Comparison of sentinel-2 and multitemporal sentinel-1 sar imagery for mapping aquaculture pond distribution in the coastal region of brebes regency, central java, Indonesia

The identification of land cover and land use is necessary to support the strategic management of coastal areas. The utilization of remote sensing technology such as synthetic aperture radar (SAR) data has been widely used for mapping the distribution of land cover and land use. This application includes the detection of aquaculture ponds in coastal areas due to SAR’s sensitivity to surface water content. In addition, multitemporal Sentinel-1 data helps to distinguish between ponds and rice fields that possess a visually similar appearance during the flooding stage. This study aims to explore the accuracy of the gray level of co-occurrence model (GCLM) textures of multitemporal Sentinel-1 data for aquaculture pond mapping in Brebes Regency, Central Java Province, Indonesia. In addition, single-date Sentinel-2 optical imagery was used to compare the results from Sentinel-1 data. The Sentinel-2 data has been identified using supervised classifications, e.g., maximum likelihood (ML), minimum distance (MD), random forest (RF), and K-nearest neighbor (KNN) algorithms, and the most accurate algorithm was selected to classify the Sentinel-1 data using GLCM textures. The results indicated that the Sentinel-1 imagery showed the best results using GLCM metrics from VH polarization with an accuracy value of 92.2 using the ML algorithm, while the best results from Sentinel-2 were also produced using ML, with an 88.4 overall accuracy. These results demonstrate that multitemporal Sentinel-1 data have higher accuracy than Sentinel-2 data when used for pond detection. This shows the potential of the combination of both sensors to increase the accuracy of aquaculture pond mapping. © 2021, Asociatia Geographia Technica. All rights reserved.