IDENTIFICATION OF LINEAR HYDROLOGIC SYSTEMS' RESPONSE FROM SHORT RECORDS IN THE PRESENCE OF NOISE.

In identifying the impulse response function of a linear hydrologic system on the basis of a short input-output record one often finds that the results are very sensitive to minor errors in the data. In particular, low-amplitude random errors in the input-output record tend to cause severe oscillations in the response function, thereby making it difficult to obtain a physically realizable solution by conventional methods. One way to obtain a physically meaningful impulse response function from a relatively short hydrologic record is to use parametric programming. This makes it possible to generate a continuous set of eventual alternative solutions to the deconvolution problem, together with a bicriterion function representing these alternatives. Refs.