Mapping zonations of mangrove species (ZMS) is important when assessing the functioning of such specific ecosystems. However, the reproducibility of remote sensing methods for discriminating and mapping mangrove habitats is often overstated due to the lack of temporal observations. Here, we investigated the potential use of temporal series of high-resolution multispectral satellite images to discriminate and map four typical Asian ZMS. This study was based on the analysis of eight images acquired between 2001 and 2014 over the mangrove area of Nusa Lembongan, Bali, Indonesia. Variations between years in the top-of-atmosphere reflectance signatures were examined as functions of the acquisition angles. We also applied maximum likelihood supervised classification to all of the images and determined the variability in the classification errors. We found that the distinction between spectral signatures of ZMS characterized by a close canopy was fairly independent of the season and sensor characteristics. By contrast, the variability in the multispectral signatures of ZMS with open canopies and associated classification errors could be attributed to variability in ground surface scattering. In both cases, sun-viewing geometry could alter the separability between ZMS classes in near-nadir viewing or frontward sun-viewing configurations, thereby explaining why the overall accuracy of ZMS classification might vary from 65% to 80%. Thus, multitemporal analysis is an important stage in the development of robust methods for ZMS mapping. It must be supported by physical-based research aiming to quantify the influences of canopy structure, species composition, ground surface properties, and viewing geometry parameters on ZMS multispectral signatures.