How does PAGER work?
In general, the shaking-related impact of an earthquake is controlled by the distribution and severity of shaking, the population exposed to each shaking intensity level, and how vulnerable that population is to building damage at each intensity level. Population vulnerability is dominated by where they are at the time of the earthquake and the degree of seismic resistance of the local building stock. The PAGER system takes all these factors into account.
At the heart of PAGER are the timely and accurate earthquake location and magnitude determinations that the USGS has been producing for decades. PAGER uses these earthquake parameters to calculate estimates of ground shaking by using the methodology and software developed for ShakeMap. The ShakeMap system produces maps of regional ground shaking using a site-specific ground-motion amplification map, seismic wave attenuation equations, and reported or recorded intensities and ground motions. The number of people exposed to each shaking intensity level is then calculated by combining the maps of estimated ground shaking with a comprehensive worldwide population database (Landscan, from the Department of Energy's Oak Ridge National Laboratory).
Next, based on the population exposed to each intensity level of shaking, the PAGER system estimates total losses based on country-specific models developed from economic and casualty data collected from past earthquakes. To calibrate the loss models, the USGS has generated an Atlas of approximately 6,000 ShakeMaps for significant global earthquakes that have occurred during the last 40 years. The calibration of loss methodologies relies on this Atlas and on fatality and damage data collected by the NEIC.
In order to estimate the empirical fatality rate for countries with few or no fatality data, Jaiswal and others (2009) proposed aggregation of fatal events from like-countries at a regional level through a scheme that focuses on likely indicators of comparable country vulnerability. Using this model, PAGER can estimate total fatalities for future earthquakes within an average factor of five to ten, with higher accuracy for more fatal events. The error estimated in hindcasting the total shaking deaths using the empirical model already incorporates i) the total variability that comes from the uncertainty in shaking hazard for each earthquake, ii) the uncertainty in the population exposure, and iii) possible errors in the number of recorded deaths in the catalog for these events.