Abstract. We consider the problem of estimating hybrid frequency moments of two dimensional data streams. In this model, data is viewed to be organized in a matrix form ( , ) 1≤ , ,≤ . The entries , are updated coordinate-wise, in arbitrary order and possibly multiple times. The updates include both increments and decrements to the current value of , . The hybrid frequency moment , ( ) is defined asand is a generalization of the frequency moment of one-dimensional data streams. We present an˜ (1) space 1 algorithm for the problem of estimating , for ∈ [0, 2] and ∈ [0, 1]. We also present a˜ ( 1−1/ ) space algorithm for estimating , for ∈ [0, 2] and ∈ (1, 2].