An efficient sampling method for fast and accurate Monte Carlo Simulations^*

* This paper was presented at the HRWS conference in Hobart, December 2015.

Abstract

Currently, there are two published Monte Carlo Simulations (MCS) methods for estimating design floods in Australia: The Cooperative Research Centre – Catchment Hydrology (CRC-CH) method and the Total Probability Theorem (TPT) method. The CRC-CH method uses variable storm durations, which makes it an attractive option compared with the TPT method. However, the CRC-CH method suffers from a large variance in design flood estimates for large to extreme events, i.e. events with low annual exceedance probabilities (AEP). This means two successive MCS runs may provide significantly different design flows, which makes the method unreliable for large to extreme events. Decreasing this variance to an acceptable level may require millions of hydrological model simulations. We introduced an alternative sampling technique, known as ‘importance sampling’, in the CRC-CH method to address this shortcoming. In a pilot study we demonstrate that this results in a significantly reduced variance in design flood estimates for large to extreme events. At the same time, computation times are significantly shortened. This creates opportunities for greater use of the CRC-CH method and similar Monte Carlo methods in applications where they were previously considered infeasible, such as dam design.

Keywords:

• Monte Carlo
• importance sampling
• flood risk

Acknowledgements

The authors want to express their gratitude for the valuable comments on the work described in this paper by representatives of the Queensland Government, by members of the Technical Working Group and by members of the Independent Panel of Experts, all of whom are involved in the Brisbane River Catchment Flood Study.

Notes

* This paper was presented at the HRWS conference in Hobart, December 2015.