The temporal aggregation in temporal databases is an extension of the conventional aggregation including the time as a range condition of the aggregation. In this paper, we propose a new tree based structure for the temporal aggregation, called the CTA-tree. In the CTA-tree, we transform the time interval of a temporal data record into a value, called the T-value, using the Corner transformation and column-scan transformation. Then, we index the data records according to their T-values. Through analyses and experiments, we evaluate and compare the performance of the proposed method with the conventional method, and show the efficiency of the proposed method for the update-intensive applications.