The Use of SPOT 6 and RapidEye Imageries for Mangrove Mapping in the Kembung River, Bengkalis Island, Indonesia

Abstract

High-resolution satellite imageries believed to map the mangrove area well. Two satellite imageries with almost similar spatial resolution were expected to achieve this objective, including SPOT (Satellite Pour l'Observation de la Terre) 6 and a RapidEye. Two images covered the same area and acquired under the equivalent season at Kembung River, Bengkalis Island, Riau Province. A comparison was tested on several objectives. First, a comparison was conducted band by band through examining their statistical spectral analysis. RapidEye had a higher variance in all visible bands but low at NIR. Second, spectral separability using Jeffrey-Matusita distance was executed with three-band composites (SPOT 6/VNIR; RAPIDEYE/VNIR and RAPIDEYE/VNIR+RedEdge). SPOT 6 had better separability than RapidEye imageries compositions (VNIR+RedEdge and VNIR), with an overall accuracy of 79.9%, 75.3%, and 67.3%, respectively. Third, mangrove land cover classification. Three pixel-based classification rules were used, including a general classifier, maximum likelihood (ML), and two advanced classifiers, i.e., neural network (NN) and support vector machine (SVM). Results indicated that the use of SPOT 6 produced better overall accuracy than the other two image composition for each classifier. SVM was promising in the mangrove mapping algorithm compare to NN and ML.