Assessment of PlanetScope images for benthic habitat and seagrass species mapping in a complex optically shallow water environment

Abstract

ABSTRACTThis paper presents the first assessment of PlanetScope image for benthic habitat and seagrass species mapping in optically shallow water. PlanetScope image is equipped with ideal resolutions for benthic habitat and seagrass mapping including high-spatial resolution (3 m), high radiometric resolution (12-bit), sufficient water penetration bands (Visible-Near-infrared) and very high temporal resolution (almost daily), which distinguishes it from other high spatial resolution images. It is necessary to assess the accuracy of this ideal system in a real-world benthic habitat and seagrass species mapping application. The optically shallow water of Karimunjawa Islands was selected as the study area. Two PlanetScope images acquired on 17 May 2017 and 15 August 2017 were tested as a control for the consistency of PlanetScope image accuracy. Several treatments were applied to both PlanetScope images including atmospheric correction, sunglint correction, Principle Component Analysis (PCA), Minimum Noise Fraction (MNF) and Linear Spectral Unmixing (LSU). Per-pixel classification algorithms (including Maximum Likelihood – ML, Support Vector Machine – SVM, and Classification Tree Analysis – CTA) and Object-based Image Analysis (OBIA) were used to perform benthic habitat and seagrass species classification. Spectra-based classifications were also applied to classify seagrass species using seagrass species spectra as input endmember. The results indicated that PlanetScope images produced 47.13–50.00% overall accuracy (OA) for benthic habitat mapping consist of five classes (coral reefs, macroalgae, seagrass, bare substratum, dead coral) and 74.03–74.31% OA for seagrass species mapping consist of five seagrass species classes. The accuracy of PlanetScope images for benthic habitat and seagrass species mapping was comparable to other high spatial resolution images. The performance of PlanetScope images was also consistent, shown by the similar accuracy obtained from May and August image. The concern regarding PlanetScope image was the low Signal-to-Noise Ratio (SNR) over homogeneous areas such as optically deep water, which led to the failure of performing sunglint correction and obtaining higher accuracy. To conclude, with the very high temporal resolution, PlanetScope image is promising for monitoring the dynamics and changes of benthic habitat and seagrass species composition, and rapid assessment of extreme events impacts, especially in coastal areas with limited accessibility.