Land cover information is a fundamental parameter in a wide range of applications like urban growth, land degradation, climate change, food security, environmental sustainability, etc. In this context, remote sensing satellites can provide valuable data to allow the generation of thematic maps. On the other hand, the recent availability of hyperspectral sensors on board aircrafts and drones offers an opportunity to improve the resolution and accuracy of land cover maps. In island territories, where land is usually a scarce resource, the need of very high spatial resolution (VHR) is essential. In this context, we have generated VHR land cover maps using multispectral Worldview data and hyperspectral airborne CASI information. In particular, after corrections and pansharpening enhancements, we have analyzed pixel-based and object-based classification approaches using different input band combinations. We have compared the performance when using multispectral or hyperspectral imagery and its robustness depending on the quality of the training samples considered.