“…where a i is the (N + 1)-dimensional CSP basis column vector and h i =b i • g(z) is the related amplitude, which is produced using the dual basis row vector b i (b i • a j = δ i j ) [25,27,47]. Interested in leading order accuracy, the CSP basis vectors a i , b i can be approximated by the right (α i ) and left (β i ), respectively, eigenvectors of the Jacobian J of g(z) [31,34,[48][49][50]. Apart from an amplitude, each CSP mode is described by a timescale which is defined as the inverse norm of the associated eigenvalue, i.e., τ i = |λ i | −1 , where [51][52][53].…”