Deteksi Batimetri Perairan Dangkal di Pulau Menjangan, Provinsi Bali Menggunakan Citra Landsat

Abstract

Remote sensing-based research in Indonesia using satellite imagery frequently faces the challenge of cloud coverage due to the tropical country. One spatial data that can be extracted from satellite imagery is bathymetry. However, cloud-covered water bathymetric extraction still needs to be examined. This study aims to understand the ability of Landsat 7 ETM+ acquired on 29 July 2013, and Landsat 8, acquired on 24 July 2020, as the representative of non-cloudy image compared to Landsat 8, acquired on 9 August 2020, as the cloudy image. Stumpf algorithm was applied, including a statistical approach of linear regression analysis with in-situ data measurement from Single Beam Echo-Sounder (SBES) to derive the absolute bathymetric map with several classes of depth ranging from 0 – 2 m up to > 10 m. To assess the accuracy, RMSE and confusion matrix was used. The result shows that Landsat 7 ETM+ yields the highest R² with 0,52, while the lowest total RMSE (8,167 m) and highest overall accuracy of about 69% from the confusion matrix was achieved by the cloudy image of Landsat 8. Nevertheless, the highest absolute depth value yield by Landsat 8 non-cloudy image with 16,1 m. This research confirms that the highest R² value does not always produce the best model, but it is still promised to be used. Furthermore, the quality of the imagery based on its percentage of cloud coverage is affecting the resulted model.