Prediction of casualties and damages of infrastructure caused by cyclone using neural network

Abstract

Natural disaster, especially cyclone, is a perennial problem for Bangladesh. Every year cyclones cause colossal damage to the people and economy of the country. With the increase of environmental deb'fadation, it is expected to increase in the future. These natural disasters can not be prevented from occurring. Rather, with proper warning and precautionary measures, the casualties and damages caused by the cyclones can be reduced greatly. Information about path and strength of the cyclone can be obtained from meteorological departments. Using this information, the casualties and damages caused by cyclone can be assessed through forecasting models. By assessing the damages, the relief and rehabilitation measures can be planned judiciously. In this study several models have been developed to forecast the damages caused by cyclones 'Neural Network' technique has been found to be most applicable in this case. Data has been collected from various sources such as CARE Bangladesh Ltd., Cyclone Preparedness Program of Bangladesh Red Crescent Society and Bangladesh Bureau of Statistics. The variables considered in the modelling are wind speed of cyclone, height of water surge, distance from the path of cyclone, population, amount of livestock, numbers of permanent and temporary houses and length of paved and unpaved roads. Back-propagation technique has been used for calibration purpose. The modelling approach has been found to be very successful. The model predictions are highly convergent with the observed results. It has been observed that the height of water surge has the most detrimental elTect on the casualties. The neural network technique has been extended further to analyze the effect of noise in data using 'Monte-Carlo Simulation'. It is observed that the prediction remains fairly consistent for noise upto twenty percent above which the predictions become quite inconsistent. The models developed in this study has a wide range of applications in disaster management planning, infrastructure planning and, relief and rehabilitation planning. The approach can be extended further to include other variables related to cyclone.