Tailored calibration of stochastic weather generators for enhanced hydrological system evaluation
Year: 2025
Type: article
Abstract: Stochastic weather generators (SWGs) are widely used in climate impact assessments for water resource management, flood protection, crop simulation, and renewable energy. However, conventional calibration methods often fail to accurately capture key climate attributes required for these assessments. This study (1) demonstrates the advantages of tailored calibration, using the Simulated Method of Moments (SMM), in better capturing important climate attributes for both historical and future climates; (2) highlights the importance of robust optimization when using SMM; and (3) shows the impact of these choices on estimating critical hydrological responses such as streamflow. The method is tested on the Abercrombie catchment in New South Wales, Australia, using the WGEN rainfall generator and GR4J rainfall-runoff model. Results show that conventional calibration underestimates historic extreme rainfall and streamflow by between 23% and 40% depending on the chosen metric, while tailored calibration with SMM eliminates these biases. A climate stress test further shows that common optimization techniques can introduce substantial errors (up to 20%) in key climate attributes, while the Robust Gauss-Newton (RGN) algorithm significantly reduces these errors and cuts uncertainty in future streamflow predictions by up to 75% for a given change in target attributes. These findings underscore the potential of tailored calibration for improving the flexibility, accuracy, and efficiency of SWG calibration.
Source: Journal of Hydrology
Institution: The University of Adelaide
Language: English
Subfield: Water Science and Technology
Field: Environmental Science
Domain: Physical Sciences
Open Access status: hybrid
Estimated APC paid: $4,030
Funder: University of Adelaide
OpenAlex ID: W7117759641