This research aims to achieve faster real time computation for predicting the failure time from natural disasters including cyclone induced flood and bushfire. High calculation efficiencies are reached through statistically summarizing simultaneous events spread in geography, into primitives allowing a distributed updating algorithm leading to parallel computing. Geographic primitives are to be identified
through offline calculations concerning behaviors of environmental driving forces (e.g. water flow patterns for flood propagation, fuel distribution and wind for bushfire propagation, etc). Real time prediction would start with minimal available data and the prediction will be refined as more data are aggregated through broadband networks where parallel processing on the geographic primitives increases calculation efficiency.
EXPECTED OUTCOME
A Model is developed for real-time prediction of cyclone induced flood propagation by focusing on:
- Making real-time prediction faster by parallel processing on identified stochastic systems (GPs)
- Starting prediction with minimal available data and refining with any data that are subsequently available.
- Allowing flexibility in data assimilation using a modularized approach
A Bayesian framework is built to generate a probabilistic prediction combining prior knowledge, including rainfall data statistics and topographical features, with any new precipitation, where the advantages of data oriented and heuristic modeling are combined.
A generalized framework will be developed which acts as a basis for calculating mean first passage times for stochastic propagation on non-homogeneous media with environmental bias.
The framework will be enhanced to calculate mean first passage time for a random spread. This will be tested on a bushfire scenario.
FLOOD PREDICTION
Three steps of flood prediction:
Cyclone Path Prediction
Rainfall Distribution Prediction
Flood Prediction
Vedio Clips:
PUBLICATIONS
I. H. Wijesundera, M. N. Halgamuge, T. Nirmalathas and T. Nanayakkara, "A case invariant modification for mean first passage time calculation in biased networks"(under review).
This page, its contents and style, are the responsibility of the author and do not necessarily represent the views, policies or opinions of The University of Melbourne.
Created: August 2013 Last Modified: August 2013 HTML by: Malka N. Halgamuge Maintained by: Malka N. Halgamuge