Abstract:
Synthetic data generation on two Indian rivers, Damodar and Cauveri was carried out using ARIMA (1,0,1) and fast fractional Gaussian noise model (ffGn). The quality of generated data regarding preserving the historical basic properties like mean, variance as well as long term properties like Hurst Coefficient (H) and run lengths of low flow generation as depicted by drought curve was investigated. The results for both type generation suggest that ffGn has a better capability of generating long low flow sequences than ARIMA model.
