# Exploring end-to-end earthquake early warning performance in large earthquakes using the February 2023 Kahramanmaraş, Türkiye sequence

**Authors:** Savvas Marcou, Angela I. Lux, Andrei Akimov, Amy L. Williamson, Richard M. Allen

PMC · DOI: 10.1038/s41598-025-29755-z · Scientific Reports · 2025-12-12

## TL;DR

This study evaluates how well earthquake early warning systems perform during large earthquakes using data from the 2023 Kahramanmaraş earthquakes in Türkiye.

## Contribution

The paper provides a realistic simulation of warning times using real-world data and a statistical model for alert delivery latency.

## Key findings

- EPIC can generate earthquake solutions within 4 seconds for the Pazarcık event and 10 seconds for the Elbistan event.
- Despite magnitude underestimation, EPIC can provide up to 20 seconds of warning time for areas with strong shaking.
- Low alerting thresholds (MMI 3 or 4) are necessary to achieve these warning times with included alert delivery latencies.

## Abstract

Earthquake early warning systems (EEWS) aim to warn end-users of impending ground shaking. They can be most impactful in large earthquakes occurring close to large population centers, where exposure to strong ground shaking is extensive. However, such earthquakes are rare, and EEWS performance expectations remain uncertain. The February 2023 Kahramanmaraş, Türkiye sequence, including the M7.8 Pazarcık and M7.5 Elbistan events, exposed millions to strong ground shaking and produced a rich waveform dataset, offering a test case. We use this data to produce a realistic simulation of warning times. We use the EPIC point source algorithm for real-time earthquake characterization, and incorporate alert delivery latency using a statistical model driven by real-world alert delivery data from California, collected by the MyShake smartphone app. We show EPIC would produce solutions very quickly (4 s for Pazarcık, 10 s for Elbistan). Despite EPIC’s expected magnitude underestimation (peak M6.7 for Pazarcık, M7.2 for Elbistan), we show that its magnitude estimate grows large quickly enough to provide areas of MMI 6+ shaking with up to 20 s of warning time, even with alert delivery latencies included, provided that low alerting thresholds of MMI 3 or 4 are used.

## Full-text entities

- **Diseases:** rupture (MESH:D012421), alert fatigue (MESH:D000071064), EPIC (MESH:D000081042)
- **Chemicals:** BA08 (-)
- **Species:** Homo sapiens (human, species) [taxon 9606]

## Full text

_Full body text omitted from this summary view._ Fetch the complete paper as Markdown: https://tomesphere.com/paper/PMC12770601/full.md

## Figures

8 figures with captions in the complete paper: https://tomesphere.com/paper/PMC12770601/full.md

## References

14 references — full list in the complete paper: https://tomesphere.com/paper/PMC12770601/full.md

---
Source: https://tomesphere.com/paper/PMC12770601