Xuwen coral reef protection zoon of state level as research area, discusses the four aspects such as underwater topography remote sensing inversion, spectrum characteristics analysis, substrate classification and image change detection. The conclusions are as follows:(1) It established the remote sensing inversion model of coral reef underwater topography based on the satellite image data and field survey data, divided the underwater ecoregion by using the water depth thematic map and landform characteristic of coral reef, and described the underwater topography on a macro point of view by remote sensing interpret. The research shows that the remote sensing inversion method of water depth has good feasibility on coral reef underwater topography.(2) After spectral measurement and feature analysis of different substrates, it shows that there are three spectral regions on reflectance diagram, the high reflection area (400-580nm), fast attenuation area (580-740nm) and low reflection area (740-900nm) in general. The water depth and quality have greater influence on the spectrum, and the spectrum feature of substrates are significant different. There are some spectrum differences between dominant species of coral,and spectrum between healthy and bleaching state of a kind of coral also have obvious difference, it can be more accurate to find the difference through the first order differential of spectrum. But for different corals of a sibling, the distinction of spectrum is subtle. It’s difficult to find the nuances even through the first order differential method. It simulated the spectral response of ETM+, Quick Bird, IKONOS and SPOT5by spectral response function, and the result shows that these sensors are suitable for simple classification. For more detailed classification, it maybe needs to use hyper-spectral data. The research laid a solid theoretical basis for coral reef remote sensing and also provided a gist for image data choice.(3) On the basic of water depth inversion method, it presented a method to identify large coral reefs after researching on coral reefs information extraction. Results show that the method can identify potential large coral reefs although the precision of this method is not high. This method has some limitations in practical application. Based on spectral characteristics, it applied QuickBird image to achieve a simple classification of substrates. It compared the results of supervised and non-supervised classification. The research shows that corals that been identified are mainly distributed in the near shore area near1km, and the shallow depth of2meters. The accuracy of supervised classification is about71%, which is higher than unsupervised classification. Influenced by water depth and remote sensing data, the precision of substrate information extraction has yet to be further improve, but it provides a new thought and technical reference for follow-up study.(4)It researched a spatial autocorrelation statistics method for multi-date images based on the spatial feature of coral reefs. It applied the spatial autocorrelation statistics method to analysis the TM image data of2004,2007and2010at Xuwen coral reef protection zone. The result shows that local area of images showed the evident change in the statistical characteristics of spatial during the year2004to2010, the change on low and high space zones are obvious, and performed the overall trends within the study area. It analyzed and explained the four field investigation points between2004and2010, and proved that the method is feasible. In addition, influenced by consecutive years of strong typhoons, a lot of dead coral debris and mud sand mixture are eroded over near shore area, and it shows obvious homogeneity trend in spatial statistic analysis. Coral reef ecosystems affected by natural disasters is obvious. It proved that the method is a simple and practical method for change detection with low cost and is not sensitive to factors such as water depth at last.This thesis studies the remote sensing monitoring method of coral reefs, discussed the feasibility of remote sensing method applied on coral reef resources investigation and dynamic monitoring. Because of the strong regional characteristics of coral reef ecosystem, in spite of achieving some preliminary results of Xuwen coral reef protection area, we still need further research in underwater spectral analysis, physical model, water quality influence, multi-source remote sensing etc.更多还原