We present a novel 28 yr dataset of Earth's black-sky surface albedo, derived from AVHRR instruments. The dataset is created using algorithms to separately derive the surface albedo for different land use areas globally. Snow, sea ice, open water and vegetation are all treated independently. The product features corrections for the atmospheric effect in satellite-observed surface radiances, a BRDF correction for the anisotropic reflectance properties of natural surfaces, and a novel topography correction of geolocation and radiometric accuracy of surface reflectance observations over mountainous areas. The dataset is based on a homogenized AVHRR radiance timeseries. The product is validated against quality-controlled in situ observations of clear-sky surface albedo at various BSRN sites around the world. Snow and ice albedo retrieval validation is given particular attention using BSRN sites over Antarctica, Greenland Climate Network stations on the Greenland Ice Sheet (GrIS), as well as sea ice albedo data from the SHEBA and Tara expeditions. The product quality is found to be comparable to other previous long-term surface albedo datasets from AVHRR.