Scale-dependent synthetic streamflow generation using a continuous wavelet transform

- A new synthetic data generation method is developed using Morlet wavelet transform.
- Reconstruction is based on random permutation of categorized wavelet coefficients.
- The scale-controlled method can be performed for individual or multiple timescales.
- The application on daily streamflow series is illustrated for the Pearl River basin, China.

SummaryThis study develops a scale-dependent synthetic data generation method for streamflow by using a continuous wavelet transform. The detailed information of streamflow variability across different timescales embedded in the data is obtained from the continuous wavelet transform. To take into account the time-dependent flow magnitudes, the wavelet coefficients are simply separated into two basic categories, namely high-flow part and low-flow part. The data reconstruction is based on the random permutation of the separated wavelet coefficients for the two categories. The synthetic generation is performed at both the individual timescales and the multiple timescales. The Morlet wavelet transform is considered as a representative continuous wavelet transform, and generation of daily streamflow data is attempted. The method is applied to a streamflow series observed in the Pearl River basin in South China. The results indicate that the proposed method: (1) is suitable for scale-controlled generation of streamflow time series and (2) provides reliable information as to the extent of spectral properties present in the original data that need to be preserved.