Learning to deal with uncertainty in Flood and Tsunami Modelling

Sullivans' creek

Dr Stephen Roberts will be presenting a special lecture for MSI's Alumni, friends and colleagues. The presentation will be followed by drinks, hors-d'oeuvres and a chance to meet and chat with the speaker and other members of the MSI.

Attendance at these special lecturers is by invitation only, so if you would like to attend, please register as a friend of MSI to recieve your invitation.


To model floods and tsunamis we have sophisticated computational models which can be used to accurately predict the outcome these complex physical processes. But the length of time and computer resources to obtain the solution of these models can be very large (hours of computer time on supercomputers). In addition we may need to adjust a number of parameters (controlling release of water from dams, or flood barriers, or blockage of structures such as culverts) or work with input data that has uncertainty (elevation and rainfall data might be inaccurate). 

To understand how the model behaves to changes in parameters and uncertainty in the data, we are forced to make many runs of our expensive computer model, which seems to remove any chance of real time predictions of model outputs for varying parameter values (in particular if we want to incorporate uncertainty). 

So this is the challenge, to provide an online system which can provide predictions along with quantification of uncertainty, in close to real time. Necessarily, this will incorporate an offline process involving the pre-computation of a number of expensive simulations. 

In this talk I will outline our progress in first developing an accurate computational models for flood and tsunami modelling, giving some demonstrations of 2011 Tohoku tsunami and some 1 in a 100 year flooding results for our local Sullivans' creek. These results are obtained using our open source package called ANUGA. From there I will present some of our recent progress in using multidimensional approximation and multi-fidelity approximation, via sparse grids and kringing, to quantify uncertainty for a tsunami modelling problem.

About the speaker

Stephen Roberts is coordinator of the Computational Mathematics research group in the Mathematical Sciences Institute (ANU). His research area is the application of efficient and robust numerical methods for the solution of partial differential equations. In particular he is the lead developer at the ANU of the ANUGA hydrodynamic modeling software. ANUGA is an open source computational tool that models the impact of dam breaks, floods and tsunamis on communities. It is used extensively by councils, governments and consultant engineers.

Lecture slides

Learning to deal with uncertainty in Flood and Tsunami Modelling (PDF, 38.2MB)