Improved understanding of land-use impacts on terrestrial carbon is a necessary part of both Irish and international efforts to mitigate climate change. Grasslands represent approximately 90% of agricultural land in Ireland and as such, any policy decisions promoting land-use change (e.g. afforestation, energy crop planting) will result in large scale grassland conversion. Within the framework of the Irish Land Mapping Observatory (ILMO) project, we present a methodology for a hierarchical land-cover classification scheme using optical remote sensing data. Although frequent cloud cover and atmospheric disturbances can strongly impede the application of optical data, the use of sensors with a high-temporal resolution (e.g. MODIS) in combination with time-series filtering can decrease the effects of cloud-contamination. In this study, state-of-the-art machine learning methods Random Forest and Support Vector Machine are used for the discrimination of different land-cover types in the Irish landscape using an 11-year MODIS 16-day composite time-series (MOD13Q1). Aside from general land-cover classes, the work is focussed on the distinction of different grassland management types. Two counties in the midlands and the North West of the island served as the primary study sites, where a detailed accuracy assessment of the classification proves the applicability of the chosen classification methods for data in homogenous areas. Nonetheless, the spatial resolution (250m) in combination with the highly fragmented Irish landscape necessitates an enhanced sub-pixel analysis. For this purpose the usage of classification probability values shows promising results. It is envisaged that an application of the developed methodology for the entire country could help to contribute to an improved estimation of greenhouse gas (GHG) emissions in the scope of different national and international requirements.