Abstract. Land-surface albedo plays a critical role in the earth's radiant energy budget studies. Satellite remote sensing provides an effective approach to acquire regional and global albedo observations. Owing to cloud coverage, seasonal snow and sensor malfunctions, spatiotemporally continuous albedo datasets are often inaccessible. The Global LAnd Surface Satellite (GLASS) project aims at providing a suite of key land surface parameter datasets with high temporal resolution and high accuracy for a global change study. The GLASS preliminary albedo datasets are global daily land-surface albedo generated by an angular bin algorithm (Qu et al., 2013). Like other products, the GLASS preliminary albedo datasets are affected by large areas of missing data; beside, sharp fluctuations exist in the time series of the GLASS preliminary albedo due to data noise and algorithm uncertainties. Based on the Bayesian theory, a statistics-based temporal filter (STF) algorithm is proposed in this paper to fill data gaps, smooth albedo time series, and generate the GLASS final albedo product. The results of the STF algorithm are smooth and gapless albedo time series, with uncertainty estimations. The performance of the STF method was tested on one tile (H25V05) and three ground stations. Results show that the STF method has greatly improved the integrity and smoothness of the GLASS final albedo product. Seasonal trends in albedo are well depicted by the GLASS final albedo product. Compared with MODerate resolution Imaging Spectroradiometer (MODIS) product, the GLASS final albedo product has a higher temporal resolution and more competence in capturing the surface albedo variations. It is recommended that the quality flag should be always checked before using the GLASS final albedo product.