Mapping Seagrass Biodiversity Indicators of Pari Island using Multiple WorldView-2 Bands Derivatives

Abstract

Comprehensive information on seagrass biodiversity indicators, such as species composition, percentage cover, and biomass carbon stock, remains limited across various regions globally. Mapping these indicators using remote sensing images requires extracting maximum information from the input images to achieve effective results. This study aims to map seagrass distribution, percent cover (PC), and aboveground carbon stock (AGC) as biodiversity indicators in the optically shallow waters surrounding Pari Island. We integrate WorldView-2 (WV2) derivatives, field seagrass data, and RF classification and regression algorithms to accomplish this objective. The WV2 image derivatives encompass surface reflectance bands, band ratios, mean and variance co-occurrence texture bands, and principle component bands. These inputs are used individually and collectively for mapping, employing a random forest algorithm trained with field seagrass data. Our results demonstrate that the most accurate benthic habitat map achieves an overall accuracy (OA) of 65.2%, with a user's accuracy of 65.2% and a producer's accuracy of 72.8% for the seagrass-dominated class. Seagrass PC mapping yields a root mean square error (RMSE) of 17.1%, with an average PC of 47.4 ± 9.9%. Seagrass AGC mapping achieves an RMSE of 5.0 g C m-2, with an average AGC range of 6.2 – 29.1 g C m-2, estimating the study area's aboveground biomass carbon stock at 27.9 tons C. Combined inputs produce the most accurate results for all biodiversity indicators, emphasizing the importance of utilizing combined bands from SR band derivatives to maximize information input for training mapping algorithms, instead of using derivative bands individually or as replacements for the initial SR bands. Keywords : Seagrass; Biodiversity; Mapping; WorldView-2; Pari Island   Copyright (c) 2023 Geosfera Indonesia and Department of Geography Education, University of Jember This work is licensed under a Creative Commons Attribution-Share A like 4.0 International License