The Potential of Crowdsourced Data for the Rapid Impact Assessment of Large Earthquakes: The 2023 M 7.8 Kahramanmara ş-Pazarc ı k, Türkiye, Earthquake

Reliable and rapid impact assessment for large earthquakes is a challenge because it is difficult to rapidly determine the fault geometry and thus the spatial distribution of shaking intensities. In this retrospective study of the M 7.8 Kahramanmara & scedil;- Pazarc & imath; k, T & uuml;rkiye, earthquake, we evaluate how eyewitness observations crowdsourced through the LastQuake system can improve such assessments. These data consist of felt reports describing the local level of shaking or damage and manually validated geolocated imagery. In the first part of this study, the methods used to derive macroseismic intensity values from felt reports, particularly for high values, are validated by comparison with independently determined intensities. This comparison confirms that the maximum intensity that can be derived from felt reports does not generally exceed VIII. A fatality estimate of 3000 could be made within a few hours by evaluating the number of people exposed to high intensities using the felt reports and assuming a point source. However, this estimate was known to be an underestimate because of the point-source approximation; this underestimate was also confirmed by the geolocated imagery showing high levels of damage at epicentral distances well beyond those predicted by circular isoseismals. However, improved estimates could have been derived from the event 's ShakeMaps using the U.S. Geological Survey Prompt Assessment of Global Earthquakes for Response (PAGER) fatality loss-modeling system, either by incorporating the felt reports into the ShakeMaps computation or using, in addition, a finite-source (here line-source) model derived from the felt reports using the Finite-fault rupture Detector software. The inclusion of fault geometry would have resulted in a fatality estimate with data collected within 10 min of the origin determination, which was consistent with the final PAGER alert level and the reported death toll that were both only known days later. Although more work would be helpful to assess the reliability of the derived fault geometry, in regions where they are collected in large numbers, felt reports collected within 10 min of the earthquake can be used to substantially improve current fatality estimates.