Scientists Create New Tool to Predict Disaster Losses
Bayesian network approach enables emergency planners to better predict, assess and manage natural and man-made catastrophes.
A group of engineering and scientific experts have developed a new model to better predict losses due to natural and man-made environmental disasters. Researchers say their approach has the potential to assist emergency planners and other disaster preparedness experts reduce negative impacts through improved prediction.
Lead researcher Lianfa Li of LREIS, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, said the model "has implications for risk assessment and management in supporting more precise information for decision-making." The approach, while geared in this instance on flooding events, "has broader applications to typhoons, hurricanes, landslides, tsunamis and other adverse situations with natural and man-made causes," Li said.
The new analysis method is detailed in the http://www3.interscience.wiley.com/cgi-bin/fulltext/123452819/HTMLSTART article "Assessment of Catastrophic Risk Using a Bayesian Network Constructed from Domain Knowledge and Spatial Data" in the July issue of the journal Risk Analysis, published by the Society for Risk Analysis. The authors include Lianfa Li, Jinfeng Wang, and Chengsheng Jiang of the Chinese Academy of Sciences and Hareton Leung of Hong Kong Polytechnic University.
The researchers say their approach is unique in that it integrates expert input, geographic data, and a host of contributing factors in predicting the likelihood of certain adverse outcomes, allowing emergency planners to pre-position resources and prepare staff based on more information than is provided by reviewing similar events that have occurred in the past.
The model in effect operates as a type of “artificial intelligence,” according to Li. The supporting computer program “can learn from existing data and users can leverage expert knowledge to revise and improve the model, make inferences about missing data, and bridge other uncertainties to enhance the predictability of natural disasters and decrease potential losses.” Li added, “Our study proposes a generic modeling framework that integrates relevant quantitative and qualitative factors within a consistent system for assessment of catastrophic risks.”
The so-called Bayesian network approach for disaster prediction makes use of information from geographers, construction engineers, ecologists and economists. It was validated against data from flood disasters along the Heihe river in northwest China from 2006 to 2008, which indicated its relatively better performance than other available known methods.