Evaluating the potential of Sentinel-2 MSI and Landsat-8 OLI data fusion for land cover mapping in Brazilian Amazon

Capturing spatial and temporal environmental dynamics in heterogeneous tropical landscapes through remote sensing data has become an important challenge in environmental monitoring studies due to cloud contamination. In Brazilian Amazon, there is a great need of data able to provide information for a wide range of decision makers. Alongside this, a wide variety of free remotely sensed data products have been made available. The combination of information from more than one sensor could maximise the number of cloud-free images as well as their temporal and spatial resolution. Based on this framework, this study investigates the potential and quality of a set of fused images obtained from the Sentinel-2 Multispectral Imager Instrument (MSI) and Landsat OLI satellites sensors over the region of Brazilian Amazon, comparing their performance according to different land cover/land use areas. In this context, the objectives of this study were: (1) assess qualitatively the two main group of image fusion approaches, namely component substitution (CS) and multiresolution analysis (MRA), as well as their suitability for the fusion of Sentinel-2 MSI and Landsat-8 OLI image pair from selected areas in Brazilian Amazon; (2) compare three different image fusion methods for measuring efficacy and performance to determine the best spatio-temporal information: the Intensity Hue Saturation (IHS) method, the Brovey transformation (BT) and the Gram-Schmidt (GS) method; (3) assess quantitatively the methods employed through a set of fusion quality metrics in order to identify the more accurate results based on different reference images.