Modeling the Rainfall-Runoff Process with a Bilinear State-Space Model
Proceedings of the 1991 European Simulation Conference
A fairly simple iterative algorithm for identifying mul-tivariable bilinear systems has been introduced, which uses input/output data directly. This new algorithm falls in the class of subspace methods and resembles well known Hankel based realization algorithms that identify a bilinear state space model on the basis of the Volterra kernels instead. Classical rainfall-runoff modeling has been limited to low order Volterra kernels due to the computational complexity involved in the identification. With the new algorithm this is no longer necessary since the parameters of the bilinear model are directly related to the Volterra kernels. The iterative subspace algorithm has been tested with a set of data from the Voer catchment in Belgium. The results indicate a better performance compared to a linear subspace algorithm, especially in tracking the base flow recession, which is where the nonlinearities are concentrtated.
Ramos, Jose A.; Moonen, Marc; and Mallants, Dirk, "Modeling the Rainfall-Runoff Process with a Bilinear State-Space Model" (1991). CEC Faculty Articles. 384.