Predicting earthquake fatalities in Nepal

Abstract

The Nepal earthquake on 25 April 2015 resulted in over 8,000 deaths. The objective of this study was to develop a model useful in predicting fatalities following an earthquake in Nepal. The data for this study were obtained from the Humanitarian Data Exchange. This included MMI (a measure of the ground shaking resulting from an earthquake), population data and house construction data. Since most data were available only at the district level, we developed models for earthquake deaths at the district level. Zero-inflated negative binomial models were used to account for excess zeros and overdispersion in the fatality data. MMI was the strongest predictor of deaths and included in all models. Several models were developed and compared. The best models included primarily roof construction variables as predictors in addition to MMI. In contrast, a model including only wall construction variables performed poorly. For the best model, the Pearson correlation between predicted and observed deaths is 0.89.