Improving the PLUM Method with DATES: Delaunay-Based Adaptive Technique for Earthquake Early Warning

Abstract

Real-time seismic intensity estimation is a key component of earthquake early warning (EEW) systems, but it remains challenging when rapid and reliable source parameter determination is difficult. The propagation of local undamped motion (PLUM) method is a robust approach for real-time seismic intensity estimation that does not rely on source parameter determination. However, its performance is sensitive to seismic network configuration, particularly in regions with sparse or uneven station distribution because it relies on a fixed selection radius to define reference stations. To address this limitation, we develop the Delaunay-based adaptive technique for earthquake station selection (DATES) method, which adaptively selects reference stations based on local station density and azimuthal coverage. DATES incorporates Delaunay triangulation to maintain sufficient spatial coverage while avoiding the use of a fixed prediction radius. This adaptive selection framework improves the robustness of seismic intensity estimation under heterogeneous network configurations. We applied DATES to strong-motion data from the 2023 Türkiye–Syria earthquake (Mw 7.8), using 232 seismic stations with a highly inhomogeneous distribution. The method was evaluated using three criteria: coverage rate, root mean square error of seismic intensity estimation, and average lead time. Compared with the original PLUM method, DATES achieved full station coverage with a shorter selection radius, improved spatial uniformity of reference stations, and extended average lead time (12.2 s versus 10.5 s) without increasing prediction error. Spatial analysis further showed that DATES suppresses overestimation in dense networks and extends warning times in sparse regions. By integrating adaptive station selection, DATES expands the applicability of the PLUM method to a broader range of seismic networks, including those with limited station density or lower technical specifications. This improvement contributes to advancing real-time ground-motion estimation for EEW systems worldwide.