Predicting the Unpredictable - Harder than Expected

An earthquake is a hazard that may cause urgent needs requiring international assistance. To ensure rapid funding for such needs-based humanitarian assistance, swift decisions are needed. However, data to guide needs-based funding decisions are often missing in the acute phase, causing delays. Instead, it may be feasible to use data building on existing indexes that capture hazard and vulnerability information to serve as a rapid tool to prioritize funding according to the scale of needs: needs-based funding. However, to date, it is not known to what extent the indicators in the indexes can predict the scale of disaster needs. The aim of this study was to identify predictors for the scale of disaster needs after earthquakes.

The predictive performance of vulnerability indicators and outcome indicators of four commonly used disaster risk and severity indexes were assessed, both individually and in different combinations, using linear regression. The number of people who reportedly died or who were affected was used as an outcome variable for the scale of needs, using data from the Emergency Events Database (EM-DAT) provided by the Centre for Research on the Epidemiology of Disasters at the Université Catholique de Louvain (CRED; Brussels, Belgium) from 2007 through 2016. Root mean square error (RMSE) was used as the performance measure.

The assessed indicators did not predict the scale of needs. This attempt to create a multivariable model that included the indicators with the lowest RMSE did not result in any substantially improved performance.

None of the indicators, nor any combination of the indicators, used in the four assessed indexes were able to predict the scale of needs in the assessed earthquakes with any precision.