A study about how to effectively achieve semi-structured spatio-temporal data storage and query in HBase was carried out. The formal description of the problem was issued; the HBase semi-structured spatio-temporal storage model was proposed by using a semi-structured approach TwigStack. On this basis, semi-structured spatio-temporal range query and kNN queries were carried out. An experiment was made in terms of real data and a comparison was made with MongoDB which needs higher hardware configuration, the results show that the performance of semistructured spatiotemporal query algorithm is similar to MongoDB in the machine with general configuration, so that it has the advantage in the practical application.