Comparison of classification algorithms on optical satellite imagery for mapping Posidonia Oceanica meadows: The case study of Limassol, Cyprus

Abstract

Posidonia Oceanica meadows are important marine ecosystems that offer habitat to fish, organisms, and shelter for threatened species as well. This study compares classification algorithms for the validity of Posidoniaoceanica mapping through optical satellite imagery in the region of Limassol-Akrotiri bay. More specifically, side-scan sonar mapped data derived from the portal of the Department of Lands and Surveys and imported to the ArcGIS WMS service, as well as a Landsat 8 satellite image for the region of Cyprus were used. Training data and regions of interest (ROI) were created, followed by supervised classification using Spectral Angle Mapper, Mahalanobis Distance, Maximum Likelihood and Minimum Distance algorithms in ENVI 5.4 software. A sample of 1,000 random points was added to the study area before conducting a relative comparison to test the performance of the algorithms used. Since, there weren't any raw data to automate the comparison of the algorithms, a random manual selection of 30 points was considered.