Getting reliable and up-to-date land cover information is essential to model the Earth system.In this respect

the observation capabilities of medium spatial resolution sensors like MODIS or MERIS are offering new land cover mapping possibilities because of the unprecedented resolutions of these sensors.This paper illustrates how the combination of the spatial（300m pixel size）

spectral（15 narrow bands in the visible and NIR）

and temporal（revisit time 2-3 days） dimensions of MERIS can be used to retrieve sub-pixel fractional land cover composition over heterogeneous areas.Three MERIS FR Level 1b scenes acquired over The Netherlands in April

July and August 2003 were used to derive fractional composition of the main land cover types present in The Netherlands.A linear spectral unmixing with an optimized number of endmembers per pixel was applied both in a mono-and in a multi-temporal way.A morphological eccentricity index（MEI） was used to explore the MERIS spatial dimension and

subsequently

to support the selection of the endmembers.The Dutch land use database（LGN5）

which has a grid structure with a pixel size of 25m

was used as a reference in this study.Classification accuracy was assessed both at per-pixel and at sub-pixel level because of the availability of this high resolution reference data.The best classification results were obtained for the combined image of April and July with a classification accuracy of about 58%.In general

sub-pixel and per-pixel classification accuracies were found to be similar.Spectral confusion was detected for several classes and dates indicating that the phenological status plays an important role in choosing the optimal combination of acquisition dates and the land cover classes that can be properly identified.