An assessment of pan-sharpening algorithms for mapping mangrove ecosystems : a hybrid approach

Abstract

Mapping mangrove (littoral and swamps) ecosystems is challenging due to the qualitative and quantitative nature of the surrounding water and mudflats. However, accurate assessment of mangroves is required to determine carbon credits. This research study explores five pan-sharpening algorithms with the aim of determining the best algorithm to map mangrove ecosystems from very high resolution satellite images. In this research, a multidimensional evaluation was employed to pinpoint the best algorithm from among five advanced algorithms, i.e. Ehler’s transformation (ET), modified intensity hue saturation (MIHS), wavelet transformation (WT), optimized high pass filter addition (OHPFA), and subtractive resolution merge (SRM). These approaches involve the calculation of spectral root mean square error (RMSE), Sobel filter RMSE and correlation coefficient (r). OHPFA and SRM provided good results during this assessment. Object-based image analysis was incorporated to further assess the best technique between these two approaches for assessing mangrove tree canopy by calculating under and over segmentation. The SRM algorithm provides the best results with a kappa coefficient (κ) of 0.875 and an accuracy of 92.3% when compared with ground data. This research is very useful in various applications such as calculation of crown projection area using high resolution satellite images for estimation of blue carbon in mangrove trees.