Faster Model Calibration Using the Fourier Transform

We propose a new methodology for faster calibration of multiparameter models using the Fourier transform. The HBV (rainfall-runoff) model is used to demonstrate the concept. It requires precipitation, temperature and potential-evapotranspiration as input time series with a total of 10 calibration parameters. By changing each parameter, the resulting runoff series is affected in a unique way. We observed that these changes are independent of others in the frequency domain. Parameters that show combined effects with others may be calibrated together. Indirectly, we can also find out model parameters that produce unique effects and the ones that are dependent on each other. The advantage is that it is easier to do multiple one-dimensional calibrations as compared to a single high-dimensional one.