摘要:
在海-气环境变化剧烈的今日,准确高效地实现珊瑚礁底质信息识别是进行珊瑚礁动态监测研究的基础。文章获取了2013—2021年4个时期西沙群岛永乐环礁的Landsat8卫星数据,结合不同底质的光谱和纹理差异,提出了一种基于光谱纹理指数的决策树分类模型,采用面向对象和基于像元的分类方法进行珊瑚信息提取,并定量统计了永乐环礁底质变化情况。结果表明: 面向对象的分类结果整体上优于基于像素的分类结果,且决策树分类结果的Kappa系数在0.631~0.681范围,分类精度高于传统监督分类精度约7~10个百分点; 珊瑚丛生带大多分布在岛礁的中部水动力较弱区域,除银屿和金银岛上的珊瑚呈面状分布外,其他岛礁上的珊瑚多呈带状分布; 总体时段内永乐环礁的珊瑚丛生带和沙洲面积变化显著,虽然珊瑚丛生带的总面积增加了1.689 km2,但石屿、晋卿岛、全富岛、珊瑚岛和羚羊礁的珊瑚丛生带退化情况严重,其面积减少了0.107~0.892 km2不等。该文证明了利用中等空间分辨率影像建立的底质指数是可靠的,可应用于珊瑚遥感信息提取,能够为珊瑚礁资源调查及科学管理提供技术支持。
Abstract:
In view of the drastic changes in the ocean-atmosphere environment, the accurate and efficient identification of coral reef substrate information is essential for the dynamic monitoring of coral reefs. Based on the Landsat8 satellite data of the Yongle Atoll in the Xisha Islands of four periods during 2013—2021, this study proposed a decision tree classification model using spectral and texture indices according to the spectral and texture differences between different substrates. Then, the coral information was extracted using object-oriented and pixel-based classification methods. In addition, the changes in the substrate of the Yongle Atoll were quantitatively analyzed. The results are as follows: ① The results of the object-oriented classification are superior to those of pixel-based classification overall. Moreover, the decision tree classification results yielded Kappa coefficients of 0.63~0.68, with classification accuracy about 7~10 percentage points higher than that of conventional supervised classification; ② Coral thickets are mostly distributed in the central, weakly-hydrodynamic parts of islands and reefs. The corals in the Yinyu Reef and the Jinyin Island exhibit a planar distribution pattern, while those in other islands and reefs mostly show a zonal distribution pattern; ③ The areas of coral thickets and sandbanks in the Yongle Atoll changed significantly overall. Although the total area of coral thickets increased by 1.689 km2, the coral thickets in the Shiyu, Jinqing, Quanfu, and Shanhu islands and the Lingyang reef were severely degraded, with areas decreasing by 0.107~0.892 km2. This study verified that the substrate index established using medium spatial resolution images is reliable and can be applied to remote sensing information extraction of corals. Therefore, this study will provide technical support for the investigation and scientific management of coral reef resources.
下载: