Accurate vegetation mapping is essential to enhance understanding of important ecosystem processes such as fire behaviour, nutrient cycling, plant community composition and mega-herbivore population dynamics in African savannahs.The inherent heterogeneity of savannah landscapes however creates significant challenges for accurate discrimination of vegetation components. Recent studies based on the use of optical remote sensing data favour very high spatial resolution (VHR) multi-spectral imagery for dealing with this challenge. However, such data are costly for use in operational management. Planet Labs, through Norway’s International Climate and Forests Initiative (NICFI) in partnership with Kongsberger Satellite Services (KSAT) and Airbus, now grant free access to high-resolution, analysis-ready mosaics of Planet imagery over the tropics, with great potential for fine-scale vegetation mapping. However, the spectral characteristics of these data are limited, with little knowledge of their ability to resolve the spectral similarity of heterogeneous savannah vegetation components. In parallel, Sentinel-2 samples a relatively high number of spectral bands, but has relatively coarse spatial resolution for dealing with the high spatial heterogeneity in African savannah landscapes. We test the hypothesis that fusing Sentinel-2 with Planet imagery leverages their spectral and spatial advantages to enhance accurate discrimination of savannah vegetation types. To achieve this, Principal Component and Gram-Schmidt transformations were compared for image fusion via pan-sharpening, where the Gram-Schmidt approach proved superior. Using this approach, we fused Sentinel-2 and Planet images and compared the three datasets (i.e. Sentinel-2, Planet and Fused images) augmented with two spectral indices and three Haralick texture features in a multi-layer perceptron neural network classification within a test site in the Lower Sabie region of Kruger National Park, South Africa. Overall, the Fused image achieved the best and most precise classification accuracy metrics (weighted F-score: 0.87±0.012) compared to Sentinel-2 (weighted F-score: 0.85±0.034) and Planet imagery (weighted F-score: 0.85±0.017). A comparison of classifications showed loss in spatial detail when using Sentinel-2 (at 10 meters spatial resolution), yet similar thematic details for vegetation classes across all three datasets. Our findings highlight the utility of Sentinel-2 and the Planet-NICFI mosaics in a heterogeneous savannah landscape, while setting a foundation for cost-effective and accurate high spatial resolution monitoring of savannah ecosystem.