Abstract

Biodiversity of benthic habitats is among the highest of all ecological communities. This study was conducted to model benthic habitat biodiversity indices using a remote sensing approach in optically shallow waters in Karimunjawa Islands-Indonesia. These islands have a wide variety of benthic environments. Two multispectral imagers, namely Sentinel-2A and Landsat 8 OLI, were used. A series of statistical tests were applied in the empirical modeling using the pixel values of both images with in situ Shannon index (H), Simpson index (D), and Shannon's Equitability (EH) calculations. The modeling inputs were sunglint-corrected bands, water column-corrected bands, PCA-transformed bands, MNF bands, and occurrence texture bands. The results indicate that multispectral remote sensing images can be used to map benthic habitat biodiversity indices. However, the difference between the concepts of H, D, and EH calculations and the reflectance value recorded by the sensor remove the possibility of obtaining higher accuracy. H, D, and EH maps derived from Sentinel-2A had varying levels of accuracy, namely 46.8%, 59.1%, and 54.5%, respectively, while Landsat 8 OLI produced these three maps with 45.81%, 57.34%, and 53.81% accuracy.

摘要

底栖生境的生物多样性是所有生态群落中最高的。本研究使用遥感方法在印度尼西亚卡里蒙贾瓦群岛的光学浅水区模拟底栖栖息地生物多样性指数。这些岛屿有各种各样的底栖环境。使用了两个多光谱成像仪，即 Sentinel-2A 和 Landsat 8 OLI。使用原位图像的像素值在经验建模中应用了一系列统计测试香农指数 (H)、辛普森指数 (D) 和香农公平 (EH) 计算。建模输入是阳光校正带、水柱校正带、PCA 转换带、MNF 带和发生纹理带。结果表明，多光谱遥感图像可用于绘制底栖生境生物多样性指数。但是，H、D 和 EH 计算的概念与传感器记录的反射率值之间的差异消除了获得更高精度的可能性。从 Sentinel-2A 导出的 H、D 和 EH 地图具有不同的准确度，分别为 46.8%、59.1% 和 54.5%，而 Landsat 8 OLI 生成这三张地图的准确度分别为 45.81%、57.34% 和 53.81% .