An Efficient Bayesian Framework for Updating PAGER Loss Estimates
Abstract
We formulate and demonstrate an application of the Bayesian framework for incorporating ground-truth damage and loss data to update near real-time loss estimates and corresponding alerts for the U.S. Geological Survey's (USGS) Prompt Assessment of Global Earthquakes for Response (PAGER) system. PAGER's fatality and economic loss estimates are currently updated only when improved estimates of earthquake shaking become available via ShakeMap (e.g., by incorporating rupture dimensions, strong motion recordings, or Did You Feel It (DYFI) or other macroseismic reports). However, the PAGER loss models are not directly updated, despite the availability of evolving reports of location-specific or the overall physical damage or casualties. The proposed Bayesian framework allows the capacity to ingest a variety of near real-time loss data, which are typically both uncertain and incomplete. Types of near real-time data examples include the growing number of total reported deaths, those reported in a certain locality/geographic region, or the number reported in specific structures that experienced collapse. Lately, such reports come quite rapidly via media outlets, governments, and crowd-sourced platforms, though they are usually uncertain, sparse, and evolving. We provide a brief background on PAGER's current loss modeling approach and discuss the mathematical formulation necessary to update: a) the total loss distribution for a given earthquake, b) the fatality rate as a function of MMI to accommodate site-specific ground-truth observations of fatalities, and c) the fatality rate conditioned on collapse of a specific structure type. The presentation will emphasize the issues related to data quality, completeness, and authoritativeness and discuss the proposed temporal component of such updating. Despite the limitations of preliminary fatality and damage reports, especially in the immediate hours of a disaster, updating loss estimates with such observations will be a key in certain situations, allowing us to assure that the evolving loss estimates are more closely aligned with the true scope of impact, which will only belatedly be determined. Such updates will still allow the USGS to inform its end-users well within the time-frame of planning international response and aid.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2018
- Bibcode:
- 2018AGUFMNH32B..03J
- Keywords:
-
- 1240 Satellite geodesy: results;
- GEODESY AND GRAVITYDE: 4331 Disaster relief;
- NATURAL HAZARDSDE: 4335 Disaster management;
- NATURAL HAZARDSDE: 4346 Emergency response and evacuations;
- NATURAL HAZARDS