1: \begin{abstract}

2: The curve time series framework provides a convenient vehicle to

3: accommodate some nonstationary features into a stationary setup. We

4: propose a new method to identify the dimensionality of curve time

5: series based on the dynamical dependence across different curves. The

6: practical implementation of our method boils down to an eigenanalysis

7: of a finite-dimensional matrix. Furthermore, the determination of the

8: dimensionality is equivalent to the identification of the nonzero

9: eigenvalues of the matrix, which we carry out in terms of some

10: bootstrap tests. Asymptotic properties of the proposed method are

11: investigated. In particular, our estimators for zero-eigenvalues enjoy

12: the fast convergence rate $n$ while the estimators for nonzero

13: eigenvalues converge at the standard $\sqrt n$-rate. The proposed

14: methodology is illustrated with both simulated and real data sets.

15: \end{abstract}