The Tibetan Plateau has the largest inland lakes, both in terms of number and size, at the highest elevation on Earth. In the background of global warming, the spatio-temporal processes of these lakes become a hot topic for researches on climate and environment changes in the plateau. However, for lack of dataset recording continuous spatio-temporal lake dynamics, past studies did not clearly characterize the inter-annual and annual lake variation. This study introduces and presents a lake water surface dataset (8-day at 250 m spatial resolution) for the Tibetan Plateau (Qinghai province and Tibet autonomous region) during 2000 - 2012, derived from the MODIS (MOD09Q1) products using the multistage segmentation and lake-by-lake automatic threshold determining extraction method. The dataset has an accuracy rate of 93.98%. Cross comparison with other similar products and research results demonstrates an accuracy rate higher than 86%. The dataset accurately reflects the inter-annual and annual variation characteristics of the lake water surface in the plateau over the past 10 years. The dataset can be used to study the dynamic changes of the lake water in the plateau, or as an input for the land surface process model or the regional climate model to study the role of lake water dynamics in the regional hydrological and climate system.