Low spatial resolution of the NOAA-AVHRR images causes that observation footprints of the pixels can overlay the surface of more than one land cover type. The pure signal can be obtained for pixels covering only one land cover class. The extraction of the vegetation index (e.g. NDVI) for one land cover class can be interfered by the presence of other classes within the surface covered by a pixel. Additionally, the inaccuracy of the geometric correction of satellite images can increase the possibility that analysed pixel covers different land cover type than could be expected based on the analysis of the land cover map overlaid on the satellite image. In this study, we presented a new agriculture mask for Poland developed from the CORINE Land Cover 2006 database. The mask of one-kilometre spatial resolution indicates the pixels of the NOAA-AVHRR, which should be used for calculation of mean vegetation indices for regions (i.e. voivodeships or provinces). The proposed mask preserves the uniform spatial distribution of pixels within each Polish region. To validate the new mask, we calculated twelve-year-long time series (1997-2008) of Vegetation Condition Index and Temperature Condition Index of agriculture areas for each voivodeship in Poland. The newly received time series of voivodeships showed higher correlation with crop yield than when using the classical agriculture mask, which classifies a pixel as agricultural if at least 50% of its area is covered by the agriculture land.